INTERACT FORUM
More => Old Versions => Media Center 15 (Development Ended) => Topic started by: Qythyx on July 27, 2010, 06:45:40 pm
-
Has the volume leveling algorithms changed much in the past few weeks/months? As I listen to my music recently I get the impression this feature is not working nearly as well as it used to. Some tracks are clearly signigicantly louder or quieter than others.
I've double checked my settings and that the tracks have been analyzed.
Any ideas?
-
There have been no changes to the Replay Gain system.
I can't think what to suggest other than to double-check settings and enabled DSPs.
-
So, I looked at this a bit more and I still think something's wrong. I was listening to a mix of diverse tracks and 2 songs back-to-back were dramatically different volume.
The first was Elton John's Rocket Man, replay gain was analyzed as -3.45 dB.
The second was INXS' Heaven Sent, replay gan was analyzed as -14.54 dB.
When these played Rocket Man was significantly louder than Heaven Sent. I tried to reanalyze the 2 tracks and the numbers came out the same.
I explored some more and tried playing tracks with a particular low replay gain (around -2 or -3) and particularly high (-14 or -15) and this behavior was consistent. The low gain tracks were much louder.
Is there anything else I can do to investigate this?
-
I've double checked my settings and that the tracks have been analyzed.
What do you have set in DSP Studio > Volume Leveling ?
The first was Elton John's Rocket Man, replay gain was analyzed as -3.45 dB.
The second was INXS' Heaven Sent, replay gan was analyzed as -14.54 dB.
Could you post also the Peak Level and Album Gain values?
-
So, I looked at this a bit more and I still think something's wrong. I was listening to a mix of diverse tracks and 2 songs back-to-back were dramatically different volume.
The first was Elton John's Rocket Man, replay gain was analyzed as -3.45 dB.
The second was INXS' Heaven Sent, replay gan was analyzed as -14.54 dB.
When these played Rocket Man was significantly louder than Heaven Sent. I tried to reanalyze the 2 tracks and the numbers came out the same.
I explored some more and tried playing tracks with a particular low replay gain (around -2 or -3) and particularly high (-14 or -15) and this behavior was consistent. The low gain tracks were much louder.
Is there anything else I can do to investigate this?
For kicks - I pulled the same two tracks. My Heaven Sent (From Greatest Hits - German Edition) checked in at -14.53 and my Rocket Man (From Honky Chateau - Mobile Fidelity) was -6.65. My RG settings are "track based" and my Adjustment is set to Fixed with a setting of +6. Both tracks are in FLAC format and I am using MC 15.0.80
These two played back just fine back to back with no jarring differences - except for the fact that Heaven Sent grinds away vs the relative calm of RM...
Cheers,
VP
-
Rocket Man: peak (69%), AG (-5.14)
Heaven Sent: peak (98%), AG (-13.45)
I was originally playing with a fixed setting of +8, but I also tried setting it to auto.
I assume it shouldn't matter which output mode I'm using. Currently it is WASAPI, but I can try others.
-
I was originally playing with a fixed setting of +8...
The "Overflow Handling: Clip Protection" setting may limit the possible positive gain.
Rocket Man (in the track based mode)
- not possible to adjust by +8 dB without causing clipping:
8 dB - 3.55 dB = 4.55 dB
If my calculation is correct, +4.55 dB multiplies the peak value by ~1.69
1.69 x 0.69 = ~1.17 aka ~117% (i.e. it would clip)
If you have Clip Protection set, the maximum possible positive gain is +3.22 dB => ~100% peak value.
However, this should make Rocket Man relatively a bit quieter than it would be when no fixed adjustment is set in the options, not louder.
Heaven Sent (in the track based mode)
- can be adjusted by +8 dB without causing clipping because the total gain change stays negative:
8 dB - 14.54 dB = -6.54 dB
I assume it shouldn't matter which output mode I'm using. Currently it is WASAPI, but I can try others.
I can't see how the selected output mode could affect the behavior. The signal processing happens before the output.
Have you tried to switch Volume Leveling off and simply adjust the volume slider by about 11 dB between the two tracks? I mean, maybe the measurement is not correct in this case and the perceptual difference is actually quite a bit less than about 11 dB.
Edit: MC's volume slider doesn't show dB values, but you can use the preamp setting in the Equalizer DSP.
-
Hmm, I think I might have the answer.
I tried turning off Volume Leveling and that actually seemed to make everything play at quite similar levels. Well, how can that be without everyone noticing? Well, here's my thought...
I'm listening via Library Server. (Almost) All of my tracks are stored as lossless WMAs, but I have Library Server configured to convert them to MP3s in order to stream them over the internet at lower bandwidth. Is it possible that MC is "applying" the existing RG information when it reencodes them? If so, then why my Library Server client machine plays the tracks it would actually be double-applying the adjustments.
Is this possible? Is there a setting for this?
-
Hmm, I think I might have the answer.
I tried turning off Volume Leveling and that actually seemed to make everything play at quite similar levels. Well, how can that be without everyone noticing? Well, here's my thought...
I'm listening via Library Server. (Almost) All of my tracks are stored as lossless WMAs, but I have Library Server configured to convert them to MP3s in order to stream them over the internet at lower bandwidth. Is it possible that MC is "applying" the existing RG information when it reencodes them? If so, then why my Library Server client machine plays the tracks it would actually be double-applying the adjustments.
Is this possible? Is there a setting for this?
Yes, I think it is applying replay gain when transcoding for serving. This is done so Replay Gain works nicely on a DLNA device.
But I don't think it should do this for Library Server, so let us take a look.
-
Hmm, I think I might have the answer.
I tried turning off Volume Leveling and that actually seemed to make everything play at quite similar levels. Well, how can that be without everyone noticing? Well, here's my thought...
I'm listening via Library Server. (Almost) All of my tracks are stored as lossless WMAs, but I have Library Server configured to convert them to MP3s in order to stream them over the internet at lower bandwidth. Is it possible that MC is "applying" the existing RG information when it reencodes them? If so, then why my Library Server client machine plays the tracks it would actually be double-applying the adjustments.
Is this possible? Is there a setting for this?
For the record: I'm playing my music via a shared HDD over my home network. While I'm experiencing the same problem described in this post (difference in volume even with RG), turning off the "Volume Leveling" option dose indeed result in bigger gaps of volume between each tracks in a playlist.
My settings are:
Mode - Album based
Ajustement - Fixed at 0.0dB
Clip protection - ON
However, i'm using a VST EQ in the DSP path and I found that reducing the output gain of this plug-in, solves the problem.
My question is: Since each file have is own RG data, maybe pluggins order in the DSP path is significant and could result in salmming by the clip protection feature ? Should we put the "Volume Leveling" plugin at the end or at the begining of the DSP path?
-
does it also happen with track based volume leveling? album gain is dependend on the other tracks on the album. when you have a silent track on an album, for instance, it will add a boost to the other tracks. or as what happened to me ones:
now this 51dB
i was just listening to a list with random songs at a low volume, it is nearly one at night here, when suddenly one song jumped at eardeafening volume out of my speakers.
the song that did this is from an album with 86 tracks from which 72 or so are silent tracks from 6 seconds. and since i play with album based volume leveling the song came with a staggering boost from 40 something instead of the -14 it had before the silent tracks were analysed. ah well, hearing the girl from above stumbling again through the house means she survived it.
i do not understand why the silent tracks need this 51dB, they are silent anyway and it makes the album gain less usefull.
thanks
gab
so check the other tracks from albums when you hear some differences, or use tracks based.
:)
gab
-
Should we put the "Volume Leveling" plugin at the end or at the begining of the DSP path?
It is always at the beginning, regardless of the order in DSP Studio. This is because it must operate on file boundaries, unlike other DSPs that operate on a stream of audio data (that may contain a cross-fade, etc.)
If 'Clip protection' is causing volume differences, it means you're overblowing the signal with the VST DSP. I would try to adjust the VST so it didn't do this. You could otherwise add a large fixed attenuation in Volume Levelling so even after the VST overblows the signal, it's still below 100%.
-
Yes, I think it is applying replay gain when transcoding for serving. This is done so Replay Gain works nicely on a DLNA device.
But I don't think it should do this for Library Server, so let us take a look.
We checked, and it is not applying replay gain when converting files and serving them over Library Server. So I don't think this is the issue. Gappie's advice about trying track instead of album gain is good.
-
We checked, and it is not applying replay gain when converting files and serving them over Library Server. So I don't think this is the issue. Gappie's advice about trying track instead of album gain is good.
Track gain works perfectly here. My two main libraries exist on a server and we connect via file share to play everything on all workstations. Each instance of MC uses "track gain" applied with Adjustment set to a Fixed quotient of +6.00. I can play the nastiest metal right next to the most mellow jazz with no reaching for the volume control ever.
Of course - replay gain cannot account for the jarring segues of two such polar opposite styles :)
Replay gain is such a vital and important part to our listening that I would notice immediately if something was off. MC total rocks in this department and has - by far - the best implementation of this feature.
Cheers!
VP
-
i play mostly with album gain, and when i play a mixed list i know where the difference might come from. it might be interesting to see what the differences are between the album gain and the replay gain. make a smartlist with audio only and add an expression collumn.
formatnumber(math([replay gain,0]-[album gain,0]),2)when you sort by this new column you can see some huge differences between the album gain and the replay gain. when you play a lot of mixed lists most tracks of those albums with a big difference will be louder or softer then a track of an other album. this would not happen when using track based gain.
:)
gab
-
I actually was trying Volume Leveling via the Track Based setting.
Since I turned off the Volume Leveling setting playback has been quite good and I haven't had any instances of a jarring volume difference between tracks. This is very strange since you're saying the Library Server shouldn't be adjusting the volume. When I have a chance I can try playback on my PC a home directly to remove the Library Server from the eqation.
-
I'm still struggling with this issue. First, for me it still seems that I get better volume leveling with this setting turned off than turned on. Obviously this should not be the case, so is there anything else I can do to investigate?