A Synthesizer Made of Only Filters

JovianPyx · Last edited by JovianPyx on Tue Jun 14, 2011 10:58 am; edited 1 time in total

I have designed a synthesizer made of only filters. There are no oscillators in the design at all. It is also up to 32 voice polyphonic. I call it PolyDrum.

It has 32 state variable filters and 32 single pole IIR low pass filters and is implemented on a $39 FPGA development board.

The state variable filters form "resonators" and the single pole filters form exponentially decaying envelope generators. The envelope generators modulate the Fc input of the state variable filters to provide a variety of pitch modulation effects.

Each resonator is frequency, Q, EG and MIDI note number programmable. More than one resonator can be programmed to respond to a given note number - but all of the resonators responding to the same note number will begin their sound with a synchronized pulse

EDIT ADD 2009-12-05:
I should have added that the instrument is played (MIDI) by sending an impulse to any resonator that receives a match between the note played and the resonator's note number map (assignable). If the Q values are large enough, it will be very tonal with a long sustain.

PolyDrum_002.mp3

Description:

A sample of the resonator synth. Each chunk of notes is done with the synth's parameters mostly the same, but one, envelope generator time, is increased for each chunk.

Filesize:

3.09 MB

Viewed:

2360 Time(s)

aerogramma · Posted: Sun Dec 06, 2009 1:43 am Post subject:

like its dirty sound... great!

Wink

aero

wmonk · Posted: Sun Dec 06, 2009 8:46 am Post subject:

Sounds great ScottG! Dirty and warm.

Smile

JovianPyx · Posted: Fri Dec 11, 2009 8:11 am Post subject:

Here's a piece of music made with the PolyDrum. I've set it up to be tonal again, but it is tuned an octave higher than the other clip I posted. It is also set to use each resonator as an individual drum (32 total drums).

I will be working on another piece of music that will demonstrate this instrument's more unusual sounds and behavior.

PolyDrum_004.mp3

Description:

A piece of music played entirely on the PolyDrum synth. No effects were used to process this sound.

Filesize:

2.63 MB

Viewed:

2166 Time(s)

blue hell · Posted: Fri Dec 11, 2009 8:50 am Post subject:

Nice organic sound!

JovianPyx · Posted: Fri Dec 11, 2009 8:53 am Post subject:

Thanks for the nice comments. I was actually sort of surprised at the "earthy" sound I got, which inspired the piece.

aerogramma · Posted: Fri Dec 11, 2009 9:03 am Post subject:

great sound again scott and nice track too

I don't know enough about it but have you considered offering this as a DIY project?

aero

JovianPyx

DIY project - hmm. Yes and no. The biggest problem for that is that the FPGA doesn't come in a package that is easy to solder like TQFP. It's a BGA device (ball grid array) which requires some sort of oven to solder it to the board.

Because of the BGA issue - which is true for all Xilinx larger FPGAs - I have used premanufactured development boards.

But I'm not done... I may see if it's possible to use several smaller devices which are easy to solder - but that creates programming logistical problems that don't exist when the entire design fits into a single FPGA.

kkissinger · Posted: Fri Dec 11, 2009 9:52 am Post subject:

Intriuging. It really has a beautiful -- almost acoustic kind of sound.

aerogramma

State Machine · Posted: Fri Dec 11, 2009 11:10 am Post subject:

Scott,

Nice work and concept. Very Happy

The timbre of the music has such a mellow effect as one would expect from your implimentation. Very nice use of the FPGA in a musical application. BRAVO again ....

Bill

JovianPyx

Wowee, thanks to all for the positive comments.

You know, when I dreamed this up, I really thought this was a hairbrained idea that would probably not work very well. And the first version didn't. The problem was that high Q filters like these need a rather wide arithmetic buss. My SVF design uses 35 bit signed values as input and output and the integrators are 70 bits wide because a multiplier is used and it's output it 70 bits signed. Since I had the realestate, I just left the RAMs at 70 bits to prevent loss of precision.

cloudscapes · Joined: Feb 09, 2004 Posts: 100 Location: Montreal

this sounds incredibly lush! thanks for sharing!

hexagon5un

JovianPyx · Posted: Wed Feb 03, 2010 11:18 am Post subject:

This is the board:

http://avnetexpress.avnet.com/store/em/EMController/Development-Kits-and-Tools/Avnet-Electronics-Marketing/AES-SP3A-EVAL400-G/_/R-8007786/A-8007786/An-0?action=part&catalogId=500201&langId=-1&storeId=500201

But it's $49 now.

Also - I hope that I wrote that it has no DAC and no MIDI electronic interface. I had to add both of those. The DAC board (CS4344) cost about $25 and I built the MIDI input on a piece of stripboard.

This code would work on a Spartan-3E board (500K gate part). But would need some modification.

The code for this project is not up... Not sure when I will do that.

v-un-v · Posted: Wed Feb 03, 2010 1:30 pm Post subject:

Beautiful track! Reminded me of little droplets of water forming into ice crystals. Like a little soundtrack to an alternative world Very Happy

More!

Inventor · Posted: Wed Feb 03, 2010 2:16 pm Post subject:

Great music, Scott, I put it on the playlist for my shows.

Les

JovianPyx · Posted: Wed Feb 03, 2010 3:09 pm Post subject:

Ok, here's the source code.

What isn't included is the VB.NET source for the programming user interface. This synth allows you to control a whole lot of parameters, so I wrote a VB.NET GUI for it that communicates through the board's USB TTY bridge. If you're interested, contact me by PM and I'll send the current version - currently it's kind of a mess and I don't want to post it the way it is. But at least interested persons can look at the Verilog code for the synth internals and get an idea of what I did.

State variable filters enhance gain when Q is more than one. That is, the peak amplitude at exactly Fc will be higher than the rest of the spectrum. This is true of both analog and digital SVFs. Because of this, I designed in a amplitude windowing feature that prevents binary overflow (sounds horrible) while preserving as much amplitude as possible. This is a cautionary note - you have to deal with this when making patches so that you get the best amplitude without causing a nasty ear ripping sound.

ver_f.zip

Description:

PolyDrum Verilog Source Code

Download (listen)

Filename:

ver_f.zip

Filesize:

449.19 KB

Downloaded:

680 Time(s)

JovianPyx · Posted: Wed Feb 03, 2010 5:22 pm Post subject:

JovianPyx · Posted: Wed Feb 03, 2010 5:26 pm Post subject:

hexagon5un · Posted: Fri Feb 05, 2010 9:58 am Post subject:

JovianPyx · Posted: Fri Feb 05, 2010 11:17 am Post subject:

Hey Elliot,

VGA - heh, I've wanted to mess with that, but I've been pre-occupied... I'd be very interested in seeing some VGA code added to a synth. I've got several VGA monitors in good working order that I'm not currently using, so it would be fun to see something happen while a synth is playing sounds.

Or a video synth...

kkissinger · Posted: Fri Feb 05, 2010 8:01 pm Post subject: You inspired me

I think the Polydrum concept is so cool that I decided to try the technique on my Aries synth.

http://electro-music.com/forum/post-282892.html

Well... it is a little different because the Aries is a mono-synth. Instead of a bank of tuned filters I just vary the cutoff frequency to get different notes.

v-un-v · Posted: Sat Feb 06, 2010 6:53 am Post subject:

JovianPyx · Posted: Fri Apr 08, 2011 4:55 pm Post subject:

I've been reworking some of my designs, the PolyDrum is now expanded to 55 voices from 32. The hardware is capable of 64, but the state variable filters misbehave when Fc is a bit too high, so I restricted it to 55. The highest pitch 3 or so notes are kind of plinky, but they are tonal and in tune so still useful.

Perhaps more banging and swearing can bring it all the way up to 64. Part of the problem is Q enhancement which creates a higher effective Q as Fc increases. I've found that there is a nonlinear curve associated with what the Q should be for a given note. I spent some time today hand adjusting the resonator Q values and got nice resonant (not plinky) sounds for all 55 notes. I think I may be able to get a few more this way. Then perhaps another piece of music from it since it's got quite a bit more range now.