Null Test: 24bit vs 16bit vs MP3

[watchnerd]

I thought this was a pretty interesting comparison that I read about in another thread elsewhere.

Perform a spectrum analysis on a track in 24bit/192khz, then transcode to 96khz, 48khz, 16bit/44.1, and LAME320 MP3, then plot the nulls compared to the original.

All credit to TheoS53 for the original graphs, the title can be found at the top of each.

24bit/192khz original:

192 khz vs 96 khz:

192khz vs 48khz:

192 khz vs 16bit/44.1khz:

192 khz vs LAME 320kbps MP3:

 

[RayDunzl]

Well, we all know there is nothing audible in the upper Stratosphere, however you did discover differences in the Ozone Layer, but, for better or worse, I live in the Troposphere exclusively.

As for the Lame comparison, I blame human-generated electrosmog.

 

[amirm]

With perceptual processing (e.g. MP3) null tests are unfair in that the codec is free to throw out tons of spectrum if it thinks it is inaudible. The fact that so much of that shows up in a null test is a testament to how well it works rather than a detriment.

The useful part is just the normal filtering that goes on and listening to the null track as a way to get trained on the lost signal.

 

[watchnerd]

With perceptual processing (e.g. MP3) null tests are unfair in that the codec is free to throw out tons of spectrum if it thinks it is inaudible. The fact that so much of that shows up in a null test is a testament to how well it works rather than a detriment.

The useful part is just the normal filtering that goes on and listening to the null track as a way to get trained on the lost signal.

I don't think anyone is claiming the MP3 scenario is a detriment to psychoacoustic methods.

Quite to the contrary, it's making the opposite case -- if most people can't differentiate MP3 320 kbps vs Redbook, with such shallow and obvious nulls, what chance do they have of ABXing high resolution with a few nulls >20khz and mostly >-90 dB?

 

[fas42]

I don't think anyone is claiming the MP3 scenario is a detriment to psychoacoustic methods.

Quite to the contrary, it's making the opposite case -- if most people can't differentiate MP3 320 kbps vs Redbook, with such shallow and obvious nulls, what chance do they have of ABXing high resolution with a few nulls >20khz and mostly >-90 dB?

The point is whether the differences are subjectively significant - an extremely irritating artifact in the sound can be heard easily, even at very low levels; other differences may be huge in measurement terms but barely perceived, because they're "benign". IOW, the 'character' of the differences is what really matters, not the levels.

 

[watchnerd]

The point is whether the differences are subjectively significant - an extremely irritating artifact in the sound can be heard easily, even at very low levels; other differences may be huge in measurement terms but barely perceived, because they're "benign". IOW, the 'character' of the differences is what really matters, not the levels.

Well the null test for 24bit vs 16bit is empty below 21khz, and the aliasing stuff is -100 dB, so I'm not sure what irritating artifacts you think the graph shows.

 

[Frank Dernie]

The error is -60dB or below, even in the MP3 nul test.

 

[watchnerd]

The error is -60dB or below, even in the MP3 nul test.

Are you referring to the threshold of inaudibility as being -60 dB?

 

[fas42]

Well the null test for 24bit vs 16bit is empty below 21khz, and the aliasing stuff is -100 dB, so I'm not sure what irritating artifacts you think the graph shows.

The graphs are not showing anything relevant to what I'm referring to - I'm just commenting on the fact that just because there is a big or small measurable difference between two variations of an audio track, that linking that to whether "it's audible" is almost meaningless.

Personally, it's what happening in the range 60 to 80dB range, that's where the real action is, in terms of whether it's a hindrance to listening pleasure. As a trivial example, tape hiss: a half decent system will "expose" this - is it a problem or not, as regards the listening; does removing the hiss matter, or doesn't it?

 

[watchnerd]

The graphs are not showing anything relevant to what I'm referring to - I'm just commenting on the fact that just because there is a big or small measurable difference between two variations of an audio track, that linking that to whether "it's audible" is almost meaningless.

Personally, it's what happening in the range 60 to 80dB range, that's where the real action is, in terms of whether it's a hindrance to listening pleasure. As a trivial example, tape hiss: a half decent system will "expose" this - is it a problem or not, as regards the listening; does removing the hiss matter, or doesn't it?

The topic is the audibility of differences amongst 24bit high resolution audio vs 16bit Redbook vs MP3.

 

[fas42]

The topic is the audibility of differences amongst 24bit high resolution audio vs 16bit Redbook vs MP3.

Which includes this point ... http://audiosciencereview.com/forum...l-test-24bit-vs-16bit-vs-mp3.1382/#post-36226

 

[watchnerd]

Which includes this point ... http://audiosciencereview.com/forum...l-test-24bit-vs-16bit-vs-mp3.1382/#post-36226

Sorry, maybe I'm dense, but that seems completely orthogonal.

Tape hiss can potentially be audible regardless of format.

 

[Thomas savage]

The graphs are not showing anything relevant to what I'm referring to - I'm just commenting on the fact that just because there is a big or small measurable difference between two variations of an audio track, that linking that to whether "it's audible" is almost meaningless.

Personally, it's what happening in the range 60 to 80dB range, that's where the real action is, in terms of whether it's a hindrance to listening pleasure. As a trivial example, tape hiss: a half decent system will "expose" this - is it a problem or not, as regards the listening; does removing the hiss matter, or doesn't it?

( moderator action)
Nor are you offering anything relevant to the thread topic, a POV is fine but your arguing ( discussing) contrary points with no evidence. The thread is being derailed as a result.

My patience has ran out, iv issued a reply ban to fas42 for this thread. You don't seem to have any real intrest in the actual topic. We have been down this road so so many times Frank , unfortunately this is now what's going to start happening.

 

[Frank Dernie]

Are you referring to the threshold of inaudibility as being -60 dB?

No, just the likelyhood of audibility
My expertise in audio is old, I gave up full time work in audio in 1976 to go into full time Formula 1 racing, but I did research on record players back then.
The ability of a record player to reject spurious vibration requires a lot of clever engineering, which almost no record player manufacturer has ever actually used, prefering styling and marketing BS to sell their products.
It is normal for a non-isolated turntable to have bass pickup via structure and air only around 10 to 15 dB below the actual music signal, and slightly delayed, so it is like adding a bit of bass reverb.
On audition few people notice this or see it as a shortcoming and of those that do most people actually prefer it.
Crosstalk of the very best, perfectly aligned pickup cartridge approaches 40dB but only in the mid-band. 15dB is more normal but the stereo effect is still strong.
So my point is really that whilst -60dB isn't much in terms of digital recording capability, and isn't enough dynamic range for recordings to give audibly hiss free silence, it really is unlikely to be audible if it is that far below the music IMHO.
60dB is around the total dynamic range of my old non-Dolby reel-to-reel recordings and the background hiss, noticable during silence, is obliterated by almost all levels of music.

 

[Blumlein 88]

No, just the likelyhood of audibility
My expertise in audio is old, I gave up full time work in audio in 1976 to go into full time Formula 1 racing, but I did research on record players back then.
The ability of a record player to reject spurious vibration requires a lot of clever engineering, which almost no record player manufacturer has ever actually used, prefering styling and marketing BS to sell their products.
It is normal for a non-isolated turntable to have bass pickup via structure and air only around 10 to 15 dB below the actual music signal, and slightly delayed, so it is like adding a bit of bass reverb.
On audition few people notice this or see it as a shortcoming and of those that do most people actually prefer it.
Crosstalk of the very best, perfectly aligned pickup cartridge approaches 40dB but only in the mid-band. 15dB is more normal but the stereo effect is still strong.
So my point is really that whilst -60dB isn't much in terms of digital recording capability, and isn't enough dynamic range for recordings to give audibly hiss free silence, it really is unlikely to be audible if it is that far below the music IMHO.
60dB is around the total dynamic range of my old non-Dolby reel-to-reel recordings and the background hiss, noticable during silence, is obliterated by almost all levels of music.

Was doing digital rips for a friend with a very nice turntable. He complained those he did himself did not sound good like the LP. I found he was doing them silently without playing over the speakers as he recorded. Told him about the air feedback with delay via the TT. He didn't believe me. He also used a tube phono preamp which I think picked up a bit more. I set his needle statically in a scrub record we had. Recorded the TT while playing a CD. You could hear the music clearly being picked up by the cartridge on the TT. FR was different, bass heavy and lower in level, but audible. He was miffed his TT was so miserable at isolation. Then we recorded his LP while playing over speakers, and result was to his liking and so he was happy. I said, "if your TT was designed to really ignore air vibration you wouldn't like it as much".


BTW, are you this Frank Dernie?

 

[Frank Dernie]

Was doing digital rips for a friend with a very nice turntable. He complained those he did himself did not sound good like the LP. I found he was doing them silently without playing over the speakers as he recorded. Told him about the air feedback with delay via the TT. He didn't believe me. He also used a tube phono preamp which I think picked up a bit more. I set his needle statically in a scrub record we had. Recorded the TT while playing a CD. You could hear the music clearly being picked up by the cartridge on the TT. FR was different, bass heavy and lower in level, but audible. He was miffed his TT was so miserable at isolation. Then we recorded his LP while playing over speakers, and result was to his liking and so he was happy. I said, "if your TT was designed to really ignore air vibration you wouldn't like it as much".


BTW, are you this Frank Dernie?

Common response.
Most of the analogue is best brigade don't realise or accept that what they like about record players is the raft of euphonic colourations they add. The compromises needed to actually cut and then make a playable LP means that the LP itself is pretty compromised, particularly at the frequency extremes before it is even put onto the platter of a record player to be further coloured.
I found that a non-isolated TT comfortably picked up buses driving by the Garrard R&D dept which was on the 4th floor of the factory with a car park between the building and the road.

That is me in the video, yes.

 

[Purité Audio]

Really fascinating film,
Keith

 

[Frank Dernie]

Really fascinating film,
Keith

They videoed a podcast being recorded.

 

[Purité Audio]

And all this time I thought you were just a good granddad!
BW Keith

 

[Frank Dernie]

And all this time I thought you were just a good granddad!
BW Keith

That is all I am now Keith, I retired from F1 at the end of 2009, otherwise I would still not have discovered any audio forums.