March 21, 2007

Replication...

by Chris Randall
 

Okay, this is turning in to more than a simple "throw a lag in there and release an update" kind of thing. There are four particular instances where clicks occur in Replicant. What they are, and what we've done, are as follows:


1. Clicking when the audio engine switches from the "live" feed to the "loop" feed and back when in duck mode. This was caused by the audio engine switching streams at something other than zero crossings. A very quick fade (on the order of a millisecond or two) on the in and out of the "live" feed solved this problem. FIXED.


2. Clicking in the looping engine itself, caused by loops that are cut at something other than zero crossing points. Apparent in mix mode, but obvious in duck and send. This was solved by recording a very quick fade at the beginning and ending of loops as they are recorded. (There was already one on the beginning; we just put one on the end.) FIXED.


3. Clicking when audio first enters a filter, and that filter is anything but wide open. This occurs in all three modes. We haven't figured out why this is happening yet. FIXED, sort of. We got the filter popping fixed, but Replicant Whack-A-Mole means that we created a new problem with this. Whee.


4. Clicking when the following conditions occur: loop length is set to anything larger than 1/16, and a retrigger occurs mid-loop. Say you have triggers on steps one and three, and the loop length is set to 1/4 note. Obviously, step three is only 1/8th note, or half of the desired loop length. So the loop will only play half-way before getting retriggered. Thus it isn't able to play the fade-out that was recorded to fix the clicks in instance #2, noted above. This one is the biggest bear of all, and we're not sure how to address it, as in order to play a crossfade, the plugin would have to know that a random trigger is going to occur on a given step. In normal programming, this click can easily be avoided. But when you're using heavy randomization of the loop length, and the sequencer mode is set to "retrigger," you get a lot of these.


So, here's what we're going to do: we're going to release an update shortly that has the first and second instances (which are currently unavoidable) fixed, but not the third and fourth (which are more or less avoidable, except in certain circumstances.) We'll continue working on this exclusively, and release another update when we have fixed the third and fourth instances.


As I mentioned on KvR, we designed this plugin with drum slicing in mind, and it didn't really occur to us that it would see such heavy use on other material. We didn't rush the release of this product; we don't rush any of our products. We release them when we feel they do what we want them to; Replicant is perfectly capable of slicing straight-time drum loops ad nauseum, and you'll never hear a click that will be more apparent than the program material. Using it on sustained material is certainly feasible in its current state, but you have to remember that it is, as I've mentioned previously, like drying your cat in the microwave. It _can_ do it, but it isn't very good at it. The final revision will basically have the entire looping engine re-written to allow it to excel at slicing sustained material, so the cat will dry un-harmed.

 
 
 

14 comments:

Page 2 of 2
 
 

 
Mar.23.2007 @ 8:02 AM
noisegeek
Thanks guys. I had a basic idea of what it was just from hearing it in music, but never knew the technical details. Good description by the way Chris. Much easier for a visual thinker like myself.
I suppose one of these days I should just bite the bullet and pay for some piano lessons or something so I can learn this stuff properly rather than just reading and asking questions online.
 
 

 
Mar.23.2007 @ 10:43 AM
wquoyle
y'know Computer Music gave Replicant 10/10?

If it, like, matters to you at all...

 
 

 
Mar.23.2007 @ 11:03 PM
Chris Randall
Yeah, we know. Generally, most major magazines send you the review they're going to publish so you can fact-check it and correct it if they got anything glaringly wrong. This is, of course, a courtesy, as they don't have to. But we know usually two to three months in advance about a review and what it's going to say.

-CR

 
 

 
Mar.24.2007 @ 4:20 PM
Adam Schabtach
...and yes, a 10/10 from CM _does_ matter to us, but it's kinda been a hectic week... thanks for mentioning it, though.

--Adam

 
 

 
Page 2 of 2
 
 

Comment:

 

Sorry, commenting is closed for this blog entry.