More data, more DR?

If the DR you are measuring is just the DR of the largest components in the signals (as the values you quote imply), then there is no reason why the degree of compression would have any correlation to the reported DR.

This is so, as DR of the largest components takes no account of the full internal DR of all the myriad other little minor components that make up the majority of the data.

The DR meter is the one supplied by Dynamic Range Database.

It's not the degree of compression. It's the different mastering of the same track on two different albums. I would have expected that the version with the greater DR would have had the larger file size when reduced to FLAC.

Flac has a range of compression options from 0 to 8, which will affect/determine file size.  This of course has nothing to do with DR compression.....

As said the dynamic range of the sampled audio bears no correlation to the number of audio samples and therefore the length of the ripped file.. Data compression and  audio dynamic range compression are two quite separate things.

Simon

I've always taken the dynamic range to be a measure of the relative difference between the "quietest" and "loudest" parts of a recording. Seems logical that reducing the DR by an emphasis on increased overall volume would require more sound/information/data in the recording. 

joe,

Firstly, considering DR to be 'a measure of the relative difference between the "quietest" and "loudest" parts of a recording' isn't as simple as it seems...  How short a time can a passage be for it to qualify as the "quietest part"?  If there's a pregnant pause with no music playing, is the DR infinite?

While the assumption that larger DR requires more data seems intuitive, that's not a safe assumption when dealing with the maths of compression algorithms. A signal that is a simple sinewave slowly going from -40dB to 0dB then back to -40dB will have a DR of 40 and yet will compress very well (i.e. give a small file size).  The complexity of the signal (numbers of harmonically unrelated sound components and harmonic richness) increases the mathematical complexity of the signal so that the FLAC compression can't be as efficient mathematically and this make for larger files.  Conversely harmonically simple signals make for smaller files.

And the above conditions don't even consider bit depth or sample rate.

In Bananahead's case I think the differences in mastering are most likely the reason behind the difference in file size.  And the higher 'superficial' DR measured by the 'Dynamic range Database' doesn't necessarily indicate the better track, as, mathematically, there's more information in the track with the lower 'superficial' DR.

as a muse - actually the information rate - or entropy - will be the same regardless of the degree of dynamic range. Surely the dynamics would need to be reduced to the point that the range is below the quantization level of the data before there would be a loss of information.. loss of information and remapping or re symboling of the information is not necessarily the same as loss of information

Simon

Simon, correct.

If there is a component within the signal (say a harmonic at -60dB) then there would be loss of information when that component of the signal is pushed down into the quantization level.  This is why the DR of the largest signal components is almost irrelevant.

agreed - now back to work...

In Bananahead's case I think the differences in mastering are most likely the reason behind the difference in file size.  And the higher 'superficial' DR measured by the 'Dynamic range Database' doesn't necessarily indicate the better track, as, mathematically, there's more information in the track with the lower 'superficial' DR.

@ Huge, how do you conclude there is more info in the track you mention ?  By the file size ?  As I said above that may be largely dictated by the flac compression level which the label has chosen to apply.  More compression doesn't affect the data content.

As to how the database does its measurements I honestly have no idea, yet I have found their findings consistent with my own subjective listening assessments for recordings I have, and which I've also referenced in their DB.

To the point about the "pregnant pause", is this not a zero signal, and therefore not part of the measurement ?  It should be the difference between the quietest musical signal, and the loudest - no ?  And surely never infinite -  rather, a specific dB range ?  Albeit if logarithms are at work, that may change things and create infinite scenarios -  beyond me I'm afraid.

Karlos, I have urgent stuff to do, but will reply later.

Cheers Huge !  Its an interesting issue.....

Karlos,

Every FLAC download I've had (so far at least!) has been compressed to FLAC compression level 5; it's possible that one of Bananahead's downloads was compressed more than the other - and that would alter the file size slightly.  Given my experience of FLACs being compression level 5, I am assuming that when I make the observation on the amount of information.  So as you point out: If the assumption doesn't hold, then neither does the conclusion.  However, neither is it correct to assume that the tracks have the SAME amount of information: In fact we know these two mixes are different as they give different results when looking at the overall 'superficial' DR measurement.

A pregnant pause is part of the composed music (it's one or more rests), and therefore should be part of the overall 'superficial' DR measurement of the piece.  Therefore it should be included in the DR of the piece.  In fact due to background noise, it won't actually be zero sound in the recording, but zero sound is what the composer actually wrote!  

I'm actually pointing out that the superficial measurement of DR as performed by the Dynamic Range Database is only valid for detecting compression on otherwise identical recordings, it doesn't actually indicate musical quality as that's a much more complex thing.

In your case, the tracks you have where greater compression has been applied almost certainly also have other differences in mastering that do not fit with your tastes.  It's likely that these arise from differences in preferred style differentiating those sound engineers who use greater compression from those who do not.  For instance those who use less compression may also generally prefer a richer mix where minor acoustic elements are allowed more prominence in the final mix, whereas those who use greater compression may think that these same elements are distracting from the main theme and mix them at a much lower level.

What surprised me was that there are three other tracks on the latest Counter Culture album that I also have from their original albums and all of the other three are within .1 MB of each other. I wonder if the record company supplied a different master for the Ezra Furman track (it is the same length). Maybe I should try with my ears.

Apologies Bananahead, I somehow overlooked that you'd done the rips, and my response was obviously based on thinking they were downloads.   Oops  :-)

So that may indeed suggest they are different masterings, and Huge's explanation above could have a bearing on the files size differences, ie other tweaks the mastering engineer has made to the different components of the recording (over & above the DR compression level).

It still remains an interesting finding, and perhaps counter intuitive as you said......

Hi Bananahead,

I also missed the point that you'd ripped different CDs to FLAC, I do try to read things carefully, but sometimes I don't succeed well enough - apologies.  However a useful discussion did come out of the error; every cloud...

Huge,

My layman's interpretation of DR comes from reading the likes of this ...

"What is dynamic range? Dynamic range is the difference between the softest and loudest sounds we can hear. Or, to put it another way, the difference between the softest and loudest sounds in a recording. Dynamic range is measured in decibels (dB)."

... from the CD Mastering Services website, a member of the Audio Engineering Society.

This notion is often promulgated on sites discussing DR, some within the industry. I'd be interested in what you or other engineers think; is the use of DR being widely bastardized or is it a case where engineers are trying to put things in more accessible terms for the mainstream reader. Maybe I'm completely off base?

Thanks for the discussion everyone.

I think that most of us would agree that more DR is a good thing ( I just bought some B-52's albums from Qobuz - DOWNLOADS  - because they have highish DR measurements and they sound lovely ).

Most of us would expect "better" to equate to larger file sizes but maybe it's not always true. At least with compression.

Agreed, good discussion. Safe to say that dynamic range compression and file compression are unrelated

Joerand, I am also a member of the AES, and I think your quoted summary is fine. I think however in this thread there is a mixture or possible confusion between audio and data compression. The two in the genereal case are not directly linked.

BTW I wouldn't get hung up to the fact audio dynamic range is measured in decibels, the relative comparison of one thing to another as opposed to the measurement of a quantity of absolute values is usually measured in decibels.

From an engineering point of viiew in audio replay, having too much dynamic range can be a bad thing, in terms of effective communication and so there are standards used in radio and TV broadcasting for example  to limit the dynamic range. Clearly in hifi we often like a lot of dynamic range, but even here I suggest its relative range within a mix as opposed to absolute dynamic range.. otherwise we would be constantly turning our volumes up and down.

Simon

Thanks Simon and I think I get it. I see no relation between data compression and DR compression. I had initially thought that more relative volume might require more data, but have readjusted my thinking. I understand the desire to standardize volume levels for broadcast, streaming, and certainly within a given album. The problem for me goes into the whole loudness war thing. Increasing compression and squeezing DR to produce a stronger signal at the user's end, the result being greater volume with reduced fidelity and differences in impact or subtlety. Hi-fi be damned, give the masses what they want, etc. Archetypical audiophile rant and a digression here, but so it goes.

Hi joerand,

I agree with the definition of DR.

However there are two meanings for that definition.

1  the 'Overall DR'

This is the obvious one; take a track and measure the total weighted amplitude against time - usually dB(A) weighted.  OK, that that sounds easy and reliable.  Hang on though...  It's easy to define the loudest point (greatest weighted amplitude), how do you determine the softest?  That's not so easy, you need to determine how long a quiet part lasts for

10 seconds - OK count it,

1 second - Hmm not so sure, I think you should probably count it

1/10 second - Hmm I think you should probably not count it

1/100th second - ?

So, exactly where do you stop counting bits of the signal as being the quietest part?  How you decide this will have a major effect on the value your algorithm gives as the 'DR' of the signal.

2 the 'Inner DR'

This is the difference between the largest total weighted amplitude - again usually dB(A) - compared to the smallest musical component in the signal.  So what is the smallest musical component in the signal?  Here are two examples of small but significant components.

Firstly take an orchestra, they play a FF tutti - it's very loud; but then there's a reflective part during which the cellos and basses provide a dark mood even though it's not particularly quiet of itself, however in this passage a triangle is struck.  Even though overall the passage isn't that quiet, the triangle can still be heard clearly, although it is a very small signal buried in a much larger one.

Second, consider just a solo cello, when bowed strongly and evenly, the overall level of the signal doesn't change much (the 'Overall DR' is only a few dB).  However at the same time there are a lot of higher frequency harmonics in the sound of a cello, some going into the kHz range.  These harmonics are at a much lower level than the fundamental note, but are very important as they form the characteristics that make a cello sound as it does.  This one instrument has a significantly large dynamic range all on it's own (it can easily exceed 30dB!).

Huge,

Sorry for the late reply. I appreciate your response and you always seem to provide me with some useful insights that help bridge the gap between the technical side and a layman's view of things. Cheers!