The DR Meter & Multichannel Files

[JediJoker]

DR14 for the Atmos mix.

TT/MAAT DRMeter readings are pretty useless for more than two channels.

 

[himey]

TT/MAAT

?

 

[JediJoker]

?

The old version and the new version. All programs that calculate DR values use one or the other of these algorithms. Neither were created to handle more than two channels with any degree of comparability. In fact, the first party tools only work with mono and stereo.

 

[himey]

The old version and the new version. All programs that calculate DR values use one or the other of these algorithms. Neither were created to handle more than two channels with any degree of comparability. In fact, the first party tools only work with mono and stereo.

I haven't heard this. I have seen it used for a long time on 5.1 files for comparing one 5.1 file to another. I know @HomerJAU uses the readings in his multichannel software.

 

[JediJoker]

I haven't heard this. I have seen it used for a long time on 5.1 files for comparing one 5.1 file to another. I know @HomerJAU uses the readings in his multichannel software.

foobar2000's implementation of the TT Meter allows for more than two channels, but that doesn't mean it gives useful readings beyond stereo. MMH's implementation attempts to mitigate the incompatibility using noise gates (I assume) for any additional channels, but it's still not a use of the algorithm intended by the developer. All mono and stereo DR readings are comparable to one another; readings for other channel configurations are not.

 

[himey]

foobar2000's implementation of the TT Meter allows for more than two channels, but that doesn't mean it gives useful readings beyond stereo. MMH's implementation attempts to mitigate the incompatibility using noise gates (I assume) for any additional channels, but it's still not a use of the algorithm intended by the developer. All mono and stereo DR readings are comparable to one another; readings for other channel configurations are not.

Comparing the same amount of channels vs another mix with the same number of channels would seem useful but who knows. I use it in addition to waveforms.

 

[JediJoker]

Comparing the same amount of channels vs another mix with the same number of channels would seem useful but who knows. I use it in addition to waveforms.

The biggest issue that throws off readings is channels with little information, like LFE. Next is the issue of louder L and R versus other channels, which get averaged out to read higher than the subjective listening experience, as the less dynamic front channels will dominate perception of DR. Unless and until a proper algorithm is developed for multichannel, it's best to avoid DR ratings beyond stereo. They just muddy the waters as far as true loudness is concerned.

 

[himey]

The biggest issue that throws off readings is channels with little information, like LFE. Next is the issue of louder L and R versus other channels, which get averaged out to read higher than the subjective listening experience, as the less dynamic front channels will dominate perception of DR. Unless and until a proper algorithm is developed for multichannel, it's best to avoid DR ratings beyond stereo. They just muddy the waters as far as true loudness is concerned.

I just did a quick test by moving the sub and center channels to the front left and right slots of a 5.1 file and it didn't change the DR reading. It isn't just reading the left and right channels at least.

 

[JediJoker]

I just did a quick test by moving the sub and center channels to the front left and right slots of a 5.1 file and it didn't change the DR reading. It isn't just reading the left and right channels at least.

That's exactly the problem: it is agnostic of which channels are which. All are treated equally, whereas to match our hearing, R and L should be weighted higher.

 

[ManWhoCan]

This problem has been around ever since the DR meter was made. It has never ever been reliable with Mch. More unreliable with Quad because of it being in 5.0 or 5.1 containers and having at least 1 or 2 totally empty channels which when averaged distort the accuracy. I have personally run the DR meter on hunreds of Mch SACDs and there it tends to inflate the Mch reading. For some years now I only rely on the 2ch reading as a guide to the actual DR value of any SACD. With vinyl rips it is also inaccurate, tending to give them around +DR2 because of surface noise etc.. Once one is familiar with these anomalies and compensates for them the DR meter then becomes more useful.

 

[par4ken]

This problem has been around ever since the DR meter was made. It has never ever been reliable with Mch. More unreliable with Quad because of it being in 5.0 or 5.1 containers and having at least 1 or 2 totally empty channels which when averaged distort the accuracy.

The DR meter works fine for Quad in a 4.0 container. I used to always convert my 5.1 files to 4.0 for compatibility with my 4.0 sound cards.
It's sad that the use of a 5.0 or 5.1 container is necessary for use with many/most modern AVR's. I would suggest converting to 4.0 then running the DR meter even if you stick with the original 5.1 files.

 

[JediJoker]

The DR meter works fine for Quad in a 4.0 container.

No, it doesn't. The algorithm was only designed for two channels. Just because third parties have made it usable on more than two channels doesn't mean the readings it gives are useful or comparable.

 

[par4ken]

No, it doesn't. The algorithm was only designed for two channels. Just because third parties have made it usable on more than two channels doesn't mean the readings it gives are useful or comparable.

Totally disagree. Quad is just stereo x 2, no reason for it not to work. Adding lfe might skew readings lower, but for four full bandwidth channels what's the problem?

 

[ssully]

I just did a quick test by moving the sub and center channels to the front left and right slots of a 5.1 file and it didn't change the DR reading. It isn't just reading the left and right channels at least.

Try *deleting* the channels -- typically C and LFE -- that typically have far less/far lower -- or zero -- content.

Then you'll see how silly these tools are for multichannel.

As regard 4.0: these tools typically output a 'DR number' for each channel in stereo, and a composite number for the track as a whole. Does it do that for 4 channels, and if so how? And what happens when the rear 2 channels just aren't as full of content as the front 2?

 

[JediJoker]

Totally disagree. Quad is just stereo x 2, no reason for it not to work. Adding lfe might skew readings lower, but for four full bandwidth channels what's the problem?

It's a perceptual algorithm. Perception of dynamic range changes when adding surround channels. The algorithm treats all channels equally, which gives inaccurate DR measurements.

 

[ssully]

Does anyone really know the full details of how TT's and later, MAAT's DR meters work?

 

[JediJoker]

Does anyone really know the full details of how TT's and later, MAAT's DR meters work?

My guess is only Friedemann Tischmeyer knows the full details. Whatever he coded, though, it's quite useful for stereo (and mono). In concert with EBU R128 loudness measurements, DR provides a great picture of how dynamic a release is.

 

[par4ken]

And what happens when the rear 2 channels just aren't as full of content as the front 2?

They always should be!

 

[Electric Moo]

Couldn't you just export the front & rear stereo pairs, then run the DR on them both, which should work as they'd be 2x separate stereo files?

Then list the front & rear DR values. Which might be more useful than a single DR value since it would supply you the front vs rear content too?

Some enterprising programmer could probably wrap that all into a script

 

[JediJoker]

Couldn't you just export the front & rear stereo pairs, then run the DR on them both, which should work as they'd be 2x separate stereo files?

Then list the front & rear DR values. Which might be more useful than a single DR value since it would supply you the front vs rear content too?

Some enterprising programmer could probably wrap that all into a script

I was thinking a binaural render would make the most sense, as it would take psychoacoustics into account. It would also work for any number of discrete channels/objects. Render your surround/immersive to binaural, run it through the DR meter, and there's your accurate and comparable measurement.