SFe 4 Spec Test
Test banks, MIDI files and reference renders for the SFe 4.0 specification. Load a bank in the synthesizer under test, play its MIDI file, and compare the result with the pass criterion of each test and with the reference render. The reference is rendered by spessasynth_core at 48 kHz.
Unless a group says otherwise, every test starts on a whole second, with a short click on channel 16, a marker that holds its id and a GS Reset. Its own notes start half a second later.
Download everything (sfe4-spec-test.zip)sfe4-10-generators Generators
The SFe generators and the changed SoundFont 2 rules, one preset per test. The bank has no default modulators, so velocity, volume and the pitch wheel do nothing unless a test says so.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| GEN-01 | 0:00 | 7.1 | initialAttenuation in true centibels Seven notes of the sine with initialAttenuation 0, 20, 40, 60, 80, 100 and 120. | Each note is 2 dB quieter than the one before, 12 dB from the first to the last. SoundFont 2's factor of 0.4 would give 0.8 dB steps. |
| GEN-02 | 0:09 | 7.1 | Modulator amounts to initialAttenuation are centibels The reference tone, then the sine with a constant modulator (No Controller to initialAttenuation) with the amount 480. | The second note is 48 dB quieter than the first: the amount of a modulator is in true centibels too. |
| GEN-03 | 0:13 | 7.1, 7.1.1 #86 | The attenuation reaches 1440 and clamps there The reference tone. Then initialAttenuation 1440 with gain 1440. Then the same with a constant modulator that adds 500 to initialAttenuation. | All three notes have the same level. 1440 cB of attenuation are allowed, and 1940 is clamped to 1440 before the gain is subtracted. |
| GEN-04 | 0:18 | 7.1.1 #86 | gain Five notes of the sine with gain -120, -60, 0, 60 and 120. | Each note is 6 dB louder than the one before. The third note has no gain, and the last two play above the level of the sample. |
| GEN-05 | 0:25 | 7.1.1 #86 | gain cancels initialAttenuation and sums across levels The reference tone. Then gain 100 with initialAttenuation 100. Then an instrument gain of 60 with a preset gain of 60. | The second note has the level of the reference tone. The third is 12 dB above it. |
| GEN-06 | 0:30 | 7.1 | Generator sums saturate The reference tone. Then an instrument gain of 20000 plus a preset gain of 20000, with initialAttenuation 1440. Both gains are stored beyond their range. | Both notes have the same level: the sum saturates at 32767 and the gain is clamped to 1440. A sum that wraps to a negative number is silent. |
| GEN-07 | 0:34 | 7.1 | The limits apply once, after the modulators The reference tone. Then a stored gain of 3000, beyond its range, with a constant modulator of -3000 to gain. | Both notes have the same level: 3000 plus -3000 is 0. Clamping the stored value to 1440 before the modulator would give -1560, clamped to -1440, which is silent. |
| GEN-08 | 0:38 | 7.1 | Preset offsets of SFe generators Two pairs. The sine on a lowpass cutoff of 440 Hz (6900) with an instrument filterQOffset of 50 plus a preset offset of 50, then with an instrument filterQOffset of 100. Then modLfoToAmplitude 250 plus a preset offset of 250, then 500 alone. | The two notes of each pair sound the same: preset values add to the instrument values for SFe generators too. |
| GEN-09 | 0:44 | 8.1 | Every SFe generator at its default A sine zone with a lowpass at 2 kHz, a vibrato of the Modulation LFO and a decaying Volume Envelope. Then the same zone with all 29 SFe generators written at their defaults. | The two notes sound the same: an SFe generator at its default leaves the SoundFont 2 voice alone. |
| GEN-10 | 0:48 | 7.1.1 #61 | The five filter types The noise unfiltered, then through the lowpass, highpass, bandpass, peak and notch filters (filterType 0 to 4), all at 1480 Hz (9000) with initialFilterQ 100. | In a spectrum: the lowpass removes the highs, the highpass removes the lows, the bandpass keeps a band around 1480 Hz, the peak raises that band by about 13 dB above the rest, and the notch cuts a narrow hole there. Away from 1480 Hz the peak and the notch sit 5 dB below the unfiltered noise. |
| GEN-11 | 0:56 | 7.1.1 #61 | The Q of a resonance of 0 The reference tone. Then the sine on the cutoff (6900, 440 Hz) with initialFilterQ 0, through the lowpass and then the highpass. | Both filtered notes are 3 dB below the reference tone: a resonance of 0 is a Butterworth response, with no peak. |
| GEN-12 | 1:01 | 7.1.1 #61 | The DC gain of the resonance The sine at key 33 (55 Hz), far below a lowpass cutoff of 13500, with initialFilterQ 0, 100, 200 and 400. | The notes are 0, 5, 10 and 20 dB below the first: the level drops by half of initialFilterQ, q / 20 dB. |
| GEN-13 | 1:07 | 7.1.1 #61 | The gain of the bandpass and the peak at the cutoff The reference tone. Then the sine on the cutoff (6900) with initialFilterQ 100, through the bandpass and then the peak filter. | The bandpass note is 2 dB above the reference tone (Q * 10^(-r/400)), the peak note 8 dB above it (2 * Q * 10^(-r/400)). |
| GEN-14 | 1:12 | 7.1.1 #61 | The far ends of the peak and the notch The sine at key 33 (55 Hz) without a filter. Then the same note through the peak and the notch filter at 9000 (1480 Hz) with initialFilterQ 200. Then the reference tone, and the sine at key 69 on the cutoff of a notch at 6900. | The peak and notch notes at key 33 are both 10 dB below the unfiltered one. The last note, on the notch, is nearly silent next to the reference tone before it. |
| GEN-15 | 1:19 | 7.1.1 #61 | A highpass at 13500 is not neutral The reference tone. Then the sine through a highpass at 13500 absolute cents (19.9 kHz) with initialFilterQ 0. Then through a lowpass with the same values. | The highpass note is nearly silent. The lowpass note has the level of the reference tone: only the lowpass may be skipped there. |
| GEN-16 | 1:24 | 8.1.2 | The lower clamp of the cutoff comes before the excursions The noise unfiltered. Then initialFilterFc 500 with a modEnvToFilterFc of 6000 and the Modulation Envelope held at its peak (its defaults). Then a plain lowpass at 7500 (622 Hz) for comparison. | The second and third notes sound the same, a lowpass at 622 Hz: 500 is clamped to 1500 before the envelope adds 6000. Without the clamp the cutoff would be 349 Hz. |
| GEN-17 | 1:29 | 8.1.2 | The upper clamp of the cutoff comes before the excursions The noise unfiltered. Then initialFilterFc 13500 with a constant modulator that adds 5000, and a held modEnvToFilterFc of -6000. Then a plain lowpass at 7500 (622 Hz). | The second and third notes sound the same, a lowpass at 622 Hz: 18500 is clamped to 13500 before the envelope takes 6000 away. Without the clamp the cutoff would be 12500 (about 11.2 kHz). |
| GEN-18 | 1:34 | 7.1.1 #62 | maxFilterFc The noise unfiltered. Then initialFilterFc 13500 with maxFilterFc 13500, 10500, 9000 and 7500. Last, initialFilterFc 13500 with no maxFilterFc at all. | 13500 sounds the same as no maxFilterFc at all: no ceiling. The others are lowpass filters at about 3.5 kHz, 1.5 kHz and 620 Hz. |
| GEN-19 | 1:42 | 7.1.1 #62, 8.1.2 | maxFilterFc stops a sweep The noise unfiltered for 2.8 s. Then a lowpass at initialFilterFc 6000 with a modEnvToFilterFc of 6000, maxFilterFc 9000, and the Modulation Envelope delayed by 0.5 s with an attack of 1 s. | For half a second the lowpass is at about 260 Hz (6000). Then the cutoff sweeps up and stops at about 1.5 kHz (9000) for the rest of the note, where 12000 would be about 8.4 kHz. |
| GEN-20 | 1:50 | 7.1.1 #62 | maxFilterFc takes a preset offset and has a minimum The noise with an instrument maxFilterFc of 13500 and a preset offset of -4500, next to a plain lowpass at 9000. Then a stored maxFilterFc of 500 next to a plain lowpass at 1500. | Each pair sounds the same: a ceiling at 9000 (1.5 kHz), then a ceiling at 1500 (20 Hz), the minimum, which leaves almost nothing of the noise. |
| GEN-21 | 1:56 | 7.1.1 #62 | No ceiling at 13500 for every type Through a bandpass with initialFilterQ 100: initialFilterFc 13000 with a held modEnvToFilterFc of -3000, next to initialFilterFc 10000 alone, both with maxFilterFc 13500. Then the first again with maxFilterFc 9000, next to initialFilterFc 9000 alone. Last, initialFilterFc 13500 with a held excursion of +300 (23.7 kHz) with maxFilterFc 13500, next to the same with maxFilterFc 13499. | The first pair sounds the same, a baseline with both cutoffs below every ceiling. The second pair sounds the same too: the ceiling of 9000 caps the final cutoff of 10000. In the last pair the first note has more of the highest frequencies (should): 13500 lets the cutoff go past 13500, as far as the synthesizer can compute, where 13499 is a ceiling. |
| GEN-22 | 2:04 | 7.1.1 #63 | filterQOffset The sine at key 69 (440 Hz) on the cutoff (6900) of a lowpass with initialFilterQ 0 and filterQOffset -100, 0 and 100. Then the sine at key 33 (55 Hz), far below a lowpass cutoff of 13500, with filterQOffset 0 and 200. | On the cutoff the first note is 10 dB quieter than the second and the third 10 dB louder, since the lowpass passes Q at its cutoff. Far below the cutoff the last two notes have the same level: filterQOffset moves the resonance and does not lower the gain. |
| GEN-23 | 2:11 | 7.1.1 #63 | Each part of the resonance is clamped by itself The sine at key 33 (55 Hz) under a lowpass at 13500 without resonance, then three pairs under the same lowpass. initialFilterQ 0 with a constant modulator of -500 to it and filterQOffset 100, next to filterQOffset 100 alone. initialFilterQ 2000 (stored beyond its range) next to 960. initialFilterQ 960 with filterQOffset -2000 (stored beyond its range), next to initialFilterQ 960 with filterQOffset -960. | Each pair has the same level: -500 is clamped to 0 before the offset is added, and each part stops at its limit on its own. The pairs are 0, 48 and 48 dB below the first note, half of the initialFilterQ part. A synthesizer that does not clamp the offset of -2000 plays the sixth note tens of dB quieter still. |
| GEN-24 | 2:20 | 7.1.1 #63 | filterQOffset drives every type The reference tone. Then the sine on the cutoff (6900) of a bandpass with initialFilterQ 0 and filterQOffset 100. | The second note is 7 dB above the reference tone: the bandpass passes Q (2.24) at its cutoff, and filterQOffset does not lower the gain the way initialFilterQ 100 does in GEN-13. |
| GEN-25 | 2:24 | 7.1.1 #64 | vibLfoToFilterFc The noise through a bandpass at 9000 with initialFilterQ 100. First the Vibrato LFO at 1 Hz with vibLfoToFilterFc 2400, then the Modulation LFO at 1 Hz with modLfoToFilterFc 2400. Each note lasts 1.8 s, and they are 2 s apart. | Both notes sweep the band the same way, two octaves up and down once a second. |
| GEN-26 | 2:30 | 8.1.7 | The cutoff excursions of both LFOs add The noise through a bandpass at 9000 with initialFilterQ 100, both LFOs at 1 Hz. modLfoToFilterFc 1200 with vibLfoToFilterFc -1200, next to no LFO. Then both at 1200, next to the Modulation LFO alone at 2400. Each note lasts 1.8 s, and they are 2 s apart. | The first pair sounds the same, a steady band: the two excursions cancel. The second pair sounds the same, a sweep of two octaves: they add. |
| GEN-27 | 2:40 | 7.1.1 #65 | vibLfoToVolume Notes of 1.8 s, 2 s apart. The sine with the Vibrato LFO at 2 Hz and vibLfoToVolume 60. Then with the Modulation LFO at 2 Hz and modLfoToVolume 60. | Both notes have the same tremolo, 6 dB up and down (12 dB from the quietest to the loudest point), twice a second. |
| GEN-28 | 2:46 | 7.1.1 #66 | modLfoToAmplitude The reference tone, then the sine with the Modulation LFO at 2 Hz and modLfoToAmplitude 500. | In a linear view the amplitude of the second note is a triangle between 0.5 and 1.5 times the reference tone, with straight sides, starting upward. |
| GEN-29 | 2:51 | 7.1.1 #66 | Full and negative amplitude depths The reference tone, then the sine with the Modulation LFO at 2 Hz and modLfoToAmplitude 1000, then -500. | The second note swings from silence to twice the amplitude of the reference tone. The third swings between 0.5 and 1.5 times, but starts downward. |
| GEN-30 | 2:58 | 7.1.1 #66, #67 | One clamp after the sum of both amplitude depths The reference tone. Then both LFOs at 2 Hz, with modLfoToAmplitude 1000 and vibLfoToAmplitude 1000. Then with 1000 and -1000. | The second note has flat tops at twice the amplitude of the reference tone and flat stretches of silence between them: the depths add to 2000 and one clamp limits the factor to 0 to 2. The third note has no tremolo at all: the two depths cancel before the clamp. |
| GEN-31 | 3:05 | 7.1.1 #67 | vibLfoToAmplitude The reference tone, then the sine with the Vibrato LFO at 3 Hz and vibLfoToAmplitude 500. | As GEN-28 at 3 Hz: a triangle between 0.5 and 1.5 times the reference tone, starting upward. |
| GEN-32 | 3:10 | 8.1.3 | The amplitude factor multiplies the other level factors Three notes with initialAttenuation 200 and a decay of 24 dB per second (decayVolEnv 2400) to a sustain of 24 dB. Without an LFO, with modLfoToAmplitude 1000 at 2 Hz, and with modLfoToAmplitude 1000 and modLfoToVolume 60 together. | Against the first note, the tops of the second are 6 dB higher at every moment, while it decays, and the tops of the third 12 dB higher: the factors multiply. |
| GEN-33 | 3:15 | 7.1.1 #68, #69 | The LFO delay and fade The reference tone. Then the sine with modLfoToAmplitude 1000 at 3 Hz, delayModLFO 0.5 s and fadeModLFO 1 s. Then the same on the Vibrato LFO (vibLfoToAmplitude, delayVibLFO, fadeVibLFO). Each note lasts 2.5 s. | No tremolo for the first 0.5 s. Then its depth grows in a straight line for one second: close to halfway through, the tops are about 1.4 times the reference tone. The delay is one and a half cycles, so an LFO that ran during it would be at zero and falling when the fade begins, the opposite of one that starts there. After that the tremolo goes from silence to twice the amplitude. |
| GEN-34 | 3:24 | 7.1.1 #68 | The fade scales every depth The Modulation LFO at 3 Hz with delayModLFO 0.5 s and fadeModLFO 1 s, on three notes of 2.5 s: modLfoToPitch 200 on the sine, modLfoToVolume 60 on the sine, and modLfoToFilterFc 2400 on the noise through a lowpass at 9000. | In each note the vibrato, the tremolo or the filter sweep is absent for 0.5 s, then grows over the same second to its full depth: 200 cents, 6 dB and two octaves. |
| GEN-35 | 3:34 | 7.1.1 #70, #71 | The four LFO waveforms The reference tone. Then the sine with modLfoToAmplitude 500 at 2 Hz and modLfoWaveform 0, 1, 2 and 3. Then the same four with the Vibrato LFO (vibLfoToAmplitude, vibLfoWaveform). | The amplitude shows, twice in each note: a triangle, a rising sawtooth, a sine and a square, each between 0.5 and 1.5 times the reference tone and starting at the middle (the square starts at the top). |
| GEN-36 | 3:45 | 7.1.1 #72, #73 | The LFO start phase The sine with modLfoToAmplitude 500 at 2 Hz and a zero-time volume delay, attack and hold, for each waveform with modLfoStartPhase 0, 90, 180, 270 and 360 (20 notes of 0.5 s). Then the same 20 with the Vibrato LFO. The reference tone comes first. | Where each note starts: triangle and sine at the middle and rising (0), at the top (90), at the middle and falling (180), at the bottom (270), and as 0 again (360). The sawtooth starts at 0, 0.5, -1, -0.5 and 0 of its swing, the square at 1, 1, -1, -1 and 1. |
| GEN-37 | 4:28 | 8.1.5 | The start phase after the delay The reference tone, then the sine with modLfoToAmplitude 500 at 3 Hz, delayModLFO 0.5 s and modLfoStartPhase 90. | The second note has the level of the reference tone for 0.5 s, then jumps to 1.5 times it and falls from there: the LFO starts at its peak when the delay ends. An LFO that had run during the delay would be at its trough there, since the delay is one and a half cycles. |
| GEN-38 | 4:33 | 7.1.1 #72 | The start phase is read once The sine with modLfoToAmplitude 500 at 2 Hz and a modulator from CC 16 to modLfoStartPhase with the amount 180. CC 16 is 0 at the note on and goes to 127 after 0.6 s. Then the same note without the controller change. | The two notes are the same: the tremolo does not jump when CC 16 moves, because the start phase is read when the voice starts. |
| GEN-39 | 4:39 | 8.1.7 | The pitch sum of both LFOs and both Modulation Envelopes The sine with modEnvToPitch 100 and modEnv2ToPitch 200, both envelopes held at their peak, the Modulation LFO at 1 Hz with modLfoToPitch 50 and the Vibrato LFO at 1.5 Hz with vibLfoToPitch 30, held for 1.9 s. | The pitch is 300 cents above 440 Hz plus the two triangles: it starts at 300 cents, peaks near 365 cents a quarter second in, and moves in the shape the sum of the two triangles gives. |
| GEN-40 | 4:43 | 7.1.1 #74, #76 to #81 | The second Modulation Envelope The sine with modEnv2ToPitch 1200 and the second Modulation Envelope at: delay 1 s, attack 1 s, hold 1 s, decay 2 s (decayModEnv2 1200), sustainModEnv2 500 (50 %), release 2 s. The note is held for 5 s, and the Volume Envelope releases over 2 s so the pitch stays audible. | The pitch has the SoundFont 2 envelope shape: 440 Hz for 1 s, a convex rise to one octave up over 1 s (the convex curve of section 7.2, about 1050 cents halfway), held for 1 s, a straight fall to half an octave up over 1 s (half of the decay time, for half of the range), held until the note off, then a straight fall back to 440 Hz over 1 s. |
| GEN-41 | 4:52 | 7.1.1 #75 | modEnv2ToFilterFc The noise through a highpass at 7000 (466 Hz) with modEnv2ToFilterFc 1200 and the envelope of GEN-40, held for 5 s. Then three short notes through plain highpass filters at 7000, 8200 and 7600 for comparison. | The cutoff follows the envelope: during the delay the first note sounds like the highpass at 7000, during the hold like the one at 8200 (one octave up), and in the sustain like the one at 7600 (half an octave up). |
| GEN-42 | 5:03 | 7.1.1 #74 to #81 | Two independent Modulation Envelopes The sine through a lowpass at 5700 (220 Hz). The first Modulation Envelope opens it by 2400 cents after a delay of 0.5 s (modEnvToFilterFc 2400). The second raises the pitch by an octave after a delay of 1.5 s (modEnv2ToPitch 1200). Both attacks are instant. | At 0.5 s the note gets louder (the filter opens) while the pitch stays at 440 Hz. At 1.5 s the pitch jumps an octave: each envelope drives only its own destination, on its own times. |
| GEN-43 | 5:07 | 7.1.2 | The defaults of the second Modulation Envelope The sine with only modEnv2ToPitch 1200 set, a note of 0.8 s, and a Volume Envelope release of 0.5 s. | The note is one octave up from its start and for the whole note (no delay, attack or hold, full sustain). It is back at 440 Hz 0.1 s after the note off, since the default release of the envelope is 1 ms, while the volume still releases. |
| GEN-44 | 5:10 | 7.1.1 #82, #83 | Key scaling of the second Modulation Envelope The sine with scaleTuning 0, so every key plays 440 Hz, and modEnv2ToPitch 1200, at keys 36, 48, 60, 72 and 84. First with a hold of 0.5 s and keyNumToModEnv2Hold 100, the envelope falling to 0 after the hold. Then with a decay of 1 s to 0 and keyNumToModEnv2Decay 100. | In the first five notes the octave lasts 2, 1, 0.5, 0.25 and 0.125 s: the hold halves per octave upward. In the last five the pitch falls in a straight line and passes half an octave after 2, 1, 0.5, 0.25 and 0.125 s: the decay halves per octave too. |
| GEN-45 | 5:37 | 7.1.1 #84, #85 | The second decay of the Volume Envelope The reference tone, then the sine with zero time for the delay, attack and hold of its Volume Envelope, decayVolEnv 4 s (24 dB per second) to sustainVolEnv 120, decay2VolEnv 8 s (12 dB per second) with sustain2VolEnv 240, held for 4 s and then released. | A decay, a second decay, a sustain and a release: the level falls 24 dB per second for the first 0.5 s, then 12 dB per second along a straight line in dB to 36 dB below the reference tone after 2 more seconds, then stays there until the note off, then releases. The slopes are checked as changes between two points of each, so smoothing does not matter. |
| GEN-46 | 5:44 | 7.1.1 #85 | A negative sustain2VolEnv The reference tone, then the sine with zero time for the delay, attack and hold of its Volume Envelope, decayVolEnv 4 s to sustainVolEnv 240, then decay2VolEnv 4 s with sustain2VolEnv -120. | The level falls 24 dB per second for the first second, then rises in a straight line, 24 dB per second for half a second, and stays 12 dB below the reference tone. |
| GEN-47 | 5:49 | 7.1.1 #84 | A sustainVolEnv above 960 counts as 960 The reference tone, then the sine with zero time for the delay, attack and hold of its Volume Envelope, decayVolEnv 1 s, sustainVolEnv 1200, decay2VolEnv 1 s and sustain2VolEnv -600. | The level falls 96 dB per second to silence in the first second, then rises back and stays 36 dB below the reference tone: the end level is 960 - 600 = 360 cB. Taking 1200 as it is would give 60 dB. |
| GEN-48 | 5:54 | 7.1.1 #84, 8.1.6 | The default decay2VolEnv, and no second decay The reference tone. Then the sine with zero time for the delay, attack and hold of its Volume Envelope, decayVolEnv 2 s (48 dB per second) to sustainVolEnv 120, and sustain2VolEnv 240 with decay2VolEnv at its default. Then decay2VolEnv 8000 with sustain2VolEnv 0. | The second note falls 48 dB per second for a quarter second (12 dB), then drops 24 dB more at once (the default second decay is 1 ms for 96 dB), and is 36 dB down 0.1 s later. The third note falls to 12 dB below the reference tone and stays there: with sustain2VolEnv 0 there is no second decay, however long decay2VolEnv is. |
| GEN-49 | 6:00 | 7.1.1 #84 | keynumToVolEnvDecay scales the second decay The reference tone. Then the sine with zero time for the delay, attack and hold of its Volume Envelope, scaleTuning 0 at keys 48, 60 and 72, sustainVolEnv 0, decay2VolEnv 4 s with sustain2VolEnv 480, keynumToVolEnvDecay 100 and a release of 2 s (48 dB per second). The notes last 2, 1 and 0.5 s, so each note off comes when its second decay is half done, 24 dB down. | The second decay falls 12, 24 and 48 dB per second: it halves in time per octave upward, so each note falls 12 dB over the middle half of its length. At each note off the release goes on from the level the note has, without a jump: 0.1 s before and after the note off differ by the two slopes alone. |
| GEN-50 | 6:15 | 7.1.1 #84 | Decay and release times are for 960 cB The sine with decayVolEnv 2 s to sustainVolEnv 240 and releaseVolEnv 2 s, a note of 2 s. | The decay and the release both fall 48 dB per second, so 24 dB take half a second, a quarter of the time given: the times are for a change of 96 dB. |
| GEN-51 | 6:21 | 7.1.1 #87 | delayVoice One note on starts four zones of the click, with delayVoice 0 (-32768), 0.25 s, 0.5 s and 1 s. | Four clicks, at the note on and 0.25, 0.5 and 1 s after it, each as close to its time as the synthesizer can (should). |
| GEN-52 | 6:24 | 7.1.1 #87 | The envelopes and LFOs start with the delayed voice The reference tone. Then the sine with delayVoice 1 s, a Volume Envelope attack of 0.5 s, and modLfoToAmplitude 500 with the Modulation LFO at 1.5 Hz. The note is held for 2 s. | The second note is silent for 1 s, then fades in over 0.5 s. 0.6 s after the voice starts the tremolo is at 0.8 times the reference tone (on its way down). An LFO that had run since the note on would be at 1.2 times there. |
| GEN-53 | 6:29 | 7.1.1 #87, 8.1 | A note off before the delayed voice starts The reference tone. Then the cosine with delayVoice 1 s, zero-time delayVolEnv and attackVolEnv (-32768), a release of 0.5 s, and a note off after 0.2 s. | The second note is silent for 1 s, then starts at full level and releases, falling 96 dB in 0.5 s: the note off counts as if it came when the voice started. A synthesizer may hold the voice for a minimum time against clicks first (spec section 8.1). The release is measured from 0.1 s after the voice start, so a minimum time up to 0.1 s passes and a longer one fails here. |
| GEN-54 | 6:34 | 7.1.1 #87, 7.1 | A delayed voice joins its exclusive class at the note on One preset with two zones in exclusiveClass 5: key 69 plays the sine at once, key 81 plays it an octave up with delayVoice 1 s. Key 69 is struck, and key 81 0.5 s later. | The 440 Hz note ends at 0.5 s, when key 81 is struck, not at 1.5 s when its voice starts. The 880 Hz note then plays from 1.5 s. |
| GEN-55 | 6:39 | 7.1.1 #88 | ignoreNoteOff The reference tone. Then a note of 0.1 s with ignoreNoteOff 1 and a Volume Envelope that decays to silence over 2 s (48 dB per second) from the first sound (zero time for its delay, attack and hold). Then the same note without ignoreNoteOff. | The second note plays its whole decay: it falls by 48 dB per second from its first sound (9.6 dB from 0.3 s to 0.5 s in) and is gone after 2 s. The third stops after its note off. |
| GEN-56 | 6:47 | 7.1.1 #88, 11.5 | What ends a voice with ignoreNoteOff Five notes of 0.2 s with ignoreNoteOff 1 and exclusiveClass 9, each ended 0.6 s after its note on by: CC 123, CC 123 while CC 64 is down, CC 120, a note of key 81 in the same class (a zone that is far too quiet to hear), and a GS Reset. The first two play key 70, whose zone has a release of 1 s, the others key 69 with a release of 0.2 s. | The first two release at once, also with the pedal down: between 0.1 and 0.3 s after CC 123 the level falls by 17 dB, the slope of the 1 s release. CC 120 cuts the third at once, silent 0.1 s later. The note of the same class ends the fourth, with a very short release (should). The GS Reset ends or releases the last one. Each note sounds until its ending message, after its note off. |
| GEN-57 | 6:54 | 7.1.1 #88 | ignoreNoteOff is read when the note off arrives The reference tone. Then the sine with a decay to silence over 2 s from the first sound (zero time for the delay, attack and hold) and a modulator from CC 16 to ignoreNoteOff with the amount 2. CC 16 is 0 at the note on and 127 0.1 s later, before the note off at 0.2 s. Then the same note with CC 16 at 0 throughout. | The second note plays its whole decay, 9.6 dB lower 0.5 s in than 0.3 s in: when the note off arrives, CC 16 has set ignoreNoteOff (2, clamped to its range of 0 to 1). The third note stops after its note off. |
| GEN-58 | 7:02 | 7.1 | Zero time and the 1 ms floor The cosine, which starts at its peak, with delayVolEnv -32768 and attackVolEnv -32768, -32767, -20000, -12001 and -12000. | The first note is at full level from its first point: the voice starts at its peak. The other four start the same way as each other: every value from -32767 to -12001 counts as -12000, which is 1 ms (a synthesizer may round it to its update period). |
| GEN-59 | 7:09 | 7.1 | Zero time by the sum of the instrument and the preset The cosine with delayVolEnv -32768 and an instrument attackVolEnv of -12000. First with a preset offset of -20768 (a sum of -32768), then with -20000 (a sum of -32000), then with no preset offset. | The first note starts at its peak: the sum reaches -32768. The second starts like the third: -32000 counts as -12000, which is 1 ms (a synthesizer may round it to its update period). |
| GEN-60 | 7:14 | 7.1 | A delay of -12000 is 1 ms The cosine with attackVolEnv -32768 and delayVolEnv -32768, then the same with delayVolEnv -12000. | The second note does not start before the first. Informative: it starts 1 ms (48 points) later after its note on, since -12000 is 1 ms and not zero time, but a synthesizer may round times shorter than its update period. |
| GEN-61 | 7:18 | 7.1 | Zero time comes before the key scaling Key 36 with scaleTuning 0, a key scaling of the hold of 400 per key (keynumToVolEnvHold), and a fall of 192 dB per second after the hold. First holdVolEnv -32768, then -20000. Then the same for the Modulation Envelope and the second Modulation Envelope, each to pitch with 1200 cents and falling to 0 after its hold. | With -32768 there is no hold at all. With -20000 the hold lasts 0.25 s: -20000 counts as -12000, and the scaling adds 400 * 24 = 9600 timecents to it. The same in the pitch of both Modulation Envelopes. |
| GEN-62 | 7:26 | 7.1 | delayVoice and delayVolEnv at -32768 and -12000 The click with delayVoice -32768, then -12000. Then the click with zero-time attack and delayVolEnv -32768, then -12000. | In each pair the second click does not come before the first. Informative: it comes 1 ms (48 points) later after its note on, but a synthesizer may round times shorter than its update period. |
| GEN-63 | 7:32 | 7.1 | A fade of zero time The sine with a square Modulation LFO (modLfoWaveform 3) at full depth (modLfoToAmplitude 1000) and no LFO delay (-32768). First fadeModLFO -32768, then -12000. | The first note starts at twice the amplitude at once. The second reaches it after a ramp of 1 ms. Informative: a synthesizer may round times shorter than its update period, so the second note may start at full depth too. |
| GEN-64 | 7:36 | 7.1 | sampleModes 0, 1 and 3 The tail (a loop of 440 Hz with 0.3 s of 880 Hz after the loop end) with a release of 2 s (48 dB per second), held for 1.5 s, with sampleModes 0, 1 and 3. | Mode 0 plays the sample through once: 440 Hz for about 1.05 s, then 880 Hz, then silence before the note off. Mode 1 loops the 440 Hz through the release. Mode 3 loops while the key is held, and after the note off it plays on to the loop end and then the 880 Hz tail. |
| GEN-65 | 7:50 | 7.1 | sampleModes 2, start on release The steps (eight falling plateaus, no loop) with sampleModes 2 and a very long release, held for 1 s. Then the same sample with sampleModes 0 for comparison. | The first note is silent while the key is held. At the note off the sample plays once from its start: the same eight plateaus, at the same level, as the second note plays at its note on. |
| GEN-66 | 7:55 | 7.1 | Start on release with ignoreNoteOff The sine with sampleModes 2, ignoreNoteOff 1 and a decay to silence over 1 s, a note of 0.5 s. | Silent: the note off is ignored, so the sample never starts. |
| GEN-67 | 7:59 | 7.1 | Start on release: the envelope ran from the note on The reference tone. Then the sine with sampleModes 2, a Volume Envelope attack of 0.5 s and a very long release, a note of 1 s. | At the note off the sample starts at once at the level of the reference tone: the attack ended while the key was held. |
| GEN-68 | 8:04 | 7.1 | sampleModes 4, the alternating loop The ramp (one rising ramp of 109 points, looped) with sampleModes 4, then with sampleModes 1. | The first note is a triangle wave with a period of 2 * (109 - 1) = 216 points, 222.2 Hz: forward through the loop, then backward, each turn point played once. The second is a sawtooth of 109 points, 440.4 Hz. |
| GEN-69 | 8:08 | 7.1 | The turn points of the alternating loop A ramp of 109 points looped with sampleModes 4, where the 8 points before the loop start are at the top of the ramp and the 8 after the loop end at the bottom: the opposite sign of the turn points next to them. It plays at key 70, so the playback position falls between points. | A clean triangle whose peaks reach the first and last point of the loop. There is no spike or dip at the turns toward the points stored outside the loop: past a turn point, interpolation uses the points the position reaches next. |
| GEN-70 | 8:11 | 7.1 | The alternating loop in the release The ramp with sampleModes 4, a note of 1 s and a release of 1 s. | The triangle of 222.2 Hz goes on through the release: the alternating loop loops in the release phase too. |
| GEN-71 | 8:15 | 7.1 | A loop of one point, and undefined sample modes The sine with a loop of one point after 1000 points, with sampleModes 4 and then 1. Then the sine (1.1 s long) with sampleModes 5, 6, 7 and 8, each held for 1.5 s. | The first two notes are the same: a loop of one point plays like the value 1, a steady level after the first 21 ms. The last four play the sample once and are silent after 1.1 s, like the value 0. |
| GEN-72 | 8:29 | 8.1.1 | Swapped and empty loops The sine (1.1 s long) with its loop end stored before its loop start and sampleModes 1. Then the sine with a loop of no points (loop start equal to loop end) and sampleModes 1, 3 and 4. Each note is held for 1.5 s. | The first note loops as if the two loop points were swapped: 440 Hz until the note off. The other three play the sample once and are silent after 1.1 s, as with sampleModes 0. |
| GEN-73 | 8:39 | 8.1.1 | Address offsets beyond the sample The steps (8000 points, no loop) plain. Then with startAddrsOffset 20000, past its end. Then with startAddrsOffset -5000, before its start. Then with endAddrOffset 20000, past its end. | Each point is limited to the sample. The second note plays nothing. The third and the fourth play the whole sample, exactly like the first: the same eight plateaus for 0.167 s. |
| GEN-74 | 8:45 | 6.2 | The sample links are not used The left (440 Hz) and right (880 Hz) samples, linked as a stereo pair, in two zones panned hard left and hard right. The right zone has fineTune 700. | The left channel plays 440 Hz. Only the right channel moves, to 880 Hz plus 700 cents (1318.5 Hz): the left zone does not follow the generators of the right one. |
| GEN-75 | 8:48 | 7.1 | The wider range of fineTune The sine at key 69 with fineTune 1200, -1200 and 150. Then at key 0 with fineTune 12700. | One octave up (880 Hz), one octave down (220 Hz), 150 cents up (479.8 Hz), and 12700 cents above key 0, the pitch of key 127 (12.5 kHz). |
| GEN-76 | 8:54 | 7.1 | exclusiveClass is wider than 8 bits A preset whose key 69 is in exclusiveClass 44 and key 81 (880 Hz) in class 300. Key 69 is held, and key 81 is struck 0.5 s later. Then a preset with classes 200 and 300, and one where both keys are in class 32767, played the same way. | In the first part both notes sound together: 300 is not 44, although its lowest 8 bits are. In the second part both sound too: 200 and 300 are different classes, which a clamp at 127 would merge. In the third part the 440 Hz note ends when key 81 is struck: class 32767 is a valid class. |
| GEN-77 | 9:03 | 7.1 | The exclusive class reaches across a Program Change A long note of the sine in exclusiveClass 7. Then a Program Change, and a note of another preset (880 Hz) in class 7, 0.5 s later. | The first note ends when the second one starts, with a very short release (should): the class is per channel, not per preset. |
| GEN-78 | 9:07 | 7.1 | One note on does not end its own voices One note on that starts two zones in exclusiveClass 7: the sine and the sine an octave up. The reference tone comes first. | Both voices sound for the whole note: 440 Hz and 880 Hz at the same level, together 3 dB above the reference tone. |
| GEN-79 | 9:11 | 7.1 | Enumerations are truncated toward zero The reference tone. Then the sine with modLfoToAmplitude 500 at 2 Hz and a unipolar linear modulator from CC 16 to modLfoWaveform with the amount 4, with CC 16 at 0, 56, 88 and 127. | The tremolo is a triangle, a sawtooth, a sine and a square: 0, 1.75, 2.76 and 4 give 0, 1, 2 and 3 (truncated toward zero, and 4 clamped to the range). Rounding would give a sine, a square and a square. |
| GEN-80 | 9:18 | 7.1 | Enumerations are clamped to their range The reference tone. Then constant modulators that push modLfoWaveform to 10 and to -2, each with modLfoToAmplitude 500 at 2 Hz. Then one that pushes filterType to 9 on the sine on a cutoff of 440 Hz. Then one that pushes ignoreNoteOff to 5 on a note of 0.1 s that decays to silence over 2 s from the first sound (zero time for its delay, attack and hold). | A square, a triangle, a nearly silent note (the notch on its cutoff) and a note that plays its whole decay (on, 9.6 dB lower 0.5 s in than 0.3 s in): each value is clamped to its range after truncation. |
| GEN-81 | 9:26 | 7.1 | Instrument-only generators in preset zones are ignored Presets whose zones set filterType 4, modLfoWaveform 3, vibLfoWaveform 3 and ignoreNoteOff 1: a lowpass on the cutoff of the sine, a triangle tremolo of each LFO, and a note of 0.1 s that decays over 2 s. The reference tone comes first. | Each preset value is ignored: the note on the cutoff is 3 dB below the reference tone (a lowpass, not a notch), both tremolos are triangles, and the short note stops after its note off. |
| GEN-82 | 9:34 | 7.1 | Preset modulators to instrument-only generators are ignored Constant preset zone modulators to instrument-only destinations, one per preset: filterType +4, modLfoWaveform +3, vibLfoWaveform +3, ignoreNoteOff +2, sampleModes -1 on a looped sine, exclusiveClass +5 on two keys struck together, overridingRootKey +12 on the sine, and startAddrsOffset +4000 on the steps. The reference tone comes first. | None acts: a lowpass on the cutoff (3 dB down), two triangle tremolos, a short note that stops, a loop that goes on past 1.1 s, two keys that both sound, 440 Hz, and the whole steps (as long as without the offset). |
| GEN-83 | 9:47 | 7.1.2 | Stored values beyond the range Pairs of notes of 1.9 s, a stored value beyond the range of the generator next to the limit: modLfoToAmplitude 1500 and 1000, vibLfoToFilterFc 20000 and 12000 (on the noise through a lowpass at 1500, with the slowest triangle Vibrato LFO started at 18 degrees, so it stays near 0.2 and the cutoff near 3900 instead of 5500), fadeModLFO 8000 and 5000 (with modLfoToAmplitude 1000), keyNumToModEnv2Hold 3000 and 1200 at key 61, sustainModEnv2 2000 and 1000, attackModEnv2 9000 and 8000, and modLfoStartPhase 400 and 360. The envelope pairs go to the pitch with modEnv2ToPitch 1200. Last, delayVoice 8000 and 5000 on two clicks struck 0.5 s apart. | The two notes of each pair are the same: a stored value is limited to the range of its generator. The two delayed clicks come 18 s after their note ons, 0.5 s apart. |
| GEN-84 | 10:37 | 7.1.1 #87, 7.1 | A delayed voice ended before it starts The preset of GEN-54. Key 81 is struck first, with its voice delayed by 1 s, and key 69 0.5 s later. | The 880 Hz voice never sounds: key 69 ends it before it starts. Only the 440 Hz note plays. |
| GEN-85 | 10:42 | 7.1 | Zero time reached through a modulator The cosine with delayVolEnv -32768 and attackVolEnv -12000. First with a constant modulator of -32768 to attackVolEnv (a sum below -32768), then with one of -20000 (a sum of about -32000), then with no modulator. | The first note starts at its peak: the time is taken after the modulators, and the sum reaches -32768. The second starts like the third: -32000 counts as -12000, which is 1 ms (a synthesizer may round it to its update period). |
| GEN-86 | 10:47 | 7.1 | The exclusive class is per channel The sine in exclusive class 5 on channels 1 and 2. Key 69 on channel 1, then key 81 on channel 2 0.5 s later, both held. | Both notes keep sounding: an exclusive class ends voices on its own channel only. The two notes together are 3 dB louder than the first alone. |
| GEN-87 | 10:51 | 7.1 | Two note ons of one class at the same time The sine in exclusive class 5. Key 69 and key 81 are struck at the same time, key 69 first, and both are held. | Only key 81 (880 Hz) keeps sounding: key 69 is a separate note on, so key 81 ends it like any other note of the class. A very short sound of key 69 at the start is allowed. |
| GEN-88 | 10:54 | 7.1 | Modulators add up before one saturation The reference tone. Then the sine with four constant modulators to gain, in this order: +32767, +32767, -32767 and -32767. They differ in their sources (the No Controller source, unipolar or bipolar, as the source and the amount source), so no two are identical, and each outputs exactly its amount. | The second note has the level of the reference tone: the four outputs add up to 0 before the sum is limited once. A synthesizer that limits after each modulator reaches 32767, then 0, then -32767, and plays the note 144 dB down, which is silence. |
| GEN-89 | 10:58 | 7.1 | A modulator on overridingRootKey is ignored The sine with overridingRootKey 69 and a modulator from CC 16 to overridingRootKey with the amount 12, at key 69. The first note with CC 16 at 0, the second with CC 16 at 127 before its note on, the third with CC 16 at 0 at its note on and at 127 0.35 s later. | 440 Hz in all three notes (should). overridingRootKey is a plain value, so a synthesizer should ignore the modulator. One that applies it is allowed but not recommended, and plays the second note lower. |
| GEN-90 | 11:03 | 8.1.1 | A start past the loop end, an end inside the loop The tail (a loop of 440 Hz from point 2400 to 50400, then 880 Hz to point 64800) with sampleModes 1, held for 1.5 s. First with startAddrsCoarseOffset 1 and startAddrsOffset 20000, so it starts at point 52768, past the loop end in the 880 Hz part. Then with endAddrOffset -30000, so the end is at point 34800, inside the loop. | Both notes play 440 Hz for as long as the key is held. The first start moves back by the loop length into the loop, so no 880 Hz plays. While a loop runs the end is not checked, so the second note goes on past 0.725 s, where it would have stopped at its end. |
| GEN-91 | 11:08 | 7.1 | Modulators to keynum and velocity are ignored The reference tone. Then the sine with the Note-On Velocity to gain with -240 and a constant modulator of 21 to velocity. Then the sine with the Note-On Key Number to gain with -240 and a constant modulator of 21 to keynum. Key 69, velocity 127. | The second note is 24 dB below the reference (velocity 127) and the third 13 dB below it (key 69). Both constant modulators are ignored. A synthesizer that applied them would read velocity or key 20 (the default -1 plus 21), 3.8 dB down. |
| GEN-92 | 11:13 | 7.1 | The address offsets take their modulators at the note on The tail (440 Hz to point 50400, then 880 Hz to point 64800, 1.35 s in all) with sampleModes 0. First with a modulator from CC 16 to startAddrsOffset with the amount 32767: one note with CC 16 at 0, one with CC 16 at 127 before its note on, one with CC 16 at 0 at its note on and at 127 0.05 s later. Then with a modulator from CC 16 to endAddrsOffset with the amount -32767: one note with CC 16 at 0 at its note on and at 127 0.05 s later, then one with CC 16 at 127 before its note on. | Between 0.45 and 0.6 s into the first three notes: 440 Hz, then 880 Hz, then 440 Hz. At CC 16 = 127 the start moves the full 32767 points (0.68 s) on, so the 880 Hz part comes after 0.37 s. The fourth note plays its whole length and is still sounding 0.8 to 1 s in. The fifth ends 32767 points early, at 0.67 s, so it is silent 0.8 to 1 s in. The offsets are read once at the note on, so the changes during the third and fourth notes do nothing. |
| GEN-93 | 11:23 | 7.1 | scaleTuning follows its modulators during the note The sine with scaleTuning 177 and a modulator from CC 16 to scaleTuning with the amount -127, at key 81, 12 keys above the root key. CC 16 is at 0 at the note on and moves to 127 0.6 s later. | The note starts 2124 cents above 440 Hz (177 cents per key) and drops to 600 cents above it when the controller moves: scaleTuning becomes 177 - 127 = 50 cents per key, since 127 is full scale. |
sfe4-20-modulators Modulators
Sections 5.3 to 5.5, 7.2, 7.2.1, 7.3, 7.4
Custom curves, the source values and the modulator rules. Most curve tests send CC 16 through a curve to gain with -240, so the curve shows as a level from 0 to 24 dB below the reference tone. A sweep pauses 0.3 s on each value a test checks, and the level is measured in the last 0.15 s of the pause, so a synthesizer may smooth level changes (spec section 8.1). Smoothing that takes longer than 0.15 s changes the result. The bank's DMOD holds only the default modulators of MOD-14, MOD-25 and MOD-31, on CC 20 to 24, which the other tests leave at 0.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| MOD-01 | 0:00 | 7.2.1 | A zigzag curve The reference tone. Then CC 16 swept from 0 to 127 over about 4 s, pausing on each checked value, through a custom curve of four rising teeth, to gain with the amount -240. | The level shows four sawtooth teeth: it falls from the reference level to 24 dB below it four times, jumping back up at CC 32, 64 and 96. No SoundFont 2 curve gives that. |
| MOD-02 | 0:11 | 7.2.1 | A staircase curve The reference tone. Then CC 16 swept from 0 to 127 over about 4 s, pausing on each checked value, through a staircase of eight equal steps ( y from 0 to 14336 in steps of 2048), to gain with -240. | Eight plateaus, each 3 dB below the one before: 0, 3, 6 and so on to 21 dB below the reference tone. |
| MOD-03 | 0:22 | 7.2.1 | A curve at 16384 is the full amount The reference tone. Then a custom curve that is 16384 everywhere on CC 16, as a unipolar source to gain with -240, at CC 16 = 0 and 127. Then the same as a bipolar source, with 24 dB of attenuation, at CC 16 = 0 and 127. | The unipolar notes are 24 dB below the reference tone and the bipolar ones 48 dB below it: in both cases the output is 1, the full amount. |
| MOD-04 | 0:29 | 7.2.1 | A bipolar curve is not mirrored The reference tone. Then CC 16 as a bipolar source through a curve that rises from 0 to 16384 over the lower half and stays there, to gain with -240 on a zone with 24 dB of attenuation, swept from 0 to 127, pausing on each checked value. | The output goes from -1 to 1 over the lower half of the sweep and stays at 1: the level falls from the reference level to 48 dB below it by CC 64 and stays there. A mirrored curve would come back up. |
| MOD-05 | 0:40 | 7.2.1 | A staircase with a negative direction The reference tone. Then the staircase of MOD-02 on CC 16 with a negative direction, swept from 0 to 127, pausing on each checked value. | The same eight plateaus in reverse order: from 21 dB below the reference tone up to the reference level. |
| MOD-06 | 0:51 | 7.2.1, 5.4 | The negative direction is 16384 minus x The reference tone. Then a curve with a narrow spike at 8192 on CC 16 with a negative direction, at CC 16 = 64. Then a curve with the points (16383, 0) and (16384, 16384) with a negative direction, at CC 16 = 0. | Both notes are 24 dB below the reference tone. 64 stays exactly 8192 and hits the spike, and 0 becomes 16384, which reaches the last point. With 16383 minus x, both would miss and play at the reference level. |
| MOD-07 | 0:56 | 7.2.1 | A vertical step The reference tone. Then the curve (0, 0), (8192, 0), (8192, 16384), (16384, 16384) on CC 16, at CC 16 = 63, 64 and 65. | Off, on, on: the reference level, then 24 dB below it twice. At an x with several points the last one gives the value. |
| MOD-08 | 1:02 | 7.2.1 | A curve that covers part of the range The reference tone. Then a curve with points only from x 4096 to 12288, on CC 16 swept from 0 to 127, pausing on each checked value. | The level is flat at 6 dB below the reference tone up to CC 32, falls in a nearly straight line to 18 dB below it at CC 96 (with a slight bend at 64, where the steps grow from 128 to 130), and stays flat there. |
| MOD-09 | 1:13 | 7.2.1 | Interpolation between points The reference tone. Then a curve of two points, (8192, 0) and (8320, 16384), on the Pitch Wheel, to gain with -240. The wheel steps 8192, 8224, 8256, 8288 and 8319, then goes to 8448. | The level falls in a straight line with the wheel: 0, 6, 12, 18 and 23.8 dB below the reference tone, then 24 dB at 8448. |
| MOD-10 | 1:22 | 5.4, 7.2 | The No Controller source is 16384 The reference tone. Then the curve (0, 16384), (16383, 16384), (16384, 0) read by the No Controller source, to gain with -240, first with the positive and then with the negative direction. Only a source value of 16384 gives 0 on it. | The second note has the level of the reference tone: No Controller is 16384, the end of the range, where the curve is 0. A synthesizer that reads it as 16383 plays 24 dB lower. The third note is 24 dB below the reference: the negative direction turns 16384 into 0, where the curve is full. |
| MOD-11 | 1:27 | 5.4 | The linear custom curve equals the SoundFont 2 linear type The custom curve (0, 0), (16384, 16384) on CC 16 next to the SoundFont 2 linear type, both to gain with -240, first with a positive and then with a negative direction, each at CC 16 = 0, 32, 64, 96 and 127. | Each custom curve note has the same level as the SoundFont 2 note at the same value: x / 16384. |
| MOD-12 | 1:50 | 7.2.1 | Custom curves on both sources The reference tone. Then CC 16 with the SoundFont 2 linear type as the source and CC 17 through a falling custom curve as the amount source. Then CC 16 through a curve that doubles and stops at 1, with the same amount source. Each at four pairs of CC 16 and CC 17 values: (32, 0), (64, 64), (127, 32) and (16, 96), all to gain with -240. | The level is 24 dB times the product of the two source outputs below the reference tone, each source read through its own curve. |
| MOD-13 | 2:01 | 7.2.1 | Custom curves on other sources The staircase of MOD-02 on the Note-On Velocity (velocities 8 to 120 in steps of 16), on the Pitch Wheel as a unipolar source (the 14-bit values in the middle of each step), and on the Pitch Wheel as a bipolar source with a negative direction, with 24 dB of attenuation (the same values). The reference tone comes first. | Each source walks the eight plateaus: 3 dB apart below the reference tone for the first two, and for the bipolar one from 42 dB below the reference tone up to the reference level as the wheel rises, in steps of 6 dB. |
| MOD-14 | 2:28 | 7.2.1, 7.3 | Custom curve modulators are identical only with the same points The bank's DMOD holds CC 20 through the staircase of MOD-02, with the No Controller source through a flat full curve as the amount source, to gain with -240. With CC 20 at 127: a zone without modulators, then zone A with the same modulator built on other curve entries with the same points and the amount -120, zone B with other points on CC 20 (the linear curve), and zone C where only the amount source's curve differs (flat at half). | The plain zone is 21 dB below the reference tone. Zone A replaces the default modulator, 10.5 dB down: curves with the same points are the same. Zones B and C add to it, 33 and 26.3 dB down. |
| MOD-15 | 2:36 | 5.4 | Curve numbering and names Six flat curves: one used by three modulators, one that no modulator uses, one with an empty name, two with the same name, and one with a name of 40 characters. Each used curve has its own level, and each note uses one of them through CC 16 (at 127) to gain with -240. The reference tone comes first. | Every note has the level of its own curve: 3, 3 and 3 dB down for the shared curve, then 6, 12, 18 and 24 dB down. An unused curve or an unusual name does not move the curves after it to other numbers. |
| MOD-16 | 2:46 | 5.5.1 | A custom curve modulator in a preset zone The staircase of MOD-02 on CC 16 at 127 to gain: first with -240 in the instrument zone only, then with -120 more in the preset zone. | The first note is 21 dB below the reference tone. The second is 31.5 dB below it: the preset modulator adds to the instrument one. |
| MOD-17 | 2:51 | 5.5.1, 7.4 | The absolute value transform on a custom curve The reference tone. Then CC 16 as a bipolar source through the linear custom curve, with the absolute value transform, to gain with 240, on a zone with 24 dB of attenuation, swept from 0 to 127. The modulator has the extended flag, so its transform is stored as 0x8002. | A V shape: the reference level at CC 0 and at CC 127, 24 dB below it at CC 64. The transform is the lower 15 bits of the enumerator. |
| MOD-18 | 3:01 | 7.2 | 7-bit scaling: 64 is exactly the center The reference tone. Then a bipolar linear modulator from CC 16 to fineTune with the amount 12700, with CC 16 at 64. | The second note has exactly the pitch of the reference tone: 64 is 8192 of 16384, the center. Scaling 64 by 127 would put it about 100 cents up. |
| MOD-19 | 3:05 | 7.2, 11.3 | The controllers 33 to 63 are sources of their own The reference tone. Then a bipolar linear modulator from CC 17 to fineTune with the amount 6350, with CC 17 at 64 and CC 49 at 127. Then a unipolar linear modulator from CC 49 to fineTune with the amount 1200: CC 49 at 64, then at 127, then CC 17 at 100 after CC 49 at 127. | 0, 600, 1200 and 1200 cents above the reference tone. CC 49 is not the LSB of CC 17, so it does not move the first modulator, and a CC 17 message does not reset it. |
| MOD-20 | 3:12 | 7.2 | Full scale at 127 The reference tone. Then a unipolar linear modulator from CC 16 to fineTune with the amount 1200, at CC 16 = 64, 96 and 127. Then a bipolar linear one with the same amount at 0 and at 127. | 600, 904.8 and 1200 cents above the reference tone, then 1200 cents below and 1200 above it. 64 is the center, 127 is exactly full scale, and the steps above 64 are 1/126 of the range. Scaling by 128 would give 900 and 1190.6 cents. |
| MOD-21 | 3:20 | 7.2 | The Pitch Wheel as a bipolar source The reference tone. Then the Pitch Wheel as a bipolar linear source to gain with -120, on a zone with 12 dB of attenuation, at 0, 4096, 8192, 12288 and 16383. | Equal steps of 6 dB, from the reference level down to 24 dB below it, with no special step at the center: the wheel is its 14-bit value, and 16383 is full scale. |
| MOD-22 | 3:28 | 7.2 | The Pitch Wheel Sensitivity as a source The Pitch Wheel Sensitivity as a unipolar linear source to fineTune with 12700. RPN 0 at 2 semitones, 24 semitones, 12 semitones and 99 cents, and 127 semitones and 127 cents. The last note is at key 0, so its pitch stays below the Nyquist frequency. The reference tone comes first. | 198.4, 2381.3 and 1288.3 cents above the reference tone (12.99 semitones times 128, 1662.7, so the cents of the LSB count, rounded down to 1662: the rounding itself is within the tolerance), and 12700 cents above key 0, about 12.5 kHz: 128.27 semitones would be 16418, and the source stops at 16383, which is full scale. |
| MOD-23 | 3:35 | 7.2 | The Modulation Depth Range as a source The Modulation Depth Range as a unipolar linear source to fineTune with 1600. RPN 5 at 0, 50 cents, 1, 8, 16 and 32 semitones. The reference tone comes first. | 0, 50, 100, 800, 1600 and 1600 cents above the reference tone: the source is min(16383, 8 * v), the center at 8 semitones and full scale from 16. |
| MOD-24 | 3:44 | 7.2 | Velocity, key, poly pressure and channel pressure as sources Each of the Note-On Velocity, the Note-On Key Number, the Poly Pressure and the Channel Pressure as a unipolar linear source to gain with -240, and as a bipolar linear source to fineTune with 1200, at 32, 64, 96 and 127. The key number zone has scaleTuning 0, so every key plays 440 Hz. The reference tone comes first. | For each source: 6, 12, 18.1 and 24 dB below the reference tone, and -600, 0, 609.5 and 1200 cents. 64 is exactly half, so the pitch is exactly 440 Hz there, and 127 is exactly full scale. |
| MOD-25 | 4:03 | 7.3 | Zone modulators against default modulators The bank's DMOD holds CC 21, CC 22 and CC 23, each linear to gain with -120. Each pair of notes moves one of them to 127, first on a zone without modulators, then on a zone with: an instrument zone modulator CC 21 to gain with -240, an instrument zone modulator CC 22 to gain with 0, and a preset zone modulator CC 23 to gain with -120. | The plain zone is 12 dB below the reference tone each time. The CC 21 zone is 24 dB down: an identical instrument modulator replaces the default. The CC 22 zone has the reference level: an amount of 0 switches the default off. The CC 23 zone is 24 dB down: a preset modulator adds to the default. |
| MOD-26 | 4:12 | 7.1 | Modulators to SFe generators CC 16 moves from 0 to 127 halfway through each held note, with a modulator to an SFe generator: gain (-240), filterQOffset (100, the sine on a lowpass cutoff), maxFilterFc (-6000, the noise with an open filter), modLfoToAmplitude (500, 2 Hz) and vibLfoToFilterFc (2400, the noise through a bandpass). | Each note changes when the controller moves: 24 dB quieter, 10 dB louder (the resonance peak at the cutoff, with no gain reduction from filterQOffset), a lowpass at about 620 Hz (7500 absolute cents), a tremolo appears, and a filter sweep appears. |
| MOD-27 | 4:24 | 7.1.2 | delayVoice is read once The click with delayVoice 0.5 s and a modulator from CC 16 to delayVoice with the amount 1200. First CC 16 is 127 at the note on and 0 just after it. Then CC 16 is 0 at the note on and 127 just after it. | The first click comes 1 s after its note on, the second 0.5 s after its note on: only the value at the note on counts. |
| MOD-28 | 4:30 | 7.4 | The absolute value transform The reference tone. Then CC 16 as a bipolar linear source with the absolute value transform (2), to gain with 240, on a zone with 24 dB of attenuation, swept from 0 to 127, pausing on each checked value. | A V shape: the reference level at CC 0 and at CC 127, 24 dB below it at CC 64. |
| MOD-29 | 4:40 | 8.1.3 | The attenuation follows its modulators in the sustain and the release The reference tone. Then the sine with a decay of 0.5 s to sustainVolEnv 120 and a release of 4 s (24 dB per second), a modulator from CC 16 to initialAttenuation with 240 and one from CC 17 to gain with 240. Both controllers are 0 at the note on. CC 16 moves to 127 in the sustain, and after the note off CC 17 moves to 64 in the release. | In the sustain the level falls by 24 dB when CC 16 moves. In the release it rises by 12 dB when CC 17 moves, on top of the falling release: 4.8 dB higher 0.3 s later instead of 7.2 dB lower. The attenuation follows its modulators in every phase of the Volume Envelope. (How a modulator to an envelope generator acts during a note is not defined, spec 8.1.6.) |
| MOD-30 | 4:46 | 7.2 | The switch at 8192, in both directions Two presets with CC 16 through the switch curve to fineTune with 1200, one with the positive direction and one with the negative direction. Each plays with CC 16 at 63, 64 and 127. | The positive switch is off at 63 (440 Hz) and on at 64 and 127 (one octave up). The negative switch is the exact opposite: on at 63, off at 64 and 127. A 7-bit 64 is 8192, where the switch turns. |
| MOD-31 | 4:54 | 7.3, 7.4 | Of two identical modulators the earlier one is ignored The reference tone. Then an instrument zone with two constant modulators to gain, -120 and then -240. Then a preset zone with the same two. Then a zone without modulators while CC 24 is at 127. The bank's DMOD holds CC 24 to gain twice, -240 and then -120. | The first two notes are 24 dB below the reference tone and the third is 12 dB below it. In a zone and in DMOD the later of two identical modulators applies and the earlier one is ignored. Adding them up would give 36 dB, and keeping the first 12 dB in a zone and 24 dB in DMOD. |
sfe4-30-dmod Default modulators
The bank holds the default modulator list of a new SFe bank, untouched. Every expected number is computed from that list, so it stays right if the list changes before the format is frozen.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| DEF-01 | 0:00 | 7.3 | Velocity The sine at velocities 127, 111, 95, 79, 63, 47, 31 and 15, with the default modulator list of a new bank. | The level follows the 96 dB concave curve of the default velocity modulator: 2.36 dB from 127 to 111, and each step below falls further than the one before it, 37 dB down at 15. |
| DEF-02 | 0:10 | 7.3 | Volume and expression The sine with CC 7 and CC 11 at 127, then CC 7 at 100, 64 and 32 with CC 11 at 127, then CC 11 at the same values with CC 7 at 127, then both at 64. | CC 7 and CC 11 give the same concave curve: one of them at 64 is 12 dB below the first note, either one. Together they add: both at 64 is 24.1 dB down. |
| DEF-03 | 0:21 | 7.3 | The pitch wheel The sine with the pitch wheel at the center, fully down, fully up, then fully up again with RPN 0 at 12 semitones. | 198.4 cents down and up at the default range of 2 semitones, then 1190.6 cents up (the note that tells the scale apart from a plain 2 semitones, and from 1200 cents): 12700 times the range over 128 semitones, with the wheel at 16383, which is full scale. |
| DEF-04 | 0:27 | 7.3, 10.5 | Pan The sine with CC 10 at 16, 32, 64, 96 and 112, each next to a zone with the pan the default modulator gives there (-375, -250, 0, 254 and 381, rounded) and CC 10 at 64. | Each pair has the same balance between the channels: the default pan modulator has the amount 500. With the 1000 of SoundFont 2, CC 10 at 32 would already be hard left. |
| DEF-05 | 0:40 | 7.3, 10.8 | The modulation wheel and the Modulation Depth Range The sine with CC 1 at 127: at the default Modulation Depth Range (50 cents), then with RPN 5 at 1 and at 2 semitones. | A vibrato of 50, 100 and 200 cents to each side, within 5 cents: the default vibrato modulator follows RPN 5. |
| DEF-06 | 0:48 | 7.3, 10.4 | Attack time from CC 73 The sine with a Volume Envelope attack of 0.2 s, with CC 73 at 56, 60, 64, 66, 69, 70, 87 and 95. | The attack follows curve A of the default list: 0.1 s at 56 and 0.14 s at 60, exactly 0.2 s from 64 to 69, then longer, about 0.75 s at 87 and 1.4 s at 95. Lower values give attacks of a few milliseconds, which a synthesizer may lengthen against clicks (spec section 8.1), so they are not checked. |
| DEF-07 | 1:06 | 7.3, 10.4 | Decay time from CC 75 and release time from CC 72 The sine with a decay of 0.5 s to silence that starts at once (zero time for the delay, attack and hold), with CC 75 at 0, 32, 56, 60, 64, 65, 67, 110, 111, 120, 126 and 127. Then a release of 0.5 s, with CC 72 at the same values. Each time is read from the levels 0.1 s into the decay or release and 20 dB further down. | Both times follow curve B of the default list: shorter below 64 (half the time at 56, 0.71 times at 60), unchanged and flat from 67 to 110, then longer, up to about 3.8 s for 96 dB at 127. The decays at CC 75 = 0 and 32 and the releases at CC 72 = 0 and 32 (2 and 31 ms for 96 dB) are not checked, since a synthesizer may lengthen such short times against clicks (spec section 8.1). |
| DEF-08 | 1:58 | 7.3, 10.7 | Brightness The noise through a lowpass at 8000 (about 830 Hz), with CC 74 at 0, 32, 64, 96 and 127, each next to a zone with the cutoff moved by what the default modulator gives there. | Each pair sounds the same: the cutoff moves 4800 cents down at 0, 2400 down at 32, not at 64, 2438 up at 96 and 4800 up at 127. |
| DEF-09 | 2:10 | 7.3, 10.3 | Resonance The noise through a lowpass at 8000 (about 830 Hz), with CC 71 at 0, 64 and 127, each next to a zone with the filterQOffset that the default CC 71 modulator gives there. | Each pair sounds the same: the default CC 71 modulator acts only above 64, so CC 71 at 0 plays like 64 (an offset of 0 cB), and at 127 the resonance moves 250 cB up, a peak at the cutoff. |
| DEF-10 | 2:18 | 7.3, 7.1.1 #63, 10.3 | CC 71 leaves the level alone The sine at key 33 (55 Hz), far below a lowpass cutoff of 13500, with initialFilterQ 300. CC 71 at 64 first, then at 63, 64, 96, 127, 32 and 0. | Every note has the level of the first: the default CC 71 modulator moves filterQOffset, which does not lower the gain, and initialFilterQ 300 lowers it by 15 dB in every note alike. There is no step at 64. |
| DEF-11 | 2:27 | 7.3 | The soft pedal The noise through a lowpass at 9000 (1.5 kHz), with CC 67 at 0, 63 and 127, each next to a zone with the cutoff the default modulator gives there. The value 64 itself is DEF-16. | Each pair sounds the same: the cutoff is one octave lower at 127 and unchanged at 0 and 63, a switch. |
| DEF-12 | 2:35 | 10.6 | No velocity to the filter cutoff The noise through a lowpass at 9000 (1.5 kHz), at velocities 127, 64 and 16. | Only the level changes: the balance between low and high frequencies is the same in the three notes, so the cutoff does not move with the velocity. |
| DEF-13 | 2:40 | 10.9 | No channel pressure vibrato The sine held for 1.8 s with the channel pressure at 127. | No vibrato: the pitch stays at 440 Hz. |
| DEF-16 | 2:44 | 7.3 | The soft pedal at exactly 64 The noise through a lowpass at 9000 (1.5 kHz), with CC 67 at 64, next to a zone with the cutoff one octave lower. | The pair sounds the same: the switch is on at 64. |
sfe4-31-dmod-empty An empty default modulator list
Section 7.3
The bank's DMOD holds only the terminal record, so it has no default modulators at all.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| DEF-14 | 0:00 | 7.3 | An empty DMOD In a bank whose DMOD holds only the terminal record: the sine at velocity 127, then at velocity 16, with CC 7 at 0, with the pitch wheel fully up, with CC 10 at 0, and with CC 1 at 127. | All six notes are the same: the same level, 440 Hz, centered, no vibrato. The synthesizer adds no default modulator of its own. |
sfe4-32-dmod-custom A default modulator list of the bank's own
Section 7.3
The bank's DMOD holds two modulators of its own in place of the usual list.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| DEF-15 | 0:00 | 7.3 | A DMOD of the bank's own In a bank whose DMOD holds Note-On Velocity to fineTune with 1200, and Note-On Velocity to initialAttenuation with 480 in place of 960: the sine at velocities 127, 96, 64 and 32. | The pitch rises with the velocity (600 cents at 64, an octave at 127), and the level follows a 48 dB concave curve: half as far down as with the usual list, 6 dB at 64. |
sfe4-40-sends Effects sends
Sections 7.1.1 #89, 7.2, 7.3, 8.1.4, 10.2
The effects are on in this render, and the preamble leaves the sends alone. The effects processors are the synthesizer's own, so the criterion is the amount of wet signal after each click, not its sound: the energy of the whole tail in both channels, from 3 ms after the click, when the dry click has ended. The bank holds the default modulator list of a new bank.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| SND-01 | 0:00 | 8.1.4, 10.2 | reverbEffectsSend The click with reverbEffectsSend at 1000, 500, 250, 0, 6 s apart, with the controllers of all three sends at 0. | The reverb tail after each click follows the send: the send scales the amplitude, so the tail is 6 dB weaker for half the amount and 12 dB for a quarter (within about 0.5 dB), and there is none at 0. 1000 is a full send. |
| SND-02 | 0:30 | 7.2, 7.3, 10.2 | CC 91 as the reverb send First the click with reverbEffectsSend at 1000 and the send controllers at 0, as the full send. Then, 6 s apart, the click with reverbEffectsSend at 0 and CC 91 at 127, 64, 32 and 0, the other send controllers at 0. The default modulator sends CC 91 to it with 1000. | Each tail is compared with the first one. The controller works in the same range as the generator: CC 91 at 127 is a full send, 64 half of it (6 dB weaker in amplitude), 32 a quarter (12 dB), 0 none, each within about 0.5 dB. The send of an SFe file is not multiplied as a SoundFont 2 one may be. |
| SND-03 | 1:06 | 8.1.4, 10.2 | chorusEffectsSend The click with chorusEffectsSend at 1000, 500, 250, 0, 6 s apart, with the controllers of all three sends at 0. | The chorus tail after each click follows the send: the send scales the amplitude, so the tail is 6 dB weaker for half the amount and 12 dB for a quarter (within about 0.5 dB), and there is none at 0. 1000 is a full send. |
| SND-04 | 1:36 | 7.2, 7.3, 10.2 | CC 93 as the chorus send First the click with chorusEffectsSend at 1000 and the send controllers at 0, as the full send. Then, 6 s apart, the click with chorusEffectsSend at 0 and CC 93 at 127, 64, 32 and 0, the other send controllers at 0. The default modulator sends CC 93 to it with 1000. | Each tail is compared with the first one. The controller works in the same range as the generator: CC 93 at 127 is a full send, 64 half of it (6 dB weaker in amplitude), 32 a quarter (12 dB), 0 none, each within about 0.5 dB. The send of an SFe file is not multiplied as a SoundFont 2 one may be. |
| SND-05 | 2:12 | 7.1.1 #89, 7.1.2, 8.1.4 | variationEffectsSend The click with variationEffectsSend at 2000, 1000, 500, 250, 0, 6 s apart, with the controllers of all three sends at 0. The value 2000 is stored beyond the range. | The variation tail after each click follows the send: the send scales the amplitude, so the tail is 6 dB weaker for half the amount and 12 dB for a quarter (within about 0.5 dB), and there is none at 0. 1000 is a full send. 2000 is limited to 1000, so the first two tails are the same. A synthesizer may have no third effects processor. It then ignores the send and plays no tail. |
| SND-06 | 2:48 | 7.1.1 #89, 7.2, 7.3, 10.2 | CC 94 as the variation send First the click with variationEffectsSend at 1000 and the send controllers at 0, as the full send. Then, 6 s apart, the click with variationEffectsSend at 0 and CC 94 at 127, 64, 32 and 0, the other send controllers at 0. The default modulator sends CC 94 to it with 1000. | Each tail is compared with the first one. The controller works in the same range as the generator: CC 94 at 127 is a full send, 64 half of it (6 dB weaker in amplitude), 32 a quarter (12 dB), 0 none, each within about 0.5 dB. The send of an SFe file is not multiplied as a SoundFont 2 one may be. A synthesizer may have no third effects processor. It then ignores the send and plays no tail. |
sfe4-50-midi MIDI behavior
The bank has no default modulators. It holds probe presets: each maps one source (a controller, the pitch wheel, a pressure, the Pitch Wheel Sensitivity or the Modulation Depth Range) to the pitch through its own modulator with the amount 6400, tuned so that the default of spec section 11.3 plays 440 Hz. A step of a 7-bit value is 50 cents below 64 and about 50.8 cents above it (spec section 7.2). The keyed drum preset has a zone per key with scaleTuning 0, 50 cents apart, so the key that plays shows in the pitch. The preamble does not zero CC 91 and CC 93, since their defaults are tested.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| MID-01 | 0:00 | 11.3 | Controller defaults after a reset After the GS Reset of the preamble, one probe note per controller of the defaults table (CC 7, 8, 10, 11, 71 to 78, 81, 83 and 91). Each probe maps its controller to the pitch, tuned so that the default of spec section 11.3 plays 440 Hz, and each step away from it 50 cents. | Every probe plays 440 Hz. |
| MID-02 | 0:06 | 11.3 | Defaults of 0, the LSBs and the other sources After the GS Reset, probes for the controllers that default to 0 (CC 1, 2, 16, 70, 93 and 94), for CC 33, 39, 42, 43 and 48 (named the LSBs of CC 1, 7, 10, 11 and 16, but sources of their own), and for the pitch wheel, the channel pressure, the poly pressure, the Pitch Wheel Sensitivity and the Modulation Depth Range. | Every probe plays 440 Hz: the controllers, CC 33 to 63 included, at 0, the pitch wheel at its center, both pressures at 0, a sensitivity of 2 semitones and a Modulation Depth Range of 50 cents. |
| MID-03 | 0:12 | 11.1 | Each of the eight mode messages is a full system reset For each of the eight messages of spec section 11.1: every controller, CC 33 to 63, the pitch wheel, both pressures and RPN 0 and 5 are moved, the message is sent, and every probe of MID-01 and MID-02 plays. | After each message every probe plays 440 Hz: each message puts every source back to its default. |
| MID-04 | 1:27 | 11.1 | GS System Mode Set CC 74 moved to 80, then GS System Mode Set with m = 0, and the CC 74 probe. The same with m = 1. Then CC 74 to 80 again and the message with m = 2, and the probe. | After m = 0 and m = 1 the probe plays 440 Hz: both reset. After m = 2 it plays 813 cents up (CC 74 at 80): that message is ignored. |
| MID-05 | 1:32 | 11 | Any device id Three times each: CC 74 moved to 80 and a GS Reset with the device id 0x10, 0x11 and 0x7F, then the CC 74 probe. CC 74 moved to 80 and GM1 System On with the device id 0x7F, 0x00 and 0x10, then the CC 74 probe. GS PART LEVEL of channel 1 at 50 with the device id 0x10, 0x11 and 0x7F, then the CC 7 probe. XG VOLUME of part 0 at 50 with the device number 0, 1 and 15, then the CC 7 probe. A GS Reset with the usual id comes before each of the last two. | Every message is received: the CC 74 probe plays 440 Hz after each GS Reset and each GM1 System On, and the CC 7 probe plays 2529 cents down (CC 7 at 50) after each part parameter. |
| MID-06 | 1:47 | 7.2, 11.3 | CC 10 at 0 and at 127 The CC 10 probe with CC 10 at 0, then at 1, then at 127. | 3200 cents down, 3150 cents down and 3200 cents up: 0 is the low end of its own and is not treated as 1, and 127 is exactly as far above 64 as 0 is below it. |
| MID-07 | 1:50 | 11.4 | A data entry applies with either half RPN 0 selected and only a data entry MSB of 4 sent, then the Pitch Wheel Sensitivity probe. Then RPN 0 selected again and only a data entry LSB of 50 sent, then the probe. | The first probe plays 100 cents up (4 semitones). The second plays 75 cents down (0 semitones and 50 cents): each half applies when it arrives, and selecting the parameter set the stored MSB to 0. |
| MID-08 | 1:53 | 11.4 | Modulation Depth Range The Modulation Depth Range probe with RPN 5 at 1 semitone and LSB 64, 0 semitones and LSB 32, and 2 semitones. | 400 cents up, 100 cents down and 600 cents up: the range is the MSB in semitones plus the LSB in 100/128 cents, read as min(16383, 8 * v). |
| MID-09 | 1:56 | 11.4 | The null parameter, and selecting resets the data entry RPN 0 selected with a data entry of MSB 12 and LSB 64, then the Pitch Wheel Sensitivity probe. Then RPN null (127, 127) selected with a data entry MSB of 5, and the probe. Then RPN 0 selected again and only a data entry MSB of 2 sent, and the probe. | 531.6 cents up twice: 12 semitones and 64 cents (12.64 times 128 is 1617.9, rounded down), and the data entry after the null parameter does nothing. Then 440 Hz: 2 semitones and 0 cents, since selecting RPN 0 set the stored LSB of 64 back to 0. |
| MID-10 | 1:59 | 11.4 | The last selected parameter gets the data entry RPN 0 selected with a data entry MSB of 4. Then the NRPN for CC 74 (1, 32) selected with a data entry MSB of 100. Then the CC 74 probe and the Pitch Wheel Sensitivity probe. | The CC 74 probe plays 1828.5 cents up (CC 74 at 100) and the sensitivity probe 100 cents up (still 4 semitones): the NRPN got the second data entry and the RPN kept its value. |
| MID-11 | 2:02 | 11.4 | The NRPNs that set a controller For each of the eight NRPNs (1, 8), (1, 9), (1, 10), (1, 32), (1, 33), (1, 99), (1, 100) and (1, 102): the NRPN with a data entry MSB of 90 and the probe of its controller. Then the controller set back to 64, the controller itself sent at 90, and the probe. | Each pair plays the same pitch, 1320.7 cents up (90): the NRPN sets CC 76, 77, 78, 74, 71, 73, 75 and 72 to its data MSB. |
| MID-12 | 2:12 | 11.5 | The sustain pedal holds from 64 A note of the sine of 0.3 s with a release of 0.1 s and CC 64 at 63. Then the same note with CC 64 at 64, and the pedal released about 0.65 s after its note off. | With 63 the note ends right after its note off. With 64 it is held after its note off and ends when the pedal goes up. |
| MID-13 | 2:17 | 11.5 | All Sound Off cuts, All Notes Off releases A held note with a release of 2 s, ended by CC 120. Then the same note ended by CC 123. | CC 120 cuts the note at once, silent 0.1 s later. CC 123 releases it: it falls 48 dB per second, 14.4 dB from 0.2 s to 0.5 s after the message. A synthesizer may smooth the cut against clicks (spec section 8.1), and a cut that takes longer than 0.1 s fails here. |
| MID-14 | 2:23 | 11.5 | All Notes Off with the pedal down A note with a release of 0.1 s and CC 64 at 127. CC 123 at 0.5 s, the pedal released at 1.5 s, the note off at 2 s. | The note goes on after CC 123 and ends when the pedal goes up: All Notes Off acts as a note off, which the pedal holds. |
| MID-15 | 2:28 | 11.5 | Omni and mono messages A held note with a release of 2 s, ended by CC 124, then by CC 125, CC 126 and CC 127. | All four release the note, as CC 123 does: it falls 48 dB per second, 14.4 dB from 0.2 s to 0.5 s after the message. CC 126 and 127 do not cut it, as MIDI 1.0 has them act as All Notes Off. |
| MID-16 | 2:38 | 11.5 | Reset All Controllers A preset of its own is selected (300 cents up), RPN 0 is set to 4 semitones, and every controller, the pitch wheel and both pressures are moved. RPN 0 stays selected. Then CC 121, a data entry MSB of 12, and a note without a Program Change. Then probes for CC 1, 11, 33, 43 and 64 to 67, for CC 7, 10, 70 to 79 and 91 to 95, for the pitch wheel, both pressures and the Pitch Wheel Sensitivity. | The note after CC 121 plays the selected preset, 300 cents up: the bank and the program are kept. The probes of CC 1, 11, 33, 43, 64 to 67, the pitch wheel and both pressures play 440 Hz: they are reset. The probes of CC 7, 10, 70 to 79 and 91 to 95 keep their moved values, and the sensitivity probe plays 100 cents up: the data entry after CC 121 went nowhere, since the selected parameter became null, and RPN 0 kept its 4 semitones. |
| MID-17 | 2:48 | 11.6.2 | The GS part parameters Each GS part parameter of spec section 11.6.2 for channel 1, then the channel message it stands for, each after a GS Reset of its own and followed by a note on a probe. TONE NUMBER and TONE MAP NUMBER select presets that only their bank values reach. USE FOR RHYTHM PART belongs to the preset selection group. | The two notes of each pair have the same pitch: each parameter does exactly what its channel message does. |
| MID-18 | 3:07 | 11.6.2 | GS part numbering For each of the 16 parts: a GS Reset, GS PART LEVEL at 50 for the part, and a note on the channel that part names, with the CC 7 probe (on channel 10 the drum preset, which shows CC 7 as well). Part 0 is channel 10, parts 1 to 9 are channels 1 to 9, and parts A to F are channels 11 to 16. | Every note plays 2529 cents down (CC 7 at 50): the parameter reached the channel its part number names. |
| MID-19 | 3:17 | 11.6.3 | The XG part parameters Each XG part parameter of spec section 11.6.3, on parts 0 to 15 in turn (part 9 is skipped, channel 10 being a drum channel), then the channel message it stands for on the same channel, each after a GS Reset of its own and followed by a note on a probe. PART MODE belongs to the preset selection group. | The two notes of each pair have the same pitch: each parameter does exactly what its channel message does, and part pp is channel pp + 1. |
| MID-20 | 3:36 | 11.6.2, 11.6.3 | The value rules of the part parameters GS MOD LFO1 PITCH DEPTH at 1, 10 and 127 and XG MW LFO PMOD DEPTH at 40 and 100, each read by the Modulation Depth Range probe. GS BEND PITCH CONTROL at 60, 64, 88 and 100, read by the Pitch Wheel Sensitivity probe. | The GS depths set RPN 5 to 6, 64 and 813 (round(6.4 * value)), and both XG depths to 256 (the value stops at 40). The bends are 0, 0, 24 and 24 semitones: value - 64, from 0 to 24. |
| MID-21 | 3:47 | 11 | What GM mode ignores After GM1 System On: CC 0 at 6 with a Program Change to 110 (only bank 6 has the preset 200 cents up, bank 0 has it 100 cents up), and CC 32 at 3 with a Program Change to 111 (bank LSB 3 has it 250 cents up, bank 0 has it 50 cents up). The NRPN for CC 74 with 100, then the CC 74 probe. GS PART LEVEL and XG VOLUME at 50, then the CC 7 probe. Then RPN 0 at 4 semitones and the sensitivity probe, and a GS Reset, after which CC 0 at 6 with Program Change 110 again. | In GM both bank selects, the NRPN and both part parameters are ignored: 100 and 50 cents up, 440 Hz and 440 Hz. The RPN works (100 cents up), and the GS Reset works: afterwards the bank select is used again (200 cents up). |
| MID-22 | 3:58 | 11.5 | A Note On with the velocity 0 is a note off The reference tone at key 69, held from 0.05 s with its own note off at 1.5 s. At 0.6 s a Note On for key 69 with the velocity 0. | The note ends at 0.6 s and is silent from 0.7 s on: the Note On with the velocity 0 is its note off. A synthesizer that ignores it, or starts a new note with it, keeps a note sounding after 0.6 s. |
sfe4-51-midi-ports MIDI ports
The bank has no default modulators. It holds probe presets: each maps one source (a controller, the pitch wheel, a pressure, the Pitch Wheel Sensitivity or the Modulation Depth Range) to the pitch through its own modulator with the amount 6400, tuned so that the default of spec section 11.3 plays 440 Hz. A step of a 7-bit value is 50 cents below 64 and about 50.8 cents above it (spec section 7.2). The keyed drum preset has a zone per key with scaleTuning 0, 50 cents apart, so the key that plays shows in the pitch.
Player: This MIDI file uses two ports. Play it with a player that maps the MIDI port meta event of each track to a group of 16 channels, which is a convention outside the specification.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| MID-23 | 0:00 | 11.6.2, 11.6.3 | The second group of 16 channels A MIDI file with two ports, channels 17 to 32 on the second. GS PART LEVEL at 50 with the address 50 11 19 (part 1 of the second group), then the CC 7 probe on channel 17 and on channel 1. After a GS Reset, XG VOLUME at 50 for part 16, then the CC 7 probe on channel 17 and on channel 1. After a GS Reset, key 70 on channel 26, channel 10 of the second group, which holds the keyed drum preset. | Channel 17 plays 2529 cents down (CC 7 at 50) both times and channel 1 stays at 440 Hz: the address 50 and the XG parts 16 to 31 reach the second group. Key 70 on channel 26 plays the drum preset, 50 cents up: channel 10 of every group of 16 channels is a drum channel after a reset. |
sfe4-52-midi-startup The starting state
The bank has no default modulators. It holds probe presets: each maps one source (a controller, the pitch wheel, a pressure, the Pitch Wheel Sensitivity or the Modulation Depth Range) to the pitch through its own modulator with the amount 6400, tuned so that the default of spec section 11.3 plays 440 Hz. A step of a 7-bit value is 50 cents below 64 and about 50.8 cents above it (spec section 7.2). The keyed drum preset has a zone per key with scaleTuning 0, 50 cents apart, so the key that plays shows in the pitch. This file sends no reset message.
Player: Play this file on a freshly started synthesizer, with nothing played on it before.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| MID-24 | 0:00 | 11.1, 11.3 | The state a synthesizer starts in No reset message at all. Every probe of MID-01 and MID-02, then a request of bank MSB 0, bank LSB 8 and program 112, then one of bank MSB 8, bank LSB 0 and program 112. The bank holds program 112 on bank 0 (100 cents up) and on bank MSB 8 (200 cents up), but nothing on LSB 8. | Every probe plays 440 Hz. The first request plays the preset on bank 0, 100 cents up (should, a fallback of rule 3): the synthesizer starts in GS, where the request falls back to the same MSB (XG would pick bank MSB 8). The second plays the preset on bank MSB 8, 200 cents up (GM would ignore the bank select). |
sfe4-60-select-full Preset selection: the full bank
Each step sends its messages on channel 1 or 10 and then plays a note there. Every preset of the bank plays the sine with scaleTuning 0 and a coarseTune of its own, so the pitch of the note names the preset. A marker in the MIDI file names the expected preset of each step, and the pass criterion lists the expected pitch in cents from 440 Hz. Rules 1 and 2 are required. The fallback lists of rule 3 are a recommendation (spec 11.2.3): without an exact match a synthesizer may select another preset, or none, and the answers marked "should" only hold for one that follows the lists. On XG channel 10 a fallback should not be a melodic preset while the file has a drum preset. This bank holds melodic presets over several programs and banks, GM/GS drum presets, XG drum presets at bank MSB 120 and 127, a preset at bank MSB 126 and two presets with the same four values. It passes the XG test.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| SEL-01 | 0:00 | 6.1 | wBank holds the bank MSB, the bank LSB and the drum flag In GS, program 1 with bank MSB 8, then with bank LSB 8. Then program 0 on channel 1 and on channel 10, which has the drum flag. Every preset of the bank plays its own pitch. | Four different presets: the MSB, the LSB and the drum flag each select a preset of their own. Expected, step by step: MSB 8, program 1: D 1 MSB 8 (-1200 cents). LSB 8, program 1: C 1 LSB 8 (-1100 cents). program 0 on channel 1: capital 0 (-1600 cents). program 0 on channel 10: GS drum 0 (-1800 cents). |
| SEL-02 | 0:04 | 11.2.3 | Rule 2, an exact match, in all four modes Program 1 requested in GM with bank MSB 8 (ignored), in GS with MSB 8, in GM2 with LSB 8, and in XG with MSB 5 and LSB 3. | Each mode plays the preset with exactly the values it uses. Expected, step by step: GM, MSB 8 (ignored): A 1 (-1500 cents). GS, MSB 8: D 1 MSB 8 (-1200 cents). GM2, LSB 8: C 1 LSB 8 (-1100 cents). XG, MSB 5 and LSB 3: F 1 MSB 5 LSB 3 (-1300 cents). |
| SEL-03 | 0:08 | 11.2.2 | Bank select only stores In GS, program 1 on bank 0. Then CC 0 at 8 without a Program Change, then a Program Change to 1. Then CC 0 at 0 and CC 32 at 8 without a Program Change, then a Program Change to 1. Last, a GS Reset and a note without a Program Change. | The second note is still the first preset: CC 0 only stores. The third plays the preset of bank MSB 8. The fourth is still that one: CC 32 only stores too. The fifth plays the preset of bank LSB 8. After a GS Reset with no Program Change, the sixth plays program 0 on bank 0: a reset selects again. Expected, step by step: program 1: A 1 (-1500 cents). CC 0 at 8 only: A 1 (-1500 cents). Program Change 1: D 1 MSB 8 (-1200 cents). CC 0 at 0 and CC 32 at 8 only: D 1 MSB 8 (-1200 cents). Program Change 1 again: C 1 LSB 8 (-1100 cents). GS Reset, no Program Change: capital 0 (-1600 cents). |
| SEL-04 | 0:14 | 11.1 | Each mode message sets its mode After each of the eight mode messages: on channel 1 bank MSB 16, LSB 3, program 1, and on channel 10 program 9. The two requests together tell GM, GS, GM2 and XG apart. | GM1 System On and GS Exit give the GM answers, GM System Off, GS Reset and GS System Mode Set the GS answers, GM2 System On the GM2 answers, and both XG messages the XG answers. Most of these answers come from the fallbacks of rule 3 (should). Expected, step by step: GM1 System On, channel 1: A 1 (-1500 cents). GM1 System On, channel 10: XG 9 MSB 120 (-2200 cents). XG System On, channel 1: F 1 MSB 5 LSB 3 (-1300 cents). XG System On, channel 10: XG 9 MSB 127 (-2300 cents). GM System Off, channel 1: B 1 MSB 16 (-1400 cents). GM System Off, channel 10: XG 9 MSB 120 (-2200 cents). XG All Parameter Reset, channel 1: F 1 MSB 5 LSB 3 (-1300 cents). XG All Parameter Reset, channel 10: XG 9 MSB 127 (-2300 cents). GM2 System On, channel 1: F 1 MSB 5 LSB 3 (-1300 cents). GM2 System On, channel 10: XG 9 MSB 120 (-2200 cents). GS Reset, channel 1: B 1 MSB 16 (-1400 cents). GS Reset, channel 10: XG 9 MSB 120 (-2200 cents). GS Exit, channel 1: A 1 (-1500 cents). GS Exit, channel 10: XG 9 MSB 120 (-2200 cents). GS System Mode Set, channel 1: B 1 MSB 16 (-1400 cents). GS System Mode Set, channel 10: XG 9 MSB 120 (-2200 cents). |
| SEL-05 | 0:26 | 11.2.2 | The default bank of each mode After each mode's reset, only a Program Change to 2. Then in GM, bank LSB 8 with program 1. | Program 2 on bank 0 in GM, GS and XG, and on bank MSB 121 in GM2: the default bank MSB of each mode. GM ignores the LSB. Expected, step by step: GM1 System On, program 2: G 2 (-1000 cents). GS Reset, program 2: G 2 (-1000 cents). GM2 System On, program 2: H 2 MSB 121 (-900 cents). XG System On, program 2: G 2 (-1000 cents). GM, LSB 8 and program 1: A 1 (-1500 cents). |
| SEL-06 | 0:31 | 11.2.2 | Channel 10 after each mode's reset After each mode's reset, program 17 on channel 10. The bank has program 17 as a GM/GS drum preset and as XG drum presets at bank MSB 120 and 127. | GM and GS play the GM/GS drum preset (the drum flag), GM2 the one at MSB 120, XG the one at MSB 127. Expected, step by step: GM1 System On, channel 10: GS drum 17 (200 cents). GS Reset, channel 10: GS drum 17 (200 cents). GM2 System On, channel 10: XG 17 MSB 120 (-2000 cents). XG System On, channel 10: XG 17 MSB 127 (-2100 cents). |
| SEL-07 | 0:35 | 11.2.3 | Rule 3.1, a request with the drum flag (the full bank) In GS on channel 10, with bank MSB 8 so no request has an exact match: program 30 with bank LSB 5, then program 17 with LSB 7 (a GM/GS drum preset and XG drum presets have program 17), then program 33 (only XG drum presets), then program 57 (only an XG drum preset at MSB 127, above 48), then program 58 (XG drum presets at MSB 127 and 120, above 48), then program 34 (no drum preset). The other banks of this test hold no GM/GS drum presets, or no drum presets at all. | Rule 3.1 in order (should): the GM/GS drum preset with the same program and LSB, then the same program, then any drum preset with the same program if the program is below 49 or the preset is at bank MSB 120 (so program 57 skips its XG drum preset and program 58 takes the one at MSB 120), then the first GM/GS drum preset, then the first drum preset, then the first preset. In the bank of XG drum presets only, program 9 is only at bank MSB 127, so the first drum preset is that one (by program before MSB), not program 17 at MSB 120. Expected, step by step: 3.1.1, MSB 8, program 30 LSB 5: GS drum 30 LSB 5 (400 cents). 3.1.2, program 17 LSB 7: GS drum 17 (200 cents). 3.1.3, program 33: XG 33 MSB 120 (800 cents). 3.1.3 skipped above 48, program 57: GS drum 0 (-1800 cents). 3.1.3 at MSB 120 above 48, program 58: XG 58 MSB 120 (1200 cents). 3.1.4, program 34: GS drum 0 (-1800 cents). |
| SEL-08 | 0:41 | 11.2.3 | Rule 3.2, drum requests in XG and GM2 (the full bank) On channel 10 in XG and in GM2: program 58 with bank LSB 5 (no exact match, and above program 48, so only rule 3.2.1 finds the XG drum preset of program 58), program 30 and program 48 (GM/GS drum presets below program 49), programs 49 and 60 (GM/GS drum presets from program 49 on, which rule 3.2.2 only takes in GM2), and program 34. Rule 1 is not reached from MIDI here: no message sets the drum flag in XG or GM2. | Rule 3.2 in order (should): the XG drum preset with the same program, then a drum preset with the same program (below program 49, or any program in GM2, which numbers its kits as GS does), then the first XG drum preset, then the first drum preset, then the first preset. Among the XG drum presets of one program, the first plays, the one at MSB 120. Expected, step by step: XG System On: 3.2.1, program 58 LSB 5: XG 58 MSB 120 (1200 cents). XG System On: 3.2.2, program 30: GS drum 30 (300 cents). XG System On: 3.2.2, program 48: GS drum 48 (500 cents). XG System On: 3.2.3, program 49: XG 9 MSB 120 (-2200 cents). XG System On: 3.2.3, program 60: XG 9 MSB 120 (-2200 cents). GM2 System On: 3.2.1, program 58 LSB 5: XG 58 MSB 120 (1200 cents). GM2 System On: 3.2.2, program 30: GS drum 30 (300 cents). GM2 System On: 3.2.2, program 48: GS drum 48 (500 cents). GM2 System On: 3.2.2, program 49: GS drum 49 (600 cents). GM2 System On: 3.2.2, program 60: GS drum 60 (-1900 cents). |
| SEL-09 | 0:49 | 11.2.3 | Rule 3.3 in GM and GS In GS, program 3 (presets at MSB 0 LSB 1, MSB 0 LSB 2, MSB 16 and MSB 20) requested with MSB 9 LSB 2, MSB 16 LSB 20 (the MSB matches, the higher value 20 matches another preset), MSB 3 LSB 20 and MSB 4 LSB 4. Then program 1 with MSB 3 LSB 1 (the preset at MSB 5 LSB 3 has the higher value as its LSB), and program 4 with MSB 10 LSB 64 (GS does not skip rule 3.3.3 for LSB 64). Then in GM, program 3 with MSB 16. | Each request is answered by its own step of rule 3.3 (should): bank MSB 0 with the same LSB, the same MSB, the MSB or LSB equal to the higher of the two, and the first candidate. GM ignores the bank and plays the first candidate on bank MSB 0, by rule 3.3.2. Expected, step by step: 3.3.1, MSB 9 LSB 2: I 3 LSB 2 (-800 cents). 3.3.2, MSB 16 LSB 20: J 3 MSB 16 (-600 cents). 3.3.3, MSB 3 LSB 20: K 3 MSB 20 (-500 cents). 3.3.4, MSB 4 LSB 4: I0 3 LSB 1 (-700 cents). 3.3.3, the higher value 3 equals an LSB, program 1 MSB 3 LSB 1: F 1 MSB 5 LSB 3 (-1300 cents). 3.3.3 with LSB 64, program 4 MSB 10: N 4 MSB 64 (-200 cents). GM, MSB 16: I0 3 LSB 1 (-700 cents). |
| SEL-10 | 0:55 | 11.2.3 | Rule 3.3 in XG and GM2 In XG, program 4 (presets at bank 0, at MSB 6 LSB 9 and at MSB 64) requested with MSB 2 LSB 9, MSB 6 LSB 1, MSB 3 LSB 2, MSB 6 LSB 0 and MSB 0 LSB 9 (the same LSB, never the same MSB), MSB 3 LSB 6 (the higher value is the LSB) and MSB 10 LSB 64. Then in GM2 with MSB 2 LSB 9 and with MSB 10 LSB 64. | The same LSB, then the MSB or LSB equal to the higher of the two, then the first candidate (should). With LSB 64 the third step is skipped, so the first candidate plays, not the preset at MSB 64. Expected, step by step: 3.3.2, MSB 2 LSB 9: M 4 MSB 6 LSB 9 (-300 cents). 3.3.3, MSB 6 LSB 1: M 4 MSB 6 LSB 9 (-300 cents). 3.3.4, MSB 3 LSB 2: L 4 (-400 cents). 3.3.2 by the LSB, MSB 6 LSB 0: L 4 (-400 cents). 3.3.2 by the LSB, MSB 0 LSB 9: M 4 MSB 6 LSB 9 (-300 cents). 3.3.3, the higher value 6 equals an MSB, MSB 3 LSB 6: M 4 MSB 6 LSB 9 (-300 cents). LSB 64 skips 3.3.3: L 4 (-400 cents). GM2: 3.3.2, MSB 2 LSB 9: M 4 MSB 6 LSB 9 (-300 cents). GM2: LSB 64 skips 3.3.3: L 4 (-400 cents). |
| SEL-11 | 1:02 | 11.2.3 | Rule 3.3 without a candidate In GS on channel 1, program 5, which has no melodic preset. | The first preset plays (should): program 0 on bank 0, although other presets come first in PHDR. Expected, step by step: program 5: capital 0 (-1600 cents). |
| SEL-12 | 1:05 | 11.2.3 | Rule 2 in XG and GM2 with MSB 127 or 120 and an exact match that is not an XG drum preset (the full bank) On channel 1, bank MSB 127 with program 9 in XG, and bank MSB 120 with program 9 in GM2. In the xgfail bank, which fails the XG test, the presets at MSB 120 and 127 are melodic, apart from its preset at MSB 127 program 10 with the drum flag set, a GM/GS drum preset. | With the full bank the exact match plays, an XG drum preset. With the xgfail bank the exact match is skipped, and rule 3.2 picks the first drum preset (should), a GM/GS drum preset. Expected, step by step: XG, MSB 127 program 9: XG 9 MSB 127 (-2300 cents). GM2, MSB 120 program 9: XG 9 MSB 120 (-2200 cents). |
| SEL-13 | 1:08 | 11.2.1, 11.2.3 | The XG test (the full bank) In GS on channel 10 (the drum flag), bank MSB 8 with program 9, which no GM/GS drum preset has. Only the presets at bank MSB 120 and 127 have program 9, and the xgfail bank holds a preset at MSB 127 with a program outside the XG list. | Rule 3.1 (should). With the full bank the presets at MSB 120 and 127 are XG drum presets, so rule 3.1.3 plays the one of program 9 at MSB 120. With the xgfail bank they are melodic, so rule 3.1.3 finds nothing and rule 3.1.4 plays the first GM/GS drum preset. Expected, step by step: GS channel 10, MSB 8 program 9: XG 9 MSB 120 (-2200 cents). |
| SEL-14 | 1:11 | 11.2.1 | MSB 126, two presets with the same values, and the order of the first In GS: bank MSB 126 with program 6. Then program 7, which two presets share with the same four values. Then program 5, which has no melodic preset. In PHDR the GM/GS drum preset of program 0 comes before the melodic program 0 on bank 0, with the same program, MSB and LSB. Last, MSB 126 with program 9, which has no melodic preset, on channel 1 in XG and on channel 10 in GM2. | The MSB 126 preset plays: an exact match selects it. Program 7 plays the first of the two presets in PHDR. Program 5 plays the first preset (should): the melodic program 0, since the GM/GS drum presets come last in that order, whatever PHDR says. MSB 126 is not a drum bank outside XG channel 10: both requests go through rule 3.3 and play the first preset (should), not an XG drum preset of program 9. Expected, step by step: MSB 126 program 6: O 6 MSB 126 (-100 cents). program 7: Q1 7 (0 cents). program 5: capital 0 (-1600 cents). XG channel 1, MSB 126 program 9: capital 0 (-1600 cents). GM2 channel 10, MSB 126 program 9: capital 0 (-1600 cents). |
| SEL-15 | 1:16 | 11.2.2 | XG channel 10 In XG on channel 10: CC 0 at 5 with program 9, then CC 0 at 126 with program 6 (an exact match), then CC 0 at 126 with program 9 (no exact match), then CC 0 at 127 and 120 with program 9, then CC 0 at 5 again with program 9. | CC 0 at 5 is ignored, so the XG drum preset at MSB 127 plays. MSB 126 with an exact match plays that melodic preset. MSB 126 without one goes through rule 3.2 and plays an XG drum preset of program 9 (should: the first, at MSB 120). CC 0 at 120 is taken, as 126 and 127 are, so that note plays the XG drum preset at MSB 120. The last CC 0 at 5 is ignored again, so the channel stays at MSB 120. Expected, step by step: CC 0 at 5, program 9: XG 9 MSB 127 (-2300 cents). CC 0 at 126, program 6: O 6 MSB 126 (-100 cents). CC 0 at 126, program 9: XG 9 MSB 120 (-2200 cents). CC 0 at 127, then 120, program 9: XG 9 MSB 120 (-2200 cents). CC 0 at 5 again, program 9: XG 9 MSB 120 (-2200 cents). |
| SEL-16 | 1:21 | 11.2.2 | GM2 channel 10, and XG drum presets in GS In GM2 on channel 10, bank MSB 121 with program 2. Then in GS on channel 1, bank MSB 127 with program 17 (an XG drum preset exists) and with program 18 (none exists). | GM2 channel 10 plays the melodic preset at MSB 121. In GS, MSB 127 program 17 plays the XG drum preset by its exact match, and MSB 127 program 18 plays the melodic program 18 (should, by rule 3.3): without an exact match GS does not select XG drum presets. Expected, step by step: GM2 channel 10, MSB 121 program 2: H 2 MSB 121 (-900 cents). GS, MSB 127 program 17: XG 17 MSB 127 (-2100 cents). GS, MSB 127 program 18: R 18 (-1700 cents). |
| SEL-17 | 1:25 | 11.6 | The part mode in GS In GS on channel 1: MSB 8 program 1. Then CC 32 at 5 (stored only) and GS USE FOR RHYTHM PART at 1. Then program 30, and USE FOR RHYTHM PART at 1 again. Then at 0, and program 1. Then XG PART MODE at 2, and USE FOR RHYTHM PART at 2. | A drum value selects program 0 with the drum flag on bank MSB and LSB 0, so program 30 then plays the drum preset of LSB 0. The same value again selects nothing, so program 30 stays. A melodic value selects program 0 on bank 0, so program 1 then plays the preset of MSB 0. XG PART MODE sets the same value as the GS message, so the last message changes nothing. Expected, step by step: MSB 8 program 1: D 1 MSB 8 (-1200 cents). CC 32 at 5, then USE FOR RHYTHM PART 1: GS drum 0 (-1800 cents). program 30, then 1 again: GS drum 30 (300 cents). USE FOR RHYTHM PART 0: capital 0 (-1600 cents). program 1: A 1 (-1500 cents). XG PART MODE 2: GS drum 0 (-1800 cents). program 17, then USE FOR RHYTHM PART 2: GS drum 17 (200 cents). |
| SEL-18 | 1:31 | 11.6 | The part mode in XG and GM2 In XG on channel 1: MSB 5 LSB 3 program 1, then XG PART MODE at 1, program 2 and XG PART MODE at 0. Then MSB 120 program 9, XG PART MODE at 1 (a note), and program 17, then CC 0 at 127 (stored only) and XG PART MODE at 1 again, then CC 0 at 5 (stored only) and XG PART MODE at 1 once more. The first part again in GM2 through GS USE FOR RHYTHM PART, with program 17 after the drum value, then USE FOR RHYTHM PART at 0 on GM2 channel 10. On XG channel 10: MSB 126 program 9 (a drum preset through rule 3.2), then XG PART MODE at 1, then at 0. In XG, LSB 8 program 1, then XG PART MODE at 0, then XG PART MODE at 1 and program 17 (the LSB 8 stays, so there is no exact match). Then the first part once more with the other message: GS USE FOR RHYTHM PART in XG and XG PART MODE in GM2, which act the same. In GM, both part mode messages at 1 after program 1. | In XG and GM2 a message acts even with the value it already has. A drum value moves the bank MSB to the drum bank of the mode, unless it already is 120 or 127, and then selects nothing. On a channel that was not a drum channel it selects program 0, on a drum channel it selects the program again. A melodic value moves the MSB back to the default of the mode, keeps the LSB and selects the program again. XG channel 10 ignores a melodic value, GM2 channel 10 takes it. GM ignores both messages. The answers that come from fallbacks are a should. Expected, step by step: XG, MSB 5 LSB 3 program 1: F 1 MSB 5 LSB 3 (-1300 cents). XG PART MODE 1: GS drum 0 (-1800 cents). program 2: XG 9 MSB 120 (-2200 cents). XG PART MODE 0: G 2 (-1000 cents). XG, MSB 120 program 9: XG 9 MSB 120 (-2200 cents). XG PART MODE 1 on a drum channel: XG 9 MSB 120 (-2200 cents). program 17: XG 17 MSB 120 (-2000 cents). CC 0 at 127, then XG PART MODE 1: XG 17 MSB 120 (-2000 cents). CC 0 at 5, then XG PART MODE 1 again: XG 17 MSB 127 (-2100 cents). GM2, MSB 5 LSB 3 program 1: F 1 MSB 5 LSB 3 (-1300 cents). GM2 USE FOR RHYTHM PART 1: GS drum 0 (-1800 cents). GM2 program 17: XG 17 MSB 120 (-2000 cents). GM2 program 2, then USE FOR RHYTHM PART 0: H 2 MSB 121 (-900 cents). GM2 channel 10, USE FOR RHYTHM PART 0: capital 0 (-1600 cents). XG channel 10, MSB 126 program 9: XG 9 MSB 120 (-2200 cents). XG channel 10, XG PART MODE 1: XG 9 MSB 127 (-2300 cents). XG channel 10, XG PART MODE 0: XG 9 MSB 127 (-2300 cents). XG, LSB 8 program 1: C 1 LSB 8 (-1100 cents). XG PART MODE 0, the LSB stays: C 1 LSB 8 (-1100 cents). XG PART MODE 1 with LSB 8, then program 17: XG 17 MSB 120 (-2000 cents). XG, MSB 5 LSB 3 program 1 again: F 1 MSB 5 LSB 3 (-1300 cents). XG, USE FOR RHYTHM PART 1, then program 2: XG 9 MSB 120 (-2200 cents). XG, USE FOR RHYTHM PART 0: G 2 (-1000 cents). GM2, MSB 5 LSB 3 program 1 again: F 1 MSB 5 LSB 3 (-1300 cents). GM2, XG PART MODE 1, then program 2: XG 9 MSB 120 (-2200 cents). GM2, XG PART MODE 0: H 2 MSB 121 (-900 cents). GM, program 1, both part mode messages at 1: A 1 (-1500 cents). |
| SEL-19 | 1:49 | 11.6 | The part mode after a reset In GS on channel 10: program 30, then GS USE FOR RHYTHM PART at 1, then at 2. On channel 1: program 1, then USE FOR RHYTHM PART at 0. | On channel 10 the value 1 is the one after the reset, so it selects nothing and program 30 stays. The value 2 differs and selects program 0. On channel 1 the value 0 is the one after the reset and selects nothing. Expected, step by step: channel 10, program 30: GS drum 30 (300 cents). channel 10, USE FOR RHYTHM PART 1: GS drum 30 (300 cents). channel 10, USE FOR RHYTHM PART 2: GS drum 0 (-1800 cents). channel 1, program 1, then 0: A 1 (-1500 cents). |
| SEL-21 | 1:53 | 11.2.2, 11.6 | A drum channel is one whose preset is a drum preset In XG on channel 1, bank MSB 127 with program 17, an XG drum preset, so channel 1 is a drum channel without any drum flag. Then XG PART MODE 1 on part 0. A drum value on a channel that was not a drum channel stores program 0 and selects it. | Both notes play the XG drum preset of program 17: the channel already was a drum channel and the bank MSB already was 127, so the part mode message selects nothing. A synthesizer that does not count the channel as a drum channel selects program 0 for the second note, which falls back to another preset. Expected, step by step: XG channel 1, MSB 127 program 17: XG 17 MSB 127 (-2100 cents). XG PART MODE 1: XG 17 MSB 127 (-2100 cents). |
sfe4-61-select-xgdrums Preset selection: XG drum presets only
Each step sends its messages on channel 1 or 10 and then plays a note there. Every preset of the bank plays the sine with scaleTuning 0 and a coarseTune of its own, so the pitch of the note names the preset. A marker in the MIDI file names the expected preset of each step, and the pass criterion lists the expected pitch in cents from 440 Hz. Rules 1 and 2 are required. The fallback lists of rule 3 are a recommendation (spec 11.2.3): without an exact match a synthesizer may select another preset, or none, and the answers marked "should" only hold for one that follows the lists. On XG channel 10 a fallback should not be a melodic preset while the file has a drum preset. This bank holds the melodic presets and the XG drum presets of the full bank, and no GM/GS drum preset.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| SEL-07 | 0:00 | 11.2.3 | Rule 3.1, a request with the drum flag (XG drum presets only) In GS on channel 10, with bank MSB 8 so no request has an exact match: program 30 with bank LSB 5, then program 17 with LSB 7 (a GM/GS drum preset and XG drum presets have program 17), then program 33 (only XG drum presets), then program 57 (only an XG drum preset at MSB 127, above 48), then program 58 (XG drum presets at MSB 127 and 120, above 48), then program 34 (no drum preset). The other banks of this test hold no GM/GS drum presets, or no drum presets at all. | Rule 3.1 in order (should): the GM/GS drum preset with the same program and LSB, then the same program, then any drum preset with the same program if the program is below 49 or the preset is at bank MSB 120 (so program 57 skips its XG drum preset and program 58 takes the one at MSB 120), then the first GM/GS drum preset, then the first drum preset, then the first preset. In the bank of XG drum presets only, program 9 is only at bank MSB 127, so the first drum preset is that one (by program before MSB), not program 17 at MSB 120. Expected, step by step: 3.1.5, program 34: XG 9 MSB 127 (-400 cents). |
sfe4-62-select-gsdrums Preset selection: GM/GS drum presets only
Each step sends its messages on channel 1 or 10 and then plays a note there. Every preset of the bank plays the sine with scaleTuning 0 and a coarseTune of its own, so the pitch of the note names the preset. A marker in the MIDI file names the expected preset of each step, and the pass criterion lists the expected pitch in cents from 440 Hz. Rules 1 and 2 are required. The fallback lists of rule 3 are a recommendation (spec 11.2.3): without an exact match a synthesizer may select another preset, or none, and the answers marked "should" only hold for one that follows the lists. On XG channel 10 a fallback should not be a melodic preset while the file has a drum preset. This bank holds the melodic presets and the GM/GS drum presets of the full bank, and no XG drum preset.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| SEL-08 | 0:00 | 11.2.3 | Rule 3.2, drum requests in XG and GM2 (GM/GS drum presets only) On channel 10 in XG and in GM2: program 58 with bank LSB 5 (no exact match, and above program 48, so only rule 3.2.1 finds the XG drum preset of program 58), program 30 and program 48 (GM/GS drum presets below program 49), programs 49 and 60 (GM/GS drum presets from program 49 on, which rule 3.2.2 only takes in GM2), and program 34. Rule 1 is not reached from MIDI here: no message sets the drum flag in XG or GM2. | Rule 3.2 in order (should): the XG drum preset with the same program, then a drum preset with the same program (below program 49, or any program in GM2, which numbers its kits as GS does), then the first XG drum preset, then the first drum preset, then the first preset. Among the XG drum presets of one program, the first plays, the one at MSB 120. Expected, step by step: XG: 3.2.4, program 34: GS drum 0 (-500 cents). |
sfe4-63-select-nodrums Preset selection: no drum presets
Each step sends its messages on channel 1 or 10 and then plays a note there. Every preset of the bank plays the sine with scaleTuning 0 and a coarseTune of its own, so the pitch of the note names the preset. A marker in the MIDI file names the expected preset of each step, and the pass criterion lists the expected pitch in cents from 440 Hz. Rules 1 and 2 are required. The fallback lists of rule 3 are a recommendation (spec 11.2.3): without an exact match a synthesizer may select another preset, or none, and the answers marked "should" only hold for one that follows the lists. On XG channel 10 a fallback should not be a melodic preset while the file has a drum preset. This bank holds only the melodic presets of the full bank.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| SEL-07 | 0:00 | 11.2.3 | Rule 3.1, a request with the drum flag (no drum presets) In GS on channel 10, with bank MSB 8 so no request has an exact match: program 30 with bank LSB 5, then program 17 with LSB 7 (a GM/GS drum preset and XG drum presets have program 17), then program 33 (only XG drum presets), then program 57 (only an XG drum preset at MSB 127, above 48), then program 58 (XG drum presets at MSB 127 and 120, above 48), then program 34 (no drum preset). The other banks of this test hold no GM/GS drum presets, or no drum presets at all. | Rule 3.1 in order (should): the GM/GS drum preset with the same program and LSB, then the same program, then any drum preset with the same program if the program is below 49 or the preset is at bank MSB 120 (so program 57 skips its XG drum preset and program 58 takes the one at MSB 120), then the first GM/GS drum preset, then the first drum preset, then the first preset. In the bank of XG drum presets only, program 9 is only at bank MSB 127, so the first drum preset is that one (by program before MSB), not program 17 at MSB 120. Expected, step by step: 3.1.6, program 34: capital 0 (-2400 cents). |
| SEL-08 | 0:03 | 11.2.3 | Rule 3.2, drum requests in XG and GM2 (no drum presets) On channel 10 in XG and in GM2: program 58 with bank LSB 5 (no exact match, and above program 48, so only rule 3.2.1 finds the XG drum preset of program 58), program 30 and program 48 (GM/GS drum presets below program 49), programs 49 and 60 (GM/GS drum presets from program 49 on, which rule 3.2.2 only takes in GM2), and program 34. Rule 1 is not reached from MIDI here: no message sets the drum flag in XG or GM2. | Rule 3.2 in order (should): the XG drum preset with the same program, then a drum preset with the same program (below program 49, or any program in GM2, which numbers its kits as GS does), then the first XG drum preset, then the first drum preset, then the first preset. Among the XG drum presets of one program, the first plays, the one at MSB 120. Expected, step by step: XG: 3.2.5, program 34: capital 0 (-2400 cents). |
sfe4-64-select-xgfail Preset selection: a bank that fails the XG test
Each step sends its messages on channel 1 or 10 and then plays a note there. Every preset of the bank plays the sine with scaleTuning 0 and a coarseTune of its own, so the pitch of the note names the preset. A marker in the MIDI file names the expected preset of each step, and the pass criterion lists the expected pitch in cents from 440 Hz. Rules 1 and 2 are required. The fallback lists of rule 3 are a recommendation (spec 11.2.3): without an exact match a synthesizer may select another preset, or none, and the answers marked "should" only hold for one that follows the lists. On XG channel 10 a fallback should not be a melodic preset while the file has a drum preset. This bank is the full bank plus a preset at bank MSB 127, program 10, with the drum flag set. Program 10 is not an XG kit, so the bank fails the XG test and its presets at bank MSB 120 and 127 are melodic.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| SEL-12 | 0:00 | 11.2.3 | Rule 2 in XG and GM2 with MSB 127 or 120 and an exact match that is not an XG drum preset (the xgfail bank) On channel 1, bank MSB 127 with program 9 in XG, and bank MSB 120 with program 9 in GM2. In the xgfail bank, which fails the XG test, the presets at MSB 120 and 127 are melodic, apart from its preset at MSB 127 program 10 with the drum flag set, a GM/GS drum preset. | With the full bank the exact match plays, an XG drum preset. With the xgfail bank the exact match is skipped, and rule 3.2 picks the first drum preset (should), a GM/GS drum preset. Expected, step by step: XG, MSB 127 program 9: GS drum 0 (-1800 cents). GM2, MSB 120 program 9: GS drum 0 (-1800 cents). |
| SEL-13 | 0:03 | 11.2.1, 11.2.3 | The XG test (the xgfail bank) In GS on channel 10 (the drum flag), bank MSB 8 with program 9, which no GM/GS drum preset has. Only the presets at bank MSB 120 and 127 have program 9, and the xgfail bank holds a preset at MSB 127 with a program outside the XG list. | Rule 3.1 (should). With the full bank the presets at MSB 120 and 127 are XG drum presets, so rule 3.1.3 plays the one of program 9 at MSB 120. With the xgfail bank they are melodic, so rule 3.1.3 finds nothing and rule 3.1.4 plays the first GM/GS drum preset. Expected, step by step: GS channel 10, MSB 8 program 9: GS drum 0 (-1800 cents). |
sfe4-65-select-empty Preset selection: a bank without presets
A bank without any preset. There is no start click either, since the click needs a preset of its own.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| SEL-20 | 0:00 | 11.2.3 | An empty bank A bank without presets, and a note with program 0. | Silence: a file without presets selects nothing. Expected, step by step: program 0: silence. |
sfe4-70-rifs 64-bit chunk sizes
Section 3.1
The probe bank written as a RIFS file.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-01 | 0:00 | 3.1 | 64-bit chunk sizes The probe bank written with 64-bit chunk sizes: the file starts with RIFS and every chunk has a 12-byte header. INFO holds an unknown sub-chunk of 7 bytes, so a pad byte follows it. Each note is followed by a plain sine at the pitch a pass should have. | The bank loads and plays the pitches below: the unknown sub-chunk and its pad byte are skipped. Expected, note by note, in cents from 440 Hz: a constant modulator: 300 cents. a custom curve modulator: 600 cents. a custom curve modulator in a preset zone: 300 cents. |
sfe4-71-pad Pad bytes
Section 3.1
The probe bank with an unknown INFO sub-chunk of odd length.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-02 | 0:00 | 3.1 | Pad bytes The probe bank with 32-bit chunk sizes. INFO holds an unknown sub-chunk of 7 bytes, so a pad byte follows it. Each note is followed by a plain sine at the pitch a pass should have. | The bank loads and plays the pitches below: the unknown sub-chunk and its pad byte are skipped. Expected, note by note, in cents from 440 Hz: a constant modulator: 300 cents. a custom curve modulator: 600 cents. a custom curve modulator in a preset zone: 300 cents. |
sfe4-72-xdta Indices past 65535
Section 5.5
A bank of more than 65536 zones at both levels.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-03 | 0:00 | 5.5 | Indices past 65535 through xdta A filler instrument with 65600 zones and a filler preset with 65600 zones come first, each zone on key 0 only, with a key range and one modulator of its own. Then a preset whose bags, generators and modulators are past the 65536th at both levels: fineTune 100 and a custom curve modulator of 600 cents in its instrument zone, and a constant modulator of 200 cents in its preset zone. The 16-bit indices of pdta only reach it with the top bits of xdta. Each note is followed by a plain sine at the pitch a pass should have. | The last preset plays its own tone, 900 cents up. A reader that ignores the top 16 bits reaches a filler zone and plays nothing or a wrong pitch. Expected, note by note, in cents from 440 Hz: the preset past 65535: 900 cents. |
sfe4-73-minor A later minor version
A bank of version 4.1 with things a reader of 4.0 doesn't know.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-04 | 0:00 | 5.1, 7.4 | A later minor version and what it may addifil is 4.1. INFO holds an unknown sub-chunk. One zone has an unknown generator (200) with the value 777. Six zones each have a modulator of CC 11 to fineTune that is unknown: source type 5 and source type 63, each once with the negative direction (a source read as 0 would act) and once with the positive one (a source read as a plain CC 11 would act), transform 4 and destination 95, next to a constant modulator that must stay. One zone has a custom curve modulator with the unknown destination 99, then a valid custom curve modulator. Each note is followed by a plain sine at the pitch a pass should have. | The bank loads. The unknown generator, sub-chunk and modulators do nothing, and the modulators next to them act. The valid custom curve modulator after the unknown one gets its own curve: the unknown one still takes its extended record. Expected, note by note, in cents from 440 Hz: an unknown generator: 0 cents. source type 5: 150 cents. source type 63: 160 cents. source type 5, positive: 155 cents. source type 63, positive: 165 cents. transform 4: 170 cents. destination 95: 180 cents. a curve modulator with destination 99, then a valid one: 915 cents. |
sfe4-74-major5 A later major version
Section 5.1
The probe bank with version 5.0.
A reader should reject this bank, so it has no reference render.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-05 | 0:00 | 5.1 | A later major version The probe bank with ifil 5.0. | The bank is rejected (should). There is no reference render. |
sfe4-75-nodmod No DMOD
Section 7.3
The probe bank without DMOD.
A reader must reject this bank, so it has no reference render.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-06 | 0:00 | 7.3 | No DMOD The probe bank without a DMOD sub-chunk. | The bank is rejected: an SFe file without DMOD is not valid. There is no reference render. |
sfe4-76-dmod-zero A DMOD of zero bytes
Section 7.3
A bank whose DMOD has no bytes at all.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-07 | 0:00 | 7.3 | A DMOD of zero bytes A bank written with the default modulator list of a new bank, then its DMOD cut to zero bytes and its XMOD removed. The sine at velocity 127, at velocity 40, and at velocity 127 after CC 7 at 0 and CC 11 at 0. | The bank loads with no default modulators: all three notes have the same level, since neither velocity nor CC 7 and CC 11 change anything. |
sfe4-77-xmod-missing A broken XMOD: no XMOD
The default modulator list of a new bank in DMOD, with no XMOD in XMOD.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-08 | 0:00 | 5.3, 7.3 | A broken XMOD (no XMOD)DMOD holds the default modulator list of a new bank, whose modulators of CC 72, 73 and 75 have custom curves and the extended flag. Their records in XMOD can't be matched. The sine with an attack of 0.1 s and CC 73 at 127. Then the sine with a release of 0.1 s, CC 72 at 127 and a note off. Then a zone with decayVolEnv -1200 and sustainVolEnv 240, with CC 75 at 127. Then the sine with CC 7 at its reset value of 100 and at 64. Last, a zone with a modulator identical to the default one of CC 73, whose own record in xdta is intact, with CC 73 at 127. | CC 72, 73 and 75 do nothing: the first note has the same level 0.25 s and 0.5 s in (its attack of 0.1 s is over), the second is at least 30 dB down 0.15 s after its note off, the third is 24 dB down 0.3 s in. CC 7 still works: -7.85 dB at 64 against 100. The zone modulator acts and gives a slow attack: at least 3 dB down 50 ms in. |
sfe4-78-xmod-few A broken XMOD: one record too few
The default modulator list of a new bank in DMOD, with one record too few in XMOD.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-08 | 0:00 | 5.3, 7.3 | A broken XMOD (one record too few)DMOD holds the default modulator list of a new bank, whose modulators of CC 72, 73 and 75 have custom curves and the extended flag. Their records in XMOD can't be matched. The sine with an attack of 0.1 s and CC 73 at 127. Then the sine with a release of 0.1 s, CC 72 at 127 and a note off. Then a zone with decayVolEnv -1200 and sustainVolEnv 240, with CC 75 at 127. Then the sine with CC 7 at its reset value of 100 and at 64. Last, a zone with a modulator identical to the default one of CC 73, whose own record in xdta is intact, with CC 73 at 127. | CC 72, 73 and 75 do nothing: the first note has the same level 0.25 s and 0.5 s in (its attack of 0.1 s is over), the second is at least 30 dB down 0.15 s after its note off, the third is 24 dB down 0.3 s in. CC 7 still works: -7.85 dB at 64 against 100. The zone modulator acts and gives a slow attack: at least 3 dB down 50 ms in. |
sfe4-79-xmod-many A broken XMOD: one record too many
The default modulator list of a new bank in DMOD, with one record too many in XMOD.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-08 | 0:00 | 5.3, 7.3 | A broken XMOD (one record too many)DMOD holds the default modulator list of a new bank, whose modulators of CC 72, 73 and 75 have custom curves and the extended flag. Their records in XMOD can't be matched. The sine with an attack of 0.1 s and CC 73 at 127. Then the sine with a release of 0.1 s, CC 72 at 127 and a note off. Then a zone with decayVolEnv -1200 and sustainVolEnv 240, with CC 75 at 127. Then the sine with CC 7 at its reset value of 100 and at 64. Last, a zone with a modulator identical to the default one of CC 73, whose own record in xdta is intact, with CC 73 at 127. | CC 72, 73 and 75 do nothing: the first note has the same level 0.25 s and 0.5 s in (its attack of 0.1 s is over), the second is at least 30 dB down 0.15 s after its note off, the third is 24 dB down 0.3 s in. CC 7 still works: -7.85 dB at 64 against 100. The zone modulator acts and gives a slow attack: at least 3 dB down 50 ms in. |
sfe4-80-xmod-size A broken XMOD: 5 bytes too long
The default modulator list of a new bank in DMOD, with 5 bytes too long in XMOD.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-08 | 0:00 | 5.3, 7.3 | A broken XMOD (5 bytes too long)DMOD holds the default modulator list of a new bank, whose modulators of CC 72, 73 and 75 have custom curves and the extended flag. Their records in XMOD can't be matched. The sine with an attack of 0.1 s and CC 73 at 127. Then the sine with a release of 0.1 s, CC 72 at 127 and a note off. Then a zone with decayVolEnv -1200 and sustainVolEnv 240, with CC 75 at 127. Then the sine with CC 7 at its reset value of 100 and at 64. Last, a zone with a modulator identical to the default one of CC 73, whose own record in xdta is intact, with CC 73 at 127. | CC 72, 73 and 75 do nothing: the first note has the same level 0.25 s and 0.5 s in (its attack of 0.1 s is over), the second is at least 30 dB down 0.15 s after its note off, the third is 24 dB down 0.3 s in. CC 7 still works: -7.85 dB at 64 against 100. The zone modulator acts and gives a slow attack: at least 3 dB down 50 ms in. |
sfe4-81-xdta-missing No xdta
Section 5.5.1
A bank without the xdta list.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-09 | 0:00 | 5.5, 5.5.1, 7.3 | No xdta The bank has no xdta list. DMOD holds one modulator, CC 16 through a linear custom curve to fineTune with 600, and its record in XMOD. Zones with a custom curve modulator, with a modulator that has the flag and no custom curve, with a custom curve modulator in the preset zone, and with a modulator identical to the default one (amount 1000, flagged), with CC 16 at 127. Each zone with a constant modulator that stays. Each note is followed by a plain sine at the pitch a pass should have. | The modulators with the flag or a custom curve are ignored, in instrument and preset zones, and the others act. The default modulator acts: the ignored zone modulator identical to it supersedes nothing. Expected, note by note, in cents from 440 Hz: a custom curve modulator: 190 cents. a modulator with the flag and no curve: 200 cents. a custom curve modulator in a preset zone: 210 cents. a default modulator, and an ignored zone modulator identical to it: 600 cents. |
sfe4-82-xdta-unsound An xdta without its pgen
Section 5.5
A bank whose xdta list has no pgen.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-10 | 0:00 | 5.5, 5.5.1, 7.3 | A structurally unsound xdta (a sub-chunk missing) The xdta list has no pgen sub-chunk, so it is structurally unsound. DMOD holds one modulator, CC 16 through a linear custom curve to fineTune with 600, and its record in XMOD. Zones with a custom curve modulator, with a modulator that has the flag and no custom curve, with a custom curve modulator in the preset zone, and with a modulator identical to the default one (amount 1000, flagged), with CC 16 at 127. Each zone with a constant modulator that stays. Each note is followed by a plain sine at the pitch a pass should have. | The whole xdta is ignored, as in CON-09. The modulators with the flag or a custom curve are ignored, in instrument and preset zones, and the others act. The default modulator acts: the ignored zone modulator identical to it supersedes nothing. Expected, note by note, in cents from 440 Hz: a custom curve modulator: 190 cents. a modulator with the flag and no curve: 200 cents. a custom curve modulator in a preset zone: 210 cents. a default modulator, and an ignored zone modulator identical to it: 600 cents. |
sfe4-83-xdta-size An xdta with an ibag of another size
Section 5.5
A bank whose xdta list has an ibag one record longer than the one of pdta.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-10 | 0:00 | 5.5, 5.5.1, 7.3 | A structurally unsound xdta (an ibag of another size) The xdta list has an ibag one record longer than the ibag of pdta, so it is structurally unsound. DMOD holds one modulator, CC 16 through a linear custom curve to fineTune with 600, and its record in XMOD. Zones with a custom curve modulator, with a modulator that has the flag and no custom curve, with a custom curve modulator in the preset zone, and with a modulator identical to the default one (amount 1000, flagged), with CC 16 at 127. Each zone with a constant modulator that stays. Each note is followed by a plain sine at the pitch a pass should have. | The whole xdta is ignored, as in CON-09. The modulators with the flag or a custom curve are ignored, in instrument and preset zones, and the others act. The default modulator acts: the ignored zone modulator identical to it supersedes nothing. Expected, note by note, in cents from 440 Hz: a custom curve modulator: 190 cents. a modulator with the flag and no curve: 200 cents. a custom curve modulator in a preset zone: 210 cents. a default modulator, and an ignored zone modulator identical to it: 600 cents. |
sfe4-84-xrec-count Too few extended records
Section 5.5.1
A bank whose xdta imod holds one record too few.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-11 | 0:00 | 5.5.1 | Extended records that can't be matched at one level (one too few) The imod of xdta holds one record too few (the last is missing), and its pmod is intact. Two zones each with a custom curve modulator and a constant one, and a preset whose preset zone has a custom curve modulator of 230 cents. Each note is followed by a plain sine at the pitch a pass should have. | Both custom curve modulators of the instrument zones are ignored, also the first, whose record is still there: the records of imod can't be matched. The one of the preset zone acts. Expected, note by note, in cents from 440 Hz: an instrument custom curve modulator: 220 cents. another instrument custom curve modulator: 225 cents. a preset custom curve modulator: 460 cents. |
sfe4-85-xrec Extended records that are taken but not used
Section 5.5.1
A bank with extended records that a reader must take and not use.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-12 | 0:00 | 5.5.1 | Extended records that are taken but not used Five zones, each with a modulator that has something wrong with its extended record, followed by a normal custom curve modulator with a curve of its own: a record with a non-zero reserved byte, a modulator with the flag and no custom curve (its record is zeros), a custom curve modulator without the flag (and without a record), a custom curve modulator with the transform 0x8001, and a custom curve modulator whose record also gives a curve index for its amount source, which is not of type 4. Each note is followed by a plain sine at the pitch a pass should have. | The first four are ignored, and the fifth acts with its own curve. Every following modulator gets its own curve: each flagged modulator takes its record, also when it is ignored. Expected, note by note, in cents from 440 Hz: a non-zero reserved byte: 945 cents. the flag without a custom curve: 635 cents. a custom curve without the flag: 968 cents. the transform 0x8001: 328 cents. a curve index for an amount source of another type: 1658 cents. |
sfe4-86-curv Invalid and missing curves
Section 5.4
A bank with invalid curves between valid ones.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-13 | 0:00 | 5.4 | Invalid and missing curvesCURV holds ten curves. Curves 1, 3, 5, 6 and 7 are invalid: no points, one point, a decreasing x, an x above 16384 and a y above 16384. The others are valid flat curves. Each curve has a modulator in its own zone, next to a constant modulator, and the modulator of curve 9 names curve 50, past the end. Each note is followed by a plain sine at the pitch a pass should have. | The modulators on the invalid curves and on the missing curve are ignored. The valid curves keep their numbers, so each modulator on a valid curve gets its own curve. Expected, note by note, in cents from 440 Hz: curve 0, valid: 575 cents. curve 1, no points: 240 cents. curve 2, valid: 1260 cents. curve 3, one point: 240 cents. curve 4, valid: 930 cents. curve 5, a decreasing x: 240 cents. curve 6, an x above 16384: 240 cents. curve 7, a y above 16384: 240 cents. curve 8, valid: 1305 cents. curve 9, valid, but its modulator names a curve past the end: 240 cents. |
sfe4-87-curv-short A CURV that ends early
Section 5.4
A bank whose CURV ends in the middle of a curve.
A reader must reject this bank, so it has no reference render.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-14 | 0:00 | 5.4 | A CURV that ends earlydwCurveCount is 5, but CURV ends in the middle of the third curve. | The bank is rejected: a CURV sub-chunk that ends before all of its curves is malformed, and an SFe file with a malformed CURV is not valid. There is no reference render. |
sfe4-88-ignored Modulators that are ignored
A bank of modulators a reader must ignore.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-15 | 0:00 | 7.2, 7.4 | Modulators that are ignored Zones each with a constant modulator of 260 cents and a modulator of CC 11 to fineTune changed into one that must be ignored: the Link source and the Link amount source (both with the negative direction), a destination with bit 15 (once linking to the other modulator, once on the number of fineTune), the transforms 1 and 3, the destinations 60, 92 and 14 (unknown), the reserved general controller 20 as source and as amount source, the illegal controllers 0, 6, 32 and 38 as source and as amount source, and the illegal CC 98 and CC 120 as source with the negative direction. First, CC 40 with the negative direction as the only modulator of a zone, which must act: the controllers 33 to 63 are valid sources. CC 0, 6, 32 and 38 are at 127 during their notes, and CC 40 stays at 0. Each note is followed by a plain sine at the pitch a pass should have. | The CC 40 modulator acts. None of the changed modulators acts: every other note plays the constant modulator alone. Spec 7.4 lists the illegal controllers of SoundFont 2 section 8.2.1, CC 0, 6, 32 and 38 among them. Expected, note by note, in cents from 440 Hz: CC 40 as the source, which is valid and acts: 460 cents. the Link source: 260 cents. the Link amount source: 260 cents. a destination with bit 15: 260 cents. a destination with bit 15 on the number of fineTune: 260 cents. the transform 1: 260 cents. the transform 3: 260 cents. the destination 60: 260 cents. the destination 92: 260 cents. the unused destination 14: 260 cents. the reserved general controller 20 as the source: 260 cents. the reserved general controller 20 as the amount source: 260 cents. CC 98 as the source: 260 cents. CC 120 as the source: 260 cents. CC 0 as the source: 260 cents. CC 0 as the amount source: 260 cents. CC 6 as the source: 260 cents. CC 6 as the amount source: 260 cents. CC 32 as the source: 260 cents. CC 32 as the amount source: 260 cents. CC 38 as the source: 260 cents. CC 38 as the amount source: 260 cents. |
sfe4-89-vorbis A compressed sample
Section 6.2
A bank with an Ogg Vorbis sample and a PCM one.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-16 | 0:00 | 6.2 | A compressed sample next to a PCM one The vorbis sample (the tail sample compressed as Ogg Vorbis: the looped 440 Hz with 880 Hz after the loop end, its loop given in decoded points) and the PCM sine in one file, each for 1.8 s. | Both play 440 Hz for the whole note at the same level, also around the first wrap of the loop at 1.05 s: the loop points of the compressed sample are decoded points from its start. A loop read in bytes, or from dwStart, reaches the 880 Hz. |
sfe4-90-enums Stored enumerations out of range
Section 7.1
A bank with stored values of the enumerations outside their values.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-17 | 0:00 | 7.1 | Stored enumerations outside their values Pairs of notes, each a stored value outside an enumeration next to the default: filterType 9 next to 0 (the sine an octave above a cutoff of 220 Hz with initialFilterQ 200, where each of the five types gives another level), modLfoWaveform 7 next to 0 (a tremolo at 2 Hz), ignoreNoteOff 5 next to 0 (a note that would decay over 2 s), and sampleModes 0x8001 next to 0 (the sine of 1.1 s played once, in notes of 1.5 s). The other notes last 0.6 s. | The two notes of each pair are the same: a stored value outside the defined values is the default. So the first pair is the lowpass, the second a triangle, the third ends at its note off and the fourth plays its sample once and is silent after 1.1 s. |
sfe4-91-wbank Bit 15 of wBank
Section 6.1
The full selection bank of group 60 with bit 15 of wBank set in every preset header. It belongs to the reader rules, and lives with the selection groups since it plays their bank.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-18 | 0:00 | 6.1 | Bit 15 of wBank is ignored The full selection bank of group 60, with bit 15 of wBank set in every preset header. Program 1 with bank MSB 8, then with bank LSB 8, program 0 on channel 1 and on channel 10, and program 17 on channel 10 in XG. | Each note plays the preset group 60 selects for the same request: bit 15 is ignored. Expected, step by step: MSB 8, program 1: D 1 MSB 8 (-1200 cents). LSB 8, program 1: C 1 LSB 8 (-1100 cents). program 0 on channel 1: capital 0 (-1600 cents). program 0 on channel 10: GS drum 0 (-1800 cents). XG, program 17 on channel 10: XG 17 MSB 127 (-2100 cents). |
sfe4-92-xrec-many One extended record too many
Section 5.5.1
A bank whose xdta imod holds one record too many.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-11 | 0:00 | 5.5.1 | Extended records that can't be matched at one level (one too many) The imod of xdta holds one record of zeros too many, before the terminal one, and its pmod is intact. Two zones each with a custom curve modulator and a constant one, and a preset whose preset zone has a custom curve modulator of 230 cents. Each note is followed by a plain sine at the pitch a pass should have. | Both custom curve modulators of the instrument zones are ignored, also the first, whose record is still there: the records of imod can't be matched. The one of the preset zone acts. Expected, note by note, in cents from 440 Hz: an instrument custom curve modulator: 220 cents. another instrument custom curve modulator: 225 cents. a preset custom curve modulator: 460 cents. |
sfe4-93-xrec-size Extended records of a size that isn't a multiple of 16
Section 5.5.1
A bank whose xdta imod is 5 bytes longer than its records.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-11 | 0:00 | 5.5.1 | Extended records that can't be matched at one level (not a multiple of 16) The imod of xdta is 5 bytes longer than its records, so its size is not a multiple of 16, and its pmod is intact. Two zones each with a custom curve modulator and a constant one, and a preset whose preset zone has a custom curve modulator of 230 cents. Each note is followed by a plain sine at the pitch a pass should have. | Both custom curve modulators of the instrument zones are ignored, also the first, whose record is still there: the records of imod can't be matched. The one of the preset zone acts. Expected, note by note, in cents from 440 Hz: an instrument custom curve modulator: 220 cents. another instrument custom curve modulator: 225 cents. a preset custom curve modulator: 460 cents. |
sfe4-94-curv-long A CURV with bytes after its curves
Section 5.4
A bank whose CURV holds 4 bytes after its last curve.
A reader must reject this bank, so it has no reference render.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-19 | 0:00 | 5.4 | A CURV with bytes after its curvesdwCurveCount is 3 and the three curves are whole, but CURV holds 4 more bytes after the last one. | The bank is rejected: a CURV sub-chunk with bytes after its last curve is malformed, and an SFe file with a malformed CURV is not valid. There is no reference render. |
sfe4-95-xdta-names Names continued in xdta
Section 5.5
A bank with a preset name continued in xdta, and a short one whose xdta name field is not empty.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-20 | 0:00 | 5.5 | Names continued in xdta Two presets of the sine. One is named "CON-20 a preset name of 32 chars": its first 20 characters fill the pdta name and the rest is in xdta. The other is named "CON-20 short", which ends with a zero byte in pdta, and its xdta name field holds "not part of the name" anyway. Both play 440 Hz. | The preset list shows "CON-20 a preset name of 32 chars" and "CON-20 short": the xdta half is appended only when the pdta half fills all 20 characters with no zero byte. A reader that always appends it shows "CON-20 shortnot part of the name", and one that ignores xdta shows "CON-20 a preset name". |
sfe4-96-index-msb instrumentMSB and sampleIDMSB
A small bank whose zones hold instrumentMSB and sampleIDMSB, written by patching placeholders, at their own level, at the wrong level, in a global zone and after the generator they extend.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-21 | 0:00 | 7.1.1, 5.5 | instrumentMSB and sampleIDMSB The reference tone, then eight presets of the sine in a small bank, each with one of the generators instrumentMSB (90) and sampleIDMSB (91): instrumentMSB 1 in the preset zone (an instrument past the bank), instrumentMSB 0 in the preset zone, sampleIDMSB 1 in the instrument zone (a sample past the bank), sampleIDMSB 1 in the preset zone (the wrong level), instrumentMSB 1 in the instrument zone (the wrong level), sampleIDMSB 1 in the global instrument zone, instrumentMSB 1 after instrument in the preset zone, sampleIDMSB 1 after sampleID in the instrument zone. | instrumentMSB 1 in the preset zone (an instrument past the bank): silent. instrumentMSB 0 in the preset zone: plays 440 Hz. sampleIDMSB 1 in the instrument zone (a sample past the bank): silent. sampleIDMSB 1 in the preset zone (the wrong level): plays 440 Hz. instrumentMSB 1 in the instrument zone (the wrong level): plays 440 Hz. sampleIDMSB 1 in the global instrument zone: plays 440 Hz. instrumentMSB 1 after instrument in the preset zone: plays 440 Hz. sampleIDMSB 1 after sampleID in the instrument zone: plays 440 Hz. An MSB of 1 makes the index 65536 plus the 16-bit value, past every instrument and sample of this bank, so the reader ignores the zone. The generators only count in a zone at their own level, before the generator they extend, and are ignored anywhere else. |
sfe4-97-index-large Instrument indices above 32767 and 65535
A bank of 65542 instruments. A preset zone refers to instrument 32768, and another reaches instrument 65541 through instrumentMSB, patched in from a placeholder.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-22 | 0:00 | 7.1.1, 5.5 | Instrument indices above 32767 and 65535 The reference tone, then two presets of a bank with 65542 instruments of the sine. The first refers to instrument 32768, which plays 300 cents up. The second refers to instrument 5 with instrumentMSB 1, so to instrument 65541, which plays 700 cents up. Instrument 5 itself plays -500 cents, and the others 440 Hz. The bank has more than 65535 instrument zones, so a reader also needs the 32-bit bag and generator indices of the xdta-list chunk (5.5). | The two presets play 300 and 700 cents above 440 Hz. instrument is read as an unsigned WORD, so 32768 is not negative, and the index is (instrumentMSB << 16) | instrument. A reader that reads 32768 as signed plays something else or nothing, and one that ignores instrumentMSB plays -500 cents. |
sfe4-98-after-index Generators after instrument and sampleID
Section 7.1.1
A small bank whose zones hold coarseTune after instrument or sampleID, written by patching placeholders, next to zones that hold it before.
| Id | Start | Section | What it plays | Pass |
|---|---|---|---|---|
| CON-23 | 0:00 | 7.1.1 | Generators after instrument and sampleID The reference tone, then four presets of the sine: coarseTune +12 after sampleID in the instrument zone, coarseTune +12 after instrument in the preset zone, coarseTune +12 before sampleID in the instrument zone (control), coarseTune +12 before instrument in the preset zone (control). | The first two play 440 Hz: a generator after instrument or sampleID is ignored, as in SoundFont 2, and the zone still plays. The two controls play an octave up (880 Hz). A reader that applies the late generators plays the first two an octave up too, and one that drops their zones plays nothing. |