Page 1 of 1

Pyramix not transparent

Posted: Fri Feb 02, 2018 01:00
by Lars Nilsson
When I put a single stereo file on timeline, set the fader to -0.1 or 0.1 dB and export it using Mix Down, the result sounds different than the source file.

The difference is subtle, but audible - I'm able to pick the Pyramix file in proper ABX double blind test with 100% confidence. The difference is most definitely more than 0.1 dB of gain.

There are no plugins on my inserts, no dithering, no other proccessing, panner is set to None and source and output formats match. I have left most of the settings on default. Both of the files have plenty of headroom and no clipping.

Anyone expirienced this? Am I doing something wrong? I'm complete newbie to the Pyramix.

Re: Pyramix not transparent

Posted: Sun Feb 04, 2018 19:39
by fl
Firstly, which version of Pyramix are you running these tests with? My comments below will refer to version 11.0.5.

Lars Nilsson wrote:When I put a single stereo file on timeline, set the fader to -0.1 or 0.1 dB and export it using Mix Down, the result sounds different than the source file.

What happens if you apply 0.0 db? Perceive any difference? Invert the polarity of this mixdown file and then mix it with your original file. Does it null properly?

The difference is subtle, but audible - I'm able to pick the Pyramix file in proper ABX double blind test with 100% confidence. The difference is most definitely more than 0.1 dB of gain.

Going back to your 0.1 dB gain change experiments, what wordlength did you select for the output file? 16 bit? 24 bit? 32 bit? Try 32 bit and see if you still hear a difference in sound quality. I'll assume that you chose to output to a WAV or an MTFF file format (you certainly don't want to use one of the data compressed formats for this experiment).

Also, what happens if you change the panning scheme used to send the source file's mixer strip output to the Mix Bus, from a Panner to a Routing to the Mix Bus? You do this by clicking on the Red "Pan" at the very top of the Strip - click on it and change from Pan to the Routing Options, where you'll set the L side of the source signal to send 100% to the Mix Bus's L input, and the R, 100% to the MB's R. (See the manual page 227.) I'm curious whether the use of the panpot vs. direct routing makes any difference to what you're hearing.

What happens if you do a Mix with the Strip gain set at -0.1 dB gain, and then do a second mix, with the gain at -0.2 dB - and then compare those two through your ABX test?

There are no plugins on my inserts, no dithering, no other proccessing, panner is set to None and source and output formats match. I have left most of the settings on default. Both of the files have plenty of headroom and no clipping.
Anyone expirienced this? Am I doing something wrong? I'm complete newbie to the Pyramix.

I believe you'd be hard pressed to find any DAW which does not impart some change to a file when you run it through its mixer, especially when you've applied some gain change. Even a minuscule gain change will result in a completely re-written file, since the change of gain requires the whole file to be re-calculated. Pyramix uses 32 bit floating point math for its mixdown, as do a lot of DAWs these days. However, some DAWs are going to be kinder to the mixdown results than others.

To be fair, mixing a two track file down through a two track Mix Bus isn't something that a lot of people are going to be doing all that often, unless they want to change it in some way (gain, effects, etc.). I'm reminded of a quotation from famed mastering engineer Bob Ohlsson which I thought was so incisive that it's printed out and hanging on my wall:

"Signal Processing is always a trade-off between sonic degradation and a perceived improvement in sound."

Re: Pyramix not transparent

Posted: Mon Feb 05, 2018 22:25
by DJS
I believe you'd be hard pressed to find any DAW which does not impart some change to a file when you run it through its mixer, especially when you've applied some gain change. Even a minuscule gain change will result in a completely re-written file, since the change of gain requires the whole file to be re-calculated.

This should definitely be transparent and inaudible. Gain is a benign change. Something must be broken if audible.

Re: Pyramix not transparent

Posted: Tue Feb 06, 2018 01:36
by Lars Nilsson
Many thanks for both of the replies!

Actually, I did a null test when doing this experiment. Pyramix nulled just to something around -100 dB. When I boosted the remaining signal, I was able to hear the source material.

I have tried this exact experiment with some other DAWs. Every time the result nulled to the source file to over -150dB, and the remaining signal was just noise, no traces of the source material when boosted.

When the fader is set to 0, the resulting file is bit-to-bit copy of original file, but -0.1 dB file DOESN'T null to the -0.2 file or +0.1 file when volume matched. Switching from a Panner to a Routing to the Mix Bus yelded just slightly better results (files nulled to 120dB, source material is still audible when boosted), but I have learned a lot about Pyramix routing while doing this. :-) I think the routing in Pyramix is brilliant.

I'm using 32bit lossless format when exporting any files and I'm using Pyramix 10.

Re: Pyramix not transparent

Posted: Sun Feb 11, 2018 21:32
by Lars Nilsson
Would someone of you proccess my audio clip in your Pyramix, to see if we get the same results, please?

I'm sure I must do something wrong.

Re: Pyramix not transparent

Posted: Mon Feb 12, 2018 17:44
by fl
Lars Nilsson wrote:Actually, I did a null test when doing this experiment. Pyramix nulled just to something around -100 dB. When I boosted the remaining signal, I was able to hear the source material.

How are you measuring this? How do you arrive at the number -100 dB?

What is/are the audio file format(s) involved? Same file format for the Mixdown as was used for the source material? Wordlength? Sample Rate?

Are you using Pyramix's Mixdown window, or are you using Album Publishing, to generate the output files?

If using the Mixdown window, are any of the Processing Options (on the right side of the window) enabled? What about Settings for the output file format (Settings button right beside the file format selector)? Do you own a FinalCheck license? If so, is it enabled?

Is there any chance that Dither is enabled in the mixer?

When the fader is set to 0, the resulting file is bit-to-bit copy of original file, but -0.1 dB file DOESN'T null to the -0.2 file or +0.1 file when volume matched.

And how are you "matching" the levels? What method are you using to invert the polarity of one of the waveforms prior to running a Null Test?

I'm using 32bit lossless format when exporting any files and I'm using Pyramix 10.

Do you mean to say that you're using 32 bit FLAC files for making the mixdowns and subsequent comparisons?

Why not use WAV? Or MTFF?

Is the original source file 32 bit Floating point? Are the mixdowns?

You say you're using Pyramix 10 - which version?

Which version of Windows?

Is this Pyramix Native or Mass Core? If Native, what is your audio interface and connection protocol (ASIO?)?

Do you have Real Time Sample Rate Conversion enabled in the Pyramix Settings?

Giving a complete rundown of your soft and hardware set-up would be helpful.

Re: Pyramix not transparent

Posted: Sat Feb 24, 2018 21:38
by ljudatervinning
Check Dithering in the mixer.
I believe dither to 16 bit is on by default on Mixbus.
Right click in mixer -> Settings -> Dithering

Re: Pyramix not transparent

Posted: Wed Mar 07, 2018 13:10
by per
Interesting test!

My bet is that the output format you are producing has a higher resolution than the original format. If the original is a 16 bit wav file and you produce a 32 bit wav file (or mtff), then the output has a higher resolution and you will hear a difference when you try to null them out. The fact that you actually hear the material undistorted when you boost the difference speaks in favour of this explanation.

Another explanation could be that dither is in fact on, but then the difference would be noise.

I'm also a bit fascinated that you can actually hear an "error" that is 100 dB down from 0 dBFS. I'm hugely impressed! :)

Per