March 23, 2007

Whack-A-Mole, Replicant Iteration...

by Chris Randall

Okay, if you're a Replicant owner, go log in to your Audio Damage account and download the installer again, and you'll be blessed with a version that has a lot of fixes. The new Windows VST version is 1.0.2, and the OSX versions are 1.0.4.

As I mentioned in my last post on the subject, there are four distinct events which caused brief audio spikes ("clicks" in the vernacular) in Replicant. We've fixed three of them. The fourth is avoidable, and requires a massive rewrite of the audio engine to fix, so we've left it in for the nonce. When Adam returns from vacation, we'll readdress it.

How it occurs is thus: if a trigger event happens while a slice is anywhere but its natural beginning and end points, and there is audio in the buffer, there will be a brief "click" as the stream switches from the old event to the new one. This occurs because the audio event is getting switched and the audio isn't at a zero crossing.

To put this another way, if you have a slice that is 1/8th note long, and you have a retrigger at 1/16th, the slice will only play half way before a new event starts. When it switches to the new event, you'll get a little click _IF_ there is audio in the buffer. With drums, this isn't a problem as there usually isn't audio in the buffer. On, say, a synth pad, there will be, so there you go.

This is easily avoidable by not having triggers midway in slices. If your slice length is 1/8, don't put any triggers on anything but 1/8 spots, and you won't hear any clicks. The trouble shows up when you have a lot of randomization on loop length or triggers via the "random beats" slider.

Now, the other caveat: there was a small pop on _EVERY_ slice joint, no matter what, in the original Replicant. There were also clicks when the filters turned on and off. We usually turn them off when they're not in use to save CPU; in Replicant we defeated this behavior, which got rid of the filter clicks and gave it maybe a quarter of a percentile more CPU usage on a normal modern machine. Since some of these clicks happened at the same time, there was some phase cancellation. Now that the only clicks are the loop cutoff clicks, they are a little bit louder than they were.

Like I said, Whack-A-Mole.

In any event, the vast majority of problems that have been giving grief to customers that want to use Replicant on sustained material have been solved. The simple rule to avoid pops on sustained material altogether is thus: don't have trigger points closer together than your maximum slice time, and you won't hear any clickin'. Nuff said. We'll address the last problem when Adam gets back from vacation.




Mar.23.2007 @ 5:46 PM
Just did a back-to-back test now that I'm home between this and the previous version, very audibly not-clicky now. I will for the time being keep both versions, since I like the clicks in some situations. They don't call it glitch for nothing.

Good job on getting the fix so quickly.


Mar.23.2007 @ 11:32 PM
Good folks here. Good folks.



Sorry, commenting is closed for this blog entry.