I didn't say they weren't different. I did say that the difference was audible, but you're taking my statement out of context. I said the difference was audible if I cranked the null signal, to the point where the actual signal would have likely been around 70db to maybe even 90db in the room. That part is kind of important, y'know. The actual question is whether or not there is a difference we should care about. In this particular context, the answer is 'probably not'. Not when the difference is more than -70dBFS. Whether or not something that is true in specific is true in general in this case is beyond me, I have no clue. Maybe there are actually cases where fixed point produces and audible difference at a normal listening level. I'm the last person to know the answer to that question.AC222 wrote:Posted by the same dude at the same db level. Very easy to check. What are you are hearing is the difference between 48-bit fixed point and the lack of precision and accumulated rounding errors of floating point.
And let's be real here. Protools isn't the only product where this phenomenon exists. It's the same deal with the TC Electronic stuff versus the native versions.
You say they null and yet acknowledge they are audibly different. I won't even elaborate on that. FLAME SUIT ON!!!! lol.
I don't think it's a question that the PTHD sounds clearly better. Even when I bump up the volume on the RTAS version, the PTHD STILL sounds better. When it comes to hearing better clarity and definition there is only so much you can do to fool the ears with bumping up the volume.
Long live this thread
But, what I do understand is that data is data and measurements are measurements. Based on what I've seen here, I see no compelling reason to believe that a difference in the -70dBFS range or more will yield an audible difference at 50dB or 60dB in the room. Based on the data, you'd only hypothetically be able to hear the differences if you were monitoring at 70dB or greater. That would be like having a leaf-blower on your desk instead of speakers.
My point is that even if the discrepancy is caused by floating-point rounding errors, it's not enough of a difference to compel a reasonable person toward actually investing in a more expensive tool. If someone comes up to you and says "I have this $1500 tool that allows for greater precision that you won't actually notice, and for that reason you should buy it over this $250 tool", you'd probably decline. You're asking me to draw an aesthetic distinction between a diamond ring and an imitation diamond ring when the difference could only be observed with a microscope. The argument just isn't compelling, and I very much doubt that you can actually discern a difference.
A null test is a null test is a null test.
it's amazing how many people miss those. a clean experiment is a prerequisite, and even the best of us sometimes mess up simple things like volume matching.
