DR : Dynamic Range meter and its limitations

[Eirikur]

As already posted and (briefly) discussed on another thread, DR measurements of CDs versus vinyl records can be misleading, as convincingly demonstrated by Ian Shepherd using his own work

Ian further investigated what might cause the difference in DR

In the video, I mention that I can hear a change in the image. More specifically, I hear it as narrower, on the vinyl. This is quite normal - there is always a certain amount of cross-talk between the left and right channels, and it's also quite common for the cutting engineer to choose to narrow the stereo image in the bass region, or even overall, as I mentioned above. I have no idea if that was done in this case - the image on the CD is quite wide, so it's perfectly possible.

Either way, I tried the same thing on the digital file, i.e. narrowing the stereo image slightly. And lo and behold, the measured DR value increased to... DR12! Looking very similar on the meter to what we see on the vinyl.

He continues to elaborate with some conjecture

So, it may be as simple as that. Either a tweak during the cut, or just normal cross-talk, is sufficient to alter the wave-forms so that their peak level increases enough to account for the "extra" DR points.
And before anybody asks, no this isn't a "real" increase in dynamics, it's just a side-effect of the audio processing.

Now to the really important part

The channel cross-talk I mentioned will certainly play some part in what we're seeing, even if the engineer also made a tweak. And it happens on every vinyl release, and just like the EQ, it depends on the playback system.

In other words, the same record played on different decks may exhibit a different degree of image width, and measure a different DR value as a result - which is a point that's been made several times in this thread already. What's interesting to me is that this change alone is sufficient to cause the spurious DR boost, and it will happen to some extent whenever a piece of vinyl is played.

OK, so far I'm just condensing Ian's findings.

Something I and many others already noticed is that many CDs from the 80s have much "better" numbers in the DR loudness database compared to most if not all remixes/remasters, for example one that I happen to own

From this you might easily conclude that the original 1986 CD is miles ahead of all the remasters... but is it?

As it turns out, this CD was mastered with pre-emphasis applied, meaning that proper playback would require applying de-emphasis.

As an avid user of foobar2000 I installed a component to do just that.

Next I made a comparison, both with DR and ReplayGain to see the difference (sit down or hold on to something)

What we can clearly see is that although the CUE sheet properly states FLAGS PRE for all tracks, it is not applied when calculating DR or ReplayGain. This sheds a new light on the entire DR database in my view.

Thoughts and elaborations are appreciated!

 

[Eirikur]

PS: de-emphasis was not applied to any of the tracks anyway until I manually added a new property "PRE_EMPHASIS" with value "yes". This could be a bug/oversight in CUE sheet processing in foobar when using DSP components, but it doesn't alter the previous results. No ABX needed to hear the difference between the raw and de-emphasized CD audio, so I'm certain that the version on top was properly processed on playback at least.

 

[maty]

Look at only without Live:

http://dr.loudness-war.info/album/list/year?artist=Dire+Straits&album=Dire+Straits

http://dr.loudness-war.info/album/list/dr/desc?artist=Dire+Straits&album=Dire+Straits

The best CD: http://dr.loudness-war.info/album/view/71169

Features the high-fidelity SHM-CD format (compatible with standard CD player) and the mastering that is used for its Platinum SHM-CD edition. Features HR (High Resolution) cutting. Mini LP with obi replica. Comes with lyrics and a description. HR cutting from the DSD master which was newly flat transferred from UK original analogue master tapes in 2013.

 

[Eirikur]

HR cutting from the DSD master which was newly flat transferred from UK original analogue master tapes in 2013

Do you have both to compare?
I don't have either CD, but I would not make a decision based on the DR database, that is my entire point!

Other considerations: why would the new one be better than the original transfer in 1983 when the tapes were fresh? Does better digital equipment make up for fidelity loss in the 30 year old tapes?
It has also been filtered by DSD encoding/decoding, so what is still flat about this CD transfer?

 

[maty]

ALL DR I had published here and in others sites are made without ReplayGain. I have many same excellent recordings in different formats (CD, SACD, Vinyl, Digital download and, in less number, cassette and R2R) and pressings. With classic and jazz from trusted brands the DR hardly varies over the years.

The DR is just one more indicative. If it is low it is a sign that things have been done very badly at the end of the music processing, when it is made available to the consumer. It is like SINAD, more information is needed to make a decision based only on measurements.

The problem is that digitalization was done badly in the eighties and more years, especially with commercial music. Even Deutsche Gramophon recordings of those years usually have bad sound, shrill, artificial and false.

 

[CTRLM]

Yes the DR database will show an inflated DR with a pre-emphasis disc but there is a comprehensive list of CDs that came out with pre-emphasis here that can give an indication of whether or not a disc on the database is showing a misleading figure:

https://forums.stevehoffman.tv/threads/pre-emphasis-list.121188/page-45

 

[Sal1950]

Other considerations: why would the new one be better than the original transfer in 1983 when the tapes were fresh? Does better digital equipment make up for fidelity loss in the 30 year old tapes?

A very good point, I wonder what the tape loss really adds up too? I know the high end gets wiped off with oxide loss from repeated playings. Everyone wants to hear those masters.

ALL DR I had published here and in others sites are made without ReplayGain.

Well dang maty, I would hope so, otherwise what would be the point?

The problem is that digitalization was done badly in the eighties and more years, especially with commercial music. Even Deutsche Gramophon recordings of those years usually have bad sound, shrill, artificial and false.

More myth than truth there IMHO

 

[Eirikur]

ALL DR I had published here and in others sites are made without ReplayGain.

I'm assuming you mean de-emphasis? I hope you did apply RIAA de-emphasis with all those vinyl recordings...

The DR is just one more indicative. If it is low it is a sign that things have been done very badly at the end of the music processing, when it is made available to the consumer. It is like SINAD, more information is needed to make a decision based only on measurements.

Indicative indeed, and obviously also easily manipulated and wrong in some cases, as demonstrated.

The problem is that digitization was done badly in the eighties and more years

That is not a useful statement without specifics. Some were done poorly, others quite OK.

 

[Eirikur]

Yes the DR database will show an inflated DR with a pre-emphasis disc but there is a comprehensive list of CDs that came out with pre-emphasis here that can give an indication of whether or not a disc on the database is showing a misleading figure:

https://forums.stevehoffman.tv/threads/pre-emphasis-list.121188/page-45

Thanks for the link - not so comprehensive though... 59 pages!

 

[DonH56]

The problem is that digitalization was done badly in the eighties and more years, especially with commercial music.

Doug Sax and Lincoln Mayorga would be sorry to hear that (Sheffield Lab). I only met them briefly, but the late conductor of our local symphony knew Lincoln well and we shared some stories after one of the concerts a few years ago.

 

[pozz]

@j_j gave an interesting talk on the dynamics, including demonstrating the measurement of certain songs using a custom Octave program:

Bottom line is that DR is not referenced to perceived loudness, so you'll always get these divergences.

 

[Eirikur]

Thanks @pozz, there is a preceding lecture to this with unfortunately poor sound quality (!)