Non-destructive upsampling?

[zgmg]

I need to upsample 16/44.1 iTunes Masters (256 kbps) to sample (i.e. hip hop sampling) in a 24/96 recording environment - is there a non-destructive medium to achieve this? And does upsampling 44.1 to 88.2, instead of 96, effectively achieve this?

 

Does the same apply with bit depth conversions? I.e. 16 to 24 and back - is that effectively non-destructive? Thank you.

[lagerfeldt]

All options are destructive.

 

All options will not improve the sound quality, to the contrary any SRC will only deteriorate the sound.

 

It makes sense to upsample in order to match the existing SR of the project, of course.

[Eriksimon]

I need to upsample 16/44.1 iTunes Masters (256 kbps) to sample (i.e. hip hop sampling) in a 24/96 recording environment - is there a non-destructive medium to achieve this? And does upsampling 44.1 to 88.2, instead of 96, effectively achieve this?

 

First: impossible: 16 bit @ 44.1 kHz (= uncompressed audio) and 256 kbps (= lossy compressed audio) are different resolutions. 16/44.1 is namely 1411 kbps (16 x 44.1 x 2 (stereo)) As you can see, the resolution of 16/44.1 is almost 6 times greater. ( 1411/256 = 5.5125 )

 

Does the same apply with bit depth conversions? I.e. 16 to 24 and back - is that effectively non-destructive? Thank you.

 

No. It is only non-destructive because the conversion (via copy/convert in the Audio Bin) will deliver a new file, leaving the source file as it is (non-destroyed).

 

If you change the bitdepth or the sampling frequency or both, then it is per definition destructive, so you'ld want to do this on a copy.

Luckily, doing this using Copy/Convert in Logics' Audio Bin, Logic will simply create a copy for you and convert that, so, effectively, it is non-destructive as you simply get an upsampled copy of your source file, leaving the original as it is.

[zgmg]

Thank you.

 

I mean to ask, can upsampling (for sampling rate) and up converting (for bit depth) be lossless? I.e., is 16/44.1 to 24/96 to 16/44.1 destructive, yet lossless?

[fastfourier]

"Lossless" isn't really the right word to describe the conversion you're talking about. Lossless/Lossy are generally used in reference to compression algorithms like FLAC or MP3. This conversion remains in the same format (PCM).

 

So "Inaccurate" would be a better way to describe this conversion. Here you will have inaccuracies in the time domain (96k is not a multiple of 44.1k).

 

Whether it's destructive or not depends on whether you delete the original audio files or not. "Destructive" refers to altering data in a file when you edit, for example in the Sample Editor in Logic. "Non-Destructive" is where the file remains intact, but the software is instructed to read it differently - like when you split an audio region in the Arrange window and move one part away from the other.

[zgmg]

So "Inaccurate" would be a better way to describe this conversion. Here you will have inaccuracies in the time domain (96k is not a multiple of 44.1k).

 

Perhaps fidelity is a better word, or eq retention with no added noise/dithering/aliasing/etc. Anyway, would 44.1 to 88.2 kHz achieve this?

 

And does 16 to 24 bit achieve this too?

[fastfourier]

Going from 16 to 24 bit will be fine. 88.2k is ok theoretically, but it depends on how the upsampling algorithm works. For shame, I have no idea what Logic's one does.

 

All these inaccuracies are going to make very little difference to the result IMO, especially as you are starting from a lossily compressed audio file. Like Lagerfeldt says, go directly from your source to your project's native sample rate, and that's the best you're gonna get.

[Eriksimon]

If you're asking whether upsampling will improve the soundquality, then no. At best (and in most situations), it will not perceptably change. But it will never 'improve' - not by going to a higher sampling frequency, nor by going to a larger bitdepth. In short: don't upsample, it won't help.

It (upsampling from 16 to 24 bits) may have some beneficial effect if those audio files are part of a mix though, because any processing (through plugins) will be more accurate, and this can add up in a mix. But for a "standalone" audiofile, it makes no difference.

The benefits from 24 bits over 16 bits (in a mix) are greater than the benefits over a doubled sampling frequency - in fact, higher sampling frequencies may have an adverse effect on soundquality, as some audio interfaces do not "handle" frequencies of 88.2 and higher very well, due to clock inaccuracies. And you have double the CPU and data bus overhead for next to no improvement.

So, in short: going from 16 to 24 bits can be sort of useful-ish, but only for audio files that are part of a (big) mix. Going from 44.1 or 48 to 88.2 or 96 is futile, so I'd advise against it. If you have a high end interface (Duet, RME firestuff, Metric Halo etcetera), then you may have some benefit from recording files at very high resolutions - but even with those, upsampling will not improve the final product.

So, in short: record in 24 bits @ 44.1 or 48 kHz, and do not upsample.

 

ps: Hey Fourier, why so fast?

 

 

[fastfourier]

I'd rather upsample one file than downsample everything else if going into a 96k recording environment though.

 

I'm feeling a bit slow-fourier today. Nyquist and I were out on the town last night.

[zgmg]

The whole purpose is to be able to record over 16/44.1 samples in a 24/96 environment.

[Eriksimon]

OK, so you have to convert them to 16/96. Going to 24 will not help, and is not necessary: you can combine different bitdepths (16 and 24) in one Logic project - but you cannot combine different sampling frequencies, therefore you have to convert any file to the projects' set sampling frequency. The sound quality won't change.

 

ps: so, slowfourier, how's Nyquists' tinnitus?

[zgmg]

OK, so you have to convert them to 16/96. Going to 24 will not help, and is not necessary: you can combine different bitdepths (16 and 24) in one Logic project - but you cannot combine different sampling frequencies, therefore you have to convert any file to the projects' set sampling frequency. The sound quality won't change.

 

ps: so, slowfourier, how's Nyquists' tinnitus?

 

OH, gotcha Eriksimon, thank you. The only follow up question I have is does 88.2 make sense, instead of 96? I really appreciate this forum's help.

[Eriksimon]

The only follow up question I have is does 88.2 make sense, instead of 96? I really appreciate this forum's help.

 

No, not if the session is recorded at 96 - you would then also have to downsample every 96 recording to 88.2, and that seems unnecessary, thpough I don't think it will sound any the worse (or better).

In short, if the project already has 96 kHz recordings and you want to introduce audio files with a lower frequency, then upconvert the frequency. If the project still needs to be started, then throttle it back to 48 (cinema & broadcast industry standard) or 44.1 kHz (iTunes, CD and portable audio device 'standard').

[fastfourier]

If your project is at 96k, then everything's going to get upsampled into that anyway. But if you had to choose, I'd say 88.2 makes more sense ... but not much!

[fastfourier]

ps: so, slowfourier, how's Nyquists' tinnitus?

 

That was the least of his problems after a few rounds of Shannon's homebrew...

[zgmg]

The project still has yet to be started - I'm designing a workflow now. Samples are at 44.1 but I want to record at 96 - so I think I'll sample in 44.1, then upsample in 96 when the sampling is done and the recording is relevant/next. Thank you all.

[zgmg]

Last question! Would a 44.1 kHz lossy master upsampled to 96 kHz have the same sample timing as its 96 kHz original master? I.e., beyond the fact that the 96 kHz original master would have at least 2x as much information, are the samples synced?

[Eriksimon]

What do you mean by "synced"? And how weird would it be if the answer was "No, they're randomly distributed."

 

O, and honestly, if you're still to start the project, use 44.1 kHz or 48 kHz - there is little benefit in going 88.2 kHz or 96 kHz, and it has one big drawback: it effectively halves your audio processing power, so you will run into overloads much sooner.

On the other hand, maybe you should just do it and see/hear for yourself how (if?) it works:

Bounce out your project, then make an identical copy but with all files at half the frequency (lots of converting to do) and with the projects' sampling rate halved (important: do not touch anything in the mix itself (faders, pans, plugin parameters)) bounce that out and then A/B both bounces to hear if there is a quality difference that justifies doublinbg your CPU overhead.

 

O, and this is not unimportant: what audio interface do you have? Can you also add that info to your signature? Some Audio IFs actually are less accurate at higher sampling frequencies, meaning that a mix of 88.2 kHz or 96 kHz files will sound worse than one of 44.1 kHz or 48 kHz files. And downsampling will not fully cure that , since the recording already happened at a higher inaccurate rate, and those inaccuracies are carried over (and can even get worse) when downsampling.