Hey all,

I've written an external to send audio data from Pd to a DAW that I'm working on. To get the signal through with no errors, I need to increase the DSP block size to 512, which I'm doing with [block~ 512 1 1]. This works fine for my external, but seems to break [dac~]; upon enabling DSP, the error int he title is printed to the console, and no audio output is heard. I also changed the block size in audio settings to 512, which I thought would solve this. My expectation was that the 'Pd vector size' is set by [block~], and the dac~ input vector size is determined by the audio setting. If both are 512, then why might I get this error?

@chvolow24 [adc~] and [dac~ only work with a block size of 64. They control the timing for Pd and control messages are processed between blocks...... so we would have problems with other timed events like midi if that were not the case.

If you put your external in a sub-patch blocked at 512 1 1 and keep your main patch un re-blocked then that should solve the problem, but you cannot have [adc~] or [dac~] in the sub-patch.

A soundcard (dac + adc) will communicate with Pd and negotiate the buffer, but I imagine that the DAW will not do that with your external, if you are having problems with a blocksize of 64 (which is the standard for digital audio in most devices).

If you are actually communicating with the DAW through Pds final output and input then you should probably increase the Delay setting in Pd Audio Settings.... rather than the block size.... if you need a bigger buffer.... https://forum.pdpatchrepo.info/topic/11834/pd-block-size-vs-audio-interface-buffer-size

Or have a look at some of the VST externals that have already been written for Pd to see whether they have implemented somehow what you are looking for.
David.

Thanks David! The subpatch solution worked perfectly, and this is all very helpful.

The DAW is my own, so I'm in control of that end of the connection, and there is a block size negotiation. The errors in the signal at blocksize 64 are probably due to shortcomings in my implementation of the synchronization between the two programs, which uses two named semaphores for writing to and reading from a shared memory block. The actual transfer is just a memcpy. Checking out other folks' VST externals is a great suggestion -- I'll definitely do that.

charlie