Is the dynamic range of CD sufficient?

[dollars2donuts]

While a discussion of math is all good, I also think it would be good to support robust investigations of how 16 bit, 44.1k compares to higher resolution music in actual controlled listening environments. While many are critical of the Meyer & Moran study, the best way to respond to it is with a study at least as robust.

Mark Waldrep appears to be trying to organize such a test per comments here: http://www.realhd-audio.com/?p=2147

Supporting Mark's efforts to conduct a robust test (financially, equipment loans, ears, etc) seems like a good thing to clarify under what conditions 16 bit 44.1k playback sounds better/the same/worse than higher resolution music (if that's his goal if I read his comments accurately). I think everyone wins if such a robust test is conducted because we all learn something from it, regardless of where people stand without such a test.

 

[LL21]

Well, first my room has a 24db noise floor. I know people with 20DB noise floor rooms.
A live Symphony can cross 120db easily for short period of times (in big venus).

120db-20db=100db vs the 96DB of a CD.

BTW:
The dynamic range of human hearing is roughly 140 dB, in my opinion this is the DR value we should strive for.

It seemed by the Ken Rockwell quote above, the de facto range of CD is about 100db...? Plus, the key is what is the recording venue noisefloor...with an orchestra of 80 people, I doubt its 20db.

 

[esldude]

Just some empirical data from a 24 bit digital recording. Done with a small number of microphones, and minimal processing. 88 khz/24 bit.

I found a quiet passage without music. Not a digital zero, and may have been only electronics other than mikes feeding it for that period. Though I doubt it.

Grabbing 100,000 samples at this quietest point, L 388 samples and R 585 samples below -120 db. I doubt anyone can put together a real system more than that at the speakers. L 4235 samples and R 8209 samples below -96 db. L 8900 samples and R 18,008 samples below -90 db. This really is not much out of 100,000 samples. Peak levels were in the mid -50 db range with the average around -68 or -69 db.

Same recording, in a quieter portion with music, in this case solo cello I grabbed 350,000 samples. L 35 samples and R 30 samples below -120 db. L 580 samples and R 473 samples below -96 db. Or about 1.5 samples below -96 db every 1000 samples. None of these samples were contiguous. Would you hear that differently if all samples below -96db were dithered and brought up to -96 db ? I don't know, but I find it doubtful. I also would think the difference even if audible is going to be really, really small.

I also grabbed an area with several instruments playing, but among the quietest areas with that. Out of 100,000 samples 2 in each channel were right at -120 db. and L 20 samples R 26 samples were below -96 db.

In one of the more energetic regions I grabbed 200,000 samples. None were as low as -120 db. L 26 samples and R 15 samples were below -96 db. Peak levels were L -5.6 db and R -6.2db. And this a recording of acoustic instruments in a quiet hall, not a rock recording by any means. Only about 1 sample in every 10,000 would test the limits of redbook CD. If dithered to 16 bit levels, seems unlikely you would notice enough difference to quibble over or worry about.

 

[Al M.]

I actually wrote an article on this very topic . Here it is: http://www.madronadigital.com/Library/RoomDynamicRange.html

Quick message is that ambient noise measurements are not correct basis for the lower floor. We need to perform a spectrum analysis and match it to our hearing.

Hi Amir,

thank you very much for that article of yours, I found it very elucidating. It certainly will lead anyone who reads it (everyone here should) to rethink the subject of noise floor. Yet I have a few issues with the observations.

Quoting Fielder, you cite measurements of 130 dB for live music peaks. Yet Fielder leaves out the frequency distributions that are at these peaks. If he analyzes frequencies for noise floor, he needs to do the same for peaks, in order to be scientifically rigorous and coherent in his argumentation. Certainly, measuring this would be much more difficult than measurements of the noise floor, but it should be done nonetheless. I am wondering, for example, which frequencies are the ones contributing the most to the maximum of 118 dB orchestral peaks. As the Galen Audio website states (I cannot verify, but it is an interesting point): "One-third of the total power of a 75-piece orchestra comes from the bass drum."

If the bass frequencies are contributing to SPL the most at orchestral peaks (my intuition tells me no, but it really would have to be measured), then the higher noise floor at low frequencies from Fielder's analysis comes into play.

Furthermore, when you quote 130 dB for live music peaks from Fielder's study, this is misleading since it leaves out important details.

(Fielder's study can be found here:
http://www.zainea.com/Dynamic range.htm)

The highest peaks of 129 dB in Fielder's study were from rock/pop concerts. But this is amplification through usually bad-sounding PA systems, and thus irrelevant for purposes of discussing high-end sound reproduction. When it comes to jazz, peaks were at 127 dB. But such loud levels are reasonably possible only in small clubs and at really close distance, and the audience present in such a small venue will ensure that even a noise floor of 40 dB is virtually impossible. More realistically, the noise floor will be around 50-60 dB minimum. This would leave a dynamic range of maximum 80 dB. And when it comes to home reproduction of such music, who in their right mind would be so crazy to shower themselves with such insane, and ear-damaging, SPL in the absence of a live audience atmosphere? Classical percussion at 122 dB as measured by Fielder is probably bass rich, which would again play into a high SPL background of low frequencies.

For the dynamic range of CD this probably leaves as the only realistic problem orchestral music, with peaks at maximum 118 dB. Yet that was the highest measured value, and in others studies cited in Fielder's paper the values were lower. Again, these must be complete exceptions, usually it does not get that loud. I once measured a live concert of a pretty loud piece (Bartok, Concerto for Orchestra) and I was 'disappointed' that my SPL meter peaked at 92 dB. I sat in the fourth row from the stage, thus pretty close -- granted, perhaps too close since a lot of the sound may literally have washed over my head and further back it might have been somewhat louder, but nonetheless. On other occasions in other venues it did become subjectively louder, but I am sure it cannot have been that much louder (I cannot imagine more than 110 dB, which would be really a heck of a lot louder!).

Yet even if we take 118 dB as the most extreme peaks, there is still the issue of noise floor:

It seemed by the Ken Rockwell quote above, the de facto range of CD is about 100db...? Plus, the key is what is the recording venue noisefloor...with an orchestra of 80 people, I doubt its 20db.

LL21 makes an excellent observation here, and the noise floor caused by the presence of the musicians even in an otherwise empty concert hall will not be restricted to just low frequencies. So the real dynamic range required for this can barely be more than 90-95 dB either.

Furthermore, who would want to listen at home at levels more than 100 dB anyway? As a routine listening level this would be ear-damaging and thus definitely not recommended. As I mentioned in my opening post, I never listen above 95-97 dB, and even at that level only in rare cases, and for short periods (I follow NIOSH recommendations). After 2 minutes of the final brass chorale of Bruckner's Fifth Symphony at that level, reproduced without appreciable distortion, I already feel pressure in my ears. And in case I listen to it, it is then always the piece that I end my listening session of the day with, since I do not want to unduly stress my ears further.

As Tim puts it well:

Well the range of a CD is from lower than I can hear above the noise in the quietest room in my house, to louder than I'd ever care to listen so, of course it is sufficient.

Tim

So yes, while I might somewhat agree that the dynamic range of CD is "on the edge" for exceptional cases, in 99.8 % or more of all real-world cases it is sufficient -- and safely so.

Yet you seem to imply that the dynamic range of CD is insufficient on a routine basis when you say:
"So it turns out we need high resolution audio (i.e. > 16 bits) after all if we want to make sure our distribution channel, i.e. recorded digital samples, does not add more noise than the rest of the chain. No cassette decks may apply."

In this case, I would disagree.

Al

 

[Al M.]

Just some empirical data from a 24 bit digital recording. Done with a small number of microphones, and minimal processing. 88 khz/24 bit.

[...]

Only about 1 sample in every 10,000 would test the limits of redbook CD. If dithered to 16 bit levels, seems unlikely you would notice enough difference to quibble over or worry about.

Thanks, very interesting.

 

[LL21]

Just some empirical data from a 24 bit digital recording. Done with a small number of microphones....
Same recording, in a quieter portion with music, in this case solo cello I grabbed 350,000 samples. L 35 samples and R 30 samples below -120 db. L 580 samples and R 473 samples below -96 db. Or about 1.5 samples below -96 db every 1000 samples. None of these samples were contiguous. Would you hear that differently if all samples below -96db were dithered and brought up to -96 db ? I don't know, but I find it doubtful. I also would think the difference even if audible is going to be really, really small.

I also grabbed an area with several instruments playing, but among the quietest areas with that. Out of 100,000 samples 2 in each channel were right at -120 db. and L 20 samples R 26 samples were below -96 db.

In one of the more energetic regions I grabbed 200,000 samples. None were as low as -120 db. L 26 samples and R 15 samples were below -96 db. Peak levels were L -5.6 db and R -6.2db. And this a recording of acoustic instruments in a quiet hall, not a rock recording by any means. Only about 1 sample in every 10,000 would test the limits of redbook CD. If dithered to 16 bit levels, seems unlikely you would notice enough difference to quibble over or worry about.

Agree with Al M. Very very interesting and helpful. Thank you!

 

[Al M.]

Any exploration of this question should include the reading the below linked study on the topic, authored by Bob Stuart of Meridian. It was published some number of years ago, I don't recall exactly how many.

https://www.meridian-audio.com/ara/coding2.pdf

Thanks for that. Figures 7 and 10 assume 120 dB and a super-quiet noise floor. For this, see my post above.

 

[Al M.]

In theory, digitally recorded red book is hi resolution.

I am glad you would agree to that. Alas, not all in the high-end community do.

 

[Sonus]

It seemed by the Ken Rockwell quote above, the de facto range of CD is about 100db...? Plus, the key is what is the recording venue noisefloor...with an orchestra of 80 people, I doubt its 20db.

Recording venue noise-floor is been handled by the microphone(s) placement and type.
The way the recording eng does it will capture the sound coming from the instrument and a different microphone captures the venue's 'hall sound'. This will give you very low noise floor.
At least this is what explained to me when I did a tour in a recording studio long time ago.

 

[thedudeabides]

Sufficient for what?

I like this forum but this thread is so subjective, that there is no right answer and often ignite spurious comments that, in the end, are so empirical, theoretical,and meaningless. Which has often lead to insulting, personal attacks, questioning members personal biases with the end result being the thread being closed.

IMHO, a waste of bandwidth and intellectual, speculative babbling.

Of course, the answer (sufficient dynamic range) is in the myriad of details. Pretty simple.

GG

 

[microstrip]

I am glad you would agree to that. Alas, not all in the high-end community do.

Al M.,

Please take no offense with my question, but I would like to know. Did you ever compare true 24 bit recordings of excellent quality with properly truncated 16 bits versions of it in an high resolution high-end system using a SOTA 24 bit a DAC? I should immediately say I did not.

 

[Alan Sircom]

Well, first my room has a 24db noise floor. I know people with 20DB noise floor rooms.
A live Symphony can cross 120db easily for short period of times (in big venus).

120db-20db=100db vs the 96DB of a CD.

BTW:
The dynamic range of human hearing is roughly 140 dB, in my opinion this is the DR value we should strive for.

Er... no.

Fortunately, even though 24-bit recording gives a notional 144dB dynamic range, reproducing this full dynamic range is impossible to achieve with our current technology. If you were using a loudspeaker with a sensitivity of a fairly typical 88dB/W/m, you were listening a fairly typical 3m from the loudspeakers and allowing for a fairly rational 3dB of amplifier headroom at 140dB, you would need an amplifier capable of delivering a fairly meaty 2,845,985W per channel. Or, you could get away with less than 600W per channel if you could design a loudspeaker that had the unfeasibly high sensitivity figure of 125dB/W/m.

I say 'fortunately' here because the human hearing mechanism only has a 140dB dynamic range if you are prepared not to have much hearing left after testing. This 140dB range goes from hearing a pin drop in an anechoic chamber to waving off F-18s on the deck of the Nimitz without ear defenders. Something with an SPL of 140dB will cause near instantaneous hearing damage without protection.

This is not to say High Resolution Audio is a bad thing for music, but many of the explanations as to why it's 'better' in the home (extended dynamic range or frequency response) don't stack up, and the one that might just stack up (less brickwall mastering, because of the expectations of the intended audience) seems to go unnoticed.

 

[LL21]

Er... no.

Fortunately, even though 24-bit recording gives a notional 144dB dynamic range, reproducing this full dynamic range is impossible to achieve with our current technology. If you were using a loudspeaker with a sensitivity of a fairly typical 88dB/W/m, you were listening a fairly typical 3m from the loudspeakers and allowing for a fairly rational 3dB of amplifier headroom at 140dB, you would need an amplifier capable of delivering a fairly meaty 2,845,985W per channel. Or, you could get away with less than 600W per channel if you could design a loudspeaker that had the unfeasibly high sensitivity figure of 125dB/W/m.

I say 'fortunately' here because the human hearing mechanism only has a 140dB dynamic range if you are prepared not to have much hearing left after testing. This 140dB range goes from hearing a pin drop in an anechoic chamber to waving off F-18s on the deck of the Nimitz without ear defenders. Something with an SPL of 140dB will cause near instantaneous hearing damage without protection.

This is not to say High Resolution Audio is a bad thing for music, but many of the explanations as to why it's 'better' in the home (extended dynamic range or frequency response) don't stack up, and the one that might just stack up (less brickwall mastering, because of the expectations of the intended audience) seems to go unnoticed.

Thanks for weighing in here. Interesting read.

 

[Phelonious Ponk]

Er... no.

Fortunately, even though 24-bit recording gives a notional 144dB dynamic range, reproducing this full dynamic range is impossible to achieve with our current technology. If you were using a loudspeaker with a sensitivity of a fairly typical 88dB/W/m, you were listening a fairly typical 3m from the loudspeakers and allowing for a fairly rational 3dB of amplifier headroom at 140dB, you would need an amplifier capable of delivering a fairly meaty 2,845,985W per channel. Or, you could get away with less than 600W per channel if you could design a loudspeaker that had the unfeasibly high sensitivity figure of 125dB/W/m.

I say 'fortunately' here because the human hearing mechanism only has a 140dB dynamic range if you are prepared not to have much hearing left after testing. This 140dB range goes from hearing a pin drop in an anechoic chamber to waving off F-18s on the deck of the Nimitz without ear defenders. Something with an SPL of 140dB will cause near instantaneous hearing damage without protection.

This is not to say High Resolution Audio is a bad thing for music, but many of the explanations as to why it's 'better' in
the home (extended dynamic range or frequency response) don't stack up, and the one that might just stack up (less brickwall mastering, because of the expectations of the intended audience) seems to go unnoticed.

Tim

 

[Al M.]

Al M.,

Please take no offense with my question, but I would like to know. Did you ever compare true 24 bit recordings of excellent quality with properly truncated 16 bits versions of it in an high resolution high-end system using a SOTA 24 bit a DAC? I should immediately say I did not.

I did not either, would be fun to know. However, I suppose there we would again be dealing more with (current) technical implementations of media, rather than their theoretical limits.

I mean I guess if you would have played to some audiophile in 1988 who was ranting against the 'woefully inadequate' theoretical limits of CD, based on his/her experience then, a SOTA CD system of 2014 they would have literally fallen off their chair.

 

[GaryProtein]

Well, first my room has a 24db noise floor. I know people with 20DB noise floor rooms.
A live Symphony can cross 120db easily for short period of times (in big venus).

120db-20db=100db vs the 96DB of a CD.

BTW:
The dynamic range of human hearing is roughly 140 dB, in my opinion this is the DR value we should strive for.

The question really is can a CD capture the full range of a performance, not the full dynamic range of sounds in the universe.

Once you hear a 140 dB sound, your subsequent threshold of hearing probably starts around 50 dB if you're lucky.

The dynamic range of a CD is plenty to capture any musical performance.

 

[thedudeabides]

Al M and others have contributed very useful information. Were a lot of post deleted in this thread, I thought I had posted here last night?

Not on this thread. Tim has shown admirable restraint.

Very useful information?

Given all the variables involved, what, pray tell, is useful within the context of connecting emotionally with the music?

That's the end goal, no?

GG

 

[JackD201]

Looking through the moderator panel, no posts have been deleted on this thread.

 

[Al M.]

well, not totally the end goal for some of us, like me, who seeks always better understanding of the technical side of the hobby. I'm weird that way and its from too much exposure to vinyl fumes...or make that soldering fumes...

No that's not weird, it's normal. In the current Trinity DAC thread (Forum: Digital) for example there is a mix going on of subjective appreciation and technical discussion, how it's implemented and why it's sounds so good. Both are normal. Just like in audio reviews that highlight both the technical side and the listening side.

Or would you burn all your Stereophile and TAS issues because of all the tech talk and measurements in there? I think not.

 

[Phelonious Ponk]

Not on this thread. Tim has shown admirable restraint.

Very useful information?

Given all the variables involved, what, pray tell, is useful within the context of connecting emotionally with the music?

That's the end goal, no?

GG

. Is that the end goal? Are you sure? Because if the end goal of discussion on WBF is connecting with the music, we'd spend at least half our time here talking about recordings, the other half talking about music. An MP3 of a really good recording, playing on a car stereo, is more involving (in terms of reproduction; great music can suck you in through a cheap jam box with a little focus) than a bad one through a state of the art system. That's the real garbage in/garbage out, apologies to Lynn.

Back on topic, how is a medium's ability to reproduce the dynamic range of recordings "useful within the context of connecting emotionally with the music?" Really?

Tim