No Need to Dither?

Hey All

I just had my first chance to master a track in WL 8.5. Some bad news, and some strange news, and some really good news.

First the bad. I tried to use the different format preview feature that I upgraded for, and I kept getting an error that it couldn’t handle the sample rate. Ok, just wish I would have known that limitation before hand.

The strange news is that I saved a 64/96 wave file to 192 kbps MP3 file with no dithering insert selected, and the mastered file sounded awesome. All these years I’ve been using the Apogee dithering and I didn’t need it?

The good news is the conversion from this HD wave file to MP3 sounds awesome. I never had such an easy time encoding a file in my life.

So, Q: What are the sample rate restrictions on the new encoding preview feature?

Thanks for reading


I wondered why you can’t audition from a 96k file too. I’m pretty sure you can with the Apple AU Round Trip audition, although it’s been a while since trying 96… I would assume it’s doing SRC to 44.1 in the background if it does.

Wavelab does nothing in the background, but if 64/96 is not a typo for 24/96 then that’s the problem, I’d guess… Otherwise, there may be limitations in certain codecs to what they accept as input?

The result of dithering usually can only be heard in very soft material, like a slow fade or ambience of the last chord in a big concert hall. So if you’re producing high power dance music or metal you may totally miss it.

What is the consensus for rendering 24-bit files to 24-bit, but when fades are added to the start and end of some clips, as well as other various plugins on clips and the master fader?

Dither set to 24-bit or no dither? I remember there were times when I was seeing higher than 24-bits on the bit meter, maybe due to some internal plugin processing so I started to use dither no matter what on 24-bit renders from 24-bit files. I want to revisit this issue.

Nope–no typo. It’s actually a 64-bit floating point, but my sound card can only go up to 24-bit, so it’s kind of a façade.

If I’m going to reduce my resolution just so I can hear how something sounds in compressed codecs, then why don’t I just encode straight to MP3 in the same amount of time and be done?

I missed that day in rock star school. :open_mouth: Very good to know though. I tried the new Izotope dither plug in, and was not pleased with the results–BTW. The genre is Punk Pop and heavy on the guitars–so, I have no problem with leaving one more process out of the…well, process.


Actually now trying the Encoder Checker again, I sometimes am able to play a 96k wav file without problem. But other times (usually after switching out a slot to a different lossy encoder), and entering Play, I get the error messages. I’ve gotten two different messages:

  1. “Error reported by plug-in Encoder Checker: Cannot handle the required sampling frequency”

  2. “Error reported by plug-in Encoder Checker: The encoding settings modify the sample rate and/or the number of channels. This configuration is not supported by the Encoder Checker.”

It would be nice if this worked without problem, but it seems that when making any of the lossy formats directly from a 96k file (not in the Encoder Checker), that Wavelab must be running Crystal in the background to get it to 44.1. I seem to remember PG saying that it uses the “High” setting when it does this. (PG, please correct me if I’m wrong). Seems to me it would be preferable to specifically place the Crystal Resampler in the Master Section and use the “Ultra” setting, whether using a 96k file in the Encoder Checker or simply rendering from 96k to lossy.

But the error messages above do seem like they shouldn’t be happening, because sometimes the 96k plays fine.

Strictly speaking, in this case you should dither using 24-bit TPDF flat dither (no noise shaping).

I never dither to 24 bits; the 24th bit is down where noone can hear it anyway, though strictly speaking you’ll be truncating the signal for every little fade that makes internal processing at 32 bit float necessary.

Well, I don’t know how you get your 64 bit fp files, but if it’s from recording I suggest you save yourself half or more of the amount of data storage and revert to 24 bit (or 32 bit fp for ease of mind). 64 bits make no sense at all, even 32 bits is way overkill for the dynamic range it can express. 24 bits stands for a dynamic range of 144 dB - try to achieve that in the real world!

Thanks. So noise shaping is only needed when changing to a lower bit-rate such as 24-bit to 16-bit?

In your example of 24-bit to 24-bit files, you are actually dithering from 32-bit float to 24-bit integer, because WL internally calculates with 32-bit float (like all current DAWs). At least with the tiniest of changes taking place, like the fades you mention. Noise shaping is simply a way of pushing the added noise (which dither is) out of the sensitive ranges.

Not really, you are changing to a lower bit rate as Arjan says: you’re going from 32-bit float to 24-bit, so you should be dithering… but noise shaping is a choice.

I was assuming that you were rendering as an interim process, and that is why I suggested dither without noise shaping. TPDF flat dither without noise shaping is usually recommended if the file is destined to be further edited / processed at a later stage.

If your resulting 24-bit files are final ‘master’ files of some kind (no more editing or processing) then you could use dither with noise shaping.

Thanks for the clarification.

They say to always use only flat TPDF if files are subject to further processing, but wouldn’t that include upsampling DACs, iTunes Soundcheck, Apple MFiT SRC, MP3 encoding. Don’t they all involve additional processing? Wouldn’t that imply that only flat TPDF should ever be used, and nothing else, because all final files are subject to final processing at the DAC itself.

I’m not sure that it does imply this but it opens up another can of worms. Yes, they all involve additional processing but, as you know, with dither we are only concerned with cases where there is bit reduction. So, for example, you are not systematically applying dither when you convert to MP3, since it involves another kind of conversion process. Experts suggest that MP3 files are best produced from high quality 32-bit float or 24-bit masters without any dither (before any bit reduction has taken place on the material). Same goes for Apple iTunes / MFiT. It’s certainly not the best option to create MP3 files from 16-bit 44.1 masters which have been processed with noise shaped dither (for example).

But if you are obliged to use bit reduction on your higher resolution master files from which you intend to create MP3 files (and similar file formats) then yes you would be better off using TPDF flat dither when you create them, and then create the MP3 files from this.

This would explain why when I went from a high definition wave straight to MP3 without dithering it sounded awesome.

My outlet is always online so I really never have to dither unless I run up against file size limitations.

Thanks [emoji2]

Thank you stingray. Lots of good information. I don’t think this has been a real problem for anyone, because there are probably thousands of major label MP3 albums that were (and are still) made from 44.1 16 bit CD masters, where noise shaping had been applied in the making of the 16 bit master. Many of those MP3 albums have thousands of clip levels because they’re still made at CD level, and even that hasn’t been a real problem. That’s just the way it’s been for 10 years.
But I’m still wondering if all end user processing (including the DAC SRC, which will affect files, CDs, everything) is considered problematic post-dither processing if the DAC is upsampling from a dithered master that was made with colored dither or noise shaping. The consequence I read of not using flat TPDF is this:
“If a colored dither is used instead at these intermediate processing stages, then frequency content may “bleed” into other frequency ranges that are more noticeable, which could become distractingly audible.”
They don’t specifically mention noise shaping in that sentence, only colored dither. It probably hasn’t been a real-world problem for anyone, but if flat TPDF is a safer option all around, for any bit-reduce stage (even the one you think is the final one), and an engineer doesn’t have a preference, I think it would be helpful to know.

I’ve thought the same thing when I’ve encoded directly from 32 bit float to MP3, but I never got around to blind tests. Maybe there’s something there, even vs. 24 bit dither, I’m not sure, but I seemed to think so at the time. If you were previously dithering to 16 bit (I don’t think you said if you were previously dithering to 24 or 16) that would probably have made a bigger difference.

True, but doesn’t make it right though.

True enough. Some day the standard procedures will change and all MP3 and AAC will be made from 24 bit, level-prepared masters. Just hasn’t happened yet.
I think one interesting thing Apple did with MFiT (for the first time ever I think), was to introduce the concept of single-sample clip levels as a problem. Previous to that most “red-light overs” in meters and analyzers had thresholds of 4,3, or 2 consecutive samples I believe. And then Apple added in even inter-sample peaks and their red light became even more strict. I guess this has encouraged some mastering engineers to bring even CD levels down, but there are probably still many new CDs out there with 1-sample or 2-sample clips and ISP.