Quantitative comparison of 16 vs 24 bit Beatles remasters

Concerning the USB 24 bit Beatles release, I decided to conduct an experiment in order to see if I could figure out the differences between the 16 bit stereo and 24 bit versions. I took the 16 and 24 bit version of Because (Abbey Road), got them lined up in audacity (interestingly the track boundary was slightly off between the two), inverted one of them, and listened. Music! Ok, well, better normalize both tracks separately first before subtracting - did that, and what I hear is: Some very light music slightly filtered (indicating the amplitudes do not match exactly, this is to be expected) peaking around -65 dB, with some grungy noise which is amplified quantization noise from the different bit depths.

The difference between 16 and 24 bit versions of Because is in the quantization near the Nyquist frequency. Its spectrum had energy with a maximum amplitude of -89 dB, barely above the -96 dB "zero" in 16 bit music. There was no energy above -96 dB below 18 kHz, in fact. So, this makes me think that you simply aren't going to hear a difference between the files for songs with no limiting. They obviously chose one of the dithering algorithms which pushes most of the quantization noise to very high frequencies (good on them).

I then took the difference file of Because and normalized *that* so I could see the waveforms better. What I saw with Because looked like white noise, and sounded somewhat like it, with the music in the background (differences in amplitude between the 16 and 24 bit are impossible to completely remove).

I then applied the same approach to Here Comes The Sun. When I normalized the differences in HCTS there are spiky pops in regions of the song where, apparently, limiting was applied. In addition, there are periodic phase changes occurring within Here Comes The Sun that are correlated with some of the peaks. This manifests itself in places where the music will suddenly clearly come out of the noise for a second or two and then fade away in the normalized difference file. This does not occur with Because, because it doesn't have any hard limiting applied, apparently - no drums. With HCTS The clicks are at the peaks, indicating, I think that a different approach was done with the limiting... not exactly sure what is going on here. I guess you could think of it as long-period jitter. But, it is clear that there is a difference with the limiting between the 16 and 24 bit files. But it's very subtle, as the largest amplitude differences even with the peaks is -55 dB with HCTS.

Remember, all these difference are very, very small. The non-normalized difference waveforms for both (even HCTS with its spikes) is A STRAIGHT LINE AT ZERO when viewed on a computer screen.

I have put small sections of each normalized difference file here:

http://gromit.orf5.com/misc/Beatles1624diff/

Beware, the noise is very nasty on HCTS. In order to hear the low-level difference on HCTS I allowed clipping when I amplified. Turn your volume to a low level before listening, and listen on your computer!

What conclusions do I draw from this experiment? Note, you may not agree:

1. You probably won't hear a difference between the 16 and 24 bit files, unless perhaps your DAC does a lot better with 24 bit files for some reason. The differences are just too far below the threshold of human hearing, even with stuff that has limiting applied. Remember, background vinyl noise is waaaay louder than the magnitude of the noise differences we're talking about here.

2. It is clear the files are digitally sourced from the same parent file, and that they have used decent algorithms for downsampling and dithering.

3. Unless another masting/mixing job is done, you might as well just stick with the Redbook CDs. If vinyl sourced from the digital files is released, you can probably assume differences in the digital vs analog are a function of your gear, not the source music itself.

 

So...how was your subjective normal listening to the two formats?

 

I have not yet done that. Today was complicated, had access to computer but not stereo.

I betcha they sound identical though

 

Wouldn't they master for vinyl differently? I was under the impression that usually the vinyl is mastered separately and it would be from the 24/192 source files wouldn't it? http://emusician.com/tutorials/mastering_vinyl/

Are the USB 24-bit files the 24/192 master files? I ask because I didn't even know the USB had 24-bit files!

 

WONDERFUL, the 24 bit tracks are limited too. Great going, EMI.
No cd in the delivery and you still managed to screw the sound up.

 

Audacity can play 24 and 16 bit tracks cocurrently, right? would it help to take a mono track, such as You Know My Name, and run 16 bit in left and 24 bit in right for a subjective listening?

 

It's unclear whether they will master differently for vinyl, as far as I know, but there is no reason other than consistency why they shouldn't master from pre-limiting sources.

The USB 24-bit files are 44.1KHz. The main USB Beatles thread covers the related issues in some considerable detail.

 

Or, damn close!

 

i'm sorry to tell, maybe Barry Diament or Steve Hoffman can help me as english is not my mother language...for me is hard to write tech infos in english.
please don't be angry with me...was a deep work but what you did don't have value!
is my phrase to call the atention of experts to read what you did and explain details of what and where is wrong.

i'm sorry but is my sincere oppinion as old technician!
i have open mind to someone correct me with tech explanations.

cheers!

 

Why assume they did not make an analog copy for vinyl? Isn't this the way most albums are done?

 

Not anymore.

 

I'm prepared to be contradicted, here, but I should have thought that normalising the 16 bit file in particular would have introduced non-trivial artefacts via truncation which may have been of sufficient magnitude to compromise the integrity of the comparison.

A safer but still imperfect method would be to normalise the 24 bit file, which will suffer less from the processing, to match the level of the 16 bit file and leave the latter intact.

 

I did the same thing last week some time, I believe the conclusion was that these are barely any close to the real 24/192 transfers.

 

But then again, it might not have. I've managed to confuse myself, now.

 

In case anybody was wondering, I finally had time to compare the 16 and 24 bit releases on my good system and I cannot detect an audible difference in the comparisons I made, other than the 24 bit files are very slightly louder (and had I not seen that with my own eyes I probably wouldn't have known by listening). See my profile for equipment if you care.

 

Can you notice a difference in the soundstage or depth? That's usually where the most noticable difference are to my ears. In addition to sounding a bit more 'brittle' on the 16 bit.

 

I can't say that I did, to be completely honest, and I was listening for it. I was able to go quickly between 16 and 24 using the web interface of the squeezebox, and they really did sound the same on my system (which includes an upsampling DAC).

 

You can not make judgments on something you haven't heard.
The USB (24/44) files are not the 24/192 master.
Therefore we still don't know about the BD or vinyl versions.

Thanks tho Leigh, for a tremendous job comparing the CD too USB.
Well done, enjoyable, & informative.

Happy New Year too all!
And maybe in 2010 we'll get too hear BD's and/or vinyl as well.

 

I realize there may be better sound for the remasters, but I am happy with the CDs. I love the Beatles, but it stops here for me. There's a lot more music to buy and listen to out there.

 

I upsampled the CDs to 192/24. Opens up the soundstage quite a bit, depending on the album.

 

you have good system, cool.
you cannot detect and audible difference in the comparisons but the 24 bit files are very slightly louder. Leigh, have contradictions!
the important is not only where you did the test but how was done the comparisons.
as you did lots of changes in the sources....no need to repeat myself as posted before.

i cannot see the link to your equipment in your profile, is there

...you're not in my ignore list, nobody is! don't seems strange your signature as others can read different words? why i read this phrase in your signature? you're free to explain or correct me where i'm wrong.
seems that when someone don't agree with you we have barriers(why?) because you must be right and perfect or... 'end of the story' you take to personal size but we are here to agree or disagree ..is one cool forum full of debates, oppinions, serious posts, posts with jokes but not for fights.

with or without barrier your complete test is not valid, i swear!
you adequate to yours 'perfects targets' in yours ex-perimentations!

cheers and sorry to be sincere, excuse!

 

!?!

there's be no content above 22050hz...

 

If you've fliped the polarity of these 24 bit files and lined them up , sample accurate to the 16 bit cd release, you'll get sigal at about -90dB dfs ish, thats all.
this biggest difference will be at the lowest levels...like -60dB dfs, the 16bit files will have been dithered correctly.
merry christmas by the way..so this is all irrelevant...time to get back to where I belong,

 

Count me as another one who is happy with the CDs.

Dumb question: Why was the USB version released as 24/44.1 and not 24/192? Seems like a waste to release it as 24/44.1.