INTERACT FORUM

Please login or register.

Login with username, password and session length
Advanced search  
Pages: [1]   Go Down

Author Topic: DSP Studio Quirks  (Read 8775 times)

tasar

  • World Citizen
  • ***
  • Posts: 148
DSP Studio Quirks
« on: February 28, 2014, 10:42:20 am »

Just updated to latest Mavericks 10.9.2 and MC 19.0.118. The affects of enhancing low SPL dynamics by engaging "night mode" is arguably one of the best features with MC. However, when enabling "volume leveling" (a must with varied file metrics), the volume/dynamics is crushed 5dB or more. This is a must fix for me.
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #1 on: February 28, 2014, 12:34:17 pm »

What do you mean by "crushed," exactly? 

The only thing that volume leveling does is attenuate volume by the amount needed to reach a reference level.  It doesn't perform any dynamic range compression or other processing, just attenuation. There's really no other way to do leveling in the digital domain because almost all material is mastered to peak at or near 0dBFS, so you can only "level" the volume by attenuating it (boosting it would cause clipping).

By contrast, night mode does perform dynamic range compression, but it sounds like that isn't the issue, necessarily?
Logged

tasar

  • World Citizen
  • ***
  • Posts: 148
Re: DSP Studio Quirks
« Reply #2 on: February 28, 2014, 03:04:13 pm »

Ok that makes sense as I did not think this was across the board "leveling ", but rather, an attenuation device to level volume disparity between tracs during play. When active, the loss created, requires too much attenuation with outside vol. Control. Not sure if DAC or digital control would fair better. Anyone find a reason to use "leveling" ?
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #3 on: February 28, 2014, 03:18:40 pm »

Ok that makes sense as I did not think this was across the board "leveling ", but rather, an attenuation device to level volume disparity between tracs during play. When active, the loss created, requires too much attenuation with outside vol. Control. Not sure if DAC or digital control would fair better. Anyone find a reason to use "leveling" ?

Volume leveling attenuates different tracks or albums by different amounts (depending on whether you're playing an album or a playlist), but it doesn't perform any kind of filtering or dynamic range compression.  It just turns the volume down by however much is needed to bring the current program material to a reference level.  

So it is an attenuation device to level volume disparity between albums or tracks.  If you're playing an album, it will not try to level each individual track because inter-track dynamics are part of the album; instead it will level the album as a whole so that it will be, as a whole, a similar loudness to any other album.  On the other hand, if you play a playlist, it will level each track individually so that all the tracks are around the same volume.

In either case, it's just turning down the volume because there's no other way to get similar volumes between albums or tracks.  Is the issue that you're getting too much attenuation with volume leveling?
Logged

tasar

  • World Citizen
  • ***
  • Posts: 148
Re: DSP Studio Quirks
« Reply #4 on: February 28, 2014, 07:26:36 pm »

Yes, as I measured over 5db loss with the SPL meter. I consider attenuation severe from a dynamics standpoint. On the other hand, other DSP enhancements are terrific, though "loudness" with internal volume setting yields little in my sensitive system.
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #5 on: March 01, 2014, 05:58:44 am »

The goal of Volume Leveling is to bring all playback to an even level, so that you don't have to touch the volume control on your amplifier - or at least, to minimize how often you have to use it.
 
Volume Leveling has a target of -23 LUFS (Loudness Units Full Scale) average volume.
-23 LUFS (dB) is used because this is about as loud as you can go before tracks with a lot of dynamic range start clipping.
 
So if the album you were testing showed a 5dB reduction in volume, its average level must be around -18 LU.
(actually, it may be that the loudest track on the album is -18 LU right now, rather than the average - a fix is in the works)
 
Some tracks/albums will be boosted in volume, and others will be reduced in volume to hit the -23 LUFS target.
 
A reduction in volume is not a reduction in "dynamics" or "dynamic range" - it is simply quieter, and turning up the amplifier will mean that playback is the same as it was before.
 
 
Using Adaptive Volume in the "Night Mode" or "Small Speaker Mode" will compress the dynamic range of the music being played though, to increase loudness at the expense of dynamics.
 
 
Yes, as I measured over 5db loss with the SPL meter. I consider attenuation severe from a dynamics standpoint. On the other hand, other DSP enhancements are terrific, though "loudness" with internal volume setting yields little in my sensitive system.
Loudness is effectively a bass boost that increases as you reduce the volume control in Media Center, to compensate for how our hearing responds to certain frequencies as playback gets quieter.
 
It should not be used without the amplfier set to a calibrated reference level, and Media Center being the only thing in control of playback volume.
 
I really think this control should be renamed, as many people enable it without knowing what they are doing.
Logged

tasar

  • World Citizen
  • ***
  • Posts: 148
Re: DSP Studio Quirks
« Reply #6 on: March 01, 2014, 09:04:51 am »

Most interesting is the comment above,'"inter track dynamics" for the album. Is this part of the "recording" or added along with studio room correction etc ? Though not part of MC discussion, it is relative to the ultimate sonics. If it is the physical recording or otherwise, I'll argue "leveling" has a detrimental affect to all red book and downloads, in a more or less degree. Mentioned above is that it does not attenuate dynamics. Perhaps not in a pure sense, but because it forces extensive analog volume compensation, I find the musicality becomes strident and dynamically distorted. FWIW, depending on origin, files purported to be "hi Rez" are in actuality, high manipulation, despite reported metrics. Leveling may work in an all digital rendering, but to my liking, quite unusable.
"Night mode" has it's attributes for classical low passages, and for those accustomed to low SPL listening. It may be a volume or gain feature, but surely adds "dynamics" where otherwise things become indiscernable. The nice thing is MC allows us this discretion, and sure beats the purchase of "hi-Rez" marketed files, which achieve nothing more. Thank you MC.
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #7 on: March 01, 2014, 09:25:32 am »

Most interesting is the comment above,'"inter track dynamics" for the album. Is this part of the "recording" or added along with studio room correction etc ?
"Inter track dynamics" with regard to leveling is that Track 3 is supposed to sound 3dB quieter than Track 2 on a disc etc.
If you used track-based leveling (adjust each track individually) rather than album-based leveling (adjust all tracks by the average gain) you would remove the 3dB difference between them.
 
Volume Leveling operates in album mode automatically when playing back albums, and preserves inter-track dynamics.

Though not part of MC discussion, it is relative to the ultimate sonics. If it is the physical recording or otherwise, I'll argue "leveling" has a detrimental affect to all red book and downloads, in a more or less degree.
Well technically you have reduced the signal to noise ratio by 5dB with digital attenuation, but that should not be audible at all with a 24-bit DAC.

Mentioned above is that it does not attenuate dynamics. Perhaps not in a pure sense, but because it forces extensive analog volume compensation, I find the musicality becomes strident and dynamically distorted.
That sounds like bad/faulty hardware if it sounds so much worse when you turn up the gain 5dB.
It should not affect anything other than being louder or quieter.

"Night mode" has it's attributes for classical low passages, and for those accustomed to low SPL listening. It may be a volume or gain feature, but surely adds "dynamics" where otherwise things become indiscernable.
Night mode makes playback louder at the expense of compressing the track dynamics. (making it less dynamic)
 
I think you are confusing volume (loudness) with dynamic range or "dynamics".
Logged

tasar

  • World Citizen
  • ***
  • Posts: 148
Re: DSP Studio Quirks
« Reply #8 on: March 01, 2014, 02:18:13 pm »

Signal to noise ratio 5dB down is certainly huge and agreed, dynamics is not part of loudness, the the "download" sites would have us believe so. Dynamics is the ability to hear subtle differences and not the domain of volume, agreed. One never knows how dynamic their system might be until the room SPLs are in the sub 70 range. I suggest the average audiophile listens in the 80 dB plus area, so never knows what their system adds. Hardware you suggest, I don't think so. Poor room FR and treatment ? Yes. Dynamic resolution resides with nuances of recording and subsequent algorithm playback. I'm most certain here is where the dots do not connect. This would be evident in even the worse hardware assembly. Sorry, but "leveling" lacks purpose(adds nothing) IMHO.
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #9 on: March 01, 2014, 02:55:44 pm »

Sorry, but "leveling" lacks purpose(adds nothing) IMHO.
I can leave my system alone and I don't have to worry about adjusting the volume +10 dB or -5 dB depending on what album is playing. Everything is played at an even level whether it's a modern "loud" album with compressed dynamics, or an older CD with good mastering.
 
I'm not sure what else you were expecting from it.
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #10 on: March 03, 2014, 07:19:44 pm »

Leveling is essential feature to me... as the previous poster stated quite simply you don't have to fiddle with your volume control on every track in a playlist which is a huge convenience. Sure ideally the full digital resolution would be passed to the DAC for any volume playback and the DAC would full scale that signal right through only then attenuating it in the final analog stage for best sound.  So the question is how do we make this happen across the entire audio chain when clearly most analog amps don't support this sort of control.  I think doing it digitally is fine considering its purpose is for convenience not necessarily best sound.

The only problem I have noticed with volume leveling and only with Media Center (iVolume + iTunes is much better) is that it does not maintain the same leveled volume across tracks in a playlist.  Perhaps this is a bug in the algorithm being used and it may be just for Play Doctor playlists which is really all I ever use.  I get quite a fluctuation of volume playing Play Doctor playlists.  Not sure though if this is fundamental to the algorithm used to determine perceived volume vs simple root squared or 'energy' analysis.  MC would do well to at the very least allow the leveling information that is provided by iVolume to tell it how 'loud' each track actually sounds.  I think this is a flaw in the algorithm being used which is not taking human perception of sound volume into account.

Would it be possible to publish the algorithm being used to determine the volume of a track for comparison?  I have a strong suspicion there are better algorithms for this then what is being used.
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #11 on: March 03, 2014, 07:33:53 pm »

Sure ideally the full digital resolution would be passed to the DAC for any volume playback and the DAC would full scale that signal right through only then attenuating it in the final analog stage for best sound.
Some people would argue that digital volume control, when properly implemented, surpasses all but the very best analog volume controls.
It's theoretically possible to create a better analog volume control, but you need extremely tight tolerances, and I don't know anything with measurably better performance than a properly implemented digital control.

The only problem I have noticed with volume leveling and only with Media Center (iVolume + iTunes is much better) is that it does not maintain the same leveled volume across tracks in a playlist.  Perhaps this is a bug in the algorithm being used and it may be just for Play Doctor playlists which is really all I ever use.  I get quite a fluctuation of volume playing Play Doctor playlists.
When playing back albums, a single value is applied to all tracks in the album. (which should be an average, but is currently based on the loudest track)
With mixed playlists, e.g. using the play doctor, tracks are leveled on an individual basis, which should put them on a more even level. The exception is when you have two tracks next to each other from the same album - it should then use the average value of the two. (though it may also use the value from the loudest track? I'd have to check)

Not sure though if this is fundamental to the algorithm used to determine perceived volume vs simple root squared or 'energy' analysis.  MC would do well to at the very least allow the leveling information that is provided by iVolume to tell it how 'loud' each track actually sounds.  I think this is a flaw in the algorithm being used which is not taking human perception of sound volume into account.

Would it be possible to publish the algorithm being used to determine the volume of a track for comparison?  I have a strong suspicion there are better algorithms for this then what is being used.
Media Center uses the EBU R128 specification (based upon ITU-R BS.1770) for volume leveling. This broadcast standard was developed after years of research on loudness perception, and should provide much better results than ReplayGain v1.
 
Have your tracks been analyzed by the Media Center analysis tool, or are you using files which already have ReplayGain data in them, or none at all? (it looks like this "iVolume" app uses ReplayGain v1)
The ReplayGain v2 spec is also based upon BS.1770, and should provide essentially the same results as Media Center's R128 implementation.
 
 
R128 has me reaching for the volume control a lot less than ReplayGain v1 did.
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #12 on: March 03, 2014, 07:34:26 pm »

Leveling is essential feature to me... as the previous poster stated quite simply you don't have to fiddle with your volume control on every track in a playlist which is a huge convenience. Sure ideally the full digital resolution would be passed to the DAC for any volume playback and the DAC would full scale that signal right through only then attenuating it in the final analog stage for best sound.  So the question is how do we make this happen across the entire audio chain when clearly most analog amps don't support this sort of control.  I think doing it digitally is fine considering its purpose is for convenience not necessarily best sound.

The only problem I have noticed with volume leveling and only with Media Center (iVolume + iTunes is much better) is that it does not maintain the same leveled volume across tracks in a playlist.  Perhaps this is a bug in the algorithm being used and it may be just for Play Doctor playlists which is really all I ever use.  I get quite a fluctuation of volume playing Play Doctor playlists.  Not sure though if this is fundamental to the algorithm used to determine perceived volume vs simple root squared or 'energy' analysis.  MC would do well to at the very least allow the leveling information that is provided by iVolume to tell it how 'loud' each track actually sounds.  I think this is a flaw in the algorithm being used which is not taking human perception of sound volume into account.

Would it be possible to publish the algorithm being used to determine the volume of a track for comparison?  I have a strong suspicion there are better algorithms for this then what is being used.

The algorithm JRiver uses is the R128 standard http://tech.ebu.ch/loudness. It's based on the current international loudness standard for broadcast, and, in my experience, does a better job of volume leveling than anything else I've tried.  The following two papers by the EBU describe it in more detail:

https://tech.ebu.ch/docs/tech/tech3341.pdf
https://tech.ebu.ch/docs/tech/tech3342.pdf

It's odd that you're experiencing significant volume variations in playlists; playlists are where the leveling is likely to work best.  You might want to make sure that you don't also have adaptive volume enabled (as that will tend to defeat volume leveling), and also make sure that all your files have been analyzed.

[EDIT: Beaten to the punch ;D, but I'll leave my post so you've got the links to the description of the algorithm]
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #13 on: March 03, 2014, 09:02:14 pm »

I do have adaptive volume enabled with Peak Level Normalization as it says 'works intelligently with volume leveling' otherwise the overall gain is too low causing me to have to crank the amp up outside of it's nominal range resulting in a lesser sound quality overall.  This may be the same issue the OP is discussing.  Perhaps it is the adaptive volume in addition to the leveling causing the issue and perhaps there is a bug in this 'Peak Level' setting.

Previously to MC all tracks in my library were processed by iVolume which with iTunes does a great job keeping my hand off the dial.  It has always given a very consistent experience.  I find with MC after playing music for some time it gets very quiet almost progressively and then I inch up the volume and invariably a track comes on up loud again.  This effect of progressive lowering of volume is not subtle either, it can be pretty dramatic over time. 

I think there is a bug somewhere in the implementation to be honest but being such a subjective thing it may never be acknowledged until it is found and fixed by accident.
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #14 on: March 04, 2014, 03:18:13 am »

I do have adaptive volume enabled with Peak Level Normalization as it says 'works intelligently with volume leveling' otherwise the overall gain is too low causing me to have to crank the amp up outside of it's nominal range resulting in a lesser sound quality overall.  This may be the same issue the OP is discussing.  Perhaps it is the adaptive volume in addition to the leveling causing the issue and perhaps there is a bug in this 'Peak Level' setting.
Volume Leveling on its own will level every track to have an average volume of -23dB. This is, if I recall correctly, about 5dB lower than the original ReplayGain spec.
The reason for this, is that the ReplayGain target was too high and would often run into clipping issues. (either you clip the peaks, or have uneven leveling and reduce the gain of a track below your target level)
 
If you have a fixed playlist, adding Adaptive Volume in Peak Normalization mode to this will then check the current playlist and maximize the volume if there is any additional headroom. In the Audio Path you will see that it is operating in "fixed" mode. This causes the volume to be different across playlists, but it should be level for the current playlist.
 
If you are using a dynamic playlist such as Play Doctor, Adaptive Volume works in the old "adaptive" mode, on a per-track basis. This effectively disables volume leveling entirely.
 
I find with MC after playing music for some time it gets very quiet almost progressively and then I inch up the volume and invariably a track comes on up loud again.  This effect of progressive lowering of volume is not subtle either, it can be pretty dramatic over time.

I think there is a bug somewhere in the implementation to be honest but being such a subjective thing it may never be acknowledged until it is found and fixed by accident.
Check the Audio Path as this happens. You will see that the adaptive mode is constantly pushing up the gain, and you will probably see clipping protection engaged as well.

…the overall gain is too low causing me to have to crank the amp up outside of it's nominal range resulting in a lesser sound quality overall.
Does it actually sound worse, or is the amp just at a higher level than you are used to? (and therefore "comfortable" with)
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #15 on: March 04, 2014, 09:46:36 am »

I do have adaptive volume enabled with Peak Level Normalization as it says 'works intelligently with volume leveling' otherwise the overall gain is too low causing me to have to crank the amp up outside of it's nominal range resulting in a lesser sound quality overall.

As 6233638 noted, in a play doctor playlist, peak level normalization will completely defeat volume leveling (i.e. no leveling will occur).  That's why you're getting inconsistent results.  

When you say you had to push your amp out of its  "nominal range" what do you mean exactly?  Regardless of what the volume setting on the dial is telling you, with most speakers a given volume output is directly equivalent to a given power input; passive speakers are just transducers.  So whether the dial is set to 1 or 10, if your speakers are outputting the same volume, the amp is putting out the same power (barring something bizarre happening).  

A concrete example:  let's say you have 90dB/1W/1 meter sensitivity speakers.  If, without volume leveling, you typically set the amp volume dial to "4" and the speakers put out about 90dB measured at 1 meter away, you know that your amp is actually putting out about 1 watt of power at that volume setting.  If you turn on volume leveling, the amp's input signal becomes much quieter, so you have to turn your amp volume dial up to "11" to get the same speaker volume (90dB).  Because your speakers are still outputting 90dB that means your amp is putting out the same 1 watt of power it was putting out when the dial was at 4.  

Barring something not functioning as intended (flaky volume knob and/or bad volume control design in the amp*) there shouldn't be any additional load on the amp or change in distortion between those two scenarios.  For most amps, if the SPL in the room from the speakers is the same, the amp's output power (and therefore distortion levels) should be identical or close to it.  

Could you elaborate on what kind of sound quality problems you're experiencing?

*With certain (non-ideal) volume control designs, turning the knob all the way up might result in an increase in the audibility of the amp's noise floor, but you can test that by listening for hiss in your speakers when nothing is playing while turning the volume knob on the amp. If there's no hiss, or what hiss is there remains the same volume regardless of where the knob is set, then that isn't an issue.
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #16 on: March 09, 2014, 10:45:31 pm »

Yes what I meant was that without the adaptive volume enable (aside the volume leveling) the volume is so low that I have to turn the knob of the amp to the 3/4 position just to get normal playback volume out of my WA7 amp.  Without any of that I can use the amp at 11 oclock position and in general the sound is much better then it is at 3/4 position.  Seems to have more dynamic range, a bit more punch in the bass and differences between soft and hard music are more significant, not as compressed persay.

I think what I am noticing is what you are now saying that during Play Doctor playlists the volume leveling is completely defeated.  This is basically what it sounds like to me and not at all what I wanted.  Ideally I want Play Doctor playlists to have the same leveling + adaptive volume as the static playlists.  I mostly listen to play doctor and 'need' the leveling and adaptive volume to give a consistent experience without the gain issue of leveling alone.  Is it possible to look ahead a dozen tracks and do your adaptive gain accordingly?  I do see it already knows what tracks are in line to be played up to a hundred ahead so it seems possible to assume those define the playlist same as static.
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #17 on: March 09, 2014, 10:48:39 pm »

... by the way JRiver sounds awesome relative to iTunes and some of the other higher end players out there.  There was recently a thread on one of the audiophile forums comparing the different players for the mac and the conclusion through various ABX tests was that while bit-perfect is more or less the same across all players the real magic is in the quality of decoding being performed for lossy track formats.  It was found that JRiver delivers the most precision in decoding higher frequencies and results in a less sibilant playback.
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #18 on: March 10, 2014, 07:06:39 am »

Yes what I meant was that without the adaptive volume enable (aside the volume leveling) the volume is so low that I have to turn the knob of the amp to the 3/4 position just to get normal playback volume out of my WA7 amp.  Without any of that I can use the amp at 11 oclock position and in general the sound is much better then it is at 3/4 position.  Seems to have more dynamic range, a bit more punch in the bass and differences between soft and hard music are more significant, not as compressed persay.
Human perception is funny in that a small difference in volume is enough to change how we perceive the music. If things are quieter, there doesn't seem to be as much bass and the sound is "flatter".
So when you introduce Volume Leveling and reduce the volume a lot, you need to be sure that you are turning up the amplifier an appropriate amount to do a properly level-matched comparison.
 
Whether your amplifier is at 11 o'clock or 3/4 of the way up, if it's producing the same volume, it's putting the same wattage into your speakers. It isn't being driven any harder.
 
You may have heard of the "loudness war" which is the fact that over time music has had its dynamic range more and more compressed.
This is done to make the track play louder, at the expense of sucking out all the dynamics. Quiet sounds are made louder, and loud sounds are made quieter so that it's just a wall of noise instead of being dynamic.
 
This is done for exactly the reason mentioned above - louder often sounds "better".
 
 
Volume Leveling directly counteracts this by making tracks play at the same volume level regardless of their dynamic range.
It has to bring the average volume down to -23 dB so that it has the headroom to do this without introducing clipping on highly dynamic tracks.
Highly dynamic tracks may actually have their level raised by volume leveling, but tracks without much dynamic range at all (modern masters) may have their volume reduced by 10 or even 15 dB if they are particularly bad.
 
Once everything is at the same level, these modern masters no longer have a loudness advantage. They are played back at the same level as well mastered music.
When you do a volume matched comparison, these tracks which have had all the dynamics sucked out to sound loud at all costs, sound completely flat compared to a good dynamic master.
 
Nothing about the volume leveling has changed how they sound, it's just that you are now able to compare things on an even level, rather than covering things up by being loud.
If you add the "Dynamic Range (DR)" field to your views, it will give you an idea of how well mastered the music is.
Good masters start about the 11-12 range. Modern, highly compressed, flat-sounding masters are often 6 or lower.

I think what I am noticing is what you are now saying that during Play Doctor playlists the volume leveling is completely defeated. 
It is completely defeated if you combine it with adaptive volume in a Play Doctor playlist, as that will put it into adaptive mode rather than fixed mode.
If you use Volume Leveling on its own, it will level each track to -23 LU individually which should do a very good job of keeping things on the same level.
If you use Adaptive Volume on its own, it will use peak level normalization in a fixed mode. (though this will not level the playlist)

Ideally I want Play Doctor playlists to have the same leveling + adaptive volume as the static playlists.
This is not really possible, because the Play Doctor list of tracks is constantly changing, so the average volume and peak level of the playlist is constantly changing too.
However, if you're looking at the average level of say the previous 10 tracks and the next 100 in the list, it probably won't change in volume nearly as much as the adaptive mode.
The problem is that one track could throw off the volume of the entire playlist.
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #19 on: March 10, 2014, 08:54:17 am »

It should just assume at any time the playlist is a static playlist of say the next 12 or more songs it has conjured up and that should give it enough songs to ramp the volume slowly up or down from the average as songs are played and added to the list.  disabling it isn't really the expected approach.
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #20 on: March 10, 2014, 09:53:06 am »

It should just assume at any time the playlist is a static playlist of say the next 12 or more songs it has conjured up and that should give it enough songs to ramp the volume slowly up or down from the average as songs are played and added to the list.  disabling it isn't really the expected approach.
You can't do that because using the average volume for all the tracks in the lists mean that tracks which are more dynamic, are likely to run into clipping.
When you use peak level normalization in the fixed mode, and are updating it based on a changing playlist, a single track can throw off the volume wildly.
 
The images attached to this post show what happens when I play an album using leveling and peak level normalization in the fixed mode, and then add an extra track from another album to it.
I had to stop and start playback for it to change, but this is what would happen if it adjusted based on X number of tracks in a constantly changing view like Play Doctor.
 
Adding one extra track from another album by the same artist dropped the volume of the entire playlist by 7.1dB!
If you want a level volume in a mixed playlist, you need to only use Volume Leveling (and not adaptive volume) and turn up your amp.

You could try boosting the volume in the parametric EQ by say 6dB, but you run the risk of clipping protection engaging and turning down the volume mid-track on very dynamic tracks.
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #21 on: March 10, 2014, 10:04:40 am »

It should just assume at any time the playlist is a static playlist of say the next 12 or more songs it has conjured up and that should give it enough songs to ramp the volume slowly up or down from the average as songs are played and added to the list.  disabling it isn't really the expected approach.

[EDIT: beaten to the punch again  ;D I'll leave the example up, though]

I don't think the rolling average would solve the problems you're describing.

An example from a recent play doctor list I generated: four tracks in a row that require the following (rounded) volume leveling to sound about "the same" volume 1.-13dB / 2.-5dB / 3.-15dB / 4. -7dB.  There's no way to make those four tracks sound close to the same volume in close succession without just reducing all of them by those amounts.  A "rolling average" peak level normalization like you're describing would still result in significant perceived volume changes between those tracks because the tracks all have vastly different volumes going in alternate directions so there's no way for the average to "inch" up or down in a meaningful way.  The "rolling average" method would also have lower overall volume than pure peak level normalization (in this example it would be close to 10dB lower than pure peak level normalization for track 3), so it might not actually help much with your low volume concerns.

Long story short, the "rolling average" might result in slightly more level volume than the current peak level normalize/volume leveling interaction, but it still wouldn't work very well, especially with play doctor lists that include music with very different average volumes.  Volume leveling without adaptive volume is the best way to get a completely random set of tracks to sound about the same volume, which is why it's implemented the way it is.  
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #22 on: March 11, 2014, 07:54:32 pm »

I get the idea but a playlist doctor playlist has hundreds of tracks not a handful.  I would be happy with a non aggressive rolling (weighted toward the song you are on obviously and tapering off the further away the 'problem' song is) then simple leveling because the analog amplifier has a fundamentally different gain linearity across frequency at different volume settings not considering the fact that with increased overall analog gain you are bumping up the noise floor of the amp considerably.  It just seems strange to have a feature such as adaptive volume if it doesn't do anything in one particular play mode when it does in another play mode that is relatively the same, static vs dynamic playlist which could have hundreds of songs in it either way.  Assuming you are convinced it doesn't make sense to make this work as expected is it possible to easy convert a play doctor playlist into a static playlist and therefore get the advantage of the adaptive volume?
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #23 on: March 11, 2014, 08:14:57 pm »

...anyway moving on I think I just found the real problem here... so I turned off adaptive volume and no matter what song i play in my library no matter how, they are all given the same -10dB leveling.  This is why I notice the issue I think in the first place volume leveling just isn't working.

Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #24 on: March 11, 2014, 08:22:35 pm »

...anyway moving on I think I just found the real problem here... so I turned off adaptive volume and no matter what song i play in my library no matter how, they are all given the same -10dB leveling.  This is why I notice the issue I think in the first place volume leveling just isn't working.



It sounds like you might not have analyzed your audio tracks?  Volume leveling only applies the -10dB default when the tracks aren't analyzed.  To check, select some tracks, right click and click analyze audio (under library tools).   It should show whether they're analyzed or not. You can set JRiver to analyze tracks when you import them so you don't have to do it manually all the time.

Quote
analog amplifier has a fundamentally different gain linearity across frequency at different volume settings not considering the fact that with increased overall analog gain you are bumping up the noise floor of the amp considerably.

On the majority of amplifiers I've personally used, the volume control does not adjust the gain of the amplifier circuit, it just applies attenuation to the audio signal, adjusting the "effective gain."  Put another way, the amp's gain is fixed, the volume knob just attenuates the audio signal somewhere along the way.  For most of the amps I've used, turning the volume knob does not affect the noise floor, because the volume control attenuator is situated before the amplification stage (or between the buffer and main amplification stage).  In that topography the volume knob will have a very limited effect on the amp's noise floor in any position (although it may make a noisey DAC prior to the amp more audible when opened up all the way).  

If the amp has the attenuator situated after the main amplification stage, it would definitely make the noise floor of the amp more audible if you opened it up, but it would also likely bork the amp's output impedance and damping factor (without some clever engineering or a buffer stage after the volume control), so I'm not aware of too many amplifiers that have the volume control positioned after the main gain stage (it's more common with pre-amps putting out a line level output for obvious reasons).  

There are other topographies that work a little differently, and some amps do have a knob that actually adjusts the gain of the circuit, but those knobs are typically labelled "Gain" and not "volume," in my experience.  Unless your amp actually has truly adjustable gain or a specific kind of volume control design you should lose very little by turning the knob to 3 o'clock rather than 9 o'clock if the output volume is the same.  

An easy test is to unplug the inputs on your amp, then turn the amp on and sweep the volume knob.  If you don't hear significant changes in the noise floor, you know your amp doesn't much care where the knob is.
Logged

tasar

  • World Citizen
  • ***
  • Posts: 148
Re: DSP Studio Quirks
« Reply #25 on: March 11, 2014, 08:50:33 pm »

Sandler once said, " that would have been nice to know.... Yesterday!" So, do we dare apply analyses to already created library files, all at once or must it be done during playback ?
Dynamically, Adaptive mode seems to do the trick alone, and I would rather hear/attenuate volume based on mood, especially classical. Remotes make one savy !
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #26 on: March 11, 2014, 08:54:24 pm »

Sandler once said, " that would have been nice to know.... Yesterday!" So, do we dare apply analyses to already created library files, all at once or must it be done during playback ?
Dynamically, Adaptive mode seems to do the trick alone, and I would rather hear/attenuate volume based on mood, especially classical. Remotes make one savy !

You need to do it prior to playback, but it's non-destructive (it just writes some tags to the files that volume leveling reads at playtime).

You can do your whole library at once if you want, it will just take a while (think hours not minutes, more for a very large library).  When the feature was introduced, I did my library in "batches" of a few thousand files at a time.  You can make a smartlist to tell you which files haven't been analyzed yet (an easy rule to show you unanalyzed files is "volume leveling (R128) is <empty>")
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #27 on: March 12, 2014, 09:02:28 am »

I get the idea but a playlist doctor playlist has hundreds of tracks not a handful.
The more tracks you introduce, and the more varied they are, the more it will trend towards -23dB.
I agree that there is possibly a solution there which is better than the current adaptive volume working in the adaptive mode, but I don't know that the JRiver team want to pursue it when the better solution is to disable adaptive volume and turn up your amplifier.

It just seems strange to have a feature such as adaptive volume if it doesn't do anything in one particular play mode when it does in another play mode that is relatively the same, static vs dynamic playlist which could have hundreds of songs in it either way.
I agree that it's confusing when it switches from static to adaptive mode, and the options don't make that clear.

Assuming you are convinced it doesn't make sense to make this work as expected is it possible to easy convert a play doctor playlist into a static playlist and therefore get the advantage of the adaptive volume?
If you only want to play 100 tracks that would probably work I suppose. I'm sure there must be an easy way to save it as a static playlist.

...anyway moving on I think I just found the real problem here... so I turned off adaptive volume and no matter what song i play in my library no matter how, they are all given the same -10dB leveling.  This is why I notice the issue I think in the first place volume leveling just isn't working.
You need to analyze your library before volume leveling will work.
 
Select your files and use: Tools > Library Tools > Analyze Audio to do this.
This might take a while. I find that analyzing 3 files at once seems to be the sweet spot on my system, but it's a quad-core PC. (more than 3 files at once and the hard drive slows things down)

Dynamically, Adaptive mode seems to do the trick alone, and I would rather hear/attenuate volume based on mood, especially classical. Remotes make one savy !
If your files are not analyzed, Adaptive Volume will still be operating in the adaptive mode and not the static mode when you disable leveling.
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #28 on: March 12, 2014, 06:45:17 pm »

hahaha... good to know now.  I guess I assumed they are analyzed during playback from something I had thought I read in the help or docs, or maybe because I already had all the tracks analyzed by both iTunes and by iVolume I maybe assumed it would use those settings.  Third tool third standard I guess.  Might should ask you if you want to analyze your files when you enable volume leveling or at the very least pop up saying "X of the Y tracks in your library have not been analyzed for volume leveling.  Would you like to do that now?"  This explains alot of confusion for me around why nothing sounds right with volume leveling or adaptive volume.  If I didn't realize this I wonder how many other people have done the same and wondered why it doesn't seem to work.
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #29 on: March 12, 2014, 06:51:24 pm »

... when bringing up the volume leveling perhaps to also have an option to use iTunes or iVolume settings instead or in the event a track has those settings but no MediaCenter analyzed setting.  Also a warning in the Audio Path box to say in bold red (volume not analyzed) next to the -10db.  Click on the link and it analyzes it.  Maybe even put the red link or button in the main window next to the volume control while playing a song if leveling is turned on but the track is not analyzed yet.

As an aside, is this analysis result stored in the music file as a tag or is it in some MediaCenter database for the library?  Wondering if any other players will be able to read this setting or perhaps even write over it while doing other tagging from say iTunes or some other player/tagger.  Sounds like my tracks now have 3 separate volume leveling tags in them which are not cross compatible between players.  For example can I get iTunes to use the MediaCenter analyzed setting like it does the iVolume one?

I found the reason why I assumed the analysis was done during playback.  It came from your description of Volume Leveling 'The industry leading R128 algorithm is used to analyze files and apply the correct volume during playback'.  The way I read that implied it did both analysis AND applying the volume 'during playback'.  You might want to re-write that to make it clear the analysis needs to be done separately and manually.

... one more thing I noticed is when you do analyze the files it defaults to 2 files at a time even though i have 4 core / 8 virtual cpus.  think it should default to the number of virtual cores the machine has.  makes the whole process much faster on a machine with more then 1 core.
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #30 on: March 12, 2014, 07:17:36 pm »

As an aside, is this analysis result stored in the music file as a tag or is it in some MediaCenter database for the library?  Wondering if any other players will be able to read this setting or perhaps even write over it while doing other tagging from say iTunes or some other player/tagger.  Sounds like my tracks now have 3 separate volume leveling tags in them which are not cross compatible between players.  For example can I get iTunes to use the MediaCenter analyzed setting like it does the iVolume one?

Media center writes the results of the audio analysis to several tags. Volume level (r128) is the important one for JRiver, but it writes five or six different analysis tags.  One of those tags is the volume level (replay gain) tag.  Replaygain is one of the most widely used volume levelling algorithms, and many other players will read that tag (see here for a list http://en.wikipedia.org/wiki/ReplayGain#Audio_players). JRiver doesn't use that tag anymore, but it still writes to it for compatibility with other players.  

IVolume is actually based on the replaygain standard, but it looks like itunes writes it to a custom tag (itunenorm) rather than the standard replaygain tag, which means iVolume may not work in any player other than iTunes.

Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #31 on: March 12, 2014, 08:50:52 pm »

... one more thing I noticed is when you do analyze the files it defaults to 2 files at a time even though i have 4 core / 8 virtual cpus.  think it should default to the number of virtual cores the machine has.  makes the whole process much faster on a machine with more then 1 core.
The analysis is multithreaded, so even if you are analyzing a single file it should use all your cores.
 
What I find is that the hard drive is usually the limiting factor.
 
On my system, anything more than three files at once actually slows down the process, because the CPU can't get data quickly enough. (3 files = 100% CPU usage, 4-8 files results in less than 100%)
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #32 on: March 13, 2014, 07:54:19 am »

Volume leveling works a lot better now that it has analyzed files hahaha  ::)
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10968
Re: DSP Studio Quirks
« Reply #33 on: March 13, 2014, 08:25:23 am »

Sometimes its the small things  ;D
Logged
~ nevcairiel
~ Author of LAV Filters

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #34 on: March 17, 2014, 09:23:07 pm »

Is there any way to turn off the fact that volume leveling keeps volume levels for songs on the same album the same?  The problem I am having is the volume can sometimes change drastically between songs on an album perhaps as the artist intended but I would rather all songs be at the level that is right for making is the same volume regardless.  In other words apply each song's volume leveling without regard for the album so all songs are at the same volume?
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #35 on: March 18, 2014, 09:14:37 am »

I believe Hendrik said he would look into this.
For now, a temporary workaround would be to set the Media Sub Type tag to "Podcast" instead of "Music"
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10968
Re: DSP Studio Quirks
« Reply #36 on: March 18, 2014, 09:42:21 am »

Sounds like he wants that for everything, not just some special mix albums, which the previous discussion was always about.
Logged
~ nevcairiel
~ Author of LAV Filters

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #37 on: March 18, 2014, 10:29:17 am »

Sounds like he wants that for everything, not just some special mix albums, which the previous discussion was always about.
The only solution I can think of for that would be a preference in the Volume Leveling DSP to switch between Automatic and forced Track-based Leveling. :-\
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5239
  • "Linux Merit Badge" Recipient
Re: DSP Studio Quirks
« Reply #38 on: March 18, 2014, 10:35:19 am »

The only solution I can think of for that would be a preference in the Volume Leveling DSP to switch between Automatic and forced Track-based Leveling. :-\

That's sort of how it used to work in MC 18, right?  The album mode versus track mode option in volume leveling was one of the options that was removed in the transition to 19, because the automatic mode was being introduced (I think).  

It's sounding like the automatic mode isn't quite one size fits all, although it certainly fits me (FWIW, I haven't missed those options, personally).
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #39 on: March 18, 2014, 11:21:39 am »

That's sort of how it used to work in MC 18, right?  The album mode versus track mode option in volume leveling was one of the options that was removed in the transition to 19, because the automatic mode was being introduced (I think).
MC18 had options for album-based leveling or track-based leveling.
MC19 is automatic, and switching between album and track-based leveling is calculated based on a number of variables.
 
The way that MC19 handles "album" playback is a lot smarter than how MC18 handled it for example.
But it seems like an option to switch between "automatic" and "track-based" may be unavoidable.
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #40 on: March 19, 2014, 05:35:05 pm »

Yeah it definitely needs the 'per track' or 'per album' options
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #41 on: March 19, 2014, 07:39:06 pm »

Yeah it definitely needs the 'per track' or 'per album' options
Media Center calculates album gain on-the-fly rather than when the files are analyzed, so your options would be "Automatic" or "Track-based" if changes are made to this.
Logged

groovyd

  • Guest
Re: DSP Studio Quirks
« Reply #42 on: March 19, 2014, 09:05:20 pm »

+1 those are excellent options, but I would recommend explaining what 'automatic' means as it is kind of a generic term.
Logged

tasar

  • World Citizen
  • ***
  • Posts: 148
Re: DSP Studio Quirks
« Reply #43 on: March 20, 2014, 08:47:16 pm »

Still a quirk in my way of thinking. The leveling attenuation technique should be file vs total file population, based on the playlist or genre or whatever one has queue up. Somehow it should find a "floor" from a buffered sampling, and playback at same level across the board. I started this thread feeling the musicality was "crushed" and still feel a sense of lifelessness in some tracks totally unrelated to gain or dynamics. Is their a "flat" base lining  organic to the algorithm responsible for attenuation ?
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: DSP Studio Quirks
« Reply #44 on: March 21, 2014, 12:15:50 am »

The leveling attenuation technique should be file vs total file population, based on the playlist or genre or whatever one has queue up.
This is exactly what happens when you are in a fixed playlist.
It has already been explained why this cannot happen in a dynamic playlist.

Somehow it should find a "floor" from a buffered sampling, and playback at same level across the board.
You do not seem to understand that the peak levels can vary hugely from track-to-track.

I do think the behavior could be improved in a dynamic playlist, by looking at the upcoming tracks and adjusting the level on a gradient rather than switching back to the old adaptive mode, but I don't know that it is something the JRiver team plans to work on.

As it is now, you can either have excellent leveling to -23 LUFS by using Volume Leveling on its own, or loud playback using Adaptive Volume on its own.
 
Using the two combined in a dynamic playlist (i.e. Play Doctor) will put it into the old adaptive mode which constantly pushes the gain up and can cause the volume to change greatly even in the same song.

I have suggested that you try enabling Leveling with a +6dB gain in the Parametric EQ - have you tried this yet?

I started this thread feeling the musicality was "crushed" and still feel a sense of lifelessness in some tracks totally unrelated to gain or dynamics. Is their a "flat" base lining  organic to the algorithm responsible for attenuation ?
All that Volume Leveling does is adjust the volume up or down. It does nothing which affects dynamics.
 
If you run out of headroom (extremely rare with the -23 LUFS target and music playback) then the playback volume will be lowered causing that track to be somewhat quieter than the target, rather than clipping or compressing its dynamic range.
Logged
Pages: [1]   Go Up