Choose a dither, or truncate in 20-24 bits?

[Falcon]

I am currently trying to compare the settings of MBit+ on Izotope RX.
My original track is a DTS Cinema Track 16 bits / 44100 Hz, which I slowed down to 44056 Hz, and resampled to 48000.
This generated a 32 bits float track that I need to reduce.

When I zoom in on the audio spectrum, I notice that there is no difference when I truncate to 24 or 20 bits (visually and listening).

When I truncate to 16, I see a difference in the audio spectrum (but I can't hear it).

Your Opinions, or your preferences ?
Dither to 16, or truncate to 24?

[deleted user]

Why not just resample directly to 16 bit? That's what I get when I resample using iZotope 64 bit SRC inside Soundforge I think.

But if that's the only option ... I think I'd just dither to 24 bits.

[Falcon]

When resample with izotope RX 7, the track is calculated to 32 bits float.


What dither setting are you using?

[deleted user]

Hah. Well, I'm using none because when I resample a 16 bit file in Soundforge using izotope 64 bit SRC, my resulting file is still 16 bit. But it's certainly an interesting question what's happening "behind closed doors" lol.

But whenever I end up with a 32 bit float file I usually dither to 24 bit. Problem with 16 bit dithering is that it can be audible and that really irks me!

[Falcon]

Have you been trying to compare multiple dither / settings ?
I can hear 16 bit dithering too, depending on the setting, with a very low setting of mbit +, I can't hear it (but I can see the difference).

[deleted user]

I haven't done extensive tests but at the higher settings it's definitely audible like you said. The noise shaping is also audible at high levels imo, you can tell the high frequencies are getting more noise.

At low settings I think it might be inaudible, but I haven't done any deeper tests I think.

Yeah that's probably the next best thing if you're at 16 bit target depth.

[pipefan413]

I've been wondering this myself lately, I don't trust my headphones so I'm not trusting what my ears are getting through them either.

Would you say it's better to use *some* noise shaping if dithering from a float back to 16-bit again (after resampling to 48 kHz from 44.1) or leave it at bog standard TPDF?

I tried both with some different settings and reckoned with the noise shaping (MBIT+) I could occasionally hear a difference in higher registers but when I left it with TPDF it sounded pretty much transparent against the original file as far as I could tell. Is it realistically likely that I just have crap headphones and/or ears so can't hear the difference?

[deleted user]

I have a suspicion that a lot of the people who like noise shaping are older people who don't hear high frequencies. Similar to how a lot of old audio recordings have a 16 kHz monitor signal in it. Personally I don't think noise shaping makes sense unless maybe if it goes into the ultrasonic range. But I'm no expert and have not done enough comparisons, it's just from 1 or 2 tries I did.

[pipefan413]

I have a suspicion that a lot of the people who like noise shaping are older people who don't hear high frequencies. Similar to how a lot of old audio recordings have a 16 kHz monitor signal in it. Personally I don't think noise shaping makes sense unless maybe if it goes into the ultrasonic range. But I'm no expert and have not done enough comparisons, it's just from 1 or 2 tries I did.

That's interesting because I seem to be oversensitive to high frequencies to the point of it being really irritating. There's a clicky thing (I... I know, but I don't know how else to describe it) that I think might be intended to keep pigeons from flocking over the doorway at one of the big train stations in a city near me that I find bloody agnoising to walk past so I have to go through a different door. Invariably, if I forget about it and end up walking underneath because somebody I'm with is walking that way, they're baffled by my reaction and have no idea what sound I'm referring to.

I guess I should probably steer clear of the noise shaping, then.

Do you reckon it's safe enough to take a 44.1 kHz track, resample in RX 7, then export with straightforward TPDF dither back to 16-bit like it was on the LaserDisc in the first place? Or is that likely to cause audible issues?

[deleted user]

Hah, that's interesting. My ears have actually gone a little worse over the years, but I can still easily hear the 16 kHz sound. I wonder if I would hear that sound.

Only one way to find out I'd say - test it. I think the thing is that Laserdiscs (and probably also CDs) already use dithering of their own, so dithering again probably increases the noise floor even further. But I'd just do a few experiments to see if it's audible at all. Crank up the volume in some quiet part where there is nothing but background noise and compare.