[Dubai]

Originally Posted by fader8

Right. Here's a little demonstration.

The 2 aux channels shown are receiving the same signal which is pink noise mixed with a needle pulse to add some nasty harmonics to the signal. One aux is reduced 3.1dB, (over half the power).

The aux's are assigned to separate hardware outputs which are fed into 2 channels of the analyzer app. The transfer function plot clearly shows the difference between the signal levels as do the mixers textual meters. One signal is -90 and the other is -93dBfs.

Now, flipping the polarity on one of the aux's and assigning both to a mono output, the project gets bounced realtime in Logic Pro 8, (or Pro Tools, same difference) to a mono file.

The resulting bounced file is shown at the right at maximum zoom in Logic's sample editor. Opening in Peak Pro at max zoom also shows the same and looks like a complete null.

You can go further with this and raise the signal levels until you start to see something in the bounced file, but the point is that host mixing engines were not designed for low level, precise instrumentation work. Nulling in any host's mix engine just doesn't cut it. You need to get up around -60 before you start seeing something, and up to -45 before seeing anything at all useful.

I'm not saying there's anything wrong with these editor apps, just that they were not designed to display that kind of low level data. Always use an "external" precision analysis application to perform this type of test.

Also, as Zaphod alluded to, you don't want to be measuring, (or being a victim of) the quantization of the host app when you really just want to see the difference between 2 processed signals.

+1.