Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[hulkss]

Background:
A significant number of multi-channel recordings are using the LFE/Subwoofer channel to compliment the bass content of the main channels. This content often extends above the bass management cross over frequency. I have personally found this to be true on many concert videos sourced on Blu-ray disk.

Issue: Bass management must carefully maintain proper signal correlation between the mains and LFE/Subwoofer channel to avoid cancellation of bass sound.

Using Room EQ Wizard (REW), WASAPI loopback, and a MOTU 16A audio interface, I measured the performance of JRiver bass management. I used 24 dB/oct filters at 100 Hz in the Room Correction DSP Studio for the test.

The test sweep signal was played into the left channel and copied to the LFE/Subwoofer channel for simultaneous play. When measured separately, everything looks fine (the left channel has -10 gain applied to match the LFE/Subwoofer channel). You can see the re-routed bass, the high-pass left channel, and the full range LFE/Subwoofer signal.

The problem arises when these signals are combined. As you can see in the next chart, there is a big bass cancellation caused by a mismatch in phase, when the LFE/Subwoofer and the left channel components combine.

Fortunately, this problem can be easily avoided. I used JRiver DSP to do bass management with fourth order Linkwitz-Riley filters (stacked Butterworth filters) and an all-pass filter again constructed from the JRiver Butterworth filters. All filters in this parametric filter set are 12 dB/oct.

This bass management filter set has perfectly flat magnitude response and smooth phase response. This change enabled a huge improvement in the sound of my home theater system.

I suggest that JRiver consider this in the design of their bass management tool.

References: "Methods to eliminate the bass cancellation between LFE and main channels", AES Conference, June 2006.

[mattkhan]

The problem arises when these signals are combined. As you can see in the next chart, there is a big bass cancellation caused by a mismatch in phase, when the LFE/Subwoofer and the left channel components combine.

I ran a simpler test and cannot reproduce this, there isn't enough detail in how you have MC configured to know what is different though but I think this is a straightforward test of whether jriver DSP is doing what it says it is

* create measurement sweep in rew, save as file
* set output format to 5.1, no JRSS
* turn off all processing
* add room correction, set L and R channels to 100 Hz / Move Bass to Subwoofer / both dropdowns at 24dB/Octave (i.e. 4th order butterworth filters)
* set output to disk writer
* play measurement sweep
* import sweep recording into REW (channels 1,2,4)
* verify that L and SW are filtered as expected
* add PEQ block after room correction
* add mix "Add Sub to Right"
* play sweep
* import R into REW
* verify that resulting measurement has expected +3dB hump at XO frequency (usual for BW filters)

IMV this is a "pure" test of MC room correction as literally no other component is involved. It works as expected for me except for one thing, the SW channel outputs 4dB too low. This must be a jriver DSP bug but no clue where that comes from.

If you see large cancellation (timing error) then there must be some delay being applied somewhere in your chain that would cause this.

Note that I don't disagree with your core point which is that jriver needs to improve how it handles low/high pass filters & also how bass management works, I think it's quite hard to recommend using it as is. Unfortunately DSP has seen little attention for some years so one might not be optimistic about that changing.

[hulkss]

Thanks for making me aware of using sweep files and "Disk Writer" output. This keeps the entire experiment free of possible signal processing issues outside of JRiver.

To be clear (I hope): I am illustrating a signal cancellation that can happen in bass management when the mixing studio puts the same bass track on the LFE/Sub channel and a main channel. This is common on concert recordings.

Here is my version of the test recipe shown in the previous post:

* create measurement sweep in REW, save as file, L channel only, -10 dBFS level
* set output format to 5.1, no JRSS
* turn off all processing
* add PEQ block before room correction
* add mix "copy left to sub -10 dB"; This simulates the same track mixed to LFE by the recording studio
* add room correction, set L channel to 100 Hz / Move Bass to Subwoofer / both dropdowns at 24dB/Octave (i.e. 4th order butterworth filters)
* add PEQ block after room correction
* add "add left to sub -10 dB"; This simulates the sound from the Left and Sub combining at the listening position and adjusts for the +10 dB acoustic gain of the sub channel.
* set output to disk writer
* play measurement sweep at 100% volume
* import sweep recording into REW (channel 4)

This test method works great and shows the problem (image below) is not in the signal path outside of JRiver.
Again, the filter solution described in the first post solves the problem.
Comments?

[mattkhan]

as an aside, do you find that such tracks have literally identical content on the LFE channel? just wondering to what extent this is a theoretical concern vs a practical one

so basically, given that jriver does not apply any filter to the LFE channel, this is just the summation of 3 signals

1) a dirac pulse
2) a BW4 high pass
3) a BW4 low pass

which do indeed sum as you indicate

an alternative might be to filter the LFE channel, as is commonly done in consumer equipment, using a similar filter (120Hz is typical) and then the dip goes away.

[hulkss]

as an aside, do you find that such tracks have literally identical content on the LFE channel? just wondering to what extent this is a theoretical concern vs a practical one

so basically, given that jriver does not apply any filter to the LFE channel, this is just the summation of 3 signals

1) a dirac pulse
2) a BW4 high pass
3) a BW4 low pass

which do indeed sum as you indicate

an alternative might be to filter the LFE channel, as is commonly done in consumer equipment, using a similar filter (120Hz is typical) and then the dip goes away.

Thanks for some great dialog on this topic. This cancellation issue can cause big errors in system room response. The errors can be big as I have shown here when using measurement signals in the room.

It seems many multi-channel music recordings put bass on the LFE and other channels. It may not be absolutely identical as done in this worst case test, however, recording engineers usually take care to keep the bass signals "in phase" on all channels.

We want bass management DSP to properly combine those signals mathematically on their way to the subwoofer(s). Then we want the subwoofer(s) and other loudspeakers to integrate well in the room.

If you use DSP tools like Audiolense, Acourate, Dirac, REW, etc for "room correction", you certainly do not want to have bass management upsetting the response you have measured and corrected.

Yes, the cancellation problem is mitigated if you low pass filter the LFE channel with the same filter order and frequency as used for the rerouted bass. Most bass is rerouted in the 80-100 Hz range. There is significant LEF content on many multi-channel music recordings above the bass management crossover and extending above the 120 Hz guideline for LFE.

I believe JRiver is correct with no low pass filter on the LFE channel. The proposed solution places an All-Pass Phase Compensator in the LFE channel.

I have read threads on other forums where people inserted low pass filters into the LFE channel and reported improved bass sound quality. They surmised that they were filtering out "bloated" upper bass. I believe that they were likely reducing the cancellation at the bass management frequency and strengthening the bass around 80 Hz.

[mattkhan]

I think whether you filter the LFE or not depends on your subs and content played, if you do have higher frequency content with an LFE channel and subs that can reproduce that then I'm not so sure letting the subs play that is a great idea (from a SQ and localisation perspective). I think it would be better if the Room Correction module gave users a choice on that.

for anyone reading along, the referenced AES paper is freely available - https://www.aes.org/technical/documentDownloads.cfm?docID=248

a few things to note from that paper

* it seems that you'd want encoder specific behaviour (e.g. Dolby Digital DVDs unlikely to need this treatment)
* the authors note the phase response of the APF adjusted version is impaired (though whether that is audible vs frequency response is another Q)

from an MC perspective, the main thing missing here for ease of use is an option for an all pass filter in the PEQ (along with easier to access lpf/hpf types) as otherwise you need to use the convolver.

I don't have many concert BDs so can't test this myself but you could investigate the degree to which you can see this effect using convert format. Run it once just with the bass management settings on and then again with a main + LFE channel summed then load the resulting files into some app that can visualise the effects and compare (my app https://beqdesigner.readthedocs.io/ can do this bit quite easily). If the content does produce a significant cancellation problem then it should be pretty clearly visible.

[hulkss]

I think whether you filter the LFE or not depends on your subs and content played, if you do have higher frequency content with an LFE channel and subs that can reproduce that then I'm not so sure letting the subs play that is a great idea (from a SQ and localisation perspective). I think it would be better if the Room Correction module gave users a choice on that.

for anyone reading along, the referenced AES paper is freely available - https://www.aes.org/technical/documentDownloads.cfm?docID=248

a few things to note from that paper

* it seems that you'd want encoder specific behaviour (e.g. Dolby Digital DVDs unlikely to need this treatment)
* the authors note the phase response of the APF adjusted version is impaired (though whether that is audible vs frequency response is another Q)

from an MC perspective, the main thing missing here for ease of use is an option for an all pass filter in the PEQ (along with easier to access lpf/hpf types) as otherwise you need to use the convolver.

I don't have many concert BDs so can't test this myself but you could investigate the degree to which you can see this effect using convert format. Run it once just with the bass management settings on and then again with a main + LFE channel summed then load the resulting files into some app that can visualise the effects and compare (my app https://beqdesigner.readthedocs.io/ can do this bit quite easily). If the content does produce a significant cancellation problem then it should be pretty clearly visible.

Yes, the subwoofers in some systems should not reproduce bass above 80 Hz due to localization or limited frequency response. Instead of a simple 80 Hz low pass on the LFE signal, a crossover filter can be used with the high pass signal going to the center main channel. I have tried this and it works quite well (it was identified in the referenced paper that Genelec employs this method in some of their systems). Again, care with signal phase alignment is required.

Some recording studio encoders do a better job than others controlling bass signal alignment, however, we still must be careful with bass management. I can tell you, bass management flaws are audible and show up in system test measurements causing problems with loudspeaker correction filters.

As far as impairment of the audio signal by the all-pass filter is concerned, in the referenced paper the authors used an 8th order LR all-pass filter. These have proven to be audible, while the 4th order filter I suggest using is not. See a perceptual study on crossover filters here: http://lib.tkk.fi/Dipl/2008/urn011933.pdf

I did develop my concern for this cancellation while using BEQDesigner. I'll attempt your suggested experiment.

[hulkss]

I completed a test using 5.1 DTS-HD Master Audio from the "John Mayall 70th Birthday Concert".

I cleared all processing in JRiver DSP, set the output to 5.1, and set the audio device to Disk Writer.

In Room Correction:
* mute all channels except Left and Sub
* add +10 dB to the sub channel
In PEQ after Room Correction:
* add Sub to Left
* mute Sub

This simulates the left channel and subwoofer sound combining at the listening position.

* I played the first ten minutes of the concert to a wave file with no Bass Management crossover
* Then a JRiver 80 Hz, 24 dB/oct Bass Management crossover with bass routed to the subwoofer.
* Then with the method in the first post @ 80 Hz and adjusted for the Sub channel level as noted above.

I then extracted the channel 1 audio from each case into BEQdesigner. The charts below show the results. There is a significant cancellation of bass with JRiver Bass Management and none with the Phase Compensated method.

[mattkhan]

that's a good illustration of the problem

can you post a pic of your settings to correct this? I could add it a link to the post from the bass management wiki for reference

[hulkss]

that's a good illustration of the problem

can you post a pic of your settings to correct this? I could add it a link to the post from the bass management wiki for reference

I have been chasing this bass cancellation problem for a long time. I just could not determine why many 5.1 and 7.1 music recordings had poor sounding bass. I worked for years improving my subwoofer system. I began to believe that there was a world of poor recordings. Now I know that the recordings were only an enabler for the problem. The real problem was in the bass management signal processing.

This cancellation issue can cause a huge errors in playback fidelity exceeding 10 dB in magnitude. It is a hidden problem that many home theater systems are living with.

The great thing about the solution for this problem is that it works with any system that reroutes bass to subwoofers and uses the LFE channel. It does not hurt playback of program material that would not experience the cancellation problem.

There is a picture of the required PEQ settings for 7.1 bass management in the first post of this thread. All the filters are 12 dB/oct. They are stacked Butterworth filters, so, the result is 4th order Linkwitz-Riley. The bass management filter frequency can be changed to any desired value, just keep them all the same.

[mattkhan]

There is a picture of the required PEQ settings for 7.1 bass management in the first post of this thread. All the filters are 12 dB/oct. They are stacked Butterworth filters, so, the result is 4th order Linkwitz-Riley. The bass management filter frequency can be changed to any desired value, just keep them all the same.

thanks, added a link to here from https://wiki.jriver.com/index.php/Room_Correction#Use_of_Butterworth_Filters_may_produce_cancellations_in_the_in_room_response

[hulkss]

Thanks for the addition of this thread to the JRiver Wiki.

The "cancellation fix" discussed previously suggests a full frequency range signal on the subwoofer channel. Why do we want that? Like the cancellation issue, the need for a full range sub signal originates in the multi-channel audio mix. On many concert videos, I find that the mix engineer has put the full bandwidth bass guitar track on the LFE channel. The bass guitar, including the overtones are no where else to be found. If you limit subwoofer bandwidth, the audio sounds horrible. The bass sounds extremely flat, undefined, and muffled. On recordings where the LFE content is bandwidth limited, you should not low pass filter it again during playback.

Now what to do?? Many subwoofers can not play midrange, let alone upper frequencies with good fidelity. In a lot of theater arrangements localization of the subwoofer(s) would be a problem if they play at much above 80 Hz.

We can fix both the cancellation problem discussed earlier and the "wide band" LFE channel problem with a crossover in the LFE (Sub) channel that sends low bass to the subwoofer and upper bass with harmonics to the center channel. This enhances bass localization to the soundstage center and maintains fidelity of the recording by not "discarding" the upper bass content. Routing to only one speaker prevents comb-filtering of the upper bass sound. Credit to Genelec for this solution concept.

Below are DSP PEQ settings for 7.1 that accomplish this. Again all filters are 12 dB/oct stacked Butterworth (Linkwitz-Riley 4th order). The X-over frequency can be changed, just keep all filters at the same frequency.

Here is an image showing only the LFE channel frequency content in the JRiver DSP analyzer. It is from the Blue-Ray "Metallica: Francais Pour Une Nuit" DTS-HD Master audio recorded in a Roman colosseum. A great show if your audio system is configured to play it. The bass guitar harmonics extend above 1kHz and there is some even higher frequency content as part of the drum kit on the LFE channel.

In summary, many music recording engineers do not follow the recommended practices for multi-channel audio that are adhered to in the Motion Picture industry.

The recommendations in this thread will make your system robust to differences in the audio mix involving the LFE(Sub) channel and address the need for bass management.

[justsomeguy]

So are these new settings you suggest in addition to or meant to replace the settings in your first post.

[hulkss]

So are these new settings you suggest in addition to or meant to replace the settings in your first post.

The new settings are a replacement and they solve both the cancellation and high frequency content issues. If you have a "capable" center speaker I recommend the crossover to center speaker method. I gave it a another listen tonight and it sounds great on my system.

My subwoofers are placed up front and they can play above 100 Hz. I tried them full range and the center speaker method was better when there were mid-bass frequencies on the LFE channel, otherwise, both methods sound the same.

[Hendrik]

The recommendations in this thread will make your system robust to differences in the audio mix involving the LFE(Sub) channel and address the need for bass management.

Unfortunately however, some movies can contain "garbage" in the upper frequencies of the LFE track, which is why we offer an option to low-pass the LFE (either if you have wider range subwoofers or  when mixing down to a format without dedicated subwoofer channel (eg. stereo)) - and your setup would run into the same problem. No one solution that works for all content, I'm afraid.

[hulkss]

Unfortunately however, some movies can contain "garbage" in the upper frequencies of the LFE track, which is why we offer an option to low-pass the LFE (either if you have wider range subwoofers or  when mixing down to a format without dedicated subwoofer channel (eg. stereo)) - and your setup would run into the same problem. No one solution that works for all content, I'm afraid.

I have looked at the signal content of many LFE tracks. Some have content above the recommended 120 Hz limit, some have content correlated with other channels. After listening using the methods I have posted, I have not found any LFE content to be "garbage". I think some people may assume it is garbage because "it is not supposed to be there" or their subwoofer & subwoofer position makes it sound like garbage.

I would like to evaluate a recording with "garbage" on the LFE track. Do you have an example of one?

[Hendrik]

This is the original thread:
https://yabb.jriver.com/interact/index.php?topic=94929.0

There was actual "garbage" in the LFE, which became evident when downmixing into stereo, ie. the LFE content was being played through full-range speakers.

Here is one of the samples originally talked about in the thread. You can hear digital distortion, eg. clipping noises, when the LFE is being mixed without low-passing, and no, we're not introducing the clipping ourselves, and the decoder for this format is proven to be bitexact.
https://files.1f0.de/samples/big/TDK-LFE.mkv

When you enable a 120Hz low-pass for the LFE, which is conservative for typical LFE speakers, the distortion is entirely gone again.

[wer]

This is a very interesting thread, on something I know little about, so I'd like to ask a question:

Is there an improvement to be made, or a beneficial technique to be implemented, based on what you all have been discussing above, for the scenario where the playback system has very full range main speakers, but no dedicated LFE speaker?

In this case the receiver/processor is seeing that LFE channel, and then either discarding it or mixing it to the mains.  So could the result be improved (or perhaps even corrected in the case of bad output) by some preemptive processing in MC?  The technique above seems to presuppose the existence of a dedicated LFE speaker with certain characteristics.

[hulkss]

This is a very interesting thread, on something I know little about, so I'd like to ask a question:

Is there an improvement to be made, or a beneficial technique to be implemented, based on what you all have been discussing above, for the scenario where the playback system has very full range main speakers, but no dedicated LFE speaker?

In this case the receiver/processor is seeing that LFE channel, and then either discarding it or mixing it to the mains.  So could the result be improved (or perhaps even corrected in the case of bad output) by some preemptive processing in MC?  The technique above seems to presuppose the existence of a dedicated LFE speaker with certain characteristics.

All that is required is to add the LFE channel content to the Main speaker channels that you want to use to play the LFE. No filtering is required if you are not rerouting any bass from non-LFE channels.

If you add to one main channel (L, R, or C) add +10 dB LFE, if you add to two channels add +4 dB LFE to each, if you add to three channels add +0.5 dB LFE to each.

You may need to mute the LFE channel after adding to the others.

[wer]

But I thought the point of the thread was cancellation issues when combining signals:

The problem arises when these signals are combined. As you can see in the next chart, there is a big bass cancellation caused by a mismatch in phase, when the LFE/Subwoofer and the left channel components combine.

Are you saying that the summation of the waveforms is not commutative?  That adding L to LFE does not produce the same result as adding LFE to L?

Or did you mean that the cancellation issue occurs when one tries to filter out of L the bass frequencies that were just added to LFE. That wasn't stated explicitly in your initial post, but perhaps it's implied by the filters you show.

As I said, my experience with this is minimal, so I want to make sure my understanding is correct.

[hulkss]

But I thought the point of the thread was cancellation issues when combining signals:
Are you saying that the summation of the waveforms is not commutative?  That adding L to LFE does not produce the same result as adding LFE to L?

Or did you mean that the cancellation issue occurs when one tries to filter out of L the bass frequencies that were just added to LFE. That wasn't stated explicitly in your initial post, but perhaps it's implied by the filters you show.

As I said, my experience with this is minimal, so I want to make sure my understanding is correct.

The cancellation problem happens during the addition of two similar signals when one of the signals has been passed through a filter and the other has not.

[mattkhan]

The problem mentioned in this thread arises from the summation of signals that have been through assorted low or high pass filters (that are not linear phase filters). Just summing a few channels *that have not been filtered* (i.e. where no bass management is in play) is fine.

[wer]

Ok, thanks.

[hulkss]

This is the original thread:
https://yabb.jriver.com/interact/index.php?topic=94929.0

There was actual "garbage" in the LFE, which became evident when downmixing into stereo, ie. the LFE content was being played through full-range speakers.

Here is one of the samples originally talked about in the thread. You can hear digital distortion, eg. clipping noises, when the LFE is being mixed without low-passing, and no, we're not introducing the clipping ourselves, and the decoder for this format is proven to be bitexact.
https://files.1f0.de/samples/big/TDK-LFE.mkv

When you enable a 120Hz low-pass for the LFE, which is conservative for typical LFE speakers, the distortion is entirely gone again.

Thanks for the link to The Dark Knight clip. I only have a few movie blu-rays and I have that disk. Years ago I did have a low-pass filter on the LFE channel.

Yes, there absolutely is LFE "garbage" in several scenes. There is also a lot of good LFE content in that movie that would be diminished by a low pass filter applied globally. The only fix for that movie is to extract the audio, filter the bad LFE sections, and remux. Yes, no simple solution will fix bad recordings.

[JimH]

Would it work to set a filter at certain times for a certain period?

[hulkss]

Would it work to set a filter at certain times for a certain period?

Yes, that would work. I think it would be possible to rip the movie to MKV and create short chapters where low pass filtering is needed to remove noise.
Would that allow some "selective" filtering to be implemented?

[wer]

As far as differentiating for MC, it wouldn't be necessary to make MKV chapters. You could do that now with Particles.

[JimH]

Particles:  https://wiki.jriver.com/index.php/Particles

[hulkss]

As far as differentiating for MC, it wouldn't be necessary to make MKV chapters. You could do that now with Particles.

If a playlist of movie particles is made, do they play without glitches between them?

[mattkhan]

I don't know why you wouldn't simply low pass the LFE channel for this sort of track

this is another example of where per item DSP settings would be extremely useful (if they were practically useable and also if it were possible to apply/override fragments of the DSP settings rather than storing a copy of an entire dsp configuration)

[mattkhan]

Yes, that would work. I think it would be possible to rip the movie to MKV and create short chapters where low pass filtering is needed to remove noise.
Would that allow some "selective" filtering to be implemented?

if you really wanted to do this then it would be possible to do such a thing via a (rather long) ffmpeg command line (to split the track into chunks, filter those chunks individually then stitch it back together)

[JimH]

If a playlist of movie particles is made, do they play without glitches between them?

You could try it.