VLC 0.8.6 Test 2 Can't, Won't or Doesn't...

[DJ]

1. Setting the EQ to the flat (or on) position radically changes the sound. This is more than a level discrepancy. No EQ is set for the test.

A little closer look at this and the gain discrepancy is 3db so the gain make up for EQ insertion is not 12db but 15db which is truly outrageous by anyones standards. Also there appears to be some EQ where there is slightly less bass and or more treble when the EQ is inserted in the flat position. It was also noted that when 15db was set in preferences, depending on where you would switch the EQ on and off it would loose the setting and revert to 12db thus forcing a restart of the player to continue the test. As a side note: I have never experienced or heard of a design for a filter set that lost 15db! Impedance matching designs of the 50s and 60s only lost 6db at the input, if the output was also impedance matching another 6db ensued. However, this was balanced transformer theory which doesn't seem has application here. In bridging active filers no loss is necessary although many designers would create a 6db pad at the input and 6db gain at the output claiming the device had better headroom.

2. I just compared some DTS 5.1 songs in VLC to the codecs I have installed for Direct Show. VLC with the EQ disabled and set to unity gain just like the Windows counterpart, VLC is dark and muffled with a general lack of definition. There is less space or air around instruments and definitely less exciting. The DRC is off on all players.

Assuming there really is a low pass filter at the output, I would guess that the Q of the cutoff filter is not right so the cutoff frequency is to low. For example : a 50KHz sample frequency the cutoff should be about 22KHz and drop about 6db per active in order to be relatively neutral (least ripple) in the passband, 18KHz should still be flat with a slight peek @ 19KHz of about .2db before beginning to roll off @ 20KHz where 22KHz is -.5db and 44Khz would be at least -6db. Of coarse in the real world there is no audio @ 44KHz and 48KHz sampleing is more common than 50KHz. This seems to work well with most 8 times or higher oversampled devices that aren't supposed to need a filter. The result is a warmer clear sound with more natural reverb and is less harsh to the ears, (specially for a female vocalist) without appearing closed down at the top end.

3. No hot keys in full screen mode. That's none, nothing functions.

Edit: This has been fixed in the nightly builds 11/25/06.

[DJ]

4. I'm not quite sure this is a bug, but seems like one and is odd. When the skins2 module is used you set the position of the skin on screen but the next time you open the player the skin is off screen, forcing a reset of preferences to get it back. Then it seems that it works again, but haven't tried all screen positions as this is quite annoying.

[The DJ]

1: File a trac ticket please
2: I think this is due to the 5.1 to 2.0 channel downsampling. (as a matter of fact, #1 might be due to this as well).
3: Fixed in test2a (not in windows test2)

4: I can't test this. can anyone else provide more detailed reports ?

[ipkiss]

For 4, I think it can happen if you use a secondary monitor (but I can hardly debug this with only 1 screen ). If this is the case, does it work better when keeping the skin on the first monitor?

[DJ]

@ ipkis, This is or did happen on the main monitor. Please keep in mind that I do not (at least haven't) install tests. I generally download the 7z file and form a new folder under VideoLAN. This leaves the current version in tact and allows me to explore questions regarding either version and make suggestions to users here or test the test. I suppose it's possible this may have something to do with the stored preferences.

Thanks for the response
DJ

@ The DJ, I will post a ticket to trac on #1 and refer to this post for info..
I suppose it is possible to have something to do with down sampling. I will explore this a bit further with some other formats and see if I can further pin it down. I do know that movies seem rather unaffected. But then in many movies the sound track is not as good and or the music track is in the background and not as noticeable.

Oh! When do you think the changes for #3 will be committed for Windows?

Thanks for the response
DJ

http://trac.videolan.org/vlc/ticket/849

[ipkiss]

@ ipkis, This is or did happen on the main monitor. Please keep in mind that I do not (at least haven't) install tests. I generally download the 7z file and form a new folder under VideoLAN. This leaves the current version in tact and allows me to explore questions regarding either version and make suggestions to users here or test the test.

It's OK, installing wouldn't change the behavior of skins at all.

I suppose it's possible this may have something to do with the stored preferences.

Indeed, the "skins2-last" setting of the vlcrc file contains the last position of the skin windows (among other things).

I would appreciate if you could find some time to answer to the following questions:
Do you have 1 desktop in each monitor, or only one in total? Does the skin 'snap' to the 4 borders of the screen on the main monitor? And on the secondary one?
Could you also give the value of the "skins2-last" setting after closing VLC while the skin window was:
1) in the top-left-hand corner of the main monitor
2) in the bottom-right-hand corner of the main monitor
3) in the top-left-hand corner of the secondary monitor
4) in the bottom-right-hand corner of the secondary monitor
In addition, I would like to know the screen resolution on each monitor...

[DJ]

Do you have 1 desktop in each monitor, or only one in total?

This machine will support three monitors, but only 1 is being used. Main monitor.

Does the skin 'snap' to the 4 borders of the screen on the main monitor?

Yes! (0.8.5)
Yes! (0.8.6 test2)
But all can be pushed off screen.

Could you also give the value of the "skins2-last" setting after closing VLC while the skin window was:

1. ( 256, 256,0)( 100, 401,0)( 100, 100,0)( 0, 0,1)( 555, 99,0) "0.8.5"

1. [creditswin theCredits 256 256 320 240 0][eqwin eqwinlayout 100 401 300 150 0][main video 100 100 454 300 0][minwin minilayout 0 0 300 100 1][plwin pllayout 555 100 300 300 0] "0.8.6 test2"

2. ( 256, 256,0)( 100, 401,0)( 100, 100,0)( 724, 638,1)( 555, 99,0)"0.8.5"

2. [creditswin theCredits 256 256 320 240 0][eqwin eqwinlayout 100 40
1 300 150 0][main video 100 100 454 300 0][minwin minilayout 724 638 300 100 1][plwin pllayout 555 100 300 300 0] "0.8.6 test2"

Screen resolution is 1024x768

I had to go into vlcrc for the last one because it is off screen and can't be retrieved. I do appreciate what you are trying to do now and am looking forward to the changes. My second monitor started acting up last week and has been sent out for repair or replace under the warranty or I would have done that set too. Sorry!

[ipkiss]

2. ( 256, 256,0)( 100, 401,0)( 100, 100,0)( 724, 638,1)( 555, 99,0)"0.8.5"

2. [creditswin theCredits 256 256 320 240 0][eqwin eqwinlayout 100 40
1 300 150 0][main video 100 100 454 300 0][minwin minilayout 724 638 300 100 1][plwin pllayout 555 100 300 300 0] "0.8.6 test2"

Screen resolution is 1024x768

These values look perfectly fine, both for 0.8.5 and 0.8.6-test2.

I had to go into vlcrc for the last one because it is off screen and can't be retrieved.

Just to confirm: you placed the skin in the bottom-right-hand corner, closed VLC, retrieved the above values in the vlcrc file, restarted VLC, and it was off screen? How do you know it is off screen, and not invisible, for instance?
When this happens, could you close VLC by right-clicking on the task in the taskbar, and choosing "Close"? I would like to know if the skins2-last setting has changed after that.

BTW, did you check the logs for error messages? I don't expect to find some, but just in case...

[DJ]

Just to confirm: you placed the skin in the bottom-right-hand corner, closed VLC, retrieved the above values in the vlcrc file, restarted VLC, and it was off screen?

Yes!

How do you know it is off screen, and not invisible, for instance?

Not quite sure I see your point here? It wasn't minimized, if that's what you mean. So I assume it's off screen.

When this happens, could you close VLC by right-clicking on the task in the taskbar, and choosing "Close"? I would like to know if the skins2-last setting has changed after that.

Hmm! I repeated the test exactly the same as last night and it has changed. 0.8.6 test2 is now holding in the lower right corner and 0.8.5 isn't when I try to go back to it. Of coarse it wasn't right last night when I went back to 0.8.5 so it seems to be acting kinda normal considering the changes.

2. (-1,0,0,1024,768)(0,48,31,355,102)(6,0,0,347,150) "0.8.5"

Had to pull this out of vlcrc as the interface could not be seen.

2. [creditswin theCredits 256 256 320 240 0][eqwin eqwinlayout 100 401 300 150 0][main video 100 100 454 300 0][minwin minilayout 724 638 300 100 1][plwin pllayout 555 100 300 300 0] "0.8.6 test2"

Not quite sure how to get Messages or logs when you can't get at the interface? Don't believe I have ever done this??

Of coarse when everything works there are NO errors.

BTW It doesn't matter where the interface was on screen when you try to go back to 0.8.5, however considering the changes I don't believe this could be considered a bug.

[DJ]

1. Setting the EQ to the flat (or on) position radically changes the sound. This is more than a level discrepancy. No EQ is set for the test.

A little closer look at this and the gain discrepancy is 3db so the gain make up for EQ insertion is not 12db but 15db which is truly outrageous by anyones standards. Also there appears to be some EQ where there is slightly less bass and or more treble when the EQ is inserted in the flat position. It was also noted that when 15db was set in preferences, depending on where you would switch the EQ on and off it would loose the setting and revert to 12db thus forcing a restart of the player to continue the test. As a side note: I have never experienced or heard of a design for a filter set that lost 15db! Impedance matching designs of the 50s and 60s only lost 6db at the input, if the output was also impedance matching another 6db ensued. However, this was balanced transformer theory which doesn't seem has application here. In bridging active filers no loss is necessary although many designers would create a 6db pad at the input and 6db gain at the output claiming the device had better headroom.

Hmm! I may have put both feet in my mouth on this one because the issue in much more complex than it first appeared. This is what happened on EQ insertion with various formats:

1. DCA (DTS) Drops 3db requiring 3db of gain.
2. A/52 (ac3) Increases 6db requiring 6db of loss.
3. MP4a (AAC) shows no change in gain but the gain from 5.1 to stereo is still incorrect as well as the channel mapping.
4. FLAC and Vorbis show no change in gain.

My statement saying that the problem is less apparent in movies and more apparent in music files is incorrect as it's the same in this regard.

If this problem is related to the fold down from 5.1 to stereo, how is this related to the EQ??? If the EQ were to force a fold down this may be understandable.

The concepts surrounding a fold down are varied.

1. Some believe that the acoustic gain should remain the same and this is based in power (3db) and this is what most of the text books teach. Therefore the sum of two channels would be reduced by 3db. In practice this doesn't work for two reasons.

a. The acoustic sum is not 3db it's 6db, just like the electrical sum and this does work acoustically. Therefore no change is necessary.

b. If the sum is 3db or 6db full scale is being changed and converters can't do that! So in order to avoid distortion the gain must be reduced.

2. In a professional recording environment headroom factors must be maintained on all pieces of equipment. In oder to do this properly in a fold down the only choice is to reduce the electrical sum of 2 channels (voltage) by 6db.

By doing this:

1. If the source is 5.1 monitored in a 5.1 environment the folded down to stereo will appear to loose gain and the loss is approximately 6db acoustically.

2. If the source is 5.1 and monitored in a stereo only environment the fold down will appear neutral only if all channels were to be equal in gain at all times. Of coarse this is very unreal and rather defeats the purpose. So in most cases it will also appear to loose some variant of gain acoustically. Unfortunately this is the only correct way of dealing with the problem and maintaining proper phase relationships and proper headroom (FS).

Formats are not relevant here as the problem is in the math as is the resolve.

If the fold down is the issue with the EQ, it is suggested that it's the improper use of the fold down that may be causing the frequency shift due to phase relationships of the channels. This seldom happens when the sum values are electrically correct. Of coarse this would also imply that the EQ was stereo only and forcing a fold down and further that the fold down for certain formats were incorrect.

For reference a doubling or halving of voltage is 6db. A linear fader set at 50% will show a 6db increase when moved to 100%. A 10 log fader set at 50% will show an approximate 20db increase when moved to 100%.

[ipkiss]

2. (-1,0,0,1024,768)(0,48,31,355,102)(6,0,0,347,150) "0.8.5"

These values are nonsense. This kind of bug is one of the reasons why the handling of saved positions was changed after 0.8.5.

2. [creditswin theCredits 256 256 320 240 0][eqwin eqwinlayout 100 401 300 150 0][main video 100 100 454 300 0][minwin minilayout 724 638 300 100 1][plwin pllayout 555 100 300 300 0] "0.8.6 test2"

This one looks OK.

Not quite sure how to get Messages or logs when you can't get at the interface? Don't believe I have ever done this??

This is not needed anymore, if you confirm that 0.8.6-test2 works fine. But for reference, one way to do it would be to add '--extraintf logger' to the command-line arguments.

[DJ]

Well, it would be kinda nice to know why twice 0.8.6 drew the interface off screen. But I well know that when it works correctly you can't fix it. Believe me when I say, I tried to get it to fail again and couldn't.

[streki]

When I use stream UDP multicast with video H264/audio mp4a codding, video is allright but no audio.