Recording advantages of 44.1kHz vs. 48kHz revisited

[WRGKMC]

I did a quick test. I took a song I has recorded at 24/48 and mixed
it down at 24/48, 24/44.1 and 16/44.1

I then pulled the 24/48 into Har Bal (green), and pulled the 24/44.1 (yellow) in as a reference file
and superimposed the two files over each other. I'd have to say there are some responce differences.
The green line has more peaks and valleys.

Har Bal isnt designed to show dynamic responce, only frequency content.
Not sure how I'd be able to show that in an audio editor besides open one at a time
and somehow create a semi transparent shot I could superimpose over another given the two files are different sample rates.



Heres another one comparing the same 24/48 mxed straight down to an MP3 file.
The loss of high frequency below 20K was expected. The slight boost in bass responce was not.

[MikeRivers]

It's nice to have a test, but what is Har-Bal really measuring? The significant thing is the difference between how the high end drops off, which would be expected. It might be more interesting to try the same experiment running the same sample twice and see if they overlay perfectly - that is, does Har-Bal display the same thing twice? It might also be more informative to use pink noise or a sine wave sweep rather than a song.

Also, how did you mix it down at the different sample rates? Were you sending an analog mix to the same A/D converter at two different sample rates? Or did you mix in the box, just specifying the output at different sample rates?

And the real test - could you hear a difference between 44.1 and 48 kHz?

[Sillypeoples]

W- Thanks for taking the time to test and posting that up.

[WRGKMC]

Originally Posted by MikeRivers

It's nice to have a test, but what is Har-Bal really measuring? The significant thing is the difference between how the high end drops off, which would be expected. It might be more interesting to try the same experiment running the same sample twice and see if they overlay perfectly - that is, does Har-Bal display the same thing twice? It might also be more informative to use pink noise or a sine wave sweep rather than a song.

Also, how did you mix it down at the different sample rates? Were you sending an analog mix to the same A/D converter at two different sample rates? Or did you mix in the box, just specifying the output at different sample rates?

And the real test - could you hear a difference between 44.1 and 48 kHz?

Har Bal is a static Frequency analizer as well as an EQ limiter.

Yes, I simply selected different sample rates to mix down. Same tracks, same interface.
Converters are not used for resampling. They are only used tracking and playing back music.
Not for mathamatical conversions of sample rates nor using tools like HarBal.
The trick is to keep it 100% digital and not inject losses caused by D/A conversions.

As a note, If I open up the same file in Har Bal, One as the main file and the same file as a reference,
the two images superimpoe over each other perfectly, so its not like you're capturing a moving target
like some frequency analizers that let you freeze the image for viewing. I cant tell you how
Har Bal actually captures a frequency image, but having used many frequency analizers dating back
to may days where I used analog test gear for sound system installations, I can say its one of the best
static analizers out there.

I mixed the same file down at the recorded sample rate of 24/48,
I downsampled one to 24/44/1 and one to 16/44.1.
Sonar doesnt mix directly down to MP3 so I had to use an audio editor for that.
I opened the 24/48 file and saved it directly to an MP3 in Cool Edit.

As far as hearing a difference, I can definately hear a difference A/B-ing an MP3 to a wave file.
It is very difficult to hear a difference between the 48 and 41.1 files even with experienced ears like mine.

Its only after mastering using tools like Waves Multiband, L2 Limiter, etc when the differences become
more apparent between the 48 and 44.1 hich is what I've said all along.

You could blame that on the mastering plugins and how they are adjusted. Theres no denying that
they manipulate what they are fed.
But the multiband does have some automation for seeking the proper threshold levels of the bands
so maybe it just isnt seeing the same digital content in both files?

Also as I mentioned in my previous posts, I was unknowingly up-sampling the wave forms recorded at 24/44.1 to 24/48
for mastering then down sampleing back to 16/44.1 for CD burning. Maybe this had more

[jbreher]

Originally Posted by MikeRivers

It's all about The Law. Professor Fourier showed us that any arbitrary waveform can be broken down into the sum of one or more sine waves

Strictly speaking, I believe Fourier's theorem states that any arbitrary continuous waveform of infinite duration with no change in periodicity can be broken down into the sum of one or more sine waves.

Not so for waveforms of finite duration.

I'm not sure how applicable this is, however, to the effect of sample rate on subjective recording quality.

[JeffLearman]

Originally Posted by jbreher

Strictly speaking, I believe Fourier's theorem states that any arbitrary continuous waveform of infinite duration with no change in periodicity can be broken down into the sum of one or more sine waves.

Not so for waveforms of finite duration.

I'm not sure how applicable this is, however, to the effect of sample rate on subjective recording quality.

Just loop the song indefinitely, and you meet the criteria. It works in perpetuity, and it also works when looking at a just the repeated segment.

The waveform is continuous, even if one sample is the lowest negative value and the next is the highest positive value.

[MikeRivers]

Originally Posted by jbreher

Strictly speaking, I believe Fourier's theorem states that any arbitrary continuous waveform of infinite duration with no change in periodicity can be broken down into the sum of one or more sine waves.

Not so for waveforms of finite duration.

Well, that certainly makes it simpler, but what's a waveform with changes in periodicity but a series of waveforms of different frequencies summed together? It's one of those things that has little practical application unless you're trying to construct a complex wave from sine waves, or break one down into its sine components. Then, at least you need something long enough to measure. But if you have a sqaure wave that's ten seconds long (starting and ending at zero) you can certainly represent all but the start and end of that segment by a series of sine waves, and you can represent the start and end as a single cycle or a few cycles of something.

[MikeRivers]

Originally Posted by learjeff

Just loop the song indefinitely, and you meet the criteria. It works in perpetuity, and it also works when looking at a just the repeated segment.

Yeah but what happens when the batteries run down?

In practice, it works well enough to be indistinguishable from magic, and that's what counts.

[henrebotha]

I scanned most of this thread, partly because it's 23:40 here and the girlfriend is going to get annoyed soon if I don't get in bed with her to watch Family Guy. But I felt I should link to this article: http://people.xiph.org/~xiphmont/demo/neil-young.html

For the tl;dr folks: the argument is basically that recording at 96 kHz requires that you have a soundcard that has a flat response up to 48 kHz. Even worse, it requires that the listener have an amp and speakers that have a flat response up to 48 kHz.

Quickly, how many of you own speakers that have that flat a frequency response?

[roomjello]

Isn't there a well established strong argument that 48k is superior over 44.1 simply for taking the LP filter by-products out of range or allowing a smoother slope?

[MikeRivers]

Originally Posted by roomjello

Isn't there a well established strong argument that 48k is superior over 44.1 simply for taking the LP filter by-products out of range or allowing a smoother slope?

It's well established, sort of, 'cause it's on the Internet somewhere, but it's not a strong argument. In either case, the "brick wall" filter is still a brick wall with all of its advantages and disadvantages. Same slope, same by-products. It just goes out 2 kHz further before it cuts off. If my calculations are correct, that's about two whole steps above whatever note is at 20 kHz. If you think this is significant, you have a better audiologist than I do.

The advantage of working at 2x sample rates is that some processes (computer processes, that is) work better because they have more samples to work with. But the netlore that there are fewer "stair steps" in the analog output when using a a higher sample frequency is simply not true.