MUFF WIGGLER Forum Index

 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

analogue neuron circuit
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2, 3  Next [all]
Author analogue neuron circuit

andrewF

Lately I have had to research a lot on neural networks, really does my head in. Anyway this circuit was in one paper and it looked simple enough, so i built it.
Guess the vid shows what it does. It is being fed two LFO tri-waves, with the frequencies being changed manually every now and then.
The output is modulating cutoff in a VCF which being fed a steady audio signal. It is also useful for doing things to audio signals.

So if you want an easy-to-build, simple source of unpredictable mad shit in your synth; this might be the one.

If using a quad op amp, perhaps the 4th stage could drive a LED?






BugBrand

It may not be related and you probably know of it already, but your post made me think of::::

http://www.musicsynthesizer.com/Neurons/Neurons1.html


andrewF

Thanks Tom, I know Grant's page but haven't tried his circuits.

I often build neuron circuits from journal papers ("research") but most are not particularly interesting, this one is among the best for synths, imho, and about the simplest.


gwaidan

Looks fascinating-thanks for posting this!


lazerkind

Oh I'm so going to perfboard that SlayerBadger!
Thanks for posting.

Is that the dual CGS utility LFO on there?


andrewF

lazerkind wrote:


Is that the dual CGS utility LFO on there?

yep CGS dual LFO


EATyourGUITAR

don't forget the jomox


and hell yes we need more of it in DIY. I dont have $4000 for a neuronium.

found this collection of resonator neuronium tracks on sound clound.


nathanxl

andrewF wrote:
Lately I have had to research a lot on neural networks, really does my head in.


eek! hihi hihi


tonyvoid

wow so simple yet wild random complex waveshaping. this is perfect. I wonder how it affect signals from my lunette. Simple enough addition.


don h

This circuit is really great, thanks!

Adding this to your reading list:
http://books.google.com/books?id=NxycaQH6PeoC&pg=PR12&lpg=PR12&dq=mark +Dolson+machine+tongues+neural+networks&source=bl&ots=AwKiswBtzC&sig=a Wh2wQ7XVH3YXMemPrbJ7HfoIcU&hl=en&ei=qVK1TvjQLoTg2QXQpIHNDQ&sa=X&oi=boo k_result&ct=result&resnum=3&ved=0CCsQ6AEwAg#v=onepage&q=mark%20Dolson% 20machine%20tongues%20neural%20networks&f=false


andrewF

cheers don h
it looks a good read and its on the shelf at my uni library so will pick it up next week.


RealDudes

this is hot shit and easy enough for me to make.
thanks Andrew!!!


decaying.sine

don h wrote:
This circuit is really great, thanks!

Adding this to your reading list:
http://books.google.com/books?id=NxycaQH6PeoC&pg=PR12&lpg=PR12&dq=mark +Dolson+machine+tongues+neural+networks&source=bl&ots=AwKiswBtzC&sig=a Wh2wQ7XVH3YXMemPrbJ7HfoIcU&hl=en&ei=qVK1TvjQLoTg2QXQpIHNDQ&sa=X&oi=boo k_result&ct=result&resnum=3&ved=0CCsQ6AEwAg#v=onepage&q=mark%20Dolson% 20machine%20tongues%20neural%20networks&f=false


There is a lot of really good stuff coming out of UCSD in computer music. I went to grad school there and was on the faculty afterwards. However, I never took advantage of the comp music resources because I wasn't in to synths back then.

Thanks for sharing that resource, and this circuit looks really cool Andrew.


DGTom

Great little cct. Andrew, thanks for posting! I would love to know what the intended use / context of it was? (I'm guessing the paper in question isn't publically available)


so, before I built it I thought I knew what I was looking at. The left & right OAs seem fairly straightforward; rectifier & mixer. I thought / am thinking the bottom is acting as a comparator?

Based on that assumption I used the 4th OA in a 074 as a 2nd mix stage & fed that into the non-inverting input of the lower OA, wired a pot between the rails & added a 4th CV input which forces the output "more negative" or "more positive" depending on the the polarity of the input signal.

But now I'm not so sure that OA actually is a comparator hmmm.....


andrewF

DGTom wrote:
I would love to know what the intended use / context of it was? (I'm guessing the paper in question isn't publically available)

The paper is called 'A general purpose analog neural computer', yep cannot post it publicly but I do have permission to send it by private message for 'the purpose of discussion'

I assumed the lower op amp is a comparator too, although the first stage is a half wave recitfier with the diodes reversed, guess I should put a probe on the output of that and see what it is spitting out MY ASS IS BLEEDING


DGTom

Guinness ftw!

the awesome sauce appears to be in tapping the junction between the diodes, I've never seen that done on purpose - tho I have sabotaged myself a few times doing it by mistake & wondering why something so simple isn't working the way "I know it should"

mine appears to be oscillating at some extreme knob settings - but not repeatably, scope time!


CLee

Interesting circuit Andrew. Could you give some hint to the "math challenged" among us (me) what the original intended purpose of the circuit is?


frijitz

Am I reading this right? It's a half-wave rectifier mixed with a squared-up version of itself? Plus an input offset? Lots of bang-for-buck!

grin

Ian


andrewF

CLee wrote:
Could you give some hint to the "math challenged" among us (me) what the original intended purpose of the circuit is?

The paper describes the neurons as having input/output relations as 'idealized versions of a typical biological neuron''
These were built into chips with programmable switching arrays to act as synapses with the intention of developing applications in vision and acoustical processing. It is from the 80s, so probably considered a bit archaic these days, but perfect for our purposes.

The original circuit had more inputs and i would assume these were 'weighted', as in different resistor values so the effects of other neurons would be different according to which combinations of signals were being received. the original circuit does not have the 22k resistor and 100k pot between the comparator and the summer, originally it was a 500k resistor and the other resistors around the summer were all 100k. So originally the comparator output had a gain of 0.2 in the summer, whereas my mod gives it a range of 0.18-1

The voltage offset pot is used to select where the linear region starts, as can be seen in the pic from the paper, it is set to 1.5V. Thats probably what makes this circuit interesting; the jumps from linear response to nonlinear.


andrewF

Spotted the Patchell CV Splitter on a recent trawl.
A very similar circuit to the neuron, would be interesting to add the neuron's comparator to it, just to crazy it up a little.
Jim P is clever!


asterisk

what kind of diodes did you use for this?
and what value is that initial pot connected to +12v & gnd? 100k also?

im looking forward to building a couple of these.


andrewF

Just use regular signal diodes, such as 1N4148
The pots on mine are both 100k

Good luck with the build, about the only tricky bit is getting the diodes pointing in the right directions.


cyklopljud

The orientation of the diodes got me for a moment too.
Looking forward to more of these.


andrewF

The build notes for Two Neurons are on this page


asterisk

thanks for the info.

so the schematic on the build notes is just for 1 neuron right?

and the second neuron is the same as the first except for the diode arrangement?

does the different diode arrangement cause the two neurons to behave in different ways?

MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2, 3  Next [all]
Page 1 of 3
Powered by phpBB © phpBB Group