Jriver mc19 volume levelling

[realysm42]

Just begun to analyse my library, should be done tonight I reckon.

Read some good things about this feature and am extra happy as my preamp won't have a remote, so levelling it should save a few trips, once I've found a volume level I like :eyebrows:

Another interesting feature is the DR assessment, so far from what I've seen the best I've got is Doug MacLeod's 'There's a time' CD. Great album too.

Anyone else use this? What are your experiences?

[realysm42]

Just tried this out, it works very well.

Library analysis took a little while but is well worth the wait.

There's basically no audible volume difference between the loudest and quietest tracks. Excellent idea, well implemented.

I recommend this if you don't have a remote for your volume or simply don't like constantly adjusting your listening level :)

[Spectral Morn]

Just tried this out, it works very well.

Library analysis took a little while but is well worth the wait.

There's basically no audible volume difference between the loudest and quietest tracks. Excellent idea, well implemented.

I recommend this if you don't have a remote for your volume or simply don't like constantly adjusting your listening level :)


Surely flattening dynamic range and levelling volume is a bad thing - not how things are in the real world. This is what's wrong with most modern CDs and recording loudness and compression. Sounds like a step backwards imho.

Regards Neil

[realysm42]

Neil, perhaps you are right; I'm not going to pretend to know the answer and (as per usual) seems to spark much debate.

I'm going to "test" it out tonight and see what I can hear, I didn't think it fair on my poor neighbours to do so this morning.

One interesting thing I picked up with this feature is that the feature uses R128, which apparently is something being used in the Loudness wars (to fight against it), here's a link for anyone intersted or with an opinion to share (I'm all ears):

http://yabb.jriver.com/interact/index.php?topic=82025.0

[WAD62]

Are you using ReplayGain? Album & Track volume offset attributes?

http://en.wikipedia.org/wiki/ReplayGain

http://wiki.hydrogenaudio.org/index.php?title=ReplayGain_1.0_specification

...a quick check would suggest that Jriver only does 'Track'/'radio' gain which is not the full ReplayGain standard, although thankfully it doesn't remaster your source data.

For a proper implementation of ReplayGain one needs 'Album', which respects/maintains the relative volume differences between tracks within an album, therefore all tracks within an album will get the same dB offset value.

In addition there should be a ReplayGain pre-amp value to compensate for the overall lowering in volume, Winamp provides this, LMS permits Album & Track but no pre-amp value.

Bare in mind that switching on ReplayGain for playback will prevent a bit-perfect signal, as 16 bit tracks are expanded to 24 in order to apply the volume modification.

I used to use it on my SBoxes, but since getting my MDAC I've turned it off, it prefers 'Bit Perfect', although I sometimes use it for parties & Djing, oh and also on my 'ambient' playlist in the bedroom...

[realysm42]

Thanks Will, I read that this feature does respect volume differences between tracks in albums.

I'm not using any other features currently, Jriver indicates that my output is bit perfect.

[Rothchild]

Surely flattening dynamic range and levelling volume is a bad thing - not how things are in the real world. This is what's wrong with most modern CDs and recording loudness and compression. Sounds like a step backwards imho.

Regards Neil

It depends what the replay gain is doing, if it's trying to turn up the 'properly' mastered tracks to reach the level of the ones that have been smashed that would be bad, but my guess is that it's turning down the overly loud ones to some more bearable RMS (or LUFS) level so that they don't leap out against the stuff that's been left with a sensible amount of dynamic range. This is good (in a way) because those overmastered tracks won't sound as good all of a sudden and there'll (hopefully) be a swing back round to using some sensible amount of headroom. IMO most current rock and pop sounds great with 14 to 12dB of dynamic range over the average RMS level.

[realysm42]

Thanks Marc.

I've put a question forward about sound quality impact with the feature.

I'll put my neck on the block and say it sounds pretty damn good with it on tbh. Makes life a lot easier too.

[Rothchild]

Notwithstanding esoteric debates about 'losing bits' by turning stuff down, in general gain changes in the digital domain are 'inaudiable' (apart from obviously being louder or quieter!).

I was musing on this the other day as one of the reasons SQ changed (from a recording perspective) with the advent of digital; previously gain was 'expensive' in that it always brought noise with it so engineers were always on the look out to for opportunities to eek a bit more gain out of ths system (to get out ahead of the higher noise floor of analogue processing), combined with the fact that going 'hot' to tape has a familiar and generally pleasant effect on sound this was all good. If you will, gain was more 'expensive' than headroom.

However, with digital it's the other way around; gain is 'cheap' (and noise floors are low) combined with the longstanding myths about 'lost bits' and a poor understanding of the dynamic range represented by digital systems this has served to push up the 'price' of headroom to the extent that for a good period now there's been a great deal of music made without any!

The likely impact on SQ of Replay Gain and other normalisation strategies is to make stuff with no headroom sound worse. 99.9% of the time if you (average) level match material with headroom against that without then, at a nice listening volume, the stuff with more headroom will sound better (deeper stage, more impact from drums etc + crecendos and the general ebb and flow of volume makes you feel more engaged by the band etc).

[realysm42]

Okay, I've had volume levelling on pretty much since I created this thread and to my ears...

(Drum roll sounds)

Actually, its a hell of a lot better without. There could be any number of reason for me thinking it sounded OK turned on (ifi preamp burning in, new interconnects doing the same) but it sounds a lot less strangled off (which is a real shame, I'd have loved it to be better/as good turned on).

This really goes to show the value in my eyes of extended testing, then removing said adjustment, its highlighted the changes clearly for me.

So, this feature could be great if you want loads of music in the background and not have to worry about level matching, but for dedicated sessions, hmm, maybe not.

[Rothchild]

That's interesting. When you're comparing on vs off are you level matching the comparators? It's a well known fact that we tend to associate louder with better (which is one of initial reasons behind the loudness wars), indeed I believe research in this area shows that even very tiny volume differences (when all other factors are equal) will yeild a preference for the louder source.

Another thought is that, maybe, by running the volume leveling it's adjusting the gain structure of your rig so that different components are being driven at different voltages and they sound different at that range in their headroom?

[realysm42]

Level matching is hard as apart from the volume levelling feature, all I've got is the knob on the itube to go by.

So, as per all of my testing, its not obsessively accurate. I did try it on several sonfgs, several times to get them as near as I could.

The difference I detected was even with the vl feature off and the pre possibly louder, the sound was just more open and expressive, less muffled.

It is nicer to have more usable range on the volume of the pre with the feature turned on, but this is more a case of equipment matching and isn't worth the loss of fidelity.

Marc, it sounds like knowledge is far greater than mine in this area.

[Rothchild]

I'm guessing your itube is one of these? http://www.audiostream.com/content/ifi-itube-active-tube-bufferpreamplifier

Might be worth trying to push the gain coming out of the dac / jriver when using the vl feature as a means to trying to determine if it's the VL that's changing the sound or the reduced gain being driven in to the itube. Given that it's a tube device the chances are that different input gains could have a fairly significant effect on the way it responds (and that potentially you've got used to / prefer the sound of it being driven harder).

It's interesting given all the interest in (what I see as) relatively minor things making a supposedly 'obvious' differences to SQ that are discussed around here there seems to be very little time or concern given to basic setup matters such as gain structure (which can give demonstrable and repeatably different outcomes depending how it's approached) - I'm not aiming this at you Martin, just musing 'out loud'

[realysm42]

That's an interesting thought.

I've only been able to apply what I've learnt in a relatively short time and only learnt what I've asked about.

Tbh I won't look too far into the itube (the one you've linked to is correct) because I'm going to get a passive pre soon, which will be a permanent part of the rig. It didn't occur to me to turn the volume levelling off when I had one on demo, which is a shame, especially if what you've said about tubes sounding different depending on how they're driven.

Of course if the sq gains remain then I'll be a happy chap as it sounded amazing anyway :)

[Peter Stockwell]

That's interesting. When you're comparing on vs off are you level matching the comparators? It's a well known fact that we tend to associate louder with better (which is one of initial reasons behind the loudness wars), indeed I believe research in this area shows that even very tiny volume differences (when all other factors are equal) will yeild a preference for the louder source.

Another thought is that, maybe, by running the volume leveling it's adjusting the gain structure of your rig so that different components are being driven at different voltages and they sound different at that range in their headroom?

That may be true, I won't argue that point. I tried the new R128 level matching feature of MC19, and hated it. I felt my system had become broken, dynamics sat upon. Overall, not to be considerd in any way "hifi".

[Rothchild]

That may be true, I won't argue that point. I tried the new R128 level matching feature of MC19, and hated it. I felt my system had become broken, dynamics sat upon. Overall, not to be considerd in any way "hifi".

Of course, the obvious corollary of 'if it sounds right, it is right' is that 'if it sounds wrong, it is wrong'.

All these volume leveling tools are making a change to what the artist/label had put out (regardless of the sense of the judgement that was made), personally I'd rather that a bunch of stuff had just been made without reliance on being smashed in to a limiter. Theoretically, at least, if something was 'over-loud' when it was made then normalising it down to R128 (LUFs) or a K-level shouldn't actually adjust the dynamics (whatever there was of them in the first place) but still the ideal solution is going to be adjusting the gain in the analogue domain to make a comfortable listening environment (where, what I notice is that, over-loud stuff encourages you to turn it down whereas well made stuff encourages you to turn it up).

I'd be interested to see anyone's take on it if they have the ability to do some decent digital level monitoring and to get a sense of what these features are actually doing to levels. My RME soundcard has great metering and (when combined with my volume calibrated setup) it gives me a good insight in to understanding how records have been mastered and which ones are 'impressive' because they're well made and which ones are just plain 'over loud'.

[realysm42]

Marc you could use a jriver demo and run the test yourself?

[Peter Stockwell]

Theoretically, at least, if something was 'over-loud' when it was made then normalising it down to R128 (LUFs) or a K-level shouldn't actually adjust the dynamics (whatever there was of them in the first place) but still the ideal solution is going to be adjusting the gain in the analogue domain to make a comfortable listening environment (where, what I notice is that, over-loud stuff encourages you to turn it down whereas well made stuff encourages you to turn it up).

I liked the old system better, not that I actually used it as such, but it gave me a heads up about how to set the volume control, was also interetsing to compare various masterings of the same original recordings.

[Rothchild]

Marc you could use a jriver demo and run the test yourself?

I've got it installing right now, will let you know how I get on.

[Rothchild]

Source file - R128 - Replay Gain - Range 128 - Range DR

Broken Social Scene, Kc Accidental - 14.6LU - -9.58dB - 9.9LU - 6 (AKA Well loud!)

I just tried a quick experiment using the old faithful 'phase flip' test.

I ripped the track as a .wav using jriver, 'scanned' it and switched on volume leveling.

I then used the feature of my RME sound card that enables me to 'loopback' outputs to inputs digitally and recorded the playback output of jriver in to my DAW (Digital Audio Workstation), alongside this I imported the .wav that I'd ripped.

Then, I flipped the phase (inverted the polarity) of one of the tracks and started drawing down the volume of the .wav file at -14.5dB the playback goes silent, which indicates that the tracks are phase cancelling each other (ergo there is no qualitative difference in SQ between the files only the volumes are different).

I'm going to re-run the test now with something that has some dynamic range and isn't horribly clipped like Broken Social Scene and see if I can make a different result, but this initial test seems to indicate that any difference people are hearing is more likely down to the gainstaging of their Hi-Fi chain and is not a product of volume change being applied to the file per-se.

[Rothchild]

Just repeated experiment with a track from the first Beautiful South album (Sail this ship alone) which has spades of headroom (indeed the highest peak is -3.1dB) and there's a good 14dB of dynamic range. Same result.

Also a salutory lesson in the precision that this test is able to muster. I sample align the files by eye, I pick an obvious feature in the wave and line it up, gradually zooming in until I'm at single sample resolution and happy that it's right. At first it wouldn't quite cancel so I thought I'd got the alignment wrong, so I nudged the track one sample along and the cancellation failed completely. When I set it back and set the volume by typing in the actual figure I wanted (rather than moving the fader which was jumping me between -3.38 and -3.42) then the cancellation happened properly.

[Peter Stockwell]

see if I can make a different result, but this initial test seems to indicate that any difference people are hearing is more likely down to the gainstaging of their Hi-Fi chain and is not a product of volume change being applied to the file per-se.

Rats! Hate to read that kind of research. I'll try it again, but be more adventurous with volume.

[Peter Stockwell]

Rats! Hate to read that kind of research. I'll try it again, but be more adventurous with volume.

With a freer hand with the volume control, I now can't hear anything untoward with R128 volume equalization. I didn't expect that!

[realysm42]

Interesting stuff Marc, thanks for doing that so quickly.

So your conclusion is that its our systems volume not being perfect, instead of the volume levelling being damaging?

[Rothchild]

More or less,

What I think it shows is that the volume levelling system is doing nothing but turning the gain down as it re-plays the file (by a fixed amount per song). Because (as I mentioned earlier) digital gain changes are 'free' this causes no difference to the quality of the file (if the tone or dynamic range had been changed I would not be able to phase cancel the file; in previous tests I've satisfied myself that tone (EQ) change of as small as 0.1dB at any frequency was detectable using this test).

So, the current 'best guesses' are that; the differences heard are caused by either the knock-on effect of driving lower gain in the rest of the system and the altered gain structure having and impact or, potentially, that: Once you put over-compressed (over-loud) material up against well mastered material in a volume matched way then the game is up for loudness wars mastering. All of a sudden the Emperor is naked! (there's no depth, there's distortion and clipping, transients are softened etc etc)

One further comment to make is that I was quite surprised at just how much gain was coming off to achieve the 23LUFs 'Broadcast' standard, the final level looked even less than K20 on my meter. If it weren't for this aspect I'd be more inclined to go with my second theory about the difference but it strikes me that the first may still be plausable in certain systems due to the reduced voltage being driven in to them.