posted 12 April 2010 11:06            

My inclination would be to suspect that a digital connection is not clocked correctly.

posted 12 April 2010 13:10            

Here is an example. It is the preset sound Synthesis-Keyboard Controlled / Formant Bank Oscillator KBD / bell voice KBD. Only modification is reduced output gain.

Again, MOTU Traveller MK3 firmware 1.02 (most recent one), set to internal clock, analog outputs used (feeding into analog ins of main DAW interface). Nothing except Pacarana and MOTU on the FW bus, everything else is on a dedicated FW controller card. Quality cabling.
All buffers etc set to maximum values to be on the safe side.

Procedure: set SR to 96k. Compile, Load & Start. Play 4 notes while recording the output (with the Metric Halo MIO console app in this case). Crackles. Apple+K to kill the sound, switch to 88.2k. Compile, Load & Start. Play 4 notes while recording. No crackling. Apple+K, switch to 176.4k, Compile & Load & Start, play 4 notes, no crackling. Apple-K, Switch to 48k (!!)..... crackling.

Here are the 96k and 88k examples as wav files: http://www.surroundsfx.com/images/Pacarana_On_Crackles.zip

Note that the errors are spike-like, not typical no-sync dropouts. Like audio data from a wrong memory location multiplexed into the output stream.

When I installed the MOTU driver on my Laptop and recorded at 192k, there was no crackling. So it's *not* the clock oscillator on the MOTU that is behaving erroneously at 48k and multiples (at least thats my current assumption).

My (un-)educated guess is that some piece of code isn't scaling to 48k and multiples thereof correctly.

posted 12 April 2010 15:21            

One thing common to most of the Sounds I use is that my sample library is in 192k and that thus most analysis type files will have been generated from that source sample rate.

posted 13 April 2010 15:29            

How many psi files are in the Tau editor when you hear the clicks? Am asking because when I tried it just now with two files at 192 kHz in Cross mode, I didn't hear the clicks. Did you have more than two files?

Another possibility is that the voiced/unvoiced markers are in the middle of a noisy segment so when it crosses over from one to the other you are hearing a click in the transition?

Could you please send us (by email attachment) the Tau file, Psi files and the FormantBankOscillator that is producing these clicks so we can try to track down the source? (We can definitely hear it in the recordings you sent but if we have the original Sounds and files we might be able to determine what's going on).

posted 14 April 2010 11:30            

(first off: the example files linked to in this thread are showing my *general* "48k andmultiples thereof" issue, NOT the TAU issue.)

"How many psi files are in the Tau editor when you hear the clicks? Am asking because when I tried it just now with two files at 192 kHz in Cross mode, I didn't hear the clicks. Did you have more than two files?"

No, just two files. Make sure that sourcemorph is a non-integer value.

"Another possibility is that the voiced/unvoiced markers are in the middle of a noisy segment so when it crosses over from one to the other you are hearing a click in the transition?"

Hmm, hard to tell, theoretically that could be --- using a piano tone and clangs on a metal gate here, so they aren't exactly noise-free sounds. BUT the unvoiced markers appear to only at transients, not in-between.

"Could you please send us (by email attachment) the Tau file, Psi files and the FormantBankOscillator that is producing these clicks so we can try to track down the source? "
Done.

Leaving the TAU be for a moment, and going back to the 48/96/192k issue : I'm seeing crackling distortion (or distorted crackling...) as heard in the files linked above on EVERY sound I play when running at 48/96/192k (not just those using formant oscillators/TAU etc). Switching to 44/96/192 immediately cures it.

Also, changing the I/O Delay parameter for the MOTU changes the density of crackles (shorter I/O latency = more crackles per second), which could indicate some buffer either getting some crap written to it at the beginning or end (by some overflow somewhere else) or running longer/shorter than it should.

Basically: maybe it is a problem with the MOTU Traveller MK3 driver in the Pacarana?

cheers,
d