Ah well said and words to live by! Many thanks. Onwards to dithering....

Yes, because the tape IS the CD and anybody hearing the tape is literally hearing the CD on tape.

'covering up stuff with the higher noise floor' is absolutely a myth. Especially with analog, it's more like a noise FOG than a 'floor'. When you get to stuff like vinyl records, that's largely rumble and extraneous noises and it's absolutely trivial to hear way into the noise floor because it's easiest to hear around 1.5k and the noise is thundering away around 30hz, plainly measurable but not in the way at all.

I think @ David Rick answered this.

and if you want to try some great DITHERS ...

check out AIRWindows [Chrisj]

His DARK Dither is one of my favorites.

Aha! Many many thanks. Very obviously I am new to the land of dither. One nagging worry that I had (one that I now know to be baseless) was that the noise of the cassette might combine with the dither in an unattractive way.

Excitingly, I get play with dithers shortly

Thank you! I have been neck deep in Chris's website and surrounding threads for the last 24hrs : ) I am super keen to play Dark and NJAD and Beam and so on. Brave new world for me...I shall report back of course

yeah ... I'm torn between NJAD and DARK. [NJAD is in the STUDIO version].

Chrisj got us all started into revisiting Dither-land.

So glad he did ... because I always heard that 'dither difference' ... and didn't like it ... but knew why it was a necessary evil.

Too bad 15 or 30ips 2-track is such a PITA ... but the Studio can spoil you.

I think that is why I'm drawn to DARK Dither. I like music emanating from the depths of total, silent, Nothingness.

Granted ... there is much more to that than just Dithering.

Firstly, many thanks to all for your excellent advice, and extra special thanks to chrisj for his careful and very patient help with installs!

I have one (hopefully final!) question. It is general enough that I reckon it might be of interest to others navigating this for the first time too...at least I hope it is.

I know that one should not dither twice on the same projects. For example, if I have Dark running last in my chain, then I want to turn Dither Options to "No Dither" in Ableton's export settings.

However, the tracks on which I am about to work have been exported with dithering and then mastered professionally. They are 24/48. We are getting cassettes made, and the duplication place records the tapes from a CD that they burn onsite. So, either I send them 16/41 for CD, or they will convert the files to 16/41 themselves at the duplication centre. I would rather do the conversion myself of course.

So, I am pretty sure that what I need to do now is the following:

1) Import our aforementioned mastered and dithered 24/48 tracks into Ableton.
2) Export at 24/41 with Dither Options set to "No Dither" in Ableton's export settings.
3) Import the new 24/41 export into a new Ableton project.
4) Set Dark to 16bit CD (or similar 16bit setting on other plugins) on this new project. There will be no other plugins on it anywhere.
5) Export from Ableton with both Bit Depth set to 16bit and Dither Options set to "No Dither" in Ableton's export settings.
6) Back away slowly and carefully.

Okay, so how many mistakes are in the above? I am nearly there people!

Many thanks again,

S

The sky does not fall if you dither twice. That is an urban myth, promulgated by poorly-informed refugees from analog tape days who thought nothing at all about bouncing a drum mix when they started running out of tracks -- that being another thing that you should theoretically "never do" because it raises the noise floor. Multiple dither passes do exactly the same thing! Bob Olhsson (that same legendary engineer who has occasionally appeared in this thread) said something years ago that I'm sorry I can only repeat in paraphrase: All music production involves trading objective degradation of audio quality for subjective artistic enhancement. Or to paraphrase the paraphrase: "Don't let 'best practice' dissuade you from doing what's best for the music."

It is true that you shouldn't dither twice in the same export step. The dither block should be the last thing in line before the bottom bits get chopped off. But if you forget that, you've just increased the noise floor a tiny bit. If someone farther down the distribution chain level-adjusts the audio and (we hope!) dithers again, that's ok because the alternatives are worse. Remember, dithering one-too-few times is much worse than dithering one-too-many times because it adds non-harmonic distortion that you can't ever get rid of. An extra instance of dithering just raises the noise floor a tiny bit, something we've been dealing with ever since we started duplicating saleable product from master copies. Rectangular dither has a 3 dB noise penalty over the "noise" level of undithered truncation. (It's 4.5 dB for triangular dither, but I'll assume rectangular here to keep the math simple.) To make it 3 dB worse again, you'd have to dither two more times. Then another four times to get lose the next 3 dB. There are much bigger problems to worry about.

The one that you should worry about is the sample-rate conversion from 48k to 44.1k. It's really hard to do this well, because 44.1k inherently sucks. (Why is the topic of a different thread.) But we're usually stuck with this fact, so we just have to choose the best SRC tools we can get access to. It's a hard job, because there is a lot of pretty terrible SRC software out there, and what's in Ableton is about as bad as it gets. (It was optimized for speed and low CPU load, not for audio quality.) At the very least, please download the free version of Voxengo r8Brain and do the SRC in that (at 24 bit resolution)*. Then dither to 16 bits and go on with your life.

* It's possible that @ chrisj offers a free SRC tool as well, but I haven't looked.

David L. Rick
Seventh String Recording

Thanks so much David. Yet again I have just learned a lot!

r8Brain looks like it is Windows only, and I am running a mac. I took one look at SoX and nearly died. For those of us not familiar with command line interfaces might there be something more cuddly around? LaTeX is one thing, but a good SRC with a GUI would make my day. I don't know how to do anything in Terminal alas : /

EDIT - dBpoweramp???

Indebted still,

S.

According to the Ableton manual as well as comparison measurements, Ableton (since 9.1) on export uses one of the highest quality sample rate conversions available. It's true this is non-neutral, as all sample rate conversion is non-neutral. But r8Brain, according to measurements, shows no apparent audible advantage over Ableton.

Looking at test graphs here: https://src.infinitewave.ca (I knew this site already but was interested to find it's recommended by Voxengo to compare r8brain to other options.)

Since you're using the export SRC in Ableton the quality should not suffer compared to r8brain - and should be high enough quality in general - unless using Live pre-9.1.

Well stay away from Vegas 17 I guess!

Seriously thanks so much. This is a super fun site in general : ) Unless I am missing something, I think I am less worried about Ableton 10x in particular than I might have been. Gosh you are right about Abletons <9.1 though, eeep.

The trick for me of course is that this learning curve is like an orbital launch. Exhilarating indeed!

Nope, I just use 'Brick'

Glad to know that Ableton have "cleaned up their act" in recent releases.

I've been using Goodhertz for the last few months with good results.

I have another question if any one can advise. If I'm using Goodhertz on my Stereo Outs when mastering a track using Ozone 9, when I go to bounce to 320 Bit Float MP3 file using Logic own algo (I'll come back to that later).. I'm assuming I should be bypassing Goodhertz at this point right? My understanding was you don't need to dither if going from Wav to MP3?

Also is there a better option as a Logic user than using Logic's MP3 option?

You're correct. Don't dither to mp3s. In fact, I make mp3s from the 32bit float and it is almost indistinguishable from the master. Definitely sounds more accurate than 16 bit wav.

Because dithering is added in the MP3 encoding itself? Or because as Robert Katz suggest "…mp3 (as well as AAC) is a special case because the LAME and Fraunhofer encoders work in 32 bit float in their encode side, and accept up to 32 bit words without penalty and without need for early reduction."?

If so, I actually didn't know this.

Yes, these mp3 encoders are able to convert directly from a 32-bit float file without the need to dither. However most decoders truncate the decoded file to 16 bits, so this is where encoding of an mp3 from a 16-bit dithered source can sound better than encoding from a 32-bit float source. Some decoders, though, are able to decode directly to a 32-bit float format.

Yes, for the same reason Mr. Katz mentioned. I was as surprised as anyone when the mp3s sounded better than the 16 bit wav in my listening tests.

Nice one. Thanks for confirming.

If this is true then it is always better to dither to 16bit first. Hoping that every decoder through which your music will be played is one of the few that can properly play a 32bit encoded mp3 isn't a good approach if quality of sound is your goal. It may sound better to you because you happen to have a decoder that can do it properly, but chances are it will sound worse to your listeners.

I can't help but wonder why the MP3 encoding doesn't include a dithering stage though. Are there encoders that do despite if the source is 32-bit float? I tried only the built-in macOS AAC encoding, and the resulting file is 16 bits.

Mp3 encoding does not need dithering, because the internal representation of mp3 is floating-point. What needs dither is mp3 decoding, when it outputs to 16 bits.

Interesting. So what has an encoder that takes a 32-bit float source and presents this after encoding as 16-bit, done to the bit depth? Reduced the 32-bit float to 16-bit float?

Yes, lossy encoders can be viewed (in a simplified way) as conversions to a 16-bit float format. Their noise level is corresponding to roughly 6 bits, but the noise is relative to the signal level, and they are capable of representing signals with very low levels (-100 dBFS and less). Then the decoder converts this 16-bit float signal back in either 16-bit (by truncation) or 32-bit float.

Hi
I'm late to the party..

But as I'm using a hybrid set up with Pro Tools (12HD no hardware), and the SSL Sigma - I started recently putting (triangular) dither on all of the outputs to the Sigma from Pro Tools. I use 16 track outputs.

.. but I hear no difference at all, I don't think - I guess I am not hearing the truncation distortion; Are there any suggestions of ways to train my ears for this?

P.

Try flat TPDF dither. Also if your monitor outs aren't dithered you might not hear any difference. You should notice improved clarity and stereo imaging. Possibly higher highs and lower lows as well.

You should dither even when you don't hear a difference because the distortion accumulates. At some point, it becomes audible and the sound becomes crunchy. Again, the dither prevents distortion. It doesn't just cover it up. It ought to not even be optional.