Analog Bit Crusher

[ExpAnonColin]

Hello fellow DIYers.

A little while back someone asked me to make them a sample rate reducer.  A little while after, someone posted on my message board about Eb9+7's nyquist aliaser.  The Nyquist Aliaser is brilliant but I have very little knowledge of transistors (my teacher last semester said "We're going to treat transistors as switches, if you're interested in other uses, please feel free to learn more outside of class).  I don't like to build things I don't understand, so I designed my own sample rate reducer using op amps.  It uses 4 op amps in total, or two duals or a quad.  There are two noninverting buffers, the standard JFET sample/hold with a cap, and a schmitt trigger/integrator oscillator being turned into a pulse generator via two diodes.  It works and sounds great, though there is some amount of bleedtrhough, which at this point I'm fairly sure is due to noise in the signal being sampled and held combined with the regular bleeding you'd expect... what was interesting was that using a quad vs using two duals did not yield much difference in bleedthrough.  The only method I have found to work at all for nulling the clock is to wire a 470k-1M resistor to the gate, to the center lugs of a 500k-1M trim whose side lugs are ground and 9v.  The problem is, this creates misbiasing either at all frequencies of clock or at some rather than others.  You might be able to reduce the noise somewhat by isolating one dual's supply voltage from the other, via 100ohm resistors or so and perhaps a filtering cap.  If I am ever to sell this commercially, I would probably make the input buffer a noninverting amplifier with a gain of maybe 4 or so, and make the output amplifier have a gain of maybe 1/4 or so, so that any clock noise introduced by the JFET switching would be removed, and any input signal noise would remaing the same.

All that said, here is the schematic.  I ahve no layout as my perf layout uses some unconventional methods.



If you are interested in designing your own, I'd recommend checking out...
http://www.musicfromouterspace.com/analogsynth/singlechipsampleandhold/singlechipsampleandhold.html
http://timara.con.oberlin.edu/~jtalbert/S&H/S&H.pdf (My teacher designed this one!)
http://www.hobby-elec.org/e_ckt17.htm

If you are interested in the thread with my building and prototyping notes and all, as well as discussion of a Low concert in the cities:
http://experimentalistsanonymous.com/board/viewtopic.php?t=647

-Colin

[KerryF]

Ok, so what exactly does this do?  Sorry  .  Also, have you built it?  Any pictures or sound clips?

[ExpAnonColin]

Ok, so what exactly does this do?  Sorry  .  Also, have you built it?  Any pictures or sound clips?

Bit crushing is an effect that is becoming more and more common, especially in electronic music.  What it does it take a signal and approximate it every so often.  So imagine you have your guitar signal lookig like a sine wave, and imagine that you have a clock signal that, at every pulse of the clock, looks at your guitar signal and outputs that voltage until the next pulse of the clock signal... over and over again.  This is done in audio rates, so you get really neat effects in the end, it is somewhat comparable to a ring mod but more crunchy and less atonal.

There are some soundclips at the Nyquist Aliaser page, as well as perhaps a less scatterbrained description.
http://www.lynx.bc.ca/~jc/nyquistAliaser.html

If you understand what sample and hold is, think of it as sample and hold at audio rates.  I have built it on perf, and am about to commit it to a box.  If I ever improve the circuit at all, I will probably build a second into a nicer looking box and try to sell it.

-Colin

P.S. - Eb7+9, the topic on my forum is the first result in google, while your page is second!  Sorry! 

[KerryF]

sounds cool.  im listening to it and it sounds like a cool type of effect to have.  if you can make a sound clip once you finish it, that would be cool.  maybe making like a soundclip with a well known song would get it more understood.  but cool.  good luck.

[Eb7+9]

Colin,

that's quite an ellegant "beginner friendly" alternative to my sub-sampler ... your idea of increasing Signal-to-(Clk)Noise in the first stage and then scaling back is a good one ... it's hard to get away from clock feedthrough happening via the small gate-to-channel stray capacitance using this straight pass circuit - even with the very low Zout of the op-amp ... I was looking at alternative sampling blocks at one point - maybe something "cleaner" will come up ...

~jc

[markusw]

Cool  Thanks a lot!

Reminds me of http://www.geofex.com/FX_images/stutter.pdf#search=%22stuttering%20pedal%22

Markus

[ExpAnonColin]

Thanks Eb7+9, let me know what you come up with.

Would adding that network around the JFET from the stuttering pedal help null the clock in mine?

-Colin

[snufkin]

cheak out BUGBRAND just google it

he is an fx pedal buider who lives quite near me he has a schem for a bit crusher and sound sampels


he also has a modular synth and loads of other cool stuff

[ExpAnonColin]

cheak out BUGBRAND just google it

he is an fx pedal buider who lives quite near me he has a schem for a bit crusher and sound sampels


he also has a modular synth and loads of other cool stuff

Yes, I met Tom once while I was in Bristol.  Awesome guy! Very much a doer rather than a planner, so he makes all kinds of neat stuff all of the time. He posts on my forum quite a bit.  The bug crusher is badass.  If I ever made a comercial crusher, it owuld have envelope controlled sample rate though.

-Colin

[snufkin]

cooooooool   i met him once or twice at gigs he played and an  demo he did in a shop in bristol hes a nice guy with some very cool wierd noize makers and pedals


your site rocks too

[ExpAnonColin]

cooooooool   i met him once or twice at gigs he played and an  demo he did in a shop in bristol hes a nice guy with some very cool wierd noize makers and pedals


your site rocks too

Thanks man.

Was it at that gallery sort of thing?  Where in the bottom he had is drone thing, the pendulum music thing and the speaker-connected-to-string-to-pen thing?  Because that's the one I was at.

-Colin

[snufkin]

no but i saw the photos of that it looked very cooooool


there was a sort of market thing set up in the here shop in britol where local artisan types sold there stuff like diy clothes comics and crazy fx pedals 

[Eb7+9]

I went back and checked that National AN32 jFET app. note and had forgot about using a diode instead of 1meg resistor on the gate - that could make a big diff ... also AN32 shows a 1meg resistor then going between Gate and Drain - possibly to help bleed that stray gate-channel cap ... hadn't seen the stuttering pedal before, it makes use of that diode ... I imagine it operates in a cyclical manner - reason why I mention that is someone once asked me if a Morse-Code kinda pedal would be possible, with randomly alternating short/long on-cycles but of fixed short/long lengths ... someone here (RG ?!) could probably figure out how to adapt the stutterer to do this ...

[ExpAnonColin]

I did try that 1M resistor with negligible results.  I will try that diode though, and then 100k and a diode.

-Colin

[gez]

Try a trimpot comming off the square output to ground (gate fed from wiper via diode/large resistor).  Less voltage = less charge injection.  It's not perfect, but it can reduce bleedthrough significantly.

[birt]

i was thinking maybe lighting control systems could be used for bit crusher effects. i might be totally wrong.

right now allmost all lighting computers use DMX (digital multiplex) to control dimmers. but before that it was AMX (analog multiplex). i think it used some sort of control voltage wich was set by the slide pots on the console and there were probably chips to sample this and make a multiplex signal of this wich would basicly be a bit chrushed signal if you would feed it with audio.  i hope what i just said makes sense.

[Paul Perry (Frostwave)]

Are you sure that's a 'bit crusher'?
My understanding is that a bitcrusher screws with the resolution of the amplitude of the signal (usually by converting to eg. a 16 bit digital representation, then losing some of the least significant bits, and reconvertig to analog). What this circuit is, is a variable rate sample and hold, putting discrete steps in the time dimension, rather than amplitude.
In either case, it's well worth doing... my commercial "Sonic Alienator" does both, plus has an analog filter that can be run eihter pre or post the sample/hold section (gives different results re the aliasing artifacts).
Another way to generate stepped amplitude might be to have a staircase generator that stops when it exceeds the signal amplitude. Certainly, it would sound rrrrough!

[ExpAnonColin]

Are you sure that's a 'bit crusher'?
My understanding is that a bitcrusher screws with the resolution of the amplitude of the signal (usually by converting to eg. a 16 bit digital representation, then losing some of the least significant bits, and reconvertig to analog). What this circuit is, is a variable rate sample and hold, putting discrete steps in the time dimension, rather than amplitude.
In either case, it's well worth doing... my commercial "Sonic Alienator" does both, plus has an analog filter that can be run eihter pre or post the sample/hold section (gives different results re the aliasing artifacts).
Another way to generate stepped amplitude might be to have a staircase generator that stops when it exceeds the signal amplitude. Certainly, it would sound rrrrough!

It's definitely a sample rate reducer, but Analog Bit Crusher = ABC

-Colin

[ExpAnonColin]

After some tinkering, I've found that a diode doesn't seem to work at all, or the diode-resistor with cap to ground or any variation.  Gez, your tip helps but it's hard to get much nulled before the entire signal drops out (what's ironic is that if you did the amplify/S+H/Deampllify thing, you'd probably have a signal drop with a much quieter carrier!  That is clearly the way to go for a more complex app).  But, what was interesting was that in this application a straight up kludge 10k resistor to ground at the output of the pulse generator got rid of all non-airial-bleedthrough.


-Colin

[gez]

After some tinkering, I've found that a diode doesn't seem to work at all, or the diode-resistor with cap to ground or any variation. 

When the square output is at ground potential the FET is off (providing pinch-off threshold is reached), but when it's at V+ potential the gate-source junction is forward biased.  A large resistance is needed to minimise current flow and a reverse bias diode does this nicely (think millennium bypass here).  Your 10k to ground is sinking current from the op-amp, probably enough to divert some away from the FET's gate and minimise bleedthrough?  I should imagine that the effectiveness of this trick would vary from type (of amp) to type. 

A RC combo isn't really appropriate: that tends to get used to slow down switching, but here your sample rate is going to be too high for you to be able to make any meaningful difference.

Gez, your tip helps but it's hard to get much nulled before the entire signal drops out

Try one of those multi-turn trimpots.  They're a pain in the a*se to use as they require 20-30 turns (something like that...who's counting!), which makes them ideal for fine tuning (also means there's less impact on an effect when they accidentally get nudged by vibration etc).