Calculated DR scores are all wrong

The calculated DR scores by Audirvana are significantly off compared to the TT Meter. I noticed that peak and the actual final DR score are typically off by a lot. Can someone explain to me how Audirvana is calculating DR scores?

Here is a sample:

Analyzed: Stranger Times / Artist: Vulture Industries

DR Peak RMS Duration Title [codec]

DR6 -0.39 dB -7.24 dB 4:36 01 - Tales of Woe [flac]
DR6 -0.00 dB -7.29 dB 5:20 02 - As the World Burns [flac]
DR6 -0.00 dB -7.84 dB 7:09 03 - Strangers [flac]
DR7 -0.37 dB -8.07 dB 5:33 04 - The Beacon [flac]
DR6 -0.50 dB -7.94 dB 4:16 05 - Something Vile [flac]
DR10 -0.78 dB -13.88 dB 2:25 06 - My Body, My Blood [flac]
DR6 -0.47 dB -7.71 dB 5:01 07 - Gentle Touch of a Killer [flac]
DR6 -0.36 dB -7.96 dB 5:30 08 - Screaming Reflections [flac]
DR6 -0.41 dB -8.16 dB 5:19 09 - Midnight Draws Near [flac]

Number of files: 9
Official DR value: DR7

Sampling rate: 44100 Hz
Average bitrate: 974kbs
Bits per sample: 16 bit

Dr14 T.meter 1.0.14

But Audrivana claims:

5, 5, 6, 6, 6, 10, 6, 5, 6 respectively.

For “Something Vile” Audirvana claims the peak is -10.4dB? Seems way off to me. All files are in FLAC format (obvious from the above output).

I have seen differences also between A+ and JRiver…
My guess is that A+ is going for the lower average of calculation instead of going at the upper average of numbers like JRiver.
Like at 5.5 average number, A+ will going for 5 and JRiver for 6… my 2 cents :slight_smile:

The difference is indeed in the rounding of this rough dynamic range computation.
This is why I also compute the one used by mastering engineers, the industry standard EBU R128. It is more accurate, especially in its ear perceived loudness filter, and also by removing the extreme outlier values that can bias the resulting value.
BTW, the -10.4dB in your example is not a peak value, but the volume change to achieve a loudness of -14dB when volume levelling.
Peak value is a decimal value between 0 and 1.
Loudness dynamics is a positive dB value.

1 Like

I appreciate the quick response.

However, I think in terms of the TT, there is something to be said about compatibility with existing scores. Folks who look at them are more than likely comparing them to numbers in the Unofficial DR Database and you coming up with your own rounding heuristic is going to confuse everyone! :frowning:

I would prefer the values to be inline with what the TT Meter spits out or at least very, very close to it. And I can’t be the only one who feels this way.

I agree the EBU R128 is more of an industry standard, but I’ll bet dollars to donuts that most Audirvana users are more familiar with the TT Meter scores more than anything else.

Thanks for the explanation about the peak. I know its a value between zero and one, that’s why I was confused by your output! :slight_smile:

Since i play files on A+, i do the DR checking first with A+.
Then i open the same CD with JRiver and redo the DR calculation, i see the changes.
Then the files have the JRiver numbers into A+ database :slight_smile:

Hello Damien,

I am one of the co–founders of MAAT, and am on the board of the Pleasurize Music Foundation. We are the keepers of DR Dynamic Range as expressed in the original TT DR Meter and now the DR range of products from MAAT.

We would love to work with you to incorporate official DRi into Audirvana. For several reasons, in this use case DRi (integrated rather than estimated DR) is a better metric than R 128, especially for pop music.

BTW, JRiver and foobar2000 are not licensed for DRi, and their measurements are not accurate and comparable to those in the DR database.