Audition Upsampled On-The-Fly During Recording?

https://forums.adobe.com/people/Eric+Holtz  wrote I realize that lower sample rates (44.1kHz, 48kHz) would suffice in regards to audible frequency content, but I have read AES engineers, and many others advocate using 96kHz (or higher) for tracking and editing, since some plugins operate more efficiently at higher rates.

Many (most) of these reports are way out of date, and I would be quite intrigued to find out exactly which plugins work more efficiently whilst they're doing twice as much work... I'm not aware of a single one that qualifies. Would somebody care to enlighten me about this apparent reversal in the Laws of Physics?

The argument in the past that held rather more water, when it came to sample rates, was the one to do with the phase effects of anti-alias filters rippling back into the audio pass-band. But with all modern converters using massive levels of over-sampling, this is entirely moot; there's no discernible difference in audio performance any more. If you sample at 96k, then at least half of all of your storage is going to consist of nothing but noise. And, you're making Audition and your computer do twice as much work too. Audition doesn't mind, but your HDs will be thrashing at twice the rate... and for what?

"...the signal cutoff appears to be much lower anyway - which is what you might expect if the OS had been playing with it." That is exactly what I'm getting at. The signal cutoff doesn't match what it should be with a 96kHz setting. I once watched the OS change the sample rate to 48kHz on the interface, right after I had set it at 88.2kHz to match a session. I wasn't paying attention to the sample rate for this project; I just set the interface and the sessions to 96kHz, and off I went. Recap: While Audition and the interface were both set to 96kHz, the operating system (Windows) changed the interface sample rate to 48kHz, then up-sampled 8-tracks on-the-fly to 96kHz for the session. File content does not match what should be there for 96kHz files. Does that sound accurate?

Several things; Firstly, your Tascam interface only has a flat response to 20kHz, and I reckon that if it extended to 25kHz in total before cutting off almost completely, it would have done quite well. So what I can see as the signal (not noise) response cutoff is what you're likely to get anyway. Secondly, and more importantly, there's no way that the OS would re-sample the output if you were using the ASIO driver for it. ASIO bypasses most of the OS, and re-sampling simply can't happen; the output from the Tascam is fed directly to Audition, which simply records what it's been sent. Yes, you have to set the record sample rate in Audition, but once you've done that, then that's all that will happen - it will record directly at that rate.

So I'd say that the way forward with this is to use the ASIO driver and sample at the rate you want the final product to be in. Even though Audition's sample rate conversion is independently rated as the best there is, you still don't want to invoke the need for it if you can avoid it.