INTERACT FORUM

Please login or register.

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

Author Topic: LFE handling of Video Files  (Read 9027 times)

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
LFE handling of Video Files
« on: March 25, 2014, 05:42:30 pm »

Hendrik,

there seems to be an issue with LAV Audio decoding. Playback of x.1 content in video mode (therefor using LAV Audio) results in an +10db boosted LFE channel. As you are aware the LFE channel is encoded with +10db headroom. Therefor a standard calibration stream has e.g. pink noise on any channel at -20dbFS but the LFE channel at -30dbFS. In video mode LAV Audio (or the MCs internal audio engine?) automatically raises the LFE channel by 10db - which will result in digital clipping with a correctly calibrated system (if MCs internal volume is not been used to get some additional headroom). Therefor such a standard calibration stream (e.g. AIX Audio BD, THX calibration disc, WOW,...) outputs -20dbFS on all channels including LFE in video mode/using LAV audio. This is not correct and has changed from previous versions - this topic was extensively discussed with Matt and we got it right since MC17/MC18 times.

In contrast playing back pure audio streams (not using video mode/LAV Audio) everything is still correct. For example MC's own internal test signals play perfectly - -20dbFS for all channels but the LFE which is at -30dbFS as encoded!

The absurd result is that a system which is calibrated spot on with MCs internal test signals, will have a +10db LFE boost as soon as video mode/LAV audio is used. Please get this sorted out. Thanks alot and Cheers!  
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: Media Center 19.0.124
« Reply #1 on: March 25, 2014, 05:58:57 pm »

Neither MC nor LAV will boost the LFE channel by default, unless you downmix into a signal without a LFE. I'm quite sure of this, I went over this whole section of the code just recently to confirm some things of how JRSS works in regards to the LFE. How do you know your test files are not mastered wrong (or intentionally louder)?

Edit:
In fact, have some pink noise. Its the -20dBFS 500Hz-2kHz 48kHz 6ch calibration file generated by MC, encoded into AC3 inside a MKV container (.mka for audio only)
It doesn't have a video image, but otherwise gets handled just the same by MC, decoded by LAV and send through the DirectShow engine.

http://files.1f0.de/samples/PinkNoise-20dbFS-ac3.mka

I can clearly see in the analyzer that the LFE channel is 10dB below all other channels, same result as the original calibration files before encoding it to AC3.

Logged
~ nevcairiel
~ Author of LAV Filters

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: Media Center 19.0.124
« Reply #2 on: March 25, 2014, 06:29:49 pm »

Thanks for the fast reply, Hendrik.

I know for a fact that my "test files" are mastered correctly, because I have used and checked them for years now. Before posting I double and triple checked with the three mentioned "calibration BDs". I also made sure it is not some exotic setting/convolution that's causing this - I disabled everything (including even output format, standard Red October playback) in DSP studio and used the internal Analyzer to see the digital levels of the decoded stream. The -30dbFS encoded LFE channel (for the pink noise channel calibration) from any of these discs is being played back at around ~ -22.5dbFS according to the Analyzer plugin (MC seems to include some kind of digital clipping headroom although "Audio Path" is showing no changes are being made...). This is the same playback level for each channel, which doesn't represent the encoded gains. I hope you agree that this is not correct.

As mentioned this has worked before, I am not sure which update broke it.

To check it yourself I guess playing back the MC internal 5.1/7.1 pink noise test clip in video mode (dummy video stream) should show the problem - if you don't trust any of the "reference calibration discs" out there ;-)

Thank you!

 
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: Media Center 19.0.124
« Reply #3 on: March 25, 2014, 06:32:21 pm »

I did test, and it does work perfectly.

For reference, here are again two test clips i created, one in AC3 and one in (lossless) DTS-HD. I simply encoded the MC calibration files, and put them into a MKV container which forces MC to use the DirectShow engine to decode them (just like a video, without video).

AC3: http://files.1f0.de/samples/PinkNoise-20dbFS-ac3.mka
DTS-HD: http://files.1f0.de/samples/PinkNoise-20dbFS-DTSHD.mka

Both files show a perfect -10dB difference between LFE and the other channels if I look at the analyzer.

Do I need to put a dummy video into those files to convince you as well? =p
Logged
~ nevcairiel
~ Author of LAV Filters

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: Media Center 19.0.124
« Reply #4 on: March 25, 2014, 07:18:16 pm »

Hendrik,

thanks for being so responsive, especially considering it's 1am where we are ;-)

I am sorry, but I don't quite understand the reasoning behind your test files. Yes, those mka's play at the correct levels (decoding=encoding), just like any other "audio only" stream I tested. But as far as I can see those don't engage MC's "video mode", although they use LAV Audio.

Long story short - may I trouble you to put in a "dummy video stream" to make it into a MKV, therefor video mode is engaged, and then I finally can agree that me and my request look especially stupid and I have used test clips with wrong mastering all the time ;-)  

THANKS alot! Gute Nacht!
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: LFE handling of Video Files
« Reply #5 on: March 26, 2014, 02:44:38 am »

You can see the DirectShow filters of such audio files in MC as well, in Player -> Display Options -> DirectShow Filters

But in any case, here with video:

AC3: http://files.1f0.de/samples/PinkNoise-20dbFS-ac3.mkv
DTS-HD: http://files.1f0.de/samples/PinkNoise-20dbFS-DTSHD.mkv

Of course this doesn't change the outcome of the test for me. :)

Anyway, your test files may not be mastered "wrong", but maybe intentionally with a louder LFE, who knows what they did.
Logged
~ nevcairiel
~ Author of LAV Filters

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: LFE handling of Video Files
« Reply #6 on: March 26, 2014, 07:48:28 am »

Thanks again and sorry for the troubles. I went back to the very basics: I extracted the PCM audio stream directly from the "AIX calibration/channel balance" m2ts using Mathias' eac3to. I then loaded the resulting wav into Audacity and took a look: Ouch...all channels including LFE are mastered with -20dbFS, and still they specifically tell users to match the measured levels of all channels (it's in the video stream ;-)...which is just plain wrong (-> e.g. http://www.sona.co.jp/pdf/M2TB_r352E.pdf)

I am very sorry for the confusion. MC does it exactly right - just as you kept telling me from the beginning. Now I am going to annoy the AIX guys ;-)



Logged

mojave

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3732
  • Requires "iTunes or better" so I installed JRiver
Re: LFE handling of Video Files
« Reply #7 on: March 26, 2014, 08:07:55 am »

Therefor a standard calibration stream has e.g. pink noise on any channel at -20dbFS but the LFE channel at -30dbFS. In video mode LAV Audio (or the MCs internal audio engine?) automatically raises the LFE channel by 10db - which will result in digital clipping with a correctly calibrated system (if MCs internal volume is not been used to get some additional headroom). Therefor such a standard calibration stream (e.g. AIX Audio BD, THX calibration disc, WOW,...) outputs -20dbFS on all channels including LFE in video mode/using LAV audio. This is not correct and has changed from previous versions - this topic was extensively discussed with Matt and we got it right since MC17/MC18 times.
All channels are mastered at -20 dbFS including the LFE. What provides the headroom is that you increase the LFE level externally by 10 dB. If the LFE was -10 dB from the channels digitally then you loose headroom. If you perform Analyze Audio on a movie, you will also see the the peak level for the LFE channel is the same as the rest of the channels.

Quote
In contrast playing back pure audio streams (not using video mode/LAV Audio) everything is still correct. For example MC's own internal test signals play perfectly - -20dbFS for all channels but the LFE which is at -30dbFS as encoded!

The absurd result is that a system which is calibrated spot on with MCs internal test signals, will have a +10db LFE boost as soon as video mode/LAV audio is used. Please get this sorted out. Thanks alot and Cheers!
It looks to me like JRiver's internal test tones are incorrect. The LFE is mastered too low.

Quote
Ouch...all channels including LFE are mastered with -20dbFS
That is the correct way they should be mastered. Using JRiver Convert Format > Convert Video to Audio tool is much easier than using eac3to. I just extracted the audio from Disney's WOW calibration disc and they are also mastered at -20 dbFS.

The Disney WOW and AIX Calibration disc are considered "forensic tools" for calibration and have been thoroughly vetted by industry experts.
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: LFE handling of Video Files
« Reply #8 on: March 26, 2014, 08:15:02 am »

I'm quite sure that Dolby specifies that the LFE channel should be encoded with a gain of -10dB, which allows for additional digital headroom (at the expense of SNR) for the LFE to become louder than the other channels without overflowing in the digital domain.
In consequence this means that -20dBFS is calibrated to 95dB reference, while all other channels are calibrated to 85dB as reference level.

But of course I could be misunderstanding how it works!
Logged
~ nevcairiel
~ Author of LAV Filters

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5162
  • "Linux Merit Badge" Recipient
Re: LFE handling of Video Files
« Reply #9 on: March 26, 2014, 08:29:06 am »

I'm quite sure that Dolby specifies that the LFE channel should be encoded with a gain of -10dB, which allows for additional digital headroom (at the expense of SNR) for the LFE to become louder than the other channels without overflowing in the digital domain.
In consequence this means that -20dBFS is calibrated to 95dB reference, while all other channels are calibrated to 85dB as reference level.

But of course I could be misunderstanding how it works!

Hendrik, your understanding is also my understanding, and seems to be Genelec's understanding as well: http://www.genelec.com/faq/multichannel/102-in-a-dolby-digital-or-dts-51-system-should-subwoofer-level-be-aligned-up-or-down-by-10-db/.  But I'm open to correction.

Mojave, the peak level doesn't really speak to the mastering target, right? You can have a 0dBFS peak on material mastered to -30dBFS just as easily as material mastered to -20dBFS.  Can you think of an easy way to extract an LFE channel from an actual film? The audio analysis of that would settle the issue firmly.  
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: LFE handling of Video Files
« Reply #10 on: March 26, 2014, 08:31:35 am »

Hm, looks like I was indeed wrong, this is confusing - but the end result for the calibration is the same, for me at least!

The +10dB gain is designed to allow the one LFE channel to compete against the other channels in volume without needing to increase the volume in the encoded audio.
This means the audio is not actually encoded quieter, but it would be perceived quieter since its only one channel, competing against a multitude of other speakers, which is where the gain comes in.

So this means, if I take a -30dBFS LFE signal (as generated by MCs calibration files), and calibrate it to the same volume level as my other (single) speakers (say 85dB reference), everything is OK.
If I used a -20dBFS LFE signal, I would have to calibrate the LFE to a reference level 10dB over the main speakers (ie. 95dB)

Glad my system is calibrated properly at least!
Logged
~ nevcairiel
~ Author of LAV Filters

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5162
  • "Linux Merit Badge" Recipient
Re: LFE handling of Video Files
« Reply #11 on: March 26, 2014, 08:36:04 am »

Hm, looks like I was indeed wrong, this is confusing - but the end result for the calibration is the same, for me at least!

The +10dB gain is designed to allow the one LFE channel to compete against the other channels in volume without needing to increase the volume in the encoded audio.
This means the audio is not actually encoded quieter, but it would be perceived quieter since its only one channel, competing against a multitude of other speakers, which is where the gain comes in.

So this means, if I take a -30dBFS LFE signal (as generated by MCs calibration files), and calibrate it to the same volume level as my other (single) speakers (say 85dB reference), everything is OK.
If I used a -20dBFS LFE signal, I would have to calibrate the LFE to a reference level 10dB over the main speakers (ie. 95dB)

Glad my system is calibrated properly at least!

Ok, now I'm thoroughly confused.  Are you saying that the LFE information is not encoded 10dB quieter than other material?  Then why does room correction maintain a 10dB differential between the LFE material and redirected bass?  Or have a I misunderstood that as well?
Logged

mojave

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3732
  • Requires "iTunes or better" so I installed JRiver
Re: LFE handling of Video Files
« Reply #12 on: March 26, 2014, 08:36:40 am »

I'm quite sure that Dolby specifies that the LFE channel should be encoded with a gain of -10dB, which allows for additional digital headroom (at the expense of SNR) for the LFE to become louder than the other channels without overflowing in the digital domain.
The LFE is encoded at -10 dB volume level, but at maximum digital signal level. This way you can have it 10 dB louder than the other channels while maximizing the digital signal.

Receivers deal with this by having an opamp in the subwoofer output signal chain that increases the signal level by 10 dB. When bass management is used, the the other channels are reduced by 10 dB, bass is copied to the LFE, and then the channels are increased by 10 dB again. The maintains the proper redirected bass/LFE volume proportion.

Those who don't use receivers can either compensate digitally by having all other channels reduced by 10 dB or by externally increasing their subwoofers amp output.

JRiver used to have a check box for whether you wanted JRiver to internally increase the subwoofer by 10 dB (by actually lowering other channels) or by having it done externally. Eventually the option was removed and JRiver just let the user add the 10 dB in Room Correction if they weren't doing it externally.

16.0.95 (5/20/2011)
5. Changed: Removed the option DSP Studio > Output Format option 'Something outside Media Center will make the subwoofer +10 dB (standard calibration)'. All level calibration should be done using Room Correction.
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: LFE handling of Video Files
« Reply #13 on: March 26, 2014, 08:42:52 am »

Then why does room correction maintain a 10dB differential between the LFE material and redirected bass?  Or have a I misunderstood that as well?

Thats done for the same reason why the LFE channel is boosted by 10dB, its assumed that redirected bass adds up from multiple channels, and makes up for the lost volume that way. If you combine the bass from 4 channels, you would recover the 10dB gain easily.

I'm kinda lucky in that regard, as my subs have their own amps, which are a tad bit stronger as the amp for my other speakers, which means I don't need to mess with the volume digitally much. :)
Logged
~ nevcairiel
~ Author of LAV Filters

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5162
  • "Linux Merit Badge" Recipient
Re: LFE handling of Video Files
« Reply #14 on: March 26, 2014, 08:51:10 am »

Thats done for the same reason why the LFE channel is boosted by 10dB, its assumed that redirected bass adds up from multiple channels, and makes up for the lost volume that way. If you combine the bass from 4 channels, you would recover the 10dB gain easily.

I'm kinda lucky in that regard, as my subs have their own amps, which are a tad bit stronger as the amp for my other speakers, which means I don't need to mess with the volume digitally much. :)

If you combine bass from four channels you only get +6dB of gain, not +10dB, right?  Even 7 channels worth of bass wouldn't add up to 10dB.  10dB is ten times the volume, so you'd need 10 speakers worth of bass to get parity.  Unless there's something special about the way sound sums in these circs?

The LFE is encoded at -10 dB volume level, but at maximum digital signal level. This way you can have it 10 dB louder than the other channels while maximizing the digital signal.

How can it be encoded at -10dB, but also at maximum digital signal level?  Do you mean that the peak is always 0dBFS, but the average level is 10dB lower?  Wouldn't that mean it's mastered to -30dBFS (instead of the normal -20dBFS standard)?  How can a channel be encoded 10dB lower, without being mastered to a 10dB lower target than the other channels (or is this a mystery of encoding/decoding?)

I'm genuinely confused, I appreciate you all taking the time to help sort me out  ?
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: LFE handling of Video Files
« Reply #15 on: March 26, 2014, 08:56:55 am »

If you combine bass from four channels you only get +6dB of gain, not +10dB.  Even 7 channels worth of bass wouldn't add up to 10dB.  10dB is ten times the volume, so you'd need 10 speakers worth of bass.

I've been told this behaves differently at low frequencies, and you get 6dB gain for a pair of channels, instead of only 3dB, because the bass signal is often strongly correlated.
Logged
~ nevcairiel
~ Author of LAV Filters

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: LFE handling of Video Files
« Reply #16 on: March 26, 2014, 08:57:20 am »

All channels are mastered at -20 dbFS including the LFE. What provides the headroom is that you increase the LFE level externally by 10 dB. If the LFE was -10 dB from the channels digitally then you loose headroom. If you perform Analyze Audio on a movie, you will also see the the peak level for the LFE channel is the same as the rest of the channels.
It looks to me like JRiver's internal test tones are incorrect. The LFE is mastered too low.
That is the correct way they should be mastered. Using JRiver Convert Format > Convert Video to Audio tool is much easier than using eac3to. I just extracted the audio from Disney's WOW calibration disc and they are also mastered at -20 dbFS.

The Disney WOW and AIX Calibration disc are considered "forensic tools" for calibration and have been thoroughly vetted by industry experts.


Thanks for chiming in, mojave.

I also trusted these "forensic tools" are right. But my understanding is different: You want to have +10db headroom for the LFE channel compared to all others. Therefor if you calibrate to -20dbFS pink noise = 85db at listening position reference you will want your LFE measure 95db (actually using a C weighted level meter like the voltcraft it will show about 89dbC as these meters are less sensitive for low frequencies...but lets not complicate things further).

This results in 105db peaks (=0dbFS) for all speaker channels, while the LFE is at 115db peak. This is exactly what complies to all THX/Dolby/DTS standards (see my THX document linked).

Calibrating the LFE (=sub) so that it outputs 10db more level at listening position (by adjusting the sub/amp gain) than the other channel for the same input signal level results in 10db more headroom.

"If you perform Analyze Audio on a movie, you will also see the the peak level for the LFE channel is the same as the rest of the channels. " - yes, sure. But this peak source level produces 10db more output at the listening position for the LFE channel.

Therefor the internal test signals in Media Center are right if we are supposed to level match all channels per measurement at listening position. Calibrated that way the +10db headroom for LFE is ensured.

That's my understanding.  

  
Logged

mojave

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3732
  • Requires "iTunes or better" so I installed JRiver
Re: LFE handling of Video Files
« Reply #17 on: March 26, 2014, 08:59:14 am »

Can you think of an easy way to extract an LFE channel from an actual film? The audio analysis of that would settle the issue firmly.
Convert Format and then check "Convert Video to Audio." I usually convert to multi-channel wav. Then you can open the and analzye each channe in Audacity or WaveShop.

Quote
How can it be encoded at -10dB, but also at maximum digital signal level?
It isn't encoded at -10dB. It is at 0 dB like the rest of the channels. However, what you hear is supposed to be 10 dB louder than the other channels. How do you make it 10 dB louder:
1.  Turn up the subwoofer by 10 dB. You can't do this digitally because you are out of headroom.
2.  Lower the other channels by 10 dB. You can do this digitally.

I've got to run. I'll be back in an hour.  :)
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5162
  • "Linux Merit Badge" Recipient
Re: LFE handling of Video Files
« Reply #18 on: March 26, 2014, 09:07:31 am »

Convert Format and then check "Convert Video to Audio." I usually convert to multi-channel wav. Then you can open the and analzye each channe in Audacity or WaveShop.

Thanks, I'll do that when I get home

Quote
It isn't encoded at -10dB. It is at 0 dB like the rest of the channels. However, what you hear is supposed to be 10 dB louder than the other channels. How do you make it 10 dB louder:
1.  Turn up the subwoofer by 10 dB. You can't do this digitally because you are out of headroom.
2.  Lower the other channels by 10 dB. You can do this digitally.

I've got to run. I'll be back in an hour.  :)

Mojave, thanks for this, but I was asking for clarification about what you said about the encoding above, you said:
Quote
The LFE is encoded at -10 dB volume level, but at maximum digital signal level.
What does it mean to be encoded at -10dB volume level, but at maximum digital signal level?  Does it mean that the peak is 0dB, and the average digital level is 10 dB lower than other material (i.e. an average level of -30dBFS)?  If so, I'm with you and I get it.  If it means something other than that, I'm still very confused.  Thanks for taking the time, I'll get it eventually  ;D

I've been told this behaves differently at low frequencies, and you get 6dB gain for a pair of channels, instead of only 3dB, because the bass signal is often strongly correlated.

If the correlation is very strong, that makes sense (and that's true anywhere in the audio band, not just bass, that's why LR crossovers sum flat at -6dB).  I was assuming the sound information was only semi-correlated at best, but it makes sense that it might be more correlated than all that.  Thanks, I'm cleared up on that at least   8)
Logged

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: LFE handling of Video Files
« Reply #19 on: March 26, 2014, 09:09:51 am »

OK, I think we all are on the same page.

What I don't quite get is the following: Take the AIX calibration Blu-Ray - level balancing. It tells you (on screen) to level match the measurement of all channels (including LFE) at listening position. That's the same procedure as done by MCs internal test signals and the test signals of any receiver/pre-Pro out there (the difference is that those generally use band limited pink noise unlike AIX, but this is of no/little consequence here). BUT the AIX disc has all 6/8 channels pink noise encoded at -20dbFS.
When you do what they say your LFE will be robbed of its headroom and will be down by -10db regarding THX/DTS/Dolby standards.

Again, have a look at an "official document" http://www.sona.co.jp/pdf/M2TB_r352E.pdf for information about the industry standards here. AIX, WOW,... might not be so forensic afterall  ?
Logged

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: LFE handling of Video Files
« Reply #20 on: March 26, 2014, 09:16:30 am »

In short: For any given signal input level the LFE channel/subwoofer must provide 10db more output at the main listening position than the other (speaker) channels.

So what's wrong with those "calibration Blu-Ray's"??
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: LFE handling of Video Files
« Reply #21 on: March 26, 2014, 09:36:08 am »

I guess it depends how those discs are meant to be used. Do they instruct you do calibrate all channels to the same level, or do they instruct you to calibrate LFE to 10dB higher?
Logged
~ nevcairiel
~ Author of LAV Filters

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: LFE handling of Video Files
« Reply #22 on: March 26, 2014, 09:55:32 am »

I guess it depends how those discs are meant to be used. Do they instruct you do calibrate all channels to the same level, or do they instruct you to calibrate LFE to 10dB higher?

Sure, Hendrik. As I said:


Take the AIX calibration Blu-Ray - level balancing. It tells you (on screen) to level match the measurement of all channels (including LFE) at listening position. That's the same procedure as done by MCs internal test signals and the test signals of any receiver/pre-Pro out there (the difference is that those generally use band limited pink noise unlike AIX, but this is of no/little consequence here). BUT the AIX disc has all 6/8 channels pink noise encoded at -20dbFS.


This is just plain wrong. The important thing is that MC does it exactly right - no digital +10db LFE boost and test signals that are encoded at 10dbFS lower for the LFE than the other channels (resulting in the same acoustic output at main listening position for all channels after calibration).

As mojave mentioned AIX, WOW,... are quite highly regarded "reference" tools, e.g. extensively discussed @ avsforum, so I trusted their mastering procedure, blaming all on you ;-) I am still puzzled...
Logged

mojave

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3732
  • Requires "iTunes or better" so I installed JRiver
Re: LFE handling of Video Files
« Reply #23 on: March 26, 2014, 11:35:28 am »

WOW recommends that you calibrate the subwoofer +3 dB (page 50 in the manual) from the other channels using an SPL Meter. I double checked the encoding level for all channels. The LFE channels average RMS levels are up 2 dBfs from the other channels. -20.2 for LFE and -22.2 for other channels. With the signal being +2 dBfs (not related to volume dB), the calibration +3 dB, C Weighting, and limited bandwidth pink noise being used, then the end result is basically that the LFE is +10 dB vs other channels.

My AIX Calibration Blu-ray is at home so I can't check it. I don't remember if it has instructions either. I had e-mailed Mark Waldrep (on On 7/20/10 12:33 PM) regarding their mixing for DVD Audio. At the time, some SACDs were improperly mixed. He verified that AIX multi-channel mixes require +10 dB for the LFE channel.

Mojave, thanks for this, but I was asking for clarification about what you said about the encoding above, you said: What does it mean to be encoded at -10dB volume level, but at maximum digital signal level?  Does it mean that the peak is 0dB, and the average digital level is 10 dB lower than other material (i.e. an average level of -30dBFS)?  If so, I'm with you and I get it.  If it means something other than that, I'm still very confused.  Thanks for taking the time, I'll get it eventually  ;D
I specifically said it isn't encoded at -10 dB.  ;) The intended maximum volume of the encoding is at -10 dB when played back the same as the other channels. It is a relative volume difference.

The +10 dB for the LFE channel refers to maximum volume level capability. You can't analyze the subwoofer soundtrack and compare the LFE to other channels. How do you know the sound mixer actually mixed sounds on the LFE channel that are 10 dB higher? There are a lot of times when the LFE channel and other channels will be exactly the same volume level. The LFE channel effects can even be a lower volume level than other channels.

The maximum level comes into play when there is something like an explosion and you want it at maximum volume on the LFE channel. The goal is for it to be 10 dB higher in volume than any other channel. The only way to do this digitally is to lower the digital level of the other channels. However, this increases noise (more an issue when the standard was set) and reduces maximum volume for the other channels. They wanted to be able to use up all digital headroom for all channels. The solution was to mix everything in the LFE channel 10 dB lower intended volume than the other channels. So, even the maximum signal in the LFE channel is 10 dB lower than intended. Not encoded 10 dB lower. It is still using all the headroom in the channel. To make up the extra 10 dB, the best solution is to do it in the analog realm by compensating with the amplifier volume. That way, the intended volume is now correct.

You can think of it this way:  The LFE channel is encoded at the same volume level as other channels, but is intended to be played back at 10 dB higher volume level.

You could say it is encoded at -10 dB , but that is relative to the other channels when played back, not relative to the digital signals. You have to separate the two in your mind. Still confused.  ;D
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5162
  • "Linux Merit Badge" Recipient
Re: LFE handling of Video Files
« Reply #24 on: March 26, 2014, 12:02:54 pm »

I specifically said it isn't encoded at -10 dB.  ;) The intended maximum volume of the encoding is at -10 dB when played back the same as the other channels. It is a relative volume difference.

Check out the first sentence of your post at reply#12, that's what I was quoting.  Based on what you're saying here, I think you may just have mistyped (or maybe I'm still missing a critical distinction!), but I read you literally.  I was reading you as saying two seemingly opposite things in reply 12 and reply 17.  Sorry for the confusion, I think I'm clear on what you're saying now ;D

Quote
The +10 dB for the LFE channel refers to maximum volume level capability. You can't analyze the subwoofer soundtrack and compare the LFE to other channels.

Just out of curiosity, assuming you can isolate it, why not?  You'd expect it to have a crest factor/dynamic range and/or R128 volume level like any other track, right?  What you're saying is that the LFE's audio analysis would look identical to that of any other track mastered for film (-20dBFS average level, etc.); that would be easy enough to see.

Quote
How do you know the sound mixer actually mixed sounds on the LFE channel that are 10 dB higher? There are a lot of times when the LFE channel and other channels will be exactly the same volume level. The LFE channel effects can even be a lower volume level than other channels.

The maximum level comes into play when there is something like an explosion and you want it at maximum volume on the LFE channel. The goal is for it to be 10 dB higher in volume than any other channel. The only way to do this digitally is to lower the digital level of the other channels. However, this increases noise (more an issue when the standard was set) and reduces maximum volume for the other channels. They wanted to be able to ue up all digital headroom for all channels. The solution was to mix everything in the LFE channel 10 dB lower intended volume than the other channels. So, even the maximum signal in the LFE channel is 10 dB lower than intended. Not encoded 10 dB lower. It is still using all the headroom in the channel. To make up the extra 10 dB, the best solution is to do it in the analog realm by compensating with the amplifier volume. That way, the intended volume is now correct.

You can think of it this way:  The LFE channel is encoded at the same volume level as other channels, but is intended to be played back at 10 dB higher volume level.

You could say it is encoded at -10 dB , but that is relative to the other channels when played back, not relative to the digital signals. You have to separate the two in your mind. Still confused.  ;D

I think I'm getting there; I think it will be easier for me once I split out some channels and do some audio analysis and actually see what we're talking about. I appreciate you sticking with me on this  ;)
Logged

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: LFE handling of Video Files
« Reply #25 on: March 26, 2014, 12:03:22 pm »

The LFE channels average RMS levels are up 2 dBfs from the other channels. -20.2 for LFE and -22.2 for other channels. With the signal being +2 dBfs (not related to volume dB), the calibration +3 dB, C Weighting, and limited bandwidth pink noise being used, then the end result is basically that the LFE is +10 dB vs other channels.

You are right, Mojave. I can now remember discussing this very issue with Richard J. Casey on avs. I was wondering about the +3db recommendation, without checking the actual encoded levels. I did that now and found this odd approach you are describing. And yes, with +2dbFS, +3db measured, and using an uncalibrated C-weighted level meter (like the Radioshack) which is down 4-6db for limited bandwidth pink noise <100Hz we are at about +10db for the LFE. Very, very odd approach but at least we arrive at the ~ correct levels.

Now for AIX - can you please confirm with your own disc? The screenshot I posted clearly states that all channels should be level matched, which is just wrong for the -20dbFS encoded LFE.

I guess another email to Mark is in order  ;)

Cheers!
Logged

mojave

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3732
  • Requires "iTunes or better" so I installed JRiver
Re: LFE handling of Video Files
« Reply #26 on: March 26, 2014, 04:01:39 pm »

Check out the first sentence of your post at reply#12, that's what I was quoting.
You got me there.  ;D

Quote
Just out of curiosity, assuming you can isolate it, why not?  You'd expect it to have a crest factor/dynamic range and/or R128 volume level like any other track, right?  What you're saying is that the LFE's audio analysis would look identical to that of any other track mastered for film (-20dBFS average level, etc.); that would be easy enough to see.
For one movie the LFE track can be almost silent for 90% of the move but have massive output the other 10%. For another movie the LFE track might have almost constant output, but never seems extremely loud. They could both have the same crest factor, dynamic range, and R128 level. Actually, since movies are supposed to follow the same standards their R128 level is much closer than in music and really doesn't matter. The analysis numbers of the whole movie and even of each channel really don't tell you anything about the output at any given time period.
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: LFE handling of Video Files
« Reply #27 on: March 26, 2014, 04:15:11 pm »

My understanding was that LFE is encoded with the reference level at -10dB so that they can push the volume up 10dB higher than the other channels.
So if you send a -20dB signal to all the surround channels, you would use a -30dB signal if you match the levels with an SPL meter. (or send a -20dB signal to all channels and calibrate LFE to measure 10dB louder than the rest)
 
 
The problems arise when you start downmixing to stereo, or playing back DVD audio, neither of which should be using a 10dB boost.
If I recall correctly, LFE should only be +4dB when downmixed to stereo, and DVD-A specifies that the LFE channel is encoded at the same level as the other channels.
 
But that assumes the DVD-A disc is mastered correctly, and I don't know what the situation is with SACDs.
Just out of curiosity, assuming you can isolate it, why not?  You'd expect it to have a crest factor/dynamic range and/or R128 volume level like any other track, right?  What you're saying is that the LFE's audio analysis would look identical to that of any other track mastered for film (-20dBFS average level, etc.); that would be easy enough to see.
I'm sure you could analyze LFE, but the R128 spec says that you should ignore the LFE channel for leveling. (assuming you don't drive it to clipping)
Logged

mojave

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 3732
  • Requires "iTunes or better" so I installed JRiver
Re: LFE handling of Video Files
« Reply #28 on: March 26, 2014, 05:17:11 pm »

and DVD-A specifies that the LFE channel is encoded at the same level as the other channels
AIX Records confirmed to me that their DVD-Audio discs require +10dB for the LFE and this large thread at AVS confirms that DVD-Audio needs the +10dB on the LFE channel (there may be a few exceptions). I've read mastering engineers also attest to this.
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: LFE handling of Video Files
« Reply #29 on: March 27, 2014, 06:25:02 am »

While we're talking about this, do you guys think its useful to cross-over the bass from the main channels for video playback?
I think my old AVR used to do that, but of course my new USB DAC doesn't - so should i tell MC to do it? I think it automatically does it when the input does not contain a LFE channel, but not if it does - so should I tell Room Correction to do it?
Logged
~ nevcairiel
~ Author of LAV Filters

TheLion

  • MC Beta Team
  • Galactic Citizen
  • *****
  • Posts: 437
Re: LFE handling of Video Files
« Reply #30 on: March 27, 2014, 06:49:38 am »

While we're talking about this, do you guys think its useful to cross-over the bass from the main channels for video playback?
I think my old AVR used to do that, but of course my new USB DAC doesn't - so should i tell MC to do it? I think it automatically does it when the input does not contain a LFE channel, but not if it does - so should I tell Room Correction to do it?

It depends... ;-)

Before undertaking such a venture I strongly recommend getting decent equipment(something like a Behringer ECM8000 + cal file should do the trick, for a German product I can recommend http://www.content.ibf-acoustic.com/catalog/product_info.php?cPath=30&products_id=35) to take measurements. This is the only way to make an informed decision in order to really optimize (XO freq., filter order, delays,...) the performance with such a step. It is much more than a black/white decision. It depends on the capability of your speakers/subs, their placement, freq. response and phase issues between all channels,...

Talking about XOs in MC19: Before Matt's tragic accident there has been some discussion of making digital XOs more flexible. It would be really of great benefit to provide e.g. the option of different XO filters: Butterworth is what MC uses now (it would be great to have the option of uneven filter orders with that - e.g. 18 and 36)
Linkwitz-Riley (this is the most often used XO - atm we have to stack 2 Butterworth filters to accomplish LR)
Bessel
perhaps even Neville-Thiele

Is that something you would be able to look into, or is it a topic for Matt's return? Thanks so much!
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5162
  • "Linux Merit Badge" Recipient
Re: LFE handling of Video Files
« Reply #31 on: March 27, 2014, 07:54:54 am »

While we're talking about this, do you guys think its useful to cross-over the bass from the main channels for video playback?
I think my old AVR used to do that, but of course my new USB DAC doesn't - so should i tell MC to do it? I think it automatically does it when the input does not contain a LFE channel, but not if it does - so should I tell Room Correction to do it?

I would say (from my perspective) that unless your mains are full range or close to it, you'd almost certainly want to cross the bass over, but TheLion is correct, it can be hard to do correctly without a calibrated microphone.  And doing it incorrectly can definitely be worse than not doing it.  If you want to go down that road, I've spent a lot of time setting up active crossovers and can offer some suggestions, (I'm currently working on a "guide" more or less on this subject, but I'm at least a few months out from finishing it).

Talking about XOs in MC19: Before Matt's tragic accident there has been some discussion of making digital XOs more flexible. It would be really of great benefit to provide e.g. the option of different XO filters: Butterworth is what MC uses now (it would be great to have the option of uneven filter orders with that - e.g. 18 and 36)
Linkwitz-Riley (this is the most often used XO - atm we have to stack 2 Butterworth filters to accomplish LR)
Bessel
perhaps even Neville-Thiele
Is that something you would be able to look into, or is it a topic for Matt's return? Thanks so much!

I would also support this 100%, I've been advocating for this for a little while because JRiver's stock filters don't actually sum very nicely out of the box.  That's no fault of JRiver's, it's a fault of even-order butterworth filters, which are the ones built into MC (with the exception of the 6dB filter). As TheLion said, we can work around by cascading two to get Linkwitz-Riley type filters, but someone just trying to use MC for Bass management isn't going to know they need to do that, and we can't currently create certain common or popular filter types (3rd order butterworth, bessel, etc.).

The problem is related to the correlated sound summing problem we discussed above.  A butterworth filter is only -3dB at the crossover point, but when you divide a signal using symmetrical even-order butterworth filters, you're dividing a perfectly correlated signal, so the two channels sum 6dB at the crossover freqeuncy.  This means that symmetrical even-order butterworths create a +3dB lump at the crossover point, or, worse yet, in the case of 2nd order butterworths (12dB), they create a huge null until you flip the polarity of one of the channels, and then they sum +3dB.

Odd-order butterworth filters (6dB, 18dB, etc.) don't have the same problem: they have partial cancellation due to phase rotation (sometimes called "summing in quadrature"), leading to a more or less flat sum. But the only odd-order option we currently have is 1st order (6dB) which is too shallow to be very useful in most setups.  Linkwitz-Riley filters don't have the summing problem because they are -6dB at the crossover frequency, and are more likely to work correctly "out of the box."  Here's a comparison chart from the wiki: http://en.wikipedia.org/wiki/File:Linkwitz_vs_Butterworth.svg

None of this is urgent, since we have a workaround, but I've always felt that (taking the long view) it was worth providing more options and maybe changing the default to linkwitz-riley, because casual users just trying to get bass management set up would be likely to wind up with wacky frequency response ripples using the built in filters.  I was actually planning to incorporate some discussion of this in my (still in development) bi-amping/active crossover/subwoofer measurement guide, as it would be easy to illustrate the problems and the ups and downs of different filter types in that context.  
Logged

6233638

  • Regular Member
  • Citizen of the Universe
  • *****
  • Posts: 5353
Re: LFE handling of Video Files
« Reply #32 on: March 27, 2014, 08:17:19 am »

While we're talking about this, do you guys think its useful to cross-over the bass from the main channels for video playback?
Is that not what the Subwoofer/Sub Clarity options in DSP Studio's Output Format section are for?
Logged

mwillems

  • MC Beta Team
  • Citizen of the Universe
  • *****
  • Posts: 5162
  • "Linux Merit Badge" Recipient
Re: LFE handling of Video Files
« Reply #33 on: March 27, 2014, 08:22:58 am »

Is that not what the Subwoofer/Sub Clarity options in DSP Studio's Output Format section are for?

The output format sub creation option requires you to select a single crossover threshold to route bass to the sub, and doesn't remove that bass from your mains (i.e. you'd get duplicate bass, which could be good or bad depending on your system). If your main speakers have better bass extension than your surrounds and center, you'd want to do the routing in room correction or PEQ to get more customization of where the crossover point is.  And none of those options can "know" and resolve some of the phase issues and frequency summing issues I mentioned above, which require in-room measurements to fully address.  

Check out this thread for more info: http://yabb.jriver.com/interact/index.php?topic=75497.0.  As mojave put it in that thread: "bass management is complicated and difficult if you don't have a way to measure the frequency response"
Logged

Hendrik

  • Administrator
  • Citizen of the Universe
  • *****
  • Posts: 10661
Re: LFE handling of Video Files
« Reply #34 on: March 27, 2014, 08:35:20 am »

Is that not what the Subwoofer/Sub Clarity options in DSP Studio's Output Format section are for?

Thats only active when the input has no LFE, iirc.
Logged
~ nevcairiel
~ Author of LAV Filters
Pages: [1]   Go Up