Skip to main content
Topic: Understanding brickwalled tracks vs dynamic range (R128) (Read 854 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Understanding brickwalled tracks vs dynamic range (R128)

I have two copies of a track from Elton John's - Elton John 1970 album.

Elton John (1970) [1985 - DJM / 827 689-2 / West Germany / CD]


and

Elton John [TIDAL / MQA]


Starting with the 1985 CD pressing - track 10 - The King Must Die you can see it's not been brickwalled:


R128 dynamic range is 15.3 LU.


Now looking at the TIDAL version - same track you can see it's pretty compressed:


R128 dynamic range is 16.0 LU.


I'm confused. I thought a 'brickwalled' track had a limiter applied during mastering to remove transient peaks so that the overall volume can be increased hence reducing dynamic range. How can a heavily brickwalled track in this case have a larger dynamic range than one which has clearly not been volume limited?


Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #1
How can a heavily brickwalled track in this case have a larger dynamic range than one which has clearly not been volume limited?
Obviously, because it really has larger dynamic range - difference between lowest and highest peak. Don't confuse real dynamic range with crest factor. For example, what you can find in DR database is not really dynamic range, it is just crest factor.

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #2
Unlike DR or peak meters, the R.128 loudness range does not respond to short peaks, to which a limiters act upon. It is a ratio between loudest to quietest sections of a recording that have lasted significant duration of around 3 seconds. This is sometimes referred as "macro" dynamics. Outliers are removed from the statistic, such as a fadeout or an intro if their duration is short relative to the total measurement time, but may be included if they are long enough. Crushed music usually does not have quiet sections, but occasinally may, and will still register a large range then. Loudness Range is most useful in film and television programming, not analysing music.

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #3
I'm confused. Is this a listening test?

Very good point about macro dynamics.  There are lots of songs with dynamic range in an artistic sense rather than a technical sense.  I'm happy to overlook the latter but YMMV.

See my avatar.
Is 24-bit/192kHz good enough for your lo-fi vinyl, or do you need 32/384?

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #4
yes, short term and long term dynamic range are indeed measuring very different things.
long term is decided more by the music composition and mixing, I think it usually isn't changed very much at later stages (except for fade in/fade out), at least not in "live" genres (as opposed to "electronic") because it may sound like someone's riding the volume knob.
Keep calm and opusenc --bitrate 128

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #5
@leonroy , have you tried the Foobar DR meter plugin for example? As far as I understand, it will show you the crest factor (short term dynamics), mentioned by @Rollin , and I think you might find it helpful.

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #6
Which one sounds best to you (level-matched)?

These results are surprising to me BUT...  You can't measure/define dynamics with a single number.  EBU R128 LRA measurement is probably "as good as anything" (and better than crest factor) and it's an agreed-upon international standard.

And, there are lots of ways to compress or otherwise alter dynamics.  We don't know what was done to either recording so this isn't much different from comparing two different recordings from two different artists.; 

Limiting or clipping ALONE will generally reduce the dynamic range.   But, clipped peaks can "sound louder" than clean peaks, so maybe  that makes the loud parts (proportionally) louder?  That might be an experiment worth doing...   Normalize a recording, then boost a couple of dB into clipping and compare the LRA results.

...I like the term dynamic contrast when we're talking about program material, and dynamic range when talking about equipment or formats, but that doesn't make it easier to quantify.  Of course, EBU R128 calls it loudness range.

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #7
@leonroy , have you tried the Foobar DR meter plugin for example?
.

Mac/Linux mostly in my household. I do use Roon though which is where I'm getting the DR128 figures from. I could fire up a Windows VM to try the Foobar DR plugin. I notice from screens it has a peak and RMS value in additon to DR.



I guess peak refers to the max volume - is RMS the average?

Obviously, because it really has larger dynamic range - difference between lowest and highest peak.

Ack, surprised I missed that, thanks for clearing that up, makes total sense.

Also if DR128 figures only give part of the picture is there something else I should be looking at when comparing different masterings of the same album to determine which one is 'best'?

Are the waveforms above a reliable indication I should be using to spot good vs bad mastering? (ie. brickwall = bad)

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #8
Both tracks here measure practically the same, less than 1 dB apart, and no conclusion can be drawn from them. A longer fadeout could be a reason why the second one has a marginally bigger difference between loud and quiet. Clipping and bright equalization may also slightly influence the measurement, as DVDdoug says, because the R.128 emphasizes "annoying" upper mid and treble frequencies with a shelf. LRA is not so much and estimate of quality, as an indication of the type of programme and whether the listener may need to adjust the volume to compensate for overly large difference in level (to hear dialogue in a movie with explosions, or with classical music).

The DR meter is a useful method with limitations. It is sensitive to instantaneous sample level (where the LRA is slow to react), and does not take into account variability within spectral bands, or varying sensitivity of the ear to them. A peak-limited version could be passed through filters (equalization, recording to tape), which almost always increase the peak level and thus the ratio between peak and average, but without an actual improvement. Louder bass, which is typically steady in level over the short time windows used, will lead to lower score, but maybe not harm or even improve the sound. DR meter contains statistical analysis, which is designed to measure only the loudest section of the recording. It can give low score to tracks that have high dynamic contrast, but maybe the limited louder section isn't subjectively harmed. Analysis of a whole album usually gives a fairly complete picture.

Consistently low DR scores (high RMS level) across the album to me indicates that it is likely bad. A high score means that it could be either. And two albums less than 3 dB apart are effectively the same, and should be compared through other means like listening. Permissible loudness depends on the style of music. Instrumental light music shouldn't have less than DR10. while pop/rock could be acceptable at DR8. The RMS level could be measured with a number of tools, such as a ReplayGain scan.

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #9
I did my little experiment -  I choose a "good sounding" song and normalized it (to 0dB peaks).   It had an EBU R128 LRA of 7.51.   Then I boosted it by 6dB and saved-as a clipped WAV.  Common sense prevails!   The LRA was reduced slightly to 7.46.
(The crest factor decreased by about 5dB.)

Quote
I guess peak refers to the max volume
Yes, it's the maximum amplitude.    But, the peaks don't correlate well with perceived loudness.   Most CDs, including quiet-sounding CDs are normalized for 0dB peaks.

Then as j7n says, there are certain processes/filtering that can increase peaks without affecting perceived loudness or perceived dynamics.    A classic example is the cutting/playing of an vinyl record.   Some peaks get higher and some lower (without affecting perception).    That gives you a higher crest factor (peak to RMS ratio) making a higher dynamic range calculation (if you do that way) making some people think the vinyl is more dynamic (less compressed) than the CD.     A similar thing can happen when you compress to MP3...  Some peaks get higher and others lower for a higher crest factor.

Quote
is RMS the average?
Yes, RMS is a kind of average* and it correlates better with perceived loudness than the peak level.  But it doesn't take into account the equal loudness curves... Our ears are most sensitive to mid-frequencies so a 2kHz tone sounds louder than a 200Hz tone or a 10kHz tone of the same RMS or peak level.   EBU R128 and ReplayGain take frequency content into consideration so these are better measures of loudness.

RMS calculations for audio are a little tricky because they usually don't calculate the RMS for the whole file at once.    Usually, RMS is calculated in short segments and then those numbers are averaged.  One reason might be that you'd be squaring and summing a huge amount of large numbers and it might get "impractical", especially for a concert-length or movie-length file.  Silence at the beginning or end or extra-quiet parts may be left-out of the calculation.   If you include silence at the beginning at the end that will lower the RMS calculation.     So, different software may give you slightly different RMS results.

Quote
Also if DR128 figures only give part of the picture is there something else I should be looking at when comparing different masterings of the same album to determine which one is 'best'?
No.   It's only one factor.   If you perceive distortion, or if it seems boring because it's all one-constant volume, that could be an explanation.  But, you have to start with your ears!

And, its a matter of taste.   I generally enjoy more dynamic music but I have some Broadway musical CDs that are "too dynamic" for listening the car.  I end-up turning-up the volume during quiet parts so I can hear above the road/car noise, and then I have to turn it down during the loud parts.    They are also "too dynamic" for listening casually or for background music at low levels.

Quote
Are the waveforms above a reliable indication I should be using to spot good vs bad mastering? (ie. brickwall = bad)
Again, it's one factor.   And remember, whatever is done during mastering is done intentionally because that's what people are buying!


* RMS is related to power.    Here in the U.S. we have 120VAC RMS at the power outlets.   You'll get the same power & brightness if you connect a light bulb to 120VDC.   120VAC has a peak of about 170V.    Since an AC waveform is positive half the time and negative half the time, the true average is zero.   The same goes for audio...  The average is zero.   The average of the absolute values is .637 x peak, so that works out to 108V for 120VAC.  Those calculations are for sine waves.    For a square wave, the peak, RMS, and average of absolute values are all the same.

Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #10
Thanks @DVDdoug and @j7n - that explains it perfectly.

When a track is brickwalled like below is audio data lost at the highest amplitudes ie. like a RAW photo with clipped highlights?


I wonder if streaming has had any effect on how recordings are mastered nowadays. Spotify for example allows us to apply the level of normalization we want for example when in the car/subway or at home:


Re: Understanding brickwalled tracks vs dynamic range (R128)

Reply #11
The amount of clipping is different on a case by case basis. Compression/limiting by itself does not clip. Instead, the amplitude is lowered for a brief moment. Some clipping may be used in addition, to reduce abrupt changes of loudness (pumping) caused by percussive sounds and to maintain their impact. In image analogy, it could be thought of as a local contrast enhancement, where every section of the image has popping "hdr" colors, sometimes with a glow at edge transitions.

I've noticed some old back catalogue recordings posted on streaming platforms totally unmastered (low level, wrong sequencing, silent gaps), possibly they were done in a hurry, not unlike old CD releases.

 
SimplePortal 1.0.0 RC1 © 2008-2019