Nyquist Aliaser Project

[Eb7+9]

While mulling over Soggybag's idea of producing a single supply S/H circuit I stumbled on the idea of using under-sampling to produce multiple sidebands as discussed in the Nyquist Sampling Theorem ... for those who don't know, sampling is the first step in converting analogue signals to digital - the next step consists of according the output to fixed levels encoded by a finite "discrete" vertical scale ... here I'm not interested in the second part, so we're staying in the analogue domain ...

The sampling theorem states that aliasing of a fixed bandwidth signal spectrum will produce frequency overlapping if the sampling frequency isn't at least twice that of the highest frequency component in the spectrum of the source material ... well, since lots of interesting music circuits act contrary to conventional fidelity principles the idea came to me of exploring what happens when you go against the principle and bring down those side bands into the audible spectrum ... personally I've never heard of this principle used as an effect so I thought I'd try it out - that is, once I understood the potential of this one clock circuit Soggybag brought up ...

http://www.lynx.net/~jc/NyquistAliaserSchem.gif
http://www.lynx.net/~jc/NyquistAliaserStrip.gif



If you're into Ring-Mods and Multiplying Octavers you'll like the sound of this thing ... the big surprise for me is when you set the sampling frequency high enough chords come out sounding not so screwed up as they do with the two other modulator category circuits ... this was a nice surprise and I find it quite musical ... of course if you bring the sampling frequency low enough then you get into mayhem type stuff ...

A few notes about the circuit:

(i) the 10k trimmer sets the reference voltage on the oscillator circuit - varying this voltage sets the frequency range ...

(ii) the 100k trimmer on the second op-amp is there because my simulator showed I might not get a tall enough sampling spike out of one op-amp alone ... in my prototype this didn't turn out to be true - I guess it depends on a number of factors - jFET and op-amp characteristics primairily ... nonetheless, it doesn't do any harm as is and it may be useful for some depending on what active componentes you use ...

(iii) the 10k gain trimmer sets the Emitter bias on the Bipolar transistor in the FET/Bipolar Darlington pair ... it's important to set this at the right value for the sampler to work ... I think I made a comment about this in Soggybag's post - nonetheless the idea will later be necessary when driving the sampler with a high-gain noise source for the S/H  ... interestingly, this trimmer also has an effect on clock feedthrough going into the sampler - in fact, a nulling point occurs within the useful gain range (!) ...

(iv) I provided simple passive isolation between the two parts of the circuit to minimize clock feeding through the rails ... I used 121ohm resistors between both grounds and rails, filtered by a 47uF cap ... I tried 2200uF and it didn't seem to make a huge difference - but these are just preliminary results anyway ... I also tried hooking things both ways, battery to clock first and then audio after the filter, and vice-versa ... I settled on the later, seemed to be a little better this way ... with gain trimmer adjusted I get reasonably low clock feedthrough - enough to make the effect quite useable ...

Pretty simple build otherwise ... and from this circuit I drew up a S/H circuit that runs on a single 9volt as well ... so far it's different from any S/H I've seen, should be interesting ...

thanks for the inspiration Soggy !

~jc

[Ry]

Wow!  That's a fascinating idea.  I love it when doing something you're taught specifically not to do is the solution to a problem.  Great work!

Ry

[britt-stinker]

If someone ever makes a PCB layout for this it's certainly going to be on my to do list.

[TELEFUNKON]

frequencymirrorconvolution 

[Peter Snowberg]

That's a very nice looking design!

[soggybag]

Thanks for the nod, but it's not really necessary. This looks like a really interesting idea. I'll have to try it out.

[Transmogrifox]

Posts like this are always fun to see.  The clock feedthrough is always a kicker, but it's good to hear it can be trimmed to a tolerable level.  I may have to try this one.  I can't think of where I have it linked...might find it in a search, but I had a self-sampling circuit diagram posted somewhere. If you can't find a link to it, but you're interested let me know and I'll post it on my putfile page.

I think it may be fun to try this with a self-sampler as opposed to a clock--or maybe use a clock enable for the self-sampling circuit.  The original concept was to sample the guitar signal on every positive peak thinking this might be useful as an envelope follower--but may be very interesting sounding as an effect itself in the intermediate stages.

[chokeyou]

so...
when can we hear samples? 

[Paul Perry (Frostwave)]

This will be a popular project, my commmercial Sonic Alienator has a low to ultrasonic sample rate knob & gives great results on percussion in particular.  (the sonic alienator has some other features as well, but the alias mode is popular, it can also give a filter something to bite on).

[DavidS]

I'm working on a PCB layout for this, wondering if you've made any modifications to it yet? And maybe a little clarification about the power rail isolation...

I can see the two separate parts of the circuit pretty clearly, so basically you're using an analog ground and V+ and a digital ground and V+, right? But your explaination of the isolation confuses me a little. Are both sections 121 ohms away from ground? And V+? And there's a 47uF cap on the analog side of the isolation resistors? I think that's what you were saying...

Oh, and is there really not supposed to be an input cap? Is it because it's feeding directly into that jfet?

Thanks!

[soggybag]

That's agood question DavidS. I don't see the 121 ohm resister on either the schem or the strip layout?

[soggybag]

Would it be a good idea to add an input cap o this or was it left off for a reason?

[Eb7+9]

Sticking an input cap is not necessary here - unless you are preceding this pedal with another that puts out non-zero DC (same thing with most tube amps) ...

good reason not to put a cap here because it would contribute/cause switch popping ... this circuit as I've drawn it does not pop at all ... notice the signal encounters only one signal cap in total as it swims to the output - the signal path is DC-coupled up until then ... surely, making the signal go through only one stage of capacitance total must be a way at aiming for top transients - reflections on gizmo land Design for FI-ness rules ...

on the rails what I did is wire a 121 ohm resistor between analogue ground and digital ground and another 121 ohm resistor between analogue +9 and digital +9 ... you could experiment with 220ohms or 470ohms depending on the op-amp and jFET you end up using there ... what I did then was clip a battery to analogue +9 and analogue ground and clipped a large cap between digital +9 and digital ground - which are now at slightly lower than +9 and slightly higher than ground ... then I'd do vice versa with the battery on the digital side and the cap on the analogue side ...

I'm curious to know what y'all find in terms of R's and C's, and which side to the battery goes on ... for those interested in elliminating the feedthrough completely I'd start by first employing two batteries to feed the analogue and digital side separately - then work on the clock feedthrough in the passing system ...

thx for pointing out your Sonic Alienator Paul - I have a buddy in Brazil who owns all your stuff, I'll ask him about it next time ... and thx for all other questions and comments ... btw, at our jam last weekend I had this thing running in the effects loop of my JCM800 along with a Yamaha delay and it was quite heavy I assure you ...

stay tooned ...

[barret77]

sorry for my ignorance, but I don't have the slightest idea on how this project should sound like... is there any similar project sound clip online?

[Eb7+9]

good question - I haven't heard anything close before myself ... not too long ago I posted a circuit that does f(x)=abs(x) (Super Fullwave Octaver) ... now I'm doing Nyquist Aliasing - arrogantly twisting the golden sampling rule ... I wonder if there's any other unexplored communications theorems or math circuits awaiting further indiscretions ??!

[Dave Eason]

Nice idea for a project!

TheApple  Logic Plugin "Bitcrusher" does the same thing; it has a control to reduce bit resolution from 1 - 24 bit and can downsample from 1x - 40x completely destroying sounds.

[Transmogrifox]

good question - I haven't heard anything close before myself ... not too long ago I posted a circuit that does f(x)=abs(x) (Super Fullwave Octaver) ... now I'm doing Nyquist Aliasing - arrogantly twisting the golden sampling rule ... I wonder if there's any other unexplored communications theorems or math circuits awaiting further indiscretions ??!

I have been toying with the thought of a chorus type effect based on the homodyne receiver:

Modulate an audio signal onto an RF carrier, then multiply it by an RF carrier of the same frequency (thus setting it back on its original frequencies).  If you can modulate the "demodulating" carrier frequency to a degree of precision to vary less than about 10-20 Hz, you could get a nice detune/chorus sound, and the effect would be quite a noise maker if the frequency deviated further.  If it deviated too much  (20-40 Hz), you would hear the carrier--and thus you would have a messed up theremin.

In fact, you could probably start with a Theremin circuit to accomplish this, and modulate the guitar signal onto one of the carriers.

EDIT:
What would be unique about this is that all frequencies of the original audio signal would all be detuned equally, thus, the audio signal would start to lose its harmonic relationships between frequencies--to to an increasingly messed up degree as the deviation increases.  It would essentially be a ring modulator for deviations of 50 Hz +, but you would hear the carrier with it unless you were able to cancel the carrier--for instance, two of these circuits in parallel, audio path into one inverted 180 degrees, then invert the output of one circuit, add, thus summing the audio components and cancelling the carrier component.

[DavidS]

I think I love you...

[CS Jones]

I spent the afternoon with this thing the other day... re-arranging the parts and pieces.

If you're into Ring-Mods and Multiplying Octavers you'll like the sound of this thing

No... odds are you'll love it.

the big surprise for me is when you set the sampling frequency high enough chords come out sounding not so screwed up as they do with the two other modulator category circuits ... this was a nice surprise and I find it quite musical ...

The "in-between" settings can be very musical, in the sense that you get the octaving/ring effect and and still have the guitar's sound blended.

I can't rightly describe what happens at certain settings, maybe someone else can help. You'll have a "rolling" ring note riding on the top, something which sounds like multi-octaves in the middle and a pleasant overdriven bottom.

I was able to almost completely cut the bleedthrough out. So far what does come through is just barely noticeable. "Robot sounds" were available too but at the price of higher bleedthrough. This was fine with me anyway since I don't have a real need for a totally detuned type of circuit. I want the blend that this circuit offers. Once I've figured out the values I want this'll be a "set it and forget it" effect for me.

JC, thanks ... I've struggled for the last 6 months or so trying to get an effect like this and I failed over and over again. You're the real deal. This is greatly appreciated. Thank you very much.

[DavidS]

I haven't made this yet, so I don't know what it actually sounds like, but I'm guessing it will be alot like the gapper/snipper effect in sound forge I was so fond of years ago...