IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
How to test/measure the quality of dynamic range compressed audio?, [was “Audio Post processing”/“Dynamic Range compression”: TOS #6]
poochi
post Aug 22 2013, 13:42
Post #1





Group: Members
Posts: 11
Joined: 12-October 12
Member No.: 103802



How do I test / measure the quality of Dynamic range compressed audio ? Are there any standards , or general test procedures .

Thanks a lot smile.gif
Go to the top of the page
+Quote Post
skamp
post Aug 22 2013, 14:11
Post #2





Group: Developer
Posts: 1343
Joined: 4-May 04
From: France
Member No.: 13875



You can use the Dynamic Range Meter for foobar2000.


--------------------
caudec.net
Go to the top of the page
+Quote Post
DVDdoug
post Aug 22 2013, 21:51
Post #3





Group: Members
Posts: 2441
Joined: 24-August 07
From: Silicon Valley
Member No.: 46454



Just be aware that it gets very tricky when to comes to human perception of dynamics. And, it's all about human enjoyment of music!

There are short-term dynamics & long-term dynamics. You could have a song with very-quiet parts and very-loud parts, with the loud parts compressed-to-death. A compressed country or jazz recording where the musicians are playing dynamically and giving each other "space" can sound more dynamic than a heavy metal recording with all of the musicians "playing at 11" full-time. I just don't think you can calculate or measure how dynamic those things sound...

Sometimes people will describe compressed sound as more dynamic! Probably because it's louder and "more exciting" as a 1st impression (before it gets boring and you turn-down the volume wink.gif ).

Or, you can do things that make some peaks bigger and other peaks smaller (all-pass filtering, the vinyl signal path, MP3 encoding), and you'll get a higher measured dynamic range without affecting the dynamics of the sound.
Go to the top of the page
+Quote Post
mjb2006
post Aug 23 2013, 07:50
Post #4





Group: Members
Posts: 706
Joined: 12-May 06
From: Colorado, USA
Member No.: 30694



QUOTE (poochi @ Aug 22 2013, 06:42) *
How do I test / measure the quality of Dynamic range compressed audio ? Are there any standards , or general test procedures .

You can use the suggested DR meter (or similar measuring functions built into general-purpose wave editors) in order to get an objective measurement of the dynamic range, but that's not the same thing as quality, which is subjective. You have to think about what you mean when you say quality. What is that, exactly? Maybe music sounds better to you when it has a moderate amount of dynamic range compression... or maybe it depends on the song and how it was mixed (some instruments may be compressed a lot, others not at all, and maybe that sounds good to you, but doesn't correlate with a particular range on the DR meter)... I mean, who is to say which measurement means better/higher quality to you? smile.gif

That said, once you get to measuring a lot of music, you're probably going to notice some trends in what numbers you prefer...

This post has been edited by mjb2006: Aug 23 2013, 07:52
Go to the top of the page
+Quote Post
poochi
post Aug 27 2013, 11:40
Post #5





Group: Members
Posts: 11
Joined: 12-October 12
Member No.: 103802



QUOTE (skamp @ Aug 22 2013, 14:11) *
You can use the Dynamic Range Meter for foobar2000.

THanks a lot that was useful smile.gif
Go to the top of the page
+Quote Post
stephan_g
post Aug 27 2013, 20:39
Post #6





Group: Members
Posts: 156
Joined: 6-June 10
From: Bavaria
Member No.: 81240



While I do use the DR Meter regularly, be warned that it can serve as no more than an indication of perceived SQ, and a rough one at that. It really depends a lot on the type of music and compressor/limiter (+ settings) used.

All I can tell is that most anything at DR4-5 or under subjectively tends to suck to varying degrees (unless it's some sort of drone stuff), and that modern (oversampling?) compressors seem to be able to squeeze rather decent to good listenability out of DR6-7 (and partly even DR5) with pop or rock music, hardly a given a decade ago. I've heard tracks from the early 2000s with annoying distortion at DR8-9, and modern-day ones that were listenable just fine at DR5. So I'm afraid you're still best off using your ears, as unsatisfying as that may be. Still, DR's not bad for comparing different masters and such.

Incidentally, album DR tends to be more reliable. Sometimes a look at RMS levels (also measured and logged by the DR meter plugin) may also be an indication. I have one (fairly new) album with a few tracks at DR5, where one at -5.6 dB sounds like it was pushed a bit too far even for a modern-day compressor, while nothing bad jumps out at me with those at -7.0 and -7.6 dB.
Go to the top of the page
+Quote Post
stephan_g
post Aug 28 2013, 00:40
Post #7





Group: Members
Posts: 156
Joined: 6-June 10
From: Bavaria
Member No.: 81240



PS: Anyone up for a little DR guessing game?
How do you think does this track fare?
I was a bit surprised...
Go to the top of the page
+Quote Post
Kohlrabi
post Aug 28 2013, 07:59
Post #8





Group: Super Moderator
Posts: 953
Joined: 12-March 05
From: Kiel, Germany
Member No.: 20561



QUOTE (stephan_g @ Aug 28 2013, 01:40) *
PS: Anyone up for a little DR guessing game?
How do you think does this track fare?
I was a bit surprised...
7-8? To me especially the drums sound like they have quite a bit room to breathe.

This post has been edited by Kohlrabi: Aug 28 2013, 08:12


--------------------
Audiophiles live in constant fear of jitter.
Go to the top of the page
+Quote Post
Dynamic
post Aug 28 2013, 12:52
Post #9





Group: Members
Posts: 793
Joined: 17-September 06
Member No.: 35307



Hmm, I'd say it has got lots of short-term dynamics (e.g. drum impacts and other transients are pretty clean) but is very flat in medium term volume level (what would be regarded as dynamics on a musical score). So I think it depends dramatically on how the two are weighted in coming up with the DR measure.


--------------------
Dynamic – the artist formerly known as DickD
Go to the top of the page
+Quote Post
Rescator
post Aug 28 2013, 18:53
Post #10





Group: Members
Posts: 73
Joined: 13-January 09
From: Trondheim
Member No.: 65515



In general dynamic range, or actually I think what you mean is headroom (or also called crest factor I think?) is more indicative of the potential for listening fatigue than actual quality.

The more the dynamics are compressed the less headroom there are the and the closer the music gets to being static/noise (headroom of 0dBFS would probably be just pure static/noise).

I did read someplace (and sadly I do not remember where) that on average a humans prefer music to be within a 30dB range, not sure if this means 30dB headroom, or if it means a RMS of +/- 15dB (30dB range total, sweet spot being in the middle somewhere?)
If the range is larger than 30 then quiet and loud are so far apart that one might feel like reaching for the control at times, but in the other direction, with a too low range (Iggy Pop made a notoriously low dynamic range album) ad it might feel tiring after a while.

Your ears prefers some peaks and valleys, flow and ebb of sound. Songs that feel lively but not tiring usually is around your sweet spot (which one would assume varies a little from person to person).
Not sure if this helps clarify anything, but hopefully it is found interesting at the very least...

I guess technically speaking one could use EBU R128, and a huge survey and try and figure out what LRA (EBU R128 loudness range) people prefer and get a statistical average or mini-range or similar from that I guess, it still would not say anything about quality though, might be interesting for automation systems though, for flagging of audio that needs manual inspection if they are too far off in either direction maybe..


--------------------
"Normality exist in the minds of others, not mine!" - Rescator
Go to the top of the page
+Quote Post
stephan_g
post Aug 29 2013, 14:51
Post #11





Group: Members
Posts: 156
Joined: 6-June 10
From: Bavaria
Member No.: 81240



QUOTE (Kohlrabi @ Aug 28 2013, 08:59) *
7-8? To me especially the drums sound like they have quite a bit room to breathe.

Fasten your seatbelts: DR4, with a -4.3/-4.5 dB RMS average! blink.gif
Must've been a SOTA compressor/limiter, and possibly a compression-friendly mix to boot. The end result certainly isn't hard on the ears, which is quite the feat at DR4.

QUOTE (Rescator @ Aug 28 2013, 19:53) *
I did read someplace (and sadly I do not remember where) that on average a humans prefer music to be within a 30dB range, not sure if this means 30dB headroom, or if it means a RMS of +/- 15dB (30dB range total, sweet spot being in the middle somewhere?)
If the range is larger than 30 then quiet and loud are so far apart that one might feel like reaching for the control at times, but in the other direction, with a too low range (Iggy Pop made a notoriously low dynamic range album) ad it might feel tiring after a while.

Your ears prefers some peaks and valleys, flow and ebb of sound. Songs that feel lively but not tiring usually is around your sweet spot (which one would assume varies a little from person to person).

I'd definitely love to see some hard numbers on this subject, as this vibes with what I'm hearing. Stuff that is overcompressed tends to give me a headache, and seemingly more quickly than the average person. (My hearing apparently also starts complaining at 100-ish dB, which would seem to be substandard as well. My preferred listening volumes in quiet surroundings are ~55-60 dB.)

This post has been edited by stephan_g: Aug 29 2013, 15:30
Go to the top of the page
+Quote Post
TomasPin
post Aug 30 2013, 02:30
Post #12





Group: Members
Posts: 167
Joined: 5-June 13
From: Argentina
Member No.: 108508



QUOTE (stephan_g @ Aug 29 2013, 10:51) *
Fasten your seatbelts: DR4, with a -4.3/-4.5 dB RMS average! blink.gif

I liked the song, got it off iTunes, and that version measures DR6 with -6,9/-7,6 RMS in average... Maybe you have the CD version? Anyway, thanks for the link.


--------------------
A man and his music: http://tubular.net/
Go to the top of the page
+Quote Post
dhromed
post Aug 30 2013, 10:28
Post #13





Group: Members
Posts: 1244
Joined: 16-February 08
From: NL
Member No.: 51347



QUOTE (stephan_g @ Aug 29 2013, 15:51) *
Fasten your seatbelts: DR4, with a -4.3/-4.5 dB RMS average! blink.gif
Must've been a SOTA compressor/limiter, and possibly a compression-friendly mix to boot. The end result certainly isn't hard on the ears, which is quite the feat at DR4.


I think that's more often the case than you think if the singing voice isn't too loud, but rather quiet and spread out like this song. Now try the same with Tegan & Sara.

This post has been edited by dhromed: Aug 30 2013, 10:29
Go to the top of the page
+Quote Post
stephan_g
post Aug 31 2013, 22:42
Post #14





Group: Members
Posts: 156
Joined: 6-June 10
From: Bavaria
Member No.: 81240



QUOTE (TomasPin @ Aug 30 2013, 03:30) *
QUOTE (stephan_g @ Aug 29 2013, 10:51) *
Fasten your seatbelts: DR4, with a -4.3/-4.5 dB RMS average! blink.gif

I liked the song, got it off iTunes, and that version measures DR6 with -6,9/-7,6 RMS in average... Maybe you have the CD version? Anyway, thanks for the link.

No, I downloaded exactly this video in 720p MP4 (which has 192 kbit AAC audio) and had that analyzed, assuming that the relatively high-bitrate encoding wouldn't mess too much with the result. Could've been wrong of course, but a difference of >2 dB in RMS levels is a little too much to make a mere difference in encodes likely.

QUOTE (dhromed @ Aug 30 2013, 11:28) *
I think that's more often the case than you think if the singing voice isn't too loud, but rather quiet and spread out like this song. Now try the same with Tegan & Sara.

Heh. Their latest record doesn't exactly seem to be an audiophile delight indeed (at least that's what my ears are telling me on the tracks I've heard), in spite of ranking a "luxurious" DR6.

But back to the topic at hand:
Another quality indication that has proven decently reliable for me is peak amplitude when converted to MP3 (Vorbis would also do, I guess). If you're seeing multiple tracks of an album peak at >~1.35 for LAME 3.98.4 @ -V 5, it'll probably sound pretty lousy. If no track ever goes beyond 1.2, chances are everything's groovy.

Obviously no rule comes without its exceptions, and so I have a few albums with audible crackling/clipping in spots that nonetheless don't peak far above 1.2 (e.g. Sweet Ride - Best of Belly was somewhat annoying, and Susanne Sundfør's last two albums have isolated crackling in partly unexpected spots that you think someone in the studio should have heard). Annoying clipped beats do not always generate overly high peak levels either, even if they already give me a headache (Heathers' Kingdom to name a recent example, sounds like the beats are running against a literal brick wall).
Some "stinkers" only end up in the grey area - the abomination of a remaster that is Cocteau Twins' Heaven or Las Vegas "only" peaks at 1.33, and the sonically questionable Die Reklamation by Wir Sind Helden maxes out at 1.31.
The "1.35 rule" seems to hold up better, though honestly I don't think The Naked And Famous' Passive Me, Aggressive You is that bad-sounding. But Ladyhawke - check. The Kills' last two albums - check. (I think that's intentional though.) Yeah Yeah Yeahs' It's Blitz - check. Marina and the Diamonds' Family Jewels - check. (This one was just borderline/spotty though; her newer material tends to be even louder and absolutely dreadful sounding.) Said Heathers' Kingdom - check. The Ting Tings' We Started Nothing - check. Clare Maguire's Light After Dark - check. (Now those are two real stinkers.) I didn't even buy La Roux.

Admittedly any rule that requires you to make a lossy encode from a lossless source isn't of too much value (since you've already bought the CD then), unless someone were to set up a public database à la DR Database. And then it would be more sensible to find a more accurate measure of Subjective Annoyingness Level beforehand, even if right now that seems to be about as easy to find as a black cat in the dark.

(Speaking of "Black Cat", that Ladytron album Velocifero was quite unbearable. I don't feel like re-ripping the darn thing just to find out whether it matches the 1.35@V5 rule, though I bet it does.)
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 17th April 2014 - 09:28