Recording LPs with greater bit depth and higher sample rate

[KurtL]

Hi, I recently bought a Focusrite Scarlett 2i2 A-to-D interface to hopefully improve my workflow when converting vinyl LP's to digital files. The new interface sports 24 bit recording, and sample rates up to 192kHz. The primary benefit I was looking for was to not have to fiddle so much to get recording levels to where they fill most of the dynamic range, yet don't clip. So my new workflow is to record the album with some headroom (not very close to clipping), then normalize after the recording is complete, and use the waveform statistics window under Analyze to inspect whether/how many samples are clipped. It still takes some undo actions and trial and error, using normalize according to RMS, to find the Normalize level that maximizes the signal without clipping. But this is still better than repeatedly recording at different level settings using "drop the needle" at different sections of the LP to find out what level settings to use.

So, my first question is, does the workflow I've described preserve as much fidelity as can be expected when I ultimately save the resulting file as 16 bit AIFF? I have set Amadeus Pro to create new files at 24 bit depth, and I've set the Focusrite to generate 24 bit data; normalizing the data is amplifying it, to some degree, which would lose resolution normally, but since I'm starting with 24 bits and going back down to 16 bits on the save operation I should still come out ahead, right?

Second question: this new workflow is very sensitive to mismatch in recording gain between the channels; should I add a step after normalization where I amplify the lesser channel to make the RMS statistics come out equal?

Third question is, is there any point to recording at higher sample rates for this job? Ultimately my output file needs to be 44.1 kHz for burning to CD; if I record at a higher rate I would expect Amadeus Pro to re-sample the higher frequency waveform to make the output 44.1 kHz, right? I'm thinking that there probably isn't a real benefit to running at the higher rate if ultimately I'm going to save at a lower rate, but I'm open to arguments...

Side Note: it appears that Amadeus Pro settings only have sample rates up to 128 kHz; I would guess that this is a limitation of Apple's Core Audio? If I were going to find a common sample rate higher than 44.1 kHz that Amadeus Pro supports it looks like 96 kHz would be the one. Or am I missing something?

Side note, bug report: in the Amadeus Pro preferences it looks like the labels for bit depth and sample rate are switched in the place where you set the defaults for new files created.

Thanks for your thoughts,

Kurt

[Jim Edgar]

Hey Kurt - 

In digital recording, there is no benefit from approaching -0 dB (peaking). Nor is there any real benefit from mastering to the nth level. Leave some headroom. Hopefully, the albums were originally well mastered so that the relative loudness between tracks is appropriate. The iterative process with RMS Normalization seems a bit indirect. Why not simply Peak Normalize to -.05 dB? That would capture all of the dynamics and relative loudness between songs. 

If  you are recording in 24 bit, then you are doing so with more dynamic range than the vinyl you are recording. 24 bit seems more common these days. I started recommending it for VO projects a couple years ago. 


Despite many articles, videos and claims to the contrary, most people cannot hear the difference between 16 and 24 bit in a double blind test. The same is true with a sample rate above 44.1kHz. Although the analogy is often invoked, comparing audio sample rate to visual pixels in a camera is not applicable. (Ian Shepherd has some excellent podcasts/articles about that - here's a good one to start with: http://productionadvice.co.uk/high-sample-rates-make-your-music-sound-worse/ He also has some excellent resources on Loudness and Mastering.)  


The biggest benefit to higher sample rate is that you can render the frequency at half of that rate. With 44.1kHz, you can capture up to 22.05 kHz, which is significantly higher than most folks can perceive.That's going to likely cover the recording captured within the grooves.


The direct benefit to higher bit depth (i.e. 24 vs 16 bit) is a lower inherent signal to noise ratio, which works out to be an ability to capture a wider usable dynamic range. I would say that 24 bit/44.1kHz is probably plenty. The raw recording will never make things sound better than the original. (Now if you had access to the individual master tracks, that's a bit different..). 


Normalizing is precisely the same as Amplifying.  You just make the change by targeting a value vs. increasing/decresing by a relative value. 
There's also no perceptible quality loss. Amplifying/Normalizing is a very transparent step and does not modify the quality of the audio. 


To answer your questions - 


downsampling (in this cast 24 bit to 16 bit) is generally better than the reverse, but will depend upon the algorithm. Things which are converted too cleanly (i.e. with no dithering) sometimes sound worse to us. It may make no perceptual difference, but at some point, 24 bit audio output may be more of a standard. If you record that way, it would be simple to maintain a higher bit depth master to give yourself more flexibility. 


As far as sample rate, 96K is about as high as any studio I know of will go for most projects. Those 96K stereo file sizes at 24 bit are pretty massive. Most don't see the benefit even for the master source recordings. 


I'd say get the cleanest raw recordings at the target resolution and be done with it in one step.


Jim Edgar

VO: JimEdgarVoices.com | @jimedgarvoices | Source-Connect - jimedgarvoices


For audio/studio help: JustAskJimVO.studio | Schedule a session













On Sun, Sep 15, 2019 at 9:47 AM KurtL <forum2mail@hairersoft.com (forum2mail@hairersoft.com)> wrote:

Hi, I recently bought a Focusrite Scarlett 2i2 A-to-D interface to hopefully improve my workflow when converting vinyl LP's to digital files.  The new interface sports 24 bit recording, and sample rates up to 192kHz.  The primary benefit I was looking for was to not have to fiddle so much to get recording levels to where they fill most of the dynamic range, yet don't clip.  So my new workflow is to record the album with some headroom (not very close to clipping), then normalize after the recording is complete, and use the waveform statistics window under Analyze to inspect whether/how many samples are clipped.  It still  takes some undo actions and trial and error, using normalize according to RMS, to find the Normalize level that maximizes the signal without clipping.  But this is still better than repeatedly recording at different level settings using "drop the needle" at different sections of the LP to find out what level settings to use.

So, my first question is, does the workflow I've described preserve as much fidelity as can be expected when I ultimately save the resulting file as 16 bit AIFF?  I have set Amadeus Pro to create new files at 24 bit depth, and I've set the Focusrite to generate 24 bit data; normalizing the data is amplifying it, to some degree, which would lose resolution normally, but since I'm starting with 24 bits and going back down to 16 bits on the save operation I should still come out ahead, right?

Second question: this new workflow is very sensitive to mismatch in recording gain between the channels; should I add a step after normalization where I amplify the lesser channel to make the RMS statistics come out equal?

Third question is, is there any point to recording at higher sample rates for this job?  Ultimately my output file needs to be 44.1 kHz for burning to CD; if I record at a higher rate I would expect Amadeus Pro to re-sample the higher frequency waveform to make the output 44.1 kHz, right?  I'm thinking that there probably isn't a real benefit to running at the higher rate if ultimately I'm going to save at a lower rate, but I'm open to arguments...

Side Note: it appears that Amadeus Pro settings only have sample rates up to 128 kHz; I would guess that this is a limitation of Apple's Core Audio?  If I were going to find a common sample rate higher than 44.1 kHz that Amadeus Pro supports it looks like 96 kHz would be the one. Or am I missing something?

Side note, bug report: in the Amadeus Pro preferences it looks like the labels for bit depth and sample rate are switched in the place where you set the defaults for new files created.

Thanks for your thoughts,

Kurt

[KurtL]

Hi Jim,

Thank you for your detailed reply; it seems to confirm pretty much what my understanding of the process is.

To address one point that you brought up: I use the RMS normalization rather than the peak normalization because very often there is a pop or click in the source LP that ends up being the loudest group of samples, and I don’t want to scale the entire recording using a sound that is noise (not the intended source recording).

One annoyance that I’ve discovered since I wrote the original entry above: the Focusrite apparently doesn’t support recording at 44.1 KHz! It supports 48 kHz, 96 kHz, 192 kHz, but no 44.1 kHz. Odd...

[Martin Hairer]

Side Note: it appears that Amadeus Pro settings only have sample rates up to 128 kHz; I would guess that this is a limitation of Apple's Core Audio? If I were going to find a common sample rate higher than 44.1 kHz that Amadeus Pro supports it looks like 96 kHz would be the one. Or am I missing something?

The sample rate is limited to 512kHz, 128kHz just happens to be the largest preset, you can enter any value in the sample rate field of the recording window.

I do however agree with Jim that anything above 44.1kHz is very unlikely to lead to any perceptible improvement, these very high sample rates are usually used by people who use Amadeus to analyse signals (not necessarily audio) for some scientific purpose. For consumer audio recordings (even for playback on very high-end hifi equipment), I would be much more worried about the recording itself (background noises, humming, etc).

Assuming you have excellent recording equipment, you probably still won't be able to hear any difference between 44.1kHz and 96kHz or even higher, but you will definitely be able to pick out any background noise when played back at high volume on a good system.
Best,

Martin

--
HairerSoft
http://www.hairersoft.com/

[Jim Edgar]

Kurt - 

Focusrite definitely records in 44.1. Should be able to change that in the Apple Audio MIDI Setup utility. What version of the Scarlett do you have? They just updated to G3 and there have been a few odd things I've encountered. 


Ahh - Didn't think of the pop/click thing.  I'd probably just expand vertically and manually delete or attenuate those first. I will say that if you are doing a lot of this, it's worth snagging the Izotope De-Click tool (it's actually in the Elements package which was on sale not long ago - though you can find it for around $80 right now.)


https://www.sweetwater.com/store/detail/RX7El--izotope-rx-elements



- J


Jim Edgar

VO: JimEdgarVoices.com | @jimedgarvoices | Source-Connect - jimedgarvoices


For audio/studio help: JustAskJimVO.studio | Schedule a session













On Sun, Sep 15, 2019 at 2:19 PM KurtL <forum2mail@hairersoft.com (forum2mail@hairersoft.com)> wrote:

Hi Jim,

Thank you for your detailed reply; it seems to confirm pretty much what my understanding of the process is.

To address one point that you brought up: I use the RMS normalization rather than the peak normalization because very often there is a pop or click in the source LP that ends up being the loudest group of samples, and I don’t want to scale the entire recording using a sound that is noise (not the intended source recording).

One annoyance that I’ve discovered since I wrote the original entry above: the Focusrite apparently doesn’t support recording at 44.1 KHz!  It supports 48 kHz, 96 kHz, 192 kHz, but no 44.1 kHz.  Odd...

[philxm]

I will say that if you are doing a lot of this, it's worth snagging the Izotope De-Click tool (it's actually in the Elements package which was on sale not long ago - though you can find it for around $80 right now.)


https://www.sweetwater.com/store/detail/RX7El--izotope-rx-elements

I've been using that too, and getting mostly good results although it seems to respond much better to high-pitched clicks than to the lower ones. Which makes sense, as the lower ones are, of course, wider, and much more difficult to knock down manually, as well.

[philxm]

To address one point that you brought up: I use the RMS normalization rather than the peak normalization because very often there is a pop or click in the source LP that ends up being the loudest group of samples, and I dont want to scale the entire recording using a sound that is noise (not the intended source recording).

For this reason I like to "tamp down" blemishes that're louder than the top level of desired signal:

Edit > Select All
Analyze > Find Maximum


If the maximum is a blemish:


make appropriate selection
Effects > Interpolate


When, after knocking down peak blemishes, the new peak signal is a desired portion of the sound, you can now Normalize.

[KurtL]

Thanks for all the advice. I have a gen 3 Focusrite Scarlett 2i2, and when I use Apple's Audio Midi Setup I can, indeed, select 44.1 kHz sampling. It seems to change what Amadeus Pro presents to me in the advanced preferences for that sound source: now instead of multiples of 48 kHz, it presents multiples of 44.1 kHz! Which is fine.