What to do with updating older recordings?

[BonDarker]

Long time MOTU user going way back to 1990. Rather then post this on a more general forum, I’ve decided to keep it “in the family” where my understanding can track all your wonderful insight—I hope! I’m presently Mac OS 10.12.6 and using DP 9.51.

I am and have been in the process of completing or reworking many original projects that have evolved over the years of writing and recording. In some ways I suppose I’m on a mission to wrap up a life’s work while I’m still able. As one might expect, over the years (dating back to analog tape - pre digital), some of the tracks are as good as I’m going to make them (I’m referring to performance more so than production or engineering). So, using what has been completed is preferred to starting over. Other tracks I will likely re-record. In order to update a wide variety of recordings and compile them into a final collection I’m in a quandary as to how render older tracks and I’d appreciate some advice. My goal would be to update all projects to a contemporary standard bit rate, etc. so that the collection meets a standard. But if a recorded track or master is at 16 bit, 44.1, where do I go from there, when the standards now are 24/96 or even 192? Do I simply convert them into a higher bit rate and that's it?

I have considered pumping the audio from older tracks out through my studio monitors, re-recording that audio with some good mics back into the DAW at a higher preferred sampling rate. What might be lost may also benefit from some actual air being moved by sound—a natural thing if I remember right . Anyway, as you can see I’m shooting in the dark. Any suggestions?
Thank you...

[mhschmieder]

Bon, I'm kind of in the same mode myself these days, but I long ago digitized everything, at least as stereo dumps (all that I was able to do at the time I did most of the work, as I didn't have a computer yet and so used an Alesis MasterLink as my conversion tool, forfeiting multi-track history on a lot of it!).

Your second paragraph is a bit dense, so I'd recommend you break it up a bit, as it's a bit difficult to follow the flow and focus and requires several re-reads to try to extract the core message, given that there is a missing piece of information: is all of this historical material already digitized, or not?

As for common sample rates/etc., I try to at least avoid 44.1 kHz sample rate on new works and on converted tracks, as I repeatedly hear the difference in blind tests, except on the crudest material with almost non-existent dynamics and frequency range.

I prefer 96 kHz as many VI's have less aliasing issues (it's a bit trickier with sample libraries, which often go back and forth between sample rates during the overall signal path). For miked stuff other than drum kit oriented music, I prefer 96 kHz and usually go for 48 kHz when it's a rock project that starts with a multi-miked drum kit -- mostly because I use ADAT for expanding my available channel count. But I also use 48 kHz if it's known 100% up-front that the material is for film (and exclusively so).

The more important spec is the bit count. No harm is done by going from 16-bits to 24-bits, and as you will work with the tracks some after the conversion or upload, it is best to ALWAYS be at 24-bits (or, alternatively, at 32-bits or 64-bits when dealing with interim files between different stages of audio restoration application). There is really no reason to use 16-bits anymore; disc space is cheap.

If you are starting from analog and converting to digital, I recommend 96k/24-bit for the most part, if you're going to work with the tracks at all. But for most material, 48k/24-bit is good enough -- the higher sample rates can indeed give a smoother feeling with less discretization, but your ears are going to be listening to reconverted analog audio anyway, so the bigger difference once you get beyond the problematic 44.1k "standard", is that 96k sample rate will preserve more detail on long sustain and decay tails on stuff like acoustic guitar, cymbals, etc. when they're exposed vs. buried.

For film, many have gone to 192k or higher, due to the ever-increasing demands for frame-syncing audio and visual cues for maximum impact. I work in the specific industry where this matters, but my own music doesn't need it so I don't bother. I only do music for indie films and don't expect that to change.

Hopefully, whatever the overall status of your current tracks as analog + digital, digital-only but on a variety of media (including outmoded ones), stereo-only or multi-track, etc., this is enough information to help and to spur next questions.

[Phil O]

I have considered pumping the audio from older tracks out through my studio monitors, re-recording that audio with some good mics back into the DAW at a higher preferred sampling rate. What might be lost may also benefit from some actual air being moved by sound—a natural thing if I remember right .

I think that might be a bad idea. If you have older tracks in the digital domain, keep them in the digital domain. Do some research on what software (at your disposal) will give you the best results for whatever conversion you wish to do, and go that route. If your tracks are in the analog domain, use your best converters with careful attention to signal levels and analog signal path. I personally don't think that air is going to be advantageous here.

Phil

[mikehalloran]

There is little evidence to suggest that there is any benefit to upsampling other than from 44.1 to 48 when an AV project requires it.

OK, there is none but this doesn’t prevent hundreds of armchair experts who hear with their eyes from recommending it. Read this:

http://productionadvice.co.uk/high-samp ... und-worse/

I am happy to argue why stereo recoding of ensembles benefits from higher sample rates but only when done as the first step from the microphone and the listening environment is done at the same sample rate (stereo locators are found in the 22k-27hHz range and the human brain can perceive them). However, once you’re at 44.1k, there’s no adding back the missing audio.

I’m not saying don’t update. Sound Designer 2 is incompatible with many modern DAWs. Also, going from 16 bit to 24 bit can be beneficial (lower noise floor) if you are adding digital processing to those old tracks such as a remix or remaster. All you are doing is stripping and rewriting the headers—the audio is unchanged. You can convert from sd2 to aif or wav/bwav at the same time. You can do this in DP but it’s much (much, much, much) faster in TwistedWave or DSP-Quattro.

[mikehalloran]

If you have problems open SD2 or AIFF, MOTU fixed this in DP 9.52. Finally.

[HCMarkus]

TL;DR: 44.1k 16 bit is still a viable and fully acceptable standard.

Although I am not sure about the "Stereo locaters above 20k" comment, I feel Mike Halloran's overall advice is excellent, and the article linked within the one he linked is especially informative:

https://people.xiph.org/~xiphmont/demo/neil-young.html

As MIke notes, if you will be remixing your older compositions, digital conversion of the individual tracks from 16 to 24 bits may be beneficial, particularly if you add new material (which should be recorded at 24 bit depth.) Recording at 24 bits allows one to worry much less about a real problem, digital overs vs noise floor, and processing 24 bit files avoids potential quantization noise issues.

The nature of your music will determine if any of this matters. The real challenge, if you are trying to create a cohesive collection of compositions in the same sonic ballpark, will be to master your tunes to match perceived levels and EQ, so the listener's ears won't be repeatedly jarred as the songs are listened to consecutively.

Edit: Changed "15" to "16" bit... duh.

[FMiguelez]

I’m in a quandary as to how render older tracks and I’d appreciate some advice. My goal would be to update all projects to a contemporary standard bit rate, etc. so that the collection meets a standard. But if a recorded track or master is at 16 bit, 44.1, where do I go from there, when the standards now are 24/96 or even 192? Do I simply convert them into a higher bit rate and that's it?

Please be careful and don't become a victim of marketing lies and empty and downright false hype.

BTW, I'm neither an engineer nor a DSP expert. I'm a composer who loves this stuff and reads a lot about it, so take the following FWIW to you, and do your own research.

IMO, your chosen sampling rate and bit depth depend on what you already have (and in what format), and what you will incorporate as new material into a revised production, and what will be your final mass-delivery formats.
For recording/processing/mixing/mastering, 24/48 KHz is perfectly fine. For final product delivery, 16/44.1 is more than enough. It can be accompanied by a 256 kb]s AAC file as well. Please see bellow regarding this **.

The ONLY difference between a sampled 24 and 16 bit signal is the relative level of the noise floor. That's it. Your signal won't be "more pure" or have more "resolution" at 24 than 16 bits, or even 8 bits, for that matter. It will be equally pure, with added noise that ranges from the un-hearable to the obvious (24 to 8 bits respectively).

With all that 16 bit and 24 bit dynamic range, imagine how loud you would have to listen to your music for you to notice the quantisation distortion or dither noise. With the former, the loudest parts would be painful to listen to, and with the latter, they would cause you hearing damage!

The only difference between sampling at 192 kHz and 44.1 kHz is the sampled high frequency content. One is for bats, the other for humans. Please don't fall pray to the myth of the "more samples = more detail of the waveforms and less steps". The DAC reconstruction filter will give you an IDENTICAL output to the original analog input signal as long as it's properly bandwidth-limited according to the Nyquist/Shannon theorem (highest frequency is less than half the sampling rate). It is almost magical!

IOW, a 14 KHz signal sampled at 100 MHz/ms won't be more "detailed" or "pure" than if sampled at 32 KHz/s. They will be identical after the required reconstruction filter when it comes out of the DAC.



Based on this, my personal recommendations are>

1.- If you already have digitized stuff, such as older recordings at 16/44.1, leave it like that and import it into a new DP session with the conversion preferences set to a) Leaving the bit depth alone and b) changing the SR of your audio files to whatever your new DP session is set to upon import (I recommend 48 KHz, especially if you won't release a CD).

DP will process everything at 32 bFP if it needs to anyway, so you wouldn't gain anything by converting those 16 bit files to 24 bits (they would still be 16 bit files with added zeroes for extra wasted space and disk bus bandwidth).

** 2.- If you are going to re-record new material, or if you digitize your old analog recordings, do it at 24/48 or 24/44.1. You'll have ample dynamic range with the 24 bits so you don't have to cram your signals up to FS, and either sample rate will faithfully capture anything a human can hear plus a little more.

DP will happily let you mix bit depths (not sample rates) in the same project. So even if you import an 8 bit file, DP will process it internally at 32 bit floating point the moment you do anything to it.

The way I understand this 192 KHz+ business is that there are 1 or 2 arguable benefits to using these higher SRs for mixing/processing audio (not for delivery), such as reducing aliasing a little when using non-linear processes, some time-stretching plugins might work better, etc., but I really think the cons outweigh the pros.

--- Unless you have a chain ending with speakers specially designed to handle all that extra HF, they will distort and stress for something you won't hear anyway. And there is the chance that artifacts will reflect back into the audible range.
Also, there's the concept that... the faster the sampling, the less the precision... or something like that (I'd have to check the source for details, which is Lavry, I think).
--- It makes no sense to double or quadruple your CPU load for all that extra unnecessary information, which you will ultimately have to resample down to 48 or 44.1, unless the masses have amazing stereo systems capable of handling ultra high SRs without distorting and sounding like crap from the converters to the speakers, so that dogs can ultimately enjoy it.
--- Also, why waste space, even if it's cheap? And processing cycles and disk extra stress? (YMMV depending on number of tracks per session, your computing power, etc.).
--- And I know there are more issues against 192, from recording to mastering, but I don't remember off the top of my head.

If you need higher rates to avoid aliasing when compressing /limiting in your stems or mix bus, it makes more sense to upsample those compressor plugins instead of the whole DP session!
Even if this completely solved the aliasing problem (which I understand it does not, it only helps a little), it might be worth using in such cases on a need-to-basis, but even then, up sampling may cause other digital problems down the road, IIRC.


Whatever you decide to use, make sure it's based upon real knowledge you can check for your self and upon your current needs, and not based on myths or hype.

And if in doubt, you can choose to believe, or you can choose to KNOW and test for yourself... Do ABX tests to determine if you can tell, consistently, the difference between bit depths and SRs. That's all that matters. If you don't, then there's no need to waste processing power and space. If you do, please post your ABX results. I'd be impressed!

I have yet to see an ABX test posted by anyone demonstrating there's a difference between properly sampled 44.1 and 192 KHz test signals or songs (such as the ones that can be done with Foobar), or if anyone can cite a particular piece of music that at 16 bits sounds like there is not enough dynamic range under normal (even loud) listening levels in a quiet room, which would typically have a noise floor of 30-40 dB SPL).

I have considered pumping the audio from older tracks out through my studio monitors, re-recording that audio with some good mics back into the DAW at a higher preferred sampling rate. What might be lost may also benefit from some actual air being moved by sound—a natural thing if I remember right . Anyway, as you can see I’m shooting in the dark. Any suggestions?
Thank you...

I wouldn't do this. Bad idea in this case, if you want to capture your music accurately.

[mhschmieder]

Mike, your link backs what I said. But there are too many variables, not knowing the details of the OP's material. If there are no virtual synths involved, and no five minute cymbal solos, then 48k is fine.

It isn't possible to generalize this stuff to simple answers. The article that Mike linked is excellent and covers a lot of ground, with the only exception being the special requirements of synthesizer-based algorithms in the digital realm, which almost without exception have huge aliasing issues below 88 kHz.

So, it matters whether the original material is analog or digital, and if digital, is it in multiple formats, in which case tough judgment calls have to be made about the likely future use and which material is the most important (that is, if one is going to settle on just one format for all of the files).

And then it's important to know if anything else new will be layered atop it, what production work will be done, etc.

The only thing that has a simple answer is bit depth, because a 16-bit file converted to 24-bits doesn't change other than for its header. But it allows a lot more headroom for anything downstream. Yet if no work at all will be done on the files production-wise (this seems unlikely from the original post), it doesn't hurt to save disc space if the originals are 16-bits. But if converting from analog, go straight to 24-bits, unquestionably.

It is good to have an excellent technical article talking about the dangers and risks of upsampling, but in the end, that also has to be balanced with other factors.

For instance, I mostly work at 48k myself, but as soon as a soft synth is involved, or becomes the dominant player, and/or the rest of the tracks end up being sample libraries vs. live miked instruments, the sample rate issues are out of the hardware realm and much more complicated due to almost never having an exact match for all of the sounds used in a project.

My impression is that those factors are not likely to be part of the OP's path, and that it boils down to unifying everything at 24-bits, and having mostly a mixture of 44.1k and 48k sample rates based on the track origins. In general, try to minimize the number of conversions in the chain.

The audio converter can play a key role in the decision as well. In spite of what is said in the linked article, there are plenty of interfaces out there that are optimized for 96k (or maybe 88.2k) and perform poorly at lower rates. Too many different design decisions from manufacturers for the pure scientific answer to automatically apply across-the-board as a simple yes/no.

[mhschmieder]

Rats; an internet glitch gobbled my entire edit buffer. Not sure if I can remember what I wrote, and I spent considerable time editing it three times for clarity and concision.

I think I was reiterating that sample rates such as 192 kHz may have adherents who claim an audio difference (I'm not one of them), but the main purpose is to support super-tight film cues for maximum impact. You need more frames per second to match what's going on with the film side of things, but movie industry competition demands the ultimate in dramatic effect.

So stay away from those high sample rates. And also from "re-recording" the audio through loudspeakers. Several others did a good job of itemizing criteria for bit depth, but for stuff that is at 16-bits already and that you keep at 16-bits, there is no risk in doing it later as only the header changes.

Similarly, Broadcast Wave files (BWAV or just plain WAV) are best for longevity and for working with the most applications, but as only the header differs (someone can correct me if I'm wrong), I don't think any audio conversion or regeneration happens when you switch audio file formats.

I have hundreds of projects still in SD2 format. It doesn't matter for now, but eventually it is safer to have things in ".wav" format. But I can't remember how I wrote this up in my lost text and never do as good a job the second time around.

Pro Tools may still only support mono files (I don't use Pro Tools and don't pay much attention to it), so those who go back and forth with that DAW often feel compelled to keep or convert everything to mono files and avoid stereo files -- hard if you're using sample libraries and soft synths, but not usually a challenge when pulling in archival material sourced from microphones, except that some people record stereo pairs as stereo files vs. individual files per mic.

My apologies for such a slop write-up; the lost post was better than anything I've written lately (clearer and more concise), so it was probably inevitable it would get swallowed into the ether.

[mikehalloran]

I don't think any audio conversion or regeneration happens when you switch audio file formats.

As long as there’s no compression (sd2, aif and wav), you’re just stripping and rewriting headers.

I don’t know how this applies to so-called “lossless” compression schemes such as Apple Lossless and FLAC.

I first read about supersonic stereo locators at AMPEX in the 1970s — the research had been done in the 1950s. My dad worked at AMPEX from 1967 or so till the 1980s. It was also widely written about during the late ‘70s-early ‘80s Direct to Disk craze and was occasionally cited as the real reason that digital was somehow flat as opposed to the 3D sound of analog. This only applied to ensemble recordings done in true stereo and not the panned mono begun in the late 1960s.

I have a record from 1957 with a technical note that one can see frequencies ranging from 16–25kHz on the grooves. It does sound quite realistic through a set of good headphones. Found it—click on more images to read the back cover.
https://www.discogs.com/The-Dukes-Of-Di ... se/1899763

Hmmmm.. Dad is in the hospital recovering from a heart attack. I’m going to put this and some of his other favorite albums on an iPod. Of course, I’ll have to teach him how to use an iPod... Oh my...

[HCMarkus]

Hope dad is back up and about soon Mike. Interesting info from the 1950's study.

FM, I think you are correct about the bit depth thing within DP. I guess converting to 24 bits would be useful only if bringing audio out, processing, then back into DP. Since I always work at 24 bits, I have always simply converted 16 bit imported projects, perhaps unnecessarily.

One question: What happens when you apply a plugin to a 16 bit file in a 24 bit project? Is the result written at 24 bits?

[FMiguelez]

FM, I think you are correct about the bit depth thing within DP. I guess converting to 24 bits would be useful only if bringing audio out, processing, then back into DP.

If you'll recapture back to DP after some analog loop, it won't hurt, but I doubt it'll make much difference. I suppose it depends on how noisy the analog chain is... Its noise floor might be higher than the 16 bit file's, rendering the point moot, and the original signal already has a16 bNF even if DP processes it at 32bFP anyway.
Also, because of this, wouldn't you have to requantize / dither the 32bFP signal down to 24 or 16 bits anyway, as it comes out of DP when you send it out to your analog chain?

Since I always work at 24 bits, I have always simply converted 16 bit imported projects, perhaps unnecessarily.

One question: What happens when you apply a plugin to a 16 bit file in a 24 bit project? Is the result written at 24 bits?

Yes, converting 16 bit files to 24 bits doesn't do much except padding with zeroes and making the file larger. You'd end up with a 24 bit representation of a 16 bit file, with the same noise floor. DP won't, of course, add any new information or reaccomodate the signal through the new dynamic range.

About your question, I'd have to double-check, but if you print plugins offline, your resulting printed file would be at whatever bit depth you have set in your current DP session, even if the file was 8 bits. In sthis case DP would first process it at 32bFP (i.e., if you move its fader or plugin), then add dither (?), and then quantize and print it to the session's bit depth. As it processes it and creates the new audio file, supposedly, this is a time where the misterious dither command in the menu at the top plays its role.

Maybe I should test this when I have time...