So, from a technical viewpoint.....

What's the difference between freezing & bouncing.

Apart for the obvious things, like bouncing doesn't remove the synth from memory, bouncing creates a brand new track etc.

But what are the mechanics of it? What exactly is going on when a synth is being frozen?

The reason I'm curious is that I've got a problem getting projects with BFD2 in them from rendering properly when freezing.
It's only happening on the cymbals and what's happening is the odd hit here and there has a "chunk" missing from the waveform.

It's not confined to a single project.

I've tried all sorts of things to get it to work, and so far, the only solution is to do a real time bounce with BFD in 16 bit mode with all samples loaded into RAM

Freezing (16 bit mode with all samples loaded into RAM) still produces the waveform glitches (and yes, they are totally audible!)

The time taken to render 15 tracks into audio is about the same for a freeze as it does for a real time bounce, so I don't think that freeze is using any sort of "fast freeze", it seems to be more or less real time, same as the bounce.

So what's the difference in the 2 sets of results?

Steve - thanks for the offer mate. If I can't work out the root cause I may well take you up on your offer.

I do now have a workaround although it's forced me into a completely different workflow, but I just want this damn song finished! This is the last thing I need to do before I can post it up.

That's a good idea of yours to strip everything else out and try a few different permutations & settings - there are SO many possibilities & parameters you have to be really methodical to eliminate them one at a time. 

I understand the confusion...in theory the two are exactly the same. One renders to another track, the other renders in place and turns the synth off after rendering. And Exporting does essentially the same.
 
What happens if you Export just the drum tracks? The nerd in a person wants to know but the musician gets tired of messing around forever. I must admit I am curious.  Likely the same a bounce.
 
Sounds like either you have a ghost in there somewhere or, barring some strange buried/linked automation snip somewhere, you've uncovered some micro-bug. The "strange buried/linked automation snip" idea seems unlikey as it would probably affect both methods. The nightmare here is it could be difficult (and time consuming) to isolate. Could be Sonar, could be the plugin. Having used specialized rendering sofware like Audio Compositor for years, which long ago rendered high-resolution audio from MIDI data before most DAWs supported audio at all and Gigasampler was still in the making, there is zero reason that a rendering engine should be subject to this type of error beyond a glitch/bug somewhere. Just thinking out loud, not criticizing Sonar or BFD2.
 
It should not require loading into RAM. It shouldn't matter if it is a quick or a realtime bounce. In theory  it shouldn't (but may?) be affected by audio buffering as it should never be buffered for audio streaming unless that is an unfortunate limitation that Sonar has by perhaps using the same engine to play as it does to render, something only a programmer/representative could truly answer as you and I can't see the rendering architecture so to speak.
 
Having said all that, have you tried raising the latency a tad to see if that helps?
 
Hmmm, intriguing and annoying. Kinda like women.  (ducks)

Hi Scott.

To answer your first point, the first time I noticed anything was awry was after I had Exported a complete mix - including the drums. I noticed then that a couple of cymbals were being "choked" or truncated.

The guys over at the Fxpansion forum are about 90% convinced it's a streaming problem from BFD.

Tonight I'm going to try my realtime bounce again, but this time without having the samples loaded into RAM - see if that makes any difference to the render. Then I'll play around with BFD's 16 bit / 24 bit options, again, see if it makes a difference.

As you say - this sort of analysis is extremely time consuming.

I have tried increasing my latency quite a bit but the freeze still messes up.

In addition to the close miked drum hits, you also get 3 stereo channels for OH, Room & Amb3 in BFD and every one of these stereo outputs also contains the truncated cymbal hits so something is propagating this anomaly across different channels.

Steve - going back to the other thread where we were discussing this, I'm happy to report that, unlike a freeze - which renders each track individually - a bounce, if you select multiple tracks, will work on them all simultaneously so I wasn't faced with a 3 hour bounce - in realtime it was 8 minutes - the length of the song (+ the time for the waveforms to draw afterwards)

Andy, your set up might be perfectly ok - I know for a fact that Steve (Twaddle) doesn't see this problem, although he routinely bounces rather than freezes.

If this is the only vsti I have to bounce rather than freeze I'll be [moderately] happy.

Do you mean the Offline button?

I always check that whenever I bounce or freeze BFD.

If it's a tail setting, I would expect the problem to manifest more randomly than it does - as of now, it happens on the same hits in the same part of the song every time I freeze - without exception.

Also, it doesn't explain why only part of the waveform is missing. You can see the transient and the beginnings of the fade - then there's a gap, and then the rest of the fade.

Thanks for your thoughts though - keep 'em coming people!

Yeah, regarding "Tails." I think this only refers to "effects tails" as opposed to the premature ending of a sample tail. At least that's how I understand it anyway.

And, again, as far as I can remember, if you are mixing down a track or complete project, it is recommended that you add an extra measure at the end of the song so that any effects tails are fully rendered (I do that by adding a single MIDI note on a  track somewhere). Slightly off topic there, but I thought I'd add that bit in...

Well it's almost as though this particular cymbal is part of a choke group - but I checked all the choke groups last night and it's definitely the only one on Group 7, which eliminates that.

something is making it cut out party way through the render, and then switches it back on 1/2 a second later