New version available now for "all" platforms.
New offline version coming end of June.
Couldn't find any info about the Foobar component.
Message from Pleasurize Music Foundation (http://dynamicrange.de/en/download)
MAAT.digital (https://www.maat.digital/drmeter/)
Thanks a lot for the info. :)
MAAT will not only maintain the DRMeter in the future but also improve it into an evolved fully ITU BS.1770 (R128/A-85) compatible universal metering tool (DRMeter MkII). Release is scheduled before end of 2017.
Thank goodness. R128 is a drastic improvement. DR Meter magically gives vinyl rips extra dynamic range, presumably due to noise. R128 is less easily tricked. I find I can trust the loudness range in LU much more.
Thanks for the post.
presumably due to noise
It's foibles were not even remotely due to noise. I'll be happy to see that crude metric die a long and agonizing death.
https://hydrogenaud.io/index.php/topic,102895.0.html
FWIW, R128 seems to have a pretty serious issue as well:
https://hydrogenaud.io/index.php/topic,110561.msg911014.html#msg911014
FWIW, R128 seems to have a pretty serious issue as well:
https://hydrogenaud.io/index.php/topic,110561.msg911014.html#msg911014
Fair enough. I hope this can be improved upon in future standards.
However, I will say in the context of my large music library, as well as our library of recordings at work, it does a great job of giving consistent playback levels (I use JRiver for this) and also giving me quick useful info to determine if one of my staff edited properly or not. It is a big step forward and the industry has really latched onto loudness standards. Here's hoping the progress continues.
I find that David Robinson's original algorithm does a better job. R128 exaggerates the loudness of modern mastering; for metal at the very least.
I find that David Robinson's original algorithm does a better job. R128 exaggerates the loudness of modern mastering; for metal at the very least.
+1, using it with logitech mediaserver. When R128 was introduced i had ambient music to low in level for my impression. I use RG in metaflac.
Case was kind enough to tell me how to continue using foobar2000 to calculate RG using the original algorithm:
https://hydrogenaud.io/index.php?topic=113986.msg938727#msg938727
Thanks but i don't use foobar for such things. I use only flac and the limited resampling/packing/RG i still do with the vintage frontah.
I used to use metaflac, but with multiple threads foobar2000 is so much faster.
I tried but i seldom keep my foobar install recent and the foo_rgscan.dll out of the linked 1.1.5 crashes my 1.3.4 version. I may try later.
VST users may try the free Youlean meter, very clean, neat and scalable UI.
http://youlean.co/youlean-loudness-meter/
DROffline MkII with R128 and A/85 has been released:
MAAT DROffline MkII (https://www.maat.digital/dro2/)
> MAAT DROffline MkII
Until the standard (algorithm description) is made publicly available, I do not approve this software and this standard, and encourage everyone to do the same (and use the "old" measurement standard, which was successfully reverse-engineered and implemented in free software).
This one is a rather sick attempt to sell proprietary software to the victims of the loudness war.
...except that the "old" measurement was crude and pretty much worthless, leading people to believe that vinyl releases were sourced from different mastering than CD releases, which in many cases weren't.
While you may be new to the discussion, this is hardly news.
It was more or less a common knowledge that it's not applicable to vinyl. Some people unfortunately didn't get that, I agree that it's bad…
But for digital records I wouldn't say that it was pretty much worthless as it allowed to estimate the overall amount of damage done to a record.
It can be improved, sure, but is it worth giving up software freedom again for this?
By the way, if someone has got an idea what the new measurement algorithm could be actually doing, I'd happily implement that in code as free software.
but is it worth giving up software freedom again for this?
It is worth no less now that is no longer free; not to me, anyway.
The
EBU Loudness Range Standard (https://tech.ebu.ch/docs/tech/tech3342.pdf) (LRA) looks good to me "on paper". I haven't done any experiments, and I don't think you can boil-down "musical dynamics" to a single number for all performances/recordings. But it does use
loudness (not just peak & average amplitude) and it is a published standard.
LRA is defined as the difference between the estimates of the 10th and the 95th percentiles of the distribution. The lower percentile of 10%, can, for example, prevent the fade-out of a music track from dominating Loudness Range. The upper percentile of 95% ensures that a single unusually loud sound, such as a gunshot in a movie, cannot by itself be responsible for a large Loudness Range.
With a long integration time of 3 seconds, LRA measures 'macro' dynamics, whereas DR Meter measures 'micro' dynamics.
A shortcoming of DR is its reliance on instantaneous peak level. While RMS power, whether perceptually weighted or not, is stable, and measures nearly the same as the signal is equalized, DC-highpassed, phase shifted, resampled, passed through crossovers and speakers, the peaks change unpredictably. A proposed Peak to Short-Term Loudness Ratio (http://www.meterplugs.com/blog/2017/05/18/crest-factor-psr-and-plr.html) is even worse, because it takes the intersample peak into account, which consists of ultra high frequencies of little significance, and inflates the measurement of squashed music.
Might a better solution be to somehow integrate a near momentary loudness power level with a short window (10-50 ms), and relate that to the short-term loudness (3 seconds)?
A discussion (http://www.kvraudio.com/forum/viewtopic.php?f=6&t=496858) of the MAAT DR meter on KVR Audio. It hasn't been well received, overpriced, of little practical usefulness, and features invasive copy protection. I do not know if the offline version has copy protection.
The conventional DR reading, in my experience, sets a limit on the pontential quality. While a recording with DR8 or less can't possibly be good, those with higher score could be either.
Some examples of LRA and DR:
Gary Moore - Still Got the Blues - LRA 3.1 dB DR13, RG(R128) -0.07 dB
Oasis - Don't Look Back in Anger - LRA 4.9 dB DR5, RG(R128) -11 dB
Oasis - Wonderwall - LRA 14.6 dB DR7, RG(R128) -9.9 dB
Scarlet - Independent Love Song - LRA 15.7 dB DR10, RG(R128) -4.6 dB
No way Oasis sounds better than Gary Moore. Independent Love Song has large loudness variation throughout the song between verse and chorus.
The conventional DR reading, in my experience, sets a limit on the pontential quality. While a recording with DR8 or less can't possibly be good, those with higher score could be either.
I'd say that the limit is at DR6 rather than 8. I have quite a few examples when DR7 records already sound just fine; in particular, the latest album by "Draconian" (a metal band from Sweden).
But with DR6 and lower I usually can hear some pumping or distortion. I can't say that this is the primary cause as there's usually no normal variant to compare (and otherwise I will just get the normal version and not bother comparing), but it seems to correlate.
Might a better solution be to somehow integrate a near momentary loudness power level with a short window (10-50 ms), and relate that to the short-term loudness (3 seconds)?
I too thought this may be a good idea. Not necessarily even 3 seconds, maybe even (10-20 ms) against 1 second. Ideally, we probably need volunteers with sensitive ears to make a reference record ranking and test the metric correlation against it.
I tried but i seldom keep my foobar install recent and the foo_rgscan.dll out of the linked 1.1.5 crashes my 1.3.4 version. I may try later.
I know it's been a while, but I finally got around to using David Robinson's original (and IMO superior) RG algorithm in fb2k and initially had the same issue and found out why...
In addition to replacing the dll, you will need to create a new dll.cfg file in your AppData directory. I originally blew the old one away, but since I'm also using the R128 version, I'm keeping a copy of each in order to retain their respective settings, which I do via batch script.
For ReplayGain scanning with the original, superior algorithm, I use an old portable Foobar version 1.0.3 (no particular reason for this one). I also have the DR scanner in it, because I disliked that it appears in the context menu in place of "Copy name", and would start processing immediately if invoked by mistake. Maybe another version (with a distinct appearance) for special tasks is a solution for people who reported issues with the DR scanner plugin in the latest beta.
The Dynamic Range DB has been promoting the DROffline scanner with a large banner for some time, since they changed to the high-fatigue flat design, well before the MkII was released. Strange that they link to the first version of the tool, instead of the new one, and claim that it already is an improvement. It seems to me that the DB has an interest in promoting MAAT products. The trial version does come with 33 MB CodeMeter protection installer, so I didn't try it, but the website gives no indication that DROffline gives anything but the same old offical DR reading.
Thanks for the tips. I don't have much new music as it used to be. Frontah still works welll for my needs and its RG with metaflac.
I never will pay a single cent for a DR meter btw.
…the website gives no indication that DROffline gives anything but the same old offical DR reading.
It does say that it's different if you go to the upload page. (And it doesn't explain in any way _why_ is is "superior"… I guess it works better to scam people.)
For ReplayGain scanning with the original, superior algorithm, I use an old portable Foobar version 1.0.3 (no particular reason for this one). I also have the DR scanner in it, because I disliked that it appears in the context menu in place of "Copy name", and would start processing immediately if invoked by mistake. Maybe another version (with a distinct appearance) for special tasks is a solution for people who reported issues with the DR scanner plugin in the latest beta.
The purpose of ReplayGain scanning and DR scanning are completely different.
First is to measure perceived loudness (to allow one to play everything at the same loudness easily). It doesn't say anything at all about the amount of dynamic compression (although when the RG goes below -9 dB, it's very likely the record had to be crushed to be this loud).
Second is to estimate how badly were the peaks in the record compressed, but it doesn't say anything about loudness. Although not very common, there are records which were compressed to death and then reduced in volume. (So they are both distorted&lacking in dynamics, and "quiet").
Although not very common, there are records which were compressed to death and then reduced in volume.
If you're including records pressed before, say 1995, then sure; otherwise a citation is needed.
Although not very common, there are records which were compressed to death and then reduced in volume.
If you're including records pressed before, say 1995, then sure; otherwise a citation is needed.
http://dr.loudness-war.info/album/view/125043
A very extreme but real example. It's crushed nearly as much as the notorious first versions of Death Magnetic, and then the gain is reduced to as much as -6 dB. It's likely done by mistake, but it's just the fact of life, a record like this got through all (if any) checks and released as it is.
I can say that this measurement is true because I used to have this record.
I also have seen other examples (but not as extreme as this), but I don't remember them.
If you're including records pressed before, say 1995, then sure
By the way, before 1995 almost no one used to destroy (already mixed) music by pathological levels of compression and clipping. It's for the most part caused by the loudness war, and it went totally out of control in 1994-1995.
(I were talking about records which are at the same time destroyed by compression AND are quiet; before 1995 it was common to be quiet but not at all common to be destroyed by compression)
I were talking about records which are at the same time destroyed by compression AND are quiet
So you aren't saying records that are destroyed but not quiet are not very common.
Citing DR figures is worthless since high DR values for vinyl are woefully inconclusive.
Yes I were of course not saying that. (It's obvious that there's a lot of them). I were only saying that
destroyed records can also be quiet and DR score only may estimate the amount of damage but not the loudness. Low DR
usually goes with high loudness (because otherwise the DR compression is completely pointless), but people aren't perfect and they do pointless things sometimes, and the link above is one extreme example of people doing pointless things.
Citing DR figures is worthless since high DR values for vinyl are woefully inconclusive.
Yeah but I were citing
low DR values and not for vinyl.
I know that there are many ways to inflate DR score without actually improving anything; phase shifting, EQ, adding noise or clicks just to name a few. You are right that high DR, strictly, doesn't guarantee anything.
But if it's low (in particular ≤ 4), there's no way it'll sound good. So citing extremely low DR may actually mean something conclusive.
http://dr.loudness-war.info/album/view/125043
"Not normalized" is not the same as "reduced", and at least the DR and peak values do not prove that it was ever
reduced in volume.
My hunch would be that it was recorded with a few dB headroom (not a bad thing), mixed to sound to the band's satisfaction (good or bad, matter of taste) and then sent off to the - quite underground - label who pressed the artist-approved master without doing anything about it.
Actually, there was a limited edition version of this release, with a bonus CD
-R.
(By the way, being self-pronounced "primitive" black metal, a low volume could just enhance a "demo" feel to it. Something I lost by using ReplayGain :-o )
Yeah but I were citing low DR values and not for vinyl.
Confirmed. It is this one: http://db.cuetools.net/?tocid=F5S_6d1bt.hAqyFLSidq6PUpgto-
And it was not released on vinyl: https://www.metal-archives.com/albums/Sorcier_des_Glaces/The_Puressence_of_Primitive_Forests/309775
"Not normalized" is not the same as "reduced", and at least the DR and peak values do not prove that it was ever reduced in volume.
My hunch would be that it was recorded with a few dB headroom (not a bad thing), mixed to sound to the band's satisfaction (good or bad, matter of taste) and then sent off to the - quite underground - label who pressed the artist-approved master without doing anything about it.
Normally the clipping in mixing happens close to 0 dB and not at some arbitrary point near -5..-6 dB. Unless their taste dictates crazy amounts of clipping, it's likely not been as you described.
It doesn't affect the result anyway, if the clipping was done intentionally at a lower level, or the gain was reduced after clipping. In both cases the end result is a clipped record with low peak levels.
If you open a track from that disk in a waveform viewer/editor, and zoom in, you'll see what I mean.
And, by the way, there's not much a label can do if they get a totally devastated master. They may perhaps refuse to release it and ask the artist for a replacement.
The DR and peak values prove (by counter-example) one thing here:
if a record is quiet, it does not follow that it must have good dynamic range.(it's already pretty obvious in theory, you can always make any record quieter (until it is buried in quantization noise, which is far away even with 16 bits) and it won't make it magically sound better/have more DR; this is just an example of an "officially released" record with such a property)
I didn't bother to look at the dr.loudness link provided, but normally "record" implies vinyl, at least in the English language, anyway; even if the definition of the word indicates that it can be used regardless of the specific type of media.
I do agree that a low number can be assumed to mean high DRC, but I've seen more than my share of posts where high DR values for vinyl were taken as meaning it was free from heavy DRC. I'll even go a step farther and say that I've seen posts where people claimed to hear more dynamic range (read: less DRC) in a vinyl version over compared to the CD version when there wasn't any.
normally "record" implies vinyl, at least in the English language, anyway; even if the definition of the word indicates that it can be used regardless of the specific type of media.
I had no idea, sorry.
I probably overreached. I'd say it's typical to call an album on vinyl a record in the US.
Turntables are often called record players, at least by older generations. I'd say the term album without specifying the medium often implies vinyl as well.
I have no stats on any of the above, though. ;)
...and if you want to buy a "record album" on amazon you get this:
https://www.amazon.com/slp/record-albums/zvy23e4vr7g23np
The purpose of ReplayGain scanning and DR scanning are completely different.
Indeed it is. I diverted into a secondary topic, offering an old version of Foobar as a scanner for two kinds of statistics, to be used as a separate tool from the main player, which is kept current. I have muscle memory with using legacy menus, like Masstagger.
Although not very common, there are records which were compressed to death and then reduced in volume. (So they are both distorted&lacking in dynamics, and "quiet").
My favorite music release that has been clipped and later turned down 2-3 dB is "Falling Into You (http://dr.loudness-war.info/album/view/103020)" by Celine Dion. Oh my god (http://i.imgur.com/HUEoavf.png). Lately, I have observed a convention of setting the limiter ceiling to -1 dB, sometimes -2 dB, as in the remastered "Death Magnetic (http://dr.loudness-war.info/album/view/112438)", with two decibels reserved "unused" for deviations of the true peak. A waste.
I was unsuccessful getting DROffline trial to run inside a virtual machine. CodeMeter is smart. I suspect the whole improvement in v.1 consists of FLAC support and the bug fixed, which treated tags as full scale samples.
Maybe LRA could have some rare applications in a music collection (apparently, JRiver collects this statistic), for example, when selecting music for a noisy environment. But it cannot measure how far drums reach above average level, and is deaf to the spaces in a dynamic, but evenly spaced regular track like "Billie Jean (http://i.imgur.com/Wc7RkvM.png)" (LRA = 2.5±0.2). The shaded area of Variance is a kind of swing of instantaneous loudness of unspecified, very short integration time (used to closely track Momentary in VisLM v2.0) decaying over the Short-term window. The decay is rectangular and crude, and there is no integrated readout of it. But crushed tracks (http://i.imgur.com/N5KgCWh.png) measure less, opposite of LRA in this case.Frontah still works welll for my needs
Frontah is a great frontend, extendable by the user, which I use on daily basis for SoX spectrograms and encoding. I'm surprised that somebody else still has it.
It's worth noting FFmpeg now has a drmeter audio filter (https://ffmpeg.org/ffmpeg-filters.html#drmeter).
I was unsuccessful getting DROffline trial to run inside a virtual machine.
When a piece of software actively prevents running itself in a VM for no good reason, you may be certain that it's a kind of
malware.
I have observed a convention of setting the limiter ceiling to -1 dB, sometimes -2 dB, as in the remastered "Death Magnetic (http://dr.loudness-war.info/album/view/112438)", with two decibels reserved "unused" for deviations of the true peak. A waste.
Also the fear of lossy codec induced clipping. Every time I suggest to use a RG compliant media player in forums other than HA people will throw out nonsense like loss of dynamic range, not "bit-perfect" and so on.
Opinions like these are pretty common in audio production related discussions:
https://www.gearslutz.com/board/mastering-forum/1200247-what-ceiling-1-dbfs-0-1-dbfs-use-master-prior-audio-normalization.html
fear of lossy codec induced clipping
It's even more funny when they "limit" data with dumb, in-your-face clipping, like the Celine Dion example above. As if baked-in
unavoidable clipping is somehow better than the other clipping which
may or may not happen.
It's worth noting FFmpeg now has a drmeter audio filter.
Also worth noting that, though we love all the contributors at FFmpeg and even use some of their code under license as noted in our user manual for DRMeter MkII, that filter only approximates DR Dynamic Range as defined by the PMF.
MAAT will not only maintain the DRMeter in the future but also improve it into an evolved fully ITU BS.1770 (R128/A-85) compatible universal metering tool (DRMeter MkII). Release is scheduled before end of 2017.
Thank goodness. R128 is a drastic improvement. DR Meter magically gives vinyl rips extra dynamic range, presumably due to noise. R128 is less easily tricked. I find I can trust the loudness range in LU much more.
Hey Moni,
To be fair, DR was not designed for vinyl rips so, you’re right not to trust it for that use case. Nothing wrong w/them per se, but a vinyl rip has gone through so many convolutions since the master, usually a digital file these days, was fed to the lathe that it’s without question a completely different beast from what was cut to the lacquer.
As with DR, R128 and A/85 metrics are standardized and thus repeatable. That said, there’s no magic in an R128 measurement except for the Gate, which is most likely responsible for your R128 measurements correlating better with subjective experience w/vinyl.
A side note: DRMeter MkII, eluded to above in the quotequote, and DROffline MkII both provide all R128 Loudness metrics, along with DR
i, Minimum PSR and a nice collection of other measurements that help engineers (and enthusiasts) make informed decisions about dynamic range.
It's worth noting FFmpeg now has a drmeter audio filter.
Also worth noting that, though we love all the contributors at FFmpeg and even use some of their code under license as noted in our user manual for DRMeter MkII, that filter only approximates DR Dynamic Range as defined by the PMF.
Also worth noting that they (or someone else) will add the "new" Dynamic Range (it's called DRi?) in no time, if you just stop keeping the formula in secret. :)
And it's a good idea to not keep it in secret anyway, otherwise nobody's really sure what exactly is this magic closed-source code is measuring.
The whole thing is a joke.
Why? Your eyes cannot hear. This thing is useful for idiots who judge music based on numbers rather than content.
To help engineers make informed decisions about dynamic range? That's absurd. It's more like help the producers see that the music isn't compressed enough.
The whole thing is a joke.
Why? Your eyes cannot hear. This thing is useful for idiots who judge music based on numbers rather than content.
To help engineers make informed decisions about dynamic range? That's absurd. It's more like help the producers see that the music isn't compressed enough.
It's not possible for us mere mortals to judge
all music based on content, because
there are time and money constraints.
And if I see (before buying) that a particular recording has suspicious numbers, then it's an increased risk of wasting time&money. Nobody says that it guarantees bad sound, but the risk is higher and for some people this can be meaningful.
If all music was free and we were immortal and never bored with life, then, sure, it would be very close to absurd.
And very few people get the opportunity to look at the digital contents of a CD without paying for it first, and most stores won't refund you money for an open package.
And very few people get the opportunity to look at the digital contents of a CD without paying for it first, and most stores won't refund you money for an open package.
This is true, but someone else could have already done it, and it can be searched on the Web. http://dr.loudness-war.info/ for example.
And if we're talking about web releases, Bandcamp usually even allows to listen to at least a few tracks from an album with poorer quality (it used to be 128k cbr mp3, not sure what it uses now, probably the same), and it can be used to estimate DR too (with less accuracy, but mp3 compression doesn't inflate DR scores as drastically as vinyl does) — in that case, sometimes it's not even necessary that someone else had bought it before and uploaded the measurement results.
(Although in the latter case, if it was only a little bit better than 128k cbr mp3, it could be judged by ear as anything higher would be close to transparent, but I guess they lower the quality intentionally to leave more incentive for buying; so it's unlikely to change, or they'll maybe use something like Opus @ 40 kbps in the future as a replacement; when you hear compression artifacts or know that they are very probable, it's hard to be sure if the defect which you hear is a result of lossy compression or a result of bad mastering, if you don't have the lossless source to compare, or at least a lossy version encoded with something that's pretty much guaranteed to be transparent)
Something like this:
https://hydrogenaud.io/index.php/topic,114892.msg947214.html#msg947214
Without telling people that "b" is a lossy version of "a", how many people will think the opus file actually sounds "better" and more "dynamic"?
IMO lossy preview at lower bitrate is a nice idea, at least it can let people know lossy codecs, like vinyl and analog tape, can change the waveform dramatically.
Precisely!
fear of lossy codec induced clipping
It's even more funny when they "limit" data with dumb, in-your-face clipping, like the Celine Dion example above. As if baked-in unavoidable clipping is somehow better than the other clipping which may or may not happen.
Someone told me iTunes rejects masters with intersample and codec induced clipping based on Apple's own analyzing tools. A bit of Googling I found it's real.
https://www.gearslutz.com/board/mastering-forum/1000408-mastered-itunes-inter-sample-peak-conundrum.html
Maybe one of the reasons why some songs are delibrately clipped at lossless and non-intersample level.
I did not see anyone say Apple rejects masters in that discussion, nor do I recall reading anything in Apple's boneheaded document saying that would be the case when they released their suite of tools. IIRC, it was only a guideline.
Maybe you have some other smoking gun?
Also, we have a dedicated discussion on Mastered for iTunes. That would have been a better place for this.
Jürgen Reis from MBL:
http://archimago.blogspot.com/2018/06/musings-thoughts-on-audio-device.html?showComment=1530717709825#c8830174310950077497
I myself cannot provide any further proof as I am not an Apple certified Mastering Engineer.
Please provide quotations of the relevant parts instead of just dropping links, especially when they appear only to be tangential at best.
I just asked him in the previous Archimago's blog link and waiting for a reply.
Is it not against the TOS to still post on this thread? If yes please point me to a relevant thread.
Clipping is one of those subjects that often gains traction, derailing the original topic.
TOS #5 covers this.
Here are two other topics:
https://hydrogenaud.io/index.php/topic,93628.0.html
https://hydrogenaud.io/index.php/topic,94214.0.html
If you haven't read Apple's Mastered for iTunes PDF you should. There is nothing in there that says less than optimal masters will be rejected. Absolutely nothing.
Hey folks,
A heads up that a new fork of the original DROffline is now available at MAAT; DROffline MkII. “DRO2” now includes complete R128 Loudness measurements plus additional informative metrics. It also interoperates with DRMeter MkII. BTW, the original DROffline is still available for those who only need the old stool stuff.
BTW, the original DROffline is still available for those who only need the old stool stuff.
Ewww :P
Please provide quotations of the relevant parts instead of just dropping links, especially when they appear only to be tangential at best.
THis, apparently:
As being also an Apple certified Mastering Engineer I can report, that for the “Mastered for iTunes” logo, you have to proof, that there are no Sample or Intersample Overs, otherwise, the files will be refused.
I'll buy that even if it's only anecdotal. It certainly isn't in the document apple released when mastered for itunes started.
I don't know if what iTunes does is relevant for "Mastered for iTunes", but FWIW: iTunes itself will/would encode AACs with peaks over >1.
Example (grab as much organ BWV as you want to for free, folks!) though only +0.0something dB:
http://www.blockmrecords.org/bach/detail.php?ID=BWV0561 . The AIFF file with fb2k's "True peak" enabled (tried three different settings) peaks at +0.0something, and so does the AAC file even without true peak enabled. (Though not the iTunes-created MP3.)