Digital Noise Synth Idea

[JovianPyx]

The block diagram depicts the basic idea I have for a digital noise synth.

The top signal chain generates white noise which is filtered by an SVF with fixed configurable Q and allows Fc to be modulated by the middle signal chain and a bias value.

The bottom signal chain generates modulation values to control an NCA in the top signal chain.

An envelope generator contours the amplitude of the top chain.

This is an experiment and may change as I develop it.

The first version will be a single voice, but if it works and is fun and useful, I will probably expand it to a polysynth of 4 or 8 voices.

---

Noise_Synth.gif
Description:	A Digital Noise Synth Idea
Filesize:	6.72 KB
Viewed:	20317 Time(s)

---

[JovianPyx]

This is an update with more detail in the drawing.

The amplitude window is a shifter mechanism that I use to accomodate widely differing amplitudes because of the increased gain at Fc with higher Q values. Without the window shifter, the signal can overflow binary wrap which sounds horrible. The detector and indicator tells me whether a "patch" is doing bad things at which point I can adjust with the amplitude window and then compensate with the level control.

---

Noise_Synth.gif
Description:	Updated Design Block Diagram
Filesize:	10.06 KB
Viewed:	579 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[JovianPyx]

Well, I finally got some sound out of the testbed for this synth. So far, I have the top row of logic running and I just got sound. The noise generator is generating noise, the filter is filtering, and the ADSR is ADSRing.

The structure is evolving into something a bit different than the block diagram, still very similar though. I'll post a new one when I finalize it.

[DrJustice]

I'm curious about the sound of this thing
Are you seeing it as being used mostly with high Q (or self oscillation even) for the filters in order to produce controllable pitched sounds?

DJ
--

[JovianPyx]

[JovianPyx]

What I have working is one chain of noise -> SVF -> NCA. NCA is modulated by an ADSR. The sample below uses the filter in lowpass mode with moderately low Q. This will sound pretty basic, noise into a resonant filter...

The noise source is a 64 bit maximal length LFSR.

I hope to get some modulation stuff done this weekend.

But at least I can get this much filtered noise out of it.

Only one voice is working at this point

Some of the overflow detection stuff is working.

---

noise_synth_01.mp3
Description:	just noise to svf to eg controlled nca	Download (listen)
Filename:	noise_synth_01.mp3
Filesize:	315.15 KB
Downloaded:	1233 Time(s)

---

[JovianPyx]

I just got Fc modulation by filtered noise working.

It seems that high frequency noise modulation of high frequency noise - still sounds mostly like noise. More interesting is when lower frequency noise is used to modulate Fc of high frequency noise. I use "high frequency" to indicate audio range.

So I will extend the range of the frequency slider. Currently I have 4 ranges, each lower setting is another octave lower Fc. I believe that 8 range settings will be quite useful, with extension to more lower frequencies.

Nice low low frequency noise modulating Fc with moderate Q (about 5) gives a nice howling wind sound.

When I get this part of the synth nailed down, hopefully later today, I will post a sample of some howling wind and perhaps other sounds I find while testing.

[JovianPyx]

Here is 3+ minutes of pure noise.

It's a wind with howl test patch on the now 2/3 functional synth.

The sound is made from 2 noise sources, each into a SVF. The audio range SVF Fc is modulated by the other SVF which is set at a low Fc itself.

So enjoy this, - even if you need to get a sweater.

---

NoiseSynth_wind_01.mp3
Description:	A 3 minute howling wind sample using a small FPGA for a DSP noise synthesizer	Download (listen)
Filename:	NoiseSynth_wind_01.mp3
Filesize:	3.01 MB
Downloaded:	1166 Time(s)

---

[JovianPyx]

another sample. an odd bassy noise thing. 3 minutes if you can bear it.

Please turn it up a bit...

---

NoiseSynth_weird_bass_sound.mp3
Description:	weird bassy patch	Download (listen)
Filename:	NoiseSynth_weird_bass_sound.mp3
Filesize:	2.98 MB
Downloaded:	1181 Time(s)

---

[JovianPyx]

Here's a screen shot of the patch editor for the NoiseSynth. It shows a fictitious patch.

I wrote this in VB.NET Express Edition (which is free from Microsoft). The program communicates with the synth hardware over a USB connection to the board. The messages are formatted exactly like system exclusive messages and could be transmit over MIDI. This is more convenient because I don't need to figure out how to interface with a MIDI connection that also supports performance data.

---

NoiseSynthPatchEditor.jpg
Description:	NoiseSynth Patch Editor screen shot.
Filesize:	114.76 KB
Viewed:	536 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[JovianPyx]

The synth now has all of the planned elements coded.

Attached is another wind sound, but this time with subtle amplitude modulation as well as Fc modulation.


EDIT ADD: I have now played with the mp3 going and another synth voice on top of it. It sounds ghostly.

It's implemented, but I haven't yet tested polyphony.... heh heh

---

noise_synth_02.mp3
Description:	another wind sound - Fc modulation and Amplitude modulation, both by separate noise streams into separate filters	Download (listen)
Filename:	noise_synth_02.mp3
Filesize:	2.88 MB
Downloaded:	1165 Time(s)

---

[Inventor]

I'm not following this in detail, but the wind I heard in the chatroom sounded very convincing. I mentioned that one could control the voices with Boolean Sequencers and got a positive response from Scott, so we'll see what happens.

Les

[JovianPyx]

Yes, Les, a sequencer could be added to the logic in place of the MIDI trigger logic for a self contained noise unit. Push a button - out comes sound that is sequenced over time for variety.

I'm thinking of a companion synth that does FM bells as wind chimes where the chimes have random timbre, random pitch and random strikes.

The current version of NoiseSynth is MIDI driven which facilitates testing in my lab/studio.

[nobody]

I'm impressed, especially with the last two samples.

Do you have this in a VST somewhere? I'd love to use it.

[JovianPyx]

[nobody]

How do you use it as it is now?

[JovianPyx]

It's a MIDI synth, so it responds to MIDI keypress note events.

Currently, the lowest 8 keys (on my wavestation) are mapped to the 8 instruments. Each instrument can be separately programmed, but all 8 have the same structure.

Another future version will eventually allow tonal melodies to be played by using higher Q values on the audio path SVF.

[nobody]

I guess what I'm confused about is whether this is hardware or software.

[JovianPyx]

Ah, heh. An FPGA is hardware. It is a system of mostly identical logic blocks, several thousand of them, who's function is defined by the bit file loaded into the device. It "assumes the identity" of the logic defined in the Verilog (or VHDL) source code. It's behavior while in operation is exactly like an ASIC (Application Specific Integrated Circuit) or any other IC that is developed for a specific purpose.

Verilog, the hardware definition language I use, looks a lot like C so it might be easy to assume that the end result is software, but it is not. It allows allocation of the FPGA resources to execute the writer's end purpose.

I think of an FPGA as a pile of "logic Lego". It has RAMs, multipliers, adders, multilplexors, and basic logic functions as well as logic function generators (that you define) that can be connected together in the same way someone might connect logic devices like CD4XXX parts.

This is not to be confused with a CPU which executes one instruction after another in a linear sequence. An FPGA _can_ do that - but a lot of it's charm comes from the fact that parallel logic is possible. For example, in a CPU, if you want to AND two data elements together and then AND two other elements together, you're usually stuck with two instruction cycle times. Something like that is easily done in an FPGA, both operations simultaneously. Arithmetically, this is attractive because an FPGA will have several hardware multipliers inside of it, each of which can execute in parallel with the others.

I do hope that made sense...

[blue hell]

Ah, this seems to be going somewhere, nice Scott!

[JovianPyx]

Thank you Blue_Hell!

And I will respond with yet more noise.

Attached is 3 minutes of wind synth from 2 voices, identically programmed.

I have polyphony!

EDIT ADD: I can see that I need to make this stereo. It should be pretty simple. I've got plenty of clocks.

EDIT ADD ALSO: Just to be clear, for this sample, I held down two keys and did nothing else.

---

NoiseSynth_2_voice_wind.mp3
Description:	Noise Synth making wind noise with 2 voices	Download (listen)
Filename:	NoiseSynth_2_voice_wind.mp3
Filesize:	2.89 MB
Downloaded:	1038 Time(s)

---

[nobody]

Groovy!

[blue hell]

Noice Jovian winds!

[JovianPyx]

Here is another sample.

About a minute of 3 wind sound voices and one low thunder voice.

---

wind_and_low_thunder.mp3
Description:		Download (listen)
Filename:	wind_and_low_thunder.mp3
Filesize:	1.26 MB
Downloaded:	1087 Time(s)

---

[blue hell]

1:22

Usable sounds though!

Edit : oops .. read "a minute or 3"