I’m running Ardour 5 on Windows 10 instead of Linux because I wasn’t able to get a Thunderbird interface working on Linux when I started this project. I’ve recorded an album and have mastered it and now need to mix down the stereo channels into a single, stereo interleaved file prior to generating the DDP and CUE/TOC files for the manufacturer. Is this possible to do on a Windows platform? JAMin isn’t Windows compatible, and I had conveniently forgotten this step was necessary this whole time. Thanks so much!
I think you can just do the typical 44.1/16-bit (add dither!) export and create the WAV+CUE files (tick the “create cue file” option in the export dialog). Various external apps can create the DDP from that but you won’t get much simpler than running it through the free DDPtools: http://ddp.andreasruge.de/
Just use the command
cue2ddp cuefile [ddpdirectory] where cuefile is the cue file generated by Ardour export and dppdirectory the directory you’ve created ahead of time to hold the DDP files.
Note: Remember to export the whole session versus individual tracks if you want the wav+cue option to work as intended!
You are correct!! Thanks so much for that! I had read Andreas’ manual and it stated the audio file had to be 16-bit stereo interleaved or it would return an error. But I tried a new export and it ran just fine and returned the proper checksums and file names.
Thank you!! I might still have to address the issue with the 24-bit files for digital distribution if they only accept stereo interleaved, but I’m not sure on that score yet.
You’d do just the same thing but switch the export profile to 24-bit (I still recommend adding dither) and render the individual tracks versus the session in this case (edit the time-span tab as part of export). Stereo-interleaved is really the only way to go for digital distribution. Why would you want anything else?
EDIT: There’s ample evidence to suggest that your lossy files should ideally be created from your 44.1/16 wave/cue due to not knowing how the MP3 is decoded on the other end. This aside from the fact that 44.1/16 is plenty goodness for every human ear.
Hmmm… I’m a tad bit confused. I can’t run the ddp with 24-bit files; that does indeed return an error of “wrong sample width in wave file (24)”. But if I simply perform an export directly from Ardour, it’s not a stereo interleaved file. (I’ve added dither in the final mastering plugin)
EDIT: I had originally though the same about lossy file generation, but I’ve started seeing admonitions to derive the submissions to things like Distrokid or CD Baby or similar distributor directly from the DAW exports at a higher bit-rate.
DDP is strictly for 44.1/16 and for creating a physical CD in a professional CD plant.
For exporting in Ardour, the default should be a stereo interleaved file. Check the “channels” tab to ensure that only the master bus is selected, and make sure 24-bit sample format is also checked. I’d be very surprised if what came out was not a series of stereo interleaved files.
As for the “admonitions”, I think these companies are wrong. 44.1/16 is plenty good enough and certainly you’d have a very hard time identifying MP3s created from 16 v 24-bit masters (again dither is a must). Dan Worrall’s point is that who knows how these files are decoded on the other end? Sound Forge apparently opens them up as 16-bit files and if there’s no 16-bit dither, you suddenly have potential for a very bad sound issue…
EDIT: Definitely check out the export part of the Ardour manual if you haven’t already: https://manual.ardour.org/exporting/export-dialog/
Well, I may not be understanding this correctly, although I’ve been using Ardour since 2009. It gets back to my original question - if I run a simple export from Ardour, 24-bit, only the Master bus, (defines two channels, output 1 and output 2), lossless PCM WAV file, it does indeed produce only a single .WAV file. But if I turn around import that same file into Ardour or any other DAW, it is split into the stereo components (R and L). Looking at this documentation for using Ardour and ddp, the final imported files, although they are stereo, occupy only a single stereo track, not two individual tracks. Perhaps my error is somewhere in the import process, but I am unable to produce a single stereo file that remains a single file (not imported as two files, one R, one L) when imported back into Ardour.
My own experience is that Ardour and Mixbus split into L and R on export behind the scenes (probably to do with the idea that on a real mixing desk individual channels are always mono?) but that other DAWs like Reaper, Samplitude etc do not (AFAIK). It shouldn’t make any difference to your actual mixing/mastering workflow. In any case, on import you should make sure “one track/region per file” is selected for the channel mapping, otherwise it will import to a series of mono tracks (as you are probably experiencing).
Given that 32 bit float only has a 24 bit mantissa, you cannot dither the signal. There are no bits to randomize.
Even for 16bit it’s unlikely to make any significant difference.
Dithering the final CD master (16bit) will add a few more dB S/N below -96dBFS. It’s just a safeguard, and unlikely to make any difference for most real-world recordings.
A nice explanation of dithering by Monty Montgomery from xiph.org:
I have always understood the need to dither any time bit depth is lowered. Is Dan Worrall incorrect in his dither video? From the 32-bit float master file I hear the truncation distortion in the non-dithered 24-bit version (yes, down at -108dB but it is there). I also clearly hear the issues in the reverb tail with the 16-bit file with no dither (down at -60dB). I guess it all depends on whether we are striving for the best possible sound versus knowing the music/listening environment will mask the issue? Delicate reverb tails and fades in and out will be susceptible to this distortion as noted here:
As a result, dithering has the most impact with music that has wide dynamics, doesn’t necessarily hit 0 very often (if at all), and is at a fairly loud playback level. For example, you won’t hear the impact of dithering on pop music that’s always kissing 0dB, but you may hear it on the reverb tails of acoustic symphonic recordings playing back from 16-bit CDs.
For what it is worth, Chris from Airwindows also has a 32-bit float dither built into every one of his plugins. While I don’t understand that myself, his video shows the science. I seem to be remember he’s concerned about stacking up of these truncation issues versus being able to hear one instance:
All I can say is that in my own recordings, I have a strong preference for dithered masters both at 24-bit (triangular) and 16-bit (noise-shaped or triangular).
I just read your explanation involving thermal noise here: https://tracker.ardour.org/view.php?id=7428
If brownian noise is higher in volume than any 24-bit dither added, what you say makes sense with regards masking of dither noise. However, my understanding is also that truncation can occasionally have noticeable effects on things/feelings/sensations above the thermal noise level in the form of “brittleness” or “crunch” as Bob Olhsson might say. I don’t know him from a hole in the wall but he seems highly respected in his craft. It sounds like he detects a correlation between undithered 24-bit files and more likely to sense “crunch” when EQing etc.
I guess my own personal thought is to dither when exporting at 24-bit as it does no harm and you might even say the process is technically more sound even if the equipment, room and/or ears mean it makes no real-world difference. Sticking on a 24-dither takes seconds so why not do it? This all said, 24-bit dither is a thing of the past for me personally as I’m finding 44.1/48 16-bit perfect for everything I do these days. And now that I have completely derailed the OP, I’ll leave it at that
Making that claim without backing it up via an ABX test score would get you banned on the HydrogenAudio forum
It is possible that your hearing and your gear are exceptionally good but it’s equally possible that your brain fools your ears into hearing “issues” because it knows the file isn’t dithered.
An ABX test will tell if you can actually hear the difference.
Let me be very clear. I’m not talking about being able to hear something -108dB or even -60dB down in this case. I’m referring to Dan Worrall’s dither video where he proves there is even truncation noise in a 24-bit undithered from 32-bit float file way down at -108dB. He compensates and raises the fader back up 108 dB. So, yes, easy to hear
I’m not claiming to have “golden ears” at all. I think I have pretty good hearing having to listen to pitch and keyboard temperament etc but I’m simply saying there’s a je-ne-sais-quoi sometimes about processing an undithered 24-bit file. Nothing more, nothing less.
It’s easy to hear when the audio is +108 dB. He should have played the 1812 Overture cannons at +108dB as well, for comparison
Dan’s video is correct from an absolutely technical point-of-view, and it’s never a bad thing to dither the end result, but to quote Monty Montgomery in Robin’s link above : “No one ever ruined a great recording by not dithering the final master”.
Agreed, aside from I think we need to separate real-world popular music squeezed into 4-5 bits and classical music as highly dynamic with delicate reverb tails not necessarily peaking anywhere near 0 dB. I’m not necessarily talking 1812 but something a little narrower like a Mozart symphony or some baroque chamber/solo music.
Given we are in agreement that dithering when moving to a lower bit depth is technically correct, I will continue to do so even at 24-bit. When it is so quick and, at the very least, does no harm…why not?
When quantizing to 16bits the lower 8 bit are thrown away, and the lowest bit (here bit 16) needs to be rounded. The data that is affected is below -90dBFS (down to -96dBFS).
Since 32bit float on has a 24bit mantissa, there is no quantization happening, and no quantization artifacts can be introduced.
Also keep in mind that thermal noise (atoms moving around, hitting each other) at room temperature (20degC) is at around -127.1 dBm. So everything in bits 20…24 is random noise and irrelevant anyway.
The only benefit of using 24bit during production is convenience. It has sufficient overhead to avoid clipping while tracking or mixing.
So you have an anechoic chamber, and equipment that can cleanly and without distortion reproduce a dynamic range >108dB? That’s amazing! I’d like to come visit You also claim your ears can hear a quiet mosquito in the far end of the room, while you are operating a jack-hammer (not necessarily at the same time, like reverb)?
Dithering 16bits should not be that loud. It’s usually only the lowest bit, max lowest two: -90dB maybe -84dB. But yes, adding some low-level noise at -60dB can be pleasant (more below).
Of course you can add noise to any signal. In some cases that actually sounds nice. Many plugins that emulate a classic tape workflow add white noise at around -60dBFS. That corresponds to what good analog master tapes were able to represent back in the day. – It can also mask artifacts introduced by some plugins (e.g. digital reverbs) and improved perceived quality (even though dynamic range is effectively lost).
That would however be a mixing decision, not a mastering one.
To be fair, he’s talking about the dither example shown at 11:50 here Dan Worrall - WTF is Dither - Youtube , where Mr Worrall attenuates a 32-bit sound by 108dB, renders it at different settings and then boosts it 108dB.
Indeed, I’m talking about Dan’s video where he creates a 32-bit float with gain down 108 dB and creates a 24-bit non-dithered file from that with fader back up 108 dB. It’s clear as day that 32-bit float to 24-bit fixed creates distortion. Whether it is audible at normal listening levels is not my point.
EDIT: And looking further into the idea of a float dither, it seems that the fixed part of the 32-float can be dithered in calibration with the floating part. However, a single instance is even technically impossible to hear but many multiple instances of float truncation does become apparent (apparently).
It would be interesting to hear if Dan’s results are reproducible using other DAWs and reverbs, or if they’re simply caused by bugs in Reaper or fabfilter.
It is absolutely reproducible given that this is simply a truncation issue at the lowest bit. My technical terms may be a little off but I hope you get my point. What is surprising to me is not that 16- or 24-bit undithered audio shows distortion really low down but that 32-bit float can also benefit from it given part of a 32-bit float is, indeed, fixed. As I said previously, I think we are talking many instances of build-up before it becomes apparent but my point all along is simply that if there is distortion (however low down), it makes sense to deliver the best master we can aided by dither. Who knows what processing people will apply to it down the line or what applications do to lossy codecs when they decode them? It may not be close to “obvious” but that little bit of magic dust adds a layer of confidence and makes mathematics that little bit more exciting.