Bizarre Export Problem, Resolved and Noted for Future Reference

Today I experienced a bizarre and entirely new behavior exporting under 8.5: there were no vocals in my exported mix. 

I had exported this same song many times before; this problem only began after changing the effects chain on the vocal bus. 

I assumed I'd muted the bus or something equally trivial so I went back in and verified that the song played back OK and that every submix was properly routed to the master bus. I examined the export options to make sure I was using my standard settings, and I was: Entire Mix, all automations enabled. Tried it with Fast Bounce disabled, then with the 64-bit option turned off, neither of which made any difference.

After eliminating routing issues and export options, I figured it must be one of the plugins on the vocal bus. Sure enough, after bypassing all bus effects the export now included vocals again. So I then started bypassing them one at a time to isolate the culprit.

I bypassed just the first plugin, FabFilter Pro-C, and the vocals came back. Hmm. This is my go-to compressor and has never given me trouble before. And it's being used here solely for de-essing. Disturbing, if this reputable plugin turns out to have some weird bug.

Not wanting to accept a problem with my beloved Pro-C, I then bypassed the second plugin, Kjaerhus Classic Compressor. To my surprise, the vocals were once again present in the export - even after re-enabling Pro-C! So it wasn't Pro-C after all.

Then I bypassed the third plugin, Ozone 4 (used for the exciter only). And oh sh*t, this is getting weird now...my vocals came back with Ozone out of circuit, too. Bypassing any of the first 3 plugins seemed to cure the problem!

So now I'm thinking maybe there's a bug in SONAR that shows up when you have 6 or more plugins on a bus (not as improbable as it sounds), or a particular combination of plugins. I saw that I had two already-bypassed plugins (Sonitus Multiband and Voxengo SPAN) in the bin so I deleted them. The vocals then came back in the export, even with all the previously-active plugins restored. I am now cursing Cakewalk but happy to have a workaround. The answer appeared to be in the number of plugins in the bin.

So now with an apparent workaround in place, I exported the entire song. You can imagine my response when the exported file again contained no vocals. Nothing had been touched since the prior export, but my workaround did not work twice in a row. SONAR seemed to be randomly muting the vocal bus, only during exports and doing so just to piss me off!

Some intuition made me go back to the export options and disable the "Bus Mute/Solo" option. As no busses were muted or soloed, it was simply a safe thing to try. Lo and behold, my vocals came back to the exported file. Maybe, at last, a real workaround!

Alas, it was not the solution. I immediately did a second export - without changing anything - and the second export had no vocals. Two exports, one after the other, one with vocals and the other without.

I currently have no explanation for this odd behavior, aside from the observation that it is probably an obscure bug. I only relate it here for purposes of documentation, in case anyone else encounters a similar problem. Or maybe, just maybe, somebody might throw out an observation that will break my mental logjam and lead to an explanation.

Nope.

Never heard of "bus corruption" before, but that's actually not a bad idea. It would actually have to be corruption in the cwp file, and perhaps creating a new bus would force SONAR to recreate the bus definition. Thanks, I'll experiment with that.

Bit, how about just 'bouncing to track'? And then export that mix. does that work?

If I remove the compressor I get vocals.

What was happening was I was driving the sidechain input on the compressor with a de-essing key track, but for some reason SONAR would occasionally spit that track out at like 99 decibels, overwhelming the compressor. That's what was killing the vocals in the export.

It only happened during the export, never during playback, and it was intermittent. I'd get a proper export about 20% of the time.

Ultimately I just reverted to a more convenient but less effective de-essing scheme, using the Pro-C's internal sidechain filter. That got me to where I could get a proper export 100% of the time, at the cost of slightly more sibilance.

This was a particularly difficult de-essing job. It wasn't that the "S"s were too hot, but that there was an awful 8KHz resonance buried within not only the "S"s but the hard consonants as well, particularly "T"s and "K"s, and even occasionally on some vowels.

This had been caused by a broken (borrowed) microphone wherein some internal component had broken loose and was resonating inside the mic. I did not catch this until 30 vocal takes had been recorded.

I resolved the problem by going old-school. I bounced all the vocals to a single mono track. Then I put an extreme filter on it to hugely emphasize the resonance, and aggressively compressed it. Now I had a track consisting of just the unwanted resonances, which I then routed to the sidechain input of the Sonitus Compressor on the vocal bus. 

That worked surprisingly well. But later on I decided I needed to automate the bus EQ. That was going to be a problem, because the Sonitus Compressor has a bug in which it will suppress automation to downstream effects if there are more than two effects after it. So the Sonitus had to go. What I did not expect was the subsequent weirdness after replacing it with the FabFilter Pro-C compressor.

I still do not know if the problem was caused by SONAR, by Pro-C, or by the specific combination of the two. But one or the other was adding 1 + 1 and getting 1,000,000.

bitflipper


What was happening was I was driving the sidechain input on the compressor with a de-essing key track, but for some reason SONAR would occasionally spit that track out at like 99 decibels, overwhelming the compressor. That's what was killing the vocals in the export.

Wow ...  The thing that came to mind while I was reading this was ....


It seems the Side Chain input would just go full scale and then trash the output, right?  So I wonder if the SC Input (again whether it is Sonar or the Plug not known yet) can consistently track any signal without making erratic processing levels? Obviously the full scale result is 'easy' to notice -  but what about a few db here and there that could alter a great mix?  I don't know the plug - but since you're a fan it definitely has my interest now.  


I just wonder if you copied a bunch of clips with needed de-essing and linearly scaled their level into the plug what would happen.


Best


-D