I got lost - dithering and resampling for DDP package

[george_vel]

Hello Masters,

I need your help, because I got lost.

I have 15 tracks in wav format - all mixed and mastered in 32 bit (float), 96 kHz in Audition (they were recorded in the same resolution and sample rate).

Now I have to prepare the DDP files for the house that will produce the CD copies.
For the purpose I got Reaper during last weekend, and following James Perrett advices in this topic I was able to put gaps and markers for the songs.

But I am totally new in Reaper and I've never used dithering so far.
I've noticed when I import wavs in Reaper (the project is set to 44100 Hz) that they had been resampled. However, just rendering the single track that contain all pieces one after another to 16 bit, 44100 seems to me very risky. I am not sure what the resampling during import did and even less sure once it goes to 16 bit what will be the effect. I am afraid to send this for production as it may have the nasty artefacts due to improper processing.

How I can handle this?
Should I go back in Audition and pre-prepare the files there for importing in Reaper (apply dither and export in 16 bit, 44.1 kHz)?
Or should I dither and resample them in a separate project in Reaper, and then import them into the DDP project?

And when I have to apply dithering? What settings to use?
If it is in Reaper and all songs are in a single track, should I apply dithering on the entire track, or on every object on the timeline?
I have Ozone 9 and probably could use its dithering options, but I can use it only in Reaper as Audition CS6 is not compatible DAW for this version.

Sorry for bombarding so many questions, but I got lost and need help to find my way again.
In general, if you share your approach and workflow with dithering, resampling and making proper DDP package, it would be invaluable.

[Hugh Robjohns]

However, just rendering the single track that contain all pieces one after another to 16 bit, 44100 seems to me very risky. I am not sure what the resampling during import did and even less sure once it goes to 16 bit what will be the effect. I am afraid to send this for production as it may have the nasty artefacts due to improper processing.

The obvious solution is to run it and then listen. If you can hear artefacts you have a problem... if you can't you haven't!

Should I go back in Audition and pre-prepare the files there for importing in Reaper (apply dither and export in 16 bit, 44.1 kHz)? Or should I dither and resample them in a separate project in Reaper, and then import them into the DDP project?

Personally, I like to remain in charge of what is going on at each stage and (assuming you're only using Reaper for gapping and DDP export -- not level changes), I would export each separate track file from Audition having done the sample rate and word-length reductions there, just so that I could ensure each track is exactly as required.

Also I know the built-in SRC in Audition performs technically much better than the built-in SRC in Reaper... (https://src.infinitewave.ca/)

And when I have to apply dithering? What settings to use?

Er... 16 bits for starters...

Most word-length reduction systems offer several different dithering options. The plain-vanilla, works-every-time 'safe' option is Triangular dither or 'TPDF'. It bestows the track with a steady flat noise floor, not unlike analogue tape (but 30dB quieter!)

Other options, such as the myriad forms of noise-shaping (either generic or branded), all try and scoop audible noise out of the ear's sensitive region (2-5kHz) and pile it up in less sensitive areas (typically above 18kHz).

The idea is to give the subjective impression of a lower noise floor than 16 bits can normally allow, and they generally work well in that respect... but the risk is that the HF dither noise might interact with any audio content already in those regions and potentially cause overloads or aliasing...

So the choice and audibility of noise-shaped dithers can be quite material-dependent and you'll just have to experiment to see if it (a) your selected dither makes any audible difference and (b) causes any audible problems.

As a general rule, the inherent (acoustic/electronic) noise within most source recordings is way higher than the 16-bit noise floor anyway, and so you might just as well keep things simple with TPDF (or a fairly gentle noise-shaped alternative) because the noise floor your audience hears won't be dither noise anyway... but YMMV.

[Drew Stephenson]

It also sounds like you haven't changed the default settings on your Reaper project.
Reaper will convert any imported media to the default settings automatically, and the default, as you've discovered, is 44.1kHz.
You'll need to go into the project settings and change the default.

[george_vel]

Thanks Hugh, blinddrew!

Hugh, ok, I’ll stick to Audition for creating the files in the format that is suitable only for DDP rendering in Reaper.
But please tell me briefly the correct order of needed steps - should I first save the files in 16 bit, 44100 Hz in Audition, and then import them back to apply the dither or...?

[James Perrett]

But please tell me briefly the correct order of needed steps - should I first save the files in 16 bit, 44100 Hz in Audition, and then import them back to apply the dither or...?

I'd convert the sample rate first, then apply dither and finally save as 16 bit. In Audition 3 you just choose Convert Sample Type from the Edit menu which will carry out both operations in one go.

[george_vel]

But please tell me briefly the correct order of needed steps - should I first save the files in 16 bit, 44100 Hz in Audition, and then import them back to apply the dither or...?

I'd convert the sample rate first, then apply dither and finally save as 16 bit. In Audition 3 you just choose Convert Sample Type from the Edit menu which will carry out both operations in one go.

[Hugh Robjohns]

But please tell me briefly the correct order of needed steps - should I first save the files in 16 bit, 44100 Hz in Audition, and then import them back to apply the dither or...?

Dither has to applied when you reduce the word-length.

So the correct order is to sample-rate convert the 32-bit floating-point files to 44.1k, then save as 16 bit files with the appropriate dither option selected.

As James says, you can perform both tasks simultaneously in Audition using the 'Convert Sample Type' option which you'll find under the Edit drop-down.

[manwilde]

Also I know the built-in SRC in Audition performs technically much better than the built-in SRC in Reaper... (https://src.infinitewave.ca/)

Sorry to go a bit off topic here, but does this also apply for Audition 1.5 vs Reaper 5 29?

[Hugh Robjohns]

Unfortunately -- but understandably -- not every iteration of DAW is listed in the SRC Comparisons site, and neither Reaper 5.29 nor Audition 1.5 are listed. But if I had to hazard a guess I'd say probably yes, AA 1.5 could well be a little better than Reaper.

SRC algorithms within a DAW are rarely updated within each software generation, and if you scroll through Sweep plots of all the different Reaper versions they're all pretty grim.

You should just see an upward curve against a pure blue/black background. In all of the Reaper plots you can see crosshatching caused by aliasing distortions. Some versions are worse than others, but they are all poor.

The earliest edition of AA tested on the site is version 2, and while that's not completely perfect, its aliasing distortion is fairly minor, and considerably better than Reaper. AA v5.5 and all the later editions are pretty much perfect.

If you're interested in your own specific system's performance, but it's unlisted, you can download the test files (details in the FAQ), run them on your own system, and send the results back for inclusion. It's easy enough to do, and there are clear instructions in the FAQ.

[manwilde]

Thanks a lot, Hugh!. Added to the "to do" list.

[James Perrett]

It may be worth saying that Reaper 6 is a big improvement over previous versions although still not as good as newer versions of Audition. I remember reading a post by Justin on the Reaper forum who explained that Reaper's SRC needed to be able to adjust input and output frequencies on the fly which is why they used the algorithm that they do rather than use a higher quality one which could only be used with fixed input and output frequencies.

[Hugh Robjohns]

Sounds like he might have been comparing synchronous and asynchronous algorithms.

It is obviously harder to convert when the input rate is varying, because the algorithm has to continually recalculate the parameters, but other manufacturers have managed it.

To be honest, I'm surprised there are still so many poor SRC algorithms in use in DAWs. It's quite a mature technology now.

I normally use the one in iZotope's RX which is the best I've found so far.