I'm trying to simulate the behaviour of a bucket brigade delay with pure data and am getting some very distorted results at the moment. The basic principle is to have a fixed array size with a clock that moves samples through the array - changing the clock speed changes the delay time and so as the delay time increases the sample rate effectively drops and you get a darker delay. Another quirk of this design are the pitch shifting glitches that you get when you change the delay time as audio is passing through. As far as I can tell, it is not possible to control the write speed to a delay or array so instead I had the idea of creating many arrays of size 1 and then controlling the read and write to each one with a metro object. At the moment I've have a patch with 64 'buckets' and the metro object should be passing them through at the sample rate, so at this point the sound should be passing through with a tiny delay but no loss of quality (I need to find an efficient way to create many buckets, I think with a subpatch and using [clone]? But I need to sort out the distortion issue first).

Here's patch so far: BBD.pd

Ok, so I've tried again using a slightly different method. Instead of creating lots of arrays, I'm now using a single array but controlling the speed of writing by using snapshot and then tabwrite with a counter. This is a more elegant solution and it does work but still has a seriously distorted output - it looks like each sample gets written several times to the array for some reason. Any help would be appreciated! BBD2.pd

@mjhbrowne Well..... interesting project..... but probably not achievable in Pd unfortunately.
I own a Klark Teknik DN34. Each channel has eight sad512 chips..... (max approx. 13ms delay). http://modularsynthesis.com/loft/450/SAD512D.pdf
This suggests that there were far more than 512 transistor/capacitor pairs in each chip......

(part of........ https://www.electrosmash.com/mn3007-bucket-brigade-devices )
That page sets out the math and the filtering necessary to achieve this back in the bad old analogue days.

Your patch bbd.pd is a start, but the number of one bit tables will be huge, and the clock will need to run much faster than 44.1 Hz. For 20kHz bandwidth we know that 44.1KHz is near perfect.

Your array shows square waveforms........ each value is being written 64 times....... because the block size of the audio is 64 samples, and the metro (data) can only be sent between audio blocks. You have to use an audio clock object (I think someone suggested that earlier in this thread).
And a clock rate of 44.1Hz is wildly slow for any chance of reasonable audio quality.
And you would need to add anti-aliasing and reconstruction filters.
And you would need 100's of thousands of arrays for only a few milliseconds of delay.

I would suggest using [delwrite~] and [vd~}
[delwrite~] sets up effectively a digital "bucket brigade" in ram.
[vd~] can read at any "bucket" output with sample accuracy....... and might give you the effect that you are looking for.......

David.

http://colinraffel.com/publications/dafx2010practical.pdf

Thanks for the replies guys. Seb - thanks, I've already read that pdf though.

David - there are BBD chips of varying lengths, 512 stages being one of the options. In effects pedals they often chain a few of them together - my choice of 512 was simply for testing purposes. I'm also very aware of how sampling works - my maths was correct in this case though - to calculate the time in ms for the [metro] object you need to divide 1000ms by 44100hz - I just simplified it to 1 / 44.1 so the 44.1 in the patch is kHz, not Hz,

Anyway, I have managed to get it working with [ipoke~] which allows variable speed writing to arrays (I couldn't get poke to work on my system). The vd object only allows for changing the read speed whereas I need to change the read and write speeds for this patch. Anyway, you can see the working patch here if you want: BBD2.1.pd (you will need to install [ipoke~] if you haven't got it)

Now I just need to add all the other features that you usually get with delays...