Hi There,

I have patched a signal generator for an active noise cancellation experiment. Basically it consits of an osc~ which feeds a delay line. I use two delread~ to drive my outputs. One is not delayed at all, the second one can be set to 0-5 ms delay via a slider.

While checking my outputs with an oscilloscope I figured out that my delread~ does not delay the signal at all until I move the slider over about 1.5 ms. Above this level the delay increases proportionally. I have compensated it by adding 1.5 to the value set by the slider but this is obviously a bad work-around.

Any Ideas where this effect comes from?

It's not a bug. delread~'s shortest delay is proportional to the blocksize. In PD the default blocksize is 64 samples. 64samples / 44100kHz = 1.45ms

You can try increasing the sampling rate or decreasing the blocksize with the object [block~]. Be careful with block~ though since putting small values in it could slow your machine down quite a bit. Check out its help.

Another option is to use [fexpr~] which would allow you to make a sample grained delay. It is computationally expensive though. You would have to set the blocksize to 256 which would allow fexpr~ to index 5.8ms of sound.

yeah.. 1.5ms is the time taken to play one block of 64 samples at normal 44.1khz samplerate.
have a look in the pd help files at G.04.control.blocksize and G.05.execution.order

Thanks!

The documentation on vd~ did mention that it could get down to 1 sample if it was executed after the delwrite~ but it never mentioned how to do that.

But form my observation the shortest possible delay is not 1.5 ms. I observed delays pretty close to 0 ms but I had to add 1.5 ms more than what delread~did.
I did not get not delay for settings below 1.5 and then it suddenly jumped to 1.5 but I got nothing for 1.5 and then proportionally more.

i.e.
2 ms (set) = 0.5 ms (measured)
10 ms (set) = 8.5 ms (measured)

There were no steps in which I could increase the delay, it was pretty much infinite (obviously my increments had to be bigger than 1/fs...).

I ran pd with default settings, so no changes of block size.