GateManPoly FPGA Synth

JovianPyx

http://www.fpga.synth.net/pmwiki/pmwiki.php?n=FPGASynth.GateManPoly

GateManPoly / 8 is an 8 voice polyphonic version of the GateMan FPGA synth.

The sound file was made using only one devboard, no effects were used.

Peake · Posted: Tue May 20, 2008 6:43 pm Post subject:

I've been to your site and read up and listened with great interest. Some excellent work, sir!

Have you investigated using three lines with cross-feedback/individual filtration? Allpass stages in the feedback?

JovianPyx · Posted: Wed May 21, 2008 5:12 am Post subject:

Peake · Posted: Wed May 21, 2008 7:38 am Post subject:

The most important thing in my experience is to have two or three separate delay lines whose feedback become common at some point. Then to add static allpass filters into the feedback loops. It's been 12 years since I've worked with Karplus-Strong so bear with me.

JovianPyx · Posted: Wed May 21, 2008 7:46 am Post subject:

Peake · Posted: Wed May 21, 2008 8:10 am Post subject:

I'd done a little work for a company using this tech and am too close to breaking NDA to continue. I'm certain that it's on the web, as I'd described.

You're doing great work. I enjoyed your solution to "microphone" placement.

JovianPyx · Posted: Wed May 21, 2008 8:16 am Post subject:

JovianPyx · Posted: Wed May 21, 2008 9:37 am Post subject:

Ok, I see the references to allpass. This is similar to the use of Lagrange interpolators, it provides a fractional delay that allows arbitrary pitches.

With my K-S synth, I did something different, some may even say wrong headed. I calculated the pitch error using _integral_ delay lengths and found that I could achieve a pitch accuracy of +/- 5 cents or better over 3.5 octaves if I used delay lines of 2048 maximum length. A ROM contains the lengths for each pitch.

Yes, this does restrict the range of the instrument, but many natural instruments have restricted range. Since I got more than 3 octaves, I was happy. The technique I used allowed for a much simpler design - it works and is musically useful.

I used fractional delay lines when I experimented with the phLUTe, but I wasn't able to eliminate or substantially reduce the transient that occurs at pitch change. Portamento actually made this worse.

At some point, I will probably revisit the PolyDaWG, using smaller waveguides and Lagrange interpolators, I may be able to make a 16 string synth in the same 16K words of RAM. It would be nice to have a larger harp. Seems I need to get busy on a 16 voice MIDI controller... It looks as if the state machine might be fast enough to do it, might be fast enough.

Peake · Posted: Wed May 21, 2008 1:37 pm Post subject:

Allpass filtration in the feed back will cause the harmonics to begin moving. The more stages, the more movement. Enough stages to present the effect of a cello being bowed will also severely impair pitch tracking...

There should be plenty of information on the web regarding the cross-feedback tip I'd mentioned. It should be easy enough to guesstimate and experiment with, at minimum.

JovianPyx · Posted: Fri May 30, 2008 11:12 am Post subject:

A new version (l) is available, see the GateManPoly link in my sig. Several fixes and improvements are included. In the pub folder of the zip of the source is a description of the MIDI system exclusive message values used to program the synth's "patch". A version history is on the page near the bottom.

JovianPyx · Posted: Fri May 30, 2008 3:28 pm Post subject:

And here is another GateManPoly sample. It's a synthetic steel drum sound.

gatemanpoly_steeldrums.wav

Description:

GateManPoly - Synthetic Steel Drum Sound Sample

Download

Filename:

gatemanpoly_steeldrums.wav

Filesize:

5.12 MB

Downloaded:

240 Time(s)

JovianPyx · Posted: Thu Jun 05, 2008 8:34 am Post subject:

I've updated the GateManPoly synth project files to contain a GUI patch editor.

If interested, see:

http://www.fpga.synth.net/pmwiki/pmwiki.php?n=FPGASynth.GateManPoly

The whole project file is in the zip file, i.e., if you download and install Microsoft VB2008 Express Edition (free), you can open and modify the patch editor if you like or you can compile and run it.

JovianPyx · Posted: Wed Dec 08, 2010 5:35 pm Post subject:

Recent updates include the addition of a pitch ADSR.

The pitch ADSR affects pitch so that harmonic glides both negative and positive are possible. The effect is a pitch effect, not a frequency effect.

I've also improved the ADSR code to use less slices.

sample attached, such as it is...

If there is interest, I will post the source code for a Spartan-3E Starter Kit.

sg_2010-12-08.mp3

Description:

This GateManPoly patch uses the pitch ADSR.

That is somewhat responsible for the phasing sound.

Pardon the clipping.

Download

Filename:

sg_2010-12-08.mp3

Filesize:

3 MB

Downloaded:

148 Time(s)

Skrog Productions · Posted: Thu Dec 09, 2010 11:00 am Post subject:

very interesting Scott, the new mp3 sounds really like a patch from my Oberheim Matrix 6R, exellent , i like it Smile

.

Dave.

JovianPyx · Posted: Thu Dec 09, 2010 11:37 am Post subject:

Thanks Dave!

This isn't a new idea by any means. I have a Roland D-110 which has a pitch envelope available to each "partial". I've always wanted to add one to GateManPoly since it was modelled loosely on that synth.

I should have described the patch too -

Each of the 4 NCOs is detuned by it's own ADSR, 2 of the NCOs have wider pitch detune than the other two. Each set of 2 NCOs have ADSR output which progresses in the opposite direction, i.e., one tunes negatively, the other positively.

I've added hardware to allow switching between slider tuning and ADSR tuning and on this patch the difference is more than subtle, without the ADSR detune, the patch is lackluster and relatively lifeless. There is a small amount of detune added by a "noise LFO", a noise source fed to a filter with a very low Fc.

I'll attach a sample that shows this. The first note is without the ADSR detune, the second is with ADSR detuning enabled.

GateManPoly_20101209.mp3

Description:

Sample of GateManPoly without ADSR detune and with ADSR detune.

Download

Filename:

GateManPoly_20101209.mp3

Filesize:

562.11 KB

Downloaded:

131 Time(s)

Dan Lavin · Posted: Thu Dec 09, 2010 7:17 pm Post subject:

Very nice Scott! I really like the way the harmonics sing out after the note "settles down".....reminds me of electric guitar feedback.....like an overtone.

Blue Hell · Posted: Sun Dec 12, 2010 8:22 am Post subject:

Sounds good!

JovianPyx · Posted: Fri Dec 24, 2010 12:29 pm Post subject:

More improvements and bug fixes.

And the ADSRs were too "fast", that is, the slowest settings for ramp times were not slow enough. I tried to reduce the sample rate for just the ADSRs, but I found that they started adding clicks to the output as the system then used larger numbers for rates (meaning the jump from the last value is larger). I've decided that it's better to sacrifice some FPGA real estate and preserve the sample rate (250 KHz in this case) for the ADSRs.

Still need to test that the ADSRs react fast enough when I want them to be fast.

I'll include a short sample of several notes of a chord being held and lifted slowly - this will demonstrate that the animation procedes mostly "hands off".

The patch consists of 4 sawtooth oscillators which are pitch controlled by the pitch ADSR to start in unison, move apart in pitch and then back to near unison. A state variable filter per voice is opening slowly and at release is closing slowly.

EDIT ADD: Oh - and I love to mention that this is produced by a 12 bit DAC Smile

Ominous_Chords.mp3

Description:

Ominous

Download

Filename:

Ominous_Chords.mp3

Filesize:

1.41 MB

Downloaded:

131 Time(s)

Ominous_Chords_01.mp3

Description:

I noticed some clipping in the other sample. Here's one where I hear no clipping.

Download

Filename:

Ominous_Chords_01.mp3

Filesize:

1.28 MB

Downloaded:

158 Time(s)

Dan Lavin · Posted: Fri Dec 24, 2010 3:40 pm Post subject:

Awesome work, as usual, Scott!

JovianPyx · Posted: Sat Nov 05, 2011 4:30 pm Post subject:

I've just added the DDR SDRAM interface to this synth giving it 16+ seconds of looped delay effect.