Noisy Fades

[JingleDjango]

Can anyone explain what I've been seeing since 2023.12?

When I throw my edited clips to RX via Connect I'm seeing noise over the boundaries of the fade from about 8kHz up. If I consolidate the edit before sending to RX the noise isn't present. See below.

Is this dither? If so I'd expect to see the noise more uniformly throughout the duration of the fade, I think. Has this always been a thing and I just never noticed?

macOS 12.5.1
PT 2024.3
sending to RX 10 Advanced via RX Connect. (*create continuous file* and *entire selection*)

[JingleDjango]

Also occurs with grouped clips as shown here:

A quirk:
The noise doesn't occur over absolutely every fade. In the screenshot below you can see that where I've put a curve on the long fade over the breath in the middle of this line, no noise had been rendered to the audio being sent to RX.

[JingleDjango]

Another demonstration, here with a bit of empty timeline consolidated, then edited with fades, then sent to RX via Connect.

I don't want to make a mountain of a molehill, but does this strike anyone else as... unacceptable? Can anyone reproduce this on their end?

[Frank Kruse]

I can only see that if I push the scale of RX all the way up so that it shows down to like -160dBFs. Level measured is around -138dBFs.
I suspect dither. I wouldn't call that "noise". More like brownian molecular movement :-D Surely WAY below of what your DAC can do I assume let alone you being able to hear unless you're monitoring at 120dBSPL ;-)

[JingleDjango]

Originally Posted by Frank Kruse

I can only see that if I push the scale of RX all the way up so that it shows down to like -160dBFs. Level measured is around -138dBFs.
I suspect dither. I wouldn't call that "noise". More like brownian molecular movement :-D Surely WAY below of what your DAC can do I assume.

Yes, it's typically inaudible. Unseemly but not the end of the world.

In the test shown above with fades over digital silence the resulting noise is audible however at louder playback levels. In contexts where the resultant clip might go through further processing or gain adjustments it can become an issue.

[Frank Kruse]

Well highest theoretically possible dynamic range at 24bit is 146dB so at -138dBFs I congratulate you to your hearing abilities AND your well isolated studio environment and your DAC that enables hearing that while other stuff is playing without ripping your ears off.

Even when calibrated to theatrical Dolby standards (85dBSPL@-20dBFs) that noise would play 120dB lower than 85dBSPL so about 35dB below the threshold of the quietest sound a human can hear which is again 15dB quieter than one of the world's quietest rooms (world record of -20dBSPL @ anechoic chamber @ Microsoft which is only 3dB "louder" than the sound of molecules colliding)

https://edition.cnn.com/style/articl...oom/index.html

(unless I got those numbers wrong)

[JingleDjango]

I don't dispute your math but what I'm measuring is an RMS level of -81dB.

I'm starting with clips levelled to -23LUFS then adding 18dB of gain through dynamics processing, sometimes more. (Don't ask. It's per client request). So after bouncing the rendered clip I have -63dB RMS of noise showing up in bursts wherever I placed a fade in the timeline. Like I say, it's not the worst thing ever but it should be avoidable.

[BScout]

Avid has traditionally used TPDF dithering everywhere when dither is used.
If they have started using shaped dithering, this should be big news. And at what version?

Maybe we can get the Avid team to chime in here?
Marianna, any chance a word from the development team on this?

Answered some of my question. Seems Pro Tools always uses shaped dither now.
Except for the one case of "Enforce Avid Compatibility" is chosen on an export then "Dither without noise shaping will be applied to files being exported from 24-bit to 16-bit."
There's no specific mention on fades.

[Frank Kruse]

Originally Posted by JingleDjango

I don't dispute your math but what I'm measuring is an RMS level of -81dB.

I'm starting with clips levelled to -23LUFS then adding 18dB of gain through dynamics processing, sometimes more. (Don't ask. It's per client request). So after bouncing the rendered clip I have -63dB RMS of noise showing up in bursts wherever I placed a fade in the timeline. Like I say, it's not the worst thing ever but it should be avoidable.

Well thats odd then. I can't reproduce. Over here the noise in fades across empty clips is at -138dBFs nowhere near the -70 that you're getting. I'm also on 24.3 and RX10

[its2loud]

This is indeed making a mountain out of a molehill, IMHO. How much billable time have you actually spent on this? You’re adding 18dB of gain to a -23 clip? That’s absurd and there’s absolutely no logical reason for that. I would advise the client of this if it were me.