Hi, it's been a while since I've used Pd and I have forgotten a lot of what I learnt, which was... not much. Hopefully someone will be able to help me make a shift register, arbitrary length, operated by a square wave (transition on rising or falling edge).

I want to take an audio signal and at the rising edge of the square wave the current value will be recorded and latched in the register, then at the next rising edge another value will be sampled and stored, the previous value will move along one value. After x amount of steps, the value appears at the output of the shift register and is listened to as audio.... so if i get a 20KHz square wave, I will get a reasonable quality audio out + a delay (im NOT just trying to make a delay, btw!)

any suggestions how I might achieve this? MUCH APPRECIATED

Not sure if I get exactly what you want, but you'll might want to check [threshold~], [snapshot~], and [list append] to get you started.

I think what you describe could be accomplished using samphold~ and a delay line if I understand correctly

@seb-harmonik.ar hmm im not sure about that - the thing is, if you take a sample with samphold~ then delay it, then between the time it goes in the delay and comes out, you have no way of altering that time; the delay is fixed. I envisage a shift register, where every clock (or bang, if you like) advances the list along 1. If the list is 10 values, at every bang the 10th value appears at the output and whatever is at the input gets entered as the first value - the queue moves along 1. The difference being that you although you can read at a set frequency - resulting in a fixed delay - you could also read 10 values in at x frequency then read them out at 2x, double the pitch and in half the time.

@alexandros yes, that sounds like it might do it, do you mean enter a list, split off the last value and append the latest value to the start? What do I have to enter in order to view the output of the append to make sure it works(i am not very experienced with pd and dont know my way around)? Also, can i enter a list of loads of 0's without doing it manually? I was thinking I could have maybe 20,000 values/stages in the list/buffer/shift register.... resulting in a 1 second delay with constant clock of 20kHz

the thought of typing 20,000 0's sort of takes my egg off the boil

I'm really not sure what you want to do, but I put together a patch that functions like an eight value shift register.
A square wave oscillator drives the patch and at its rising edge its signal is being sampled and put in the register, where the previous values are being shifted by one position.
When the register is full the first value that came in is being output and the register frees that spot and moves the rest of the values by one position, leaving the first position (the last to be output) empty for the next entry.
The problem is that with a square wave, all the values are 1, which I guess is not what you want. Still the patch might give you some ideas, if it's any close to what you want to achieve...
Let me know if it works for you.
shift_register.pd

@alexandros thanks! that is definitely in the ball park of what i want, but for the purposes of testing I modified it so the value is taken from the phasor and operated by a bang. Could you tell me how I can view all the values in the register at once? So i can see the values ripple through....

Connect a [print] to [list]

I modified the patch to sample a sin wave of 500Hz at 20kHz, to check if it was working, changing the output from the list to an audio signal using sig~, drawing that with an array and listening to it. Visually the output doesn't seem to be sampling at a high enough frequency (the original wave and the output look pretty different) and the sound sounds pretty distorted and weird. The patch works fine when banging values through it, but when you try and sample a wave in this way it seems to have problems. Anyone have an idea why?

shift_register2.pd

I guess it's because [threshold~] will output a bang only after a sample block has finished. For a 44,100 Hz sampling rate and a 64 sample block size, this will take about 1.45 milliseconds, whereas a period of a 20,000 Hz frequency lasts only 0.05 milliseconds (correct me if I'm wrong).
What I understand is that the control signal (the square wave) runs much faster than the sampling of [threshold~]. I changed your patch and put the sampling part in a subpatch with [block~ 1], even used a [lop~ 100] instead of [sig~], but still it didn't work.

I don't think that sampling in the control domain will really work. Why don't you try [samphold~] instead? You could probably use a one element array and send the sample to it with [tabsend~] inside a subpatch with [block~ 1] or something (just brainstorming here).

yes that makes sense, i will try that

but - this may save me a bit of time - is it realistic or feasible for me to be sampling at, say, 20kHz - 100kHz in Pd? I am starting to think maybe i need to simulate this with something else..

I haven't really understood what you're trying to do, but sampling at 20kHz in the control domain doesn't seem to work. I don't think the frequency of the signal you're sampling has to do with the sampling quality, but the frequency that controls the sampling (in your case 20kHz) is maybe what you should tweak.
Also, you'll probably need to make a ramp between consecutive values if you're to play the samples in the signal domain, otherwise you're very likely to get clicks.
For what I know, if you want to work on the signal domain, stay in that domain and don't mix the two. Unless you want to use the sampling data in the control domain, which is another story. In this case you're probably confusing yourself by trying to project the data onto the signal domain and play it as an oscillator.

Well, these are my two cents, maybe I'm wrong.