DIGITAL At Least a Dozen 2002/2003 Elliot Scheiner 5.1 Mixes Have a LFE Issue (info/list/fixes inside)

[svtmike]

Was thinking about this a bit today...

When a mastering or authoring error results in a channel being inverted - and we correct that by inverting it back, we shift the phase of every frequency in the channel by 180 degrees. This brings the listening experience back to the mixing engineer's intent.

But what happens when we time shift an LFE channel?

Assuming that the LFE delay is in what the mixing engineer heard... I will assume that the sound with the delay is the engineer's intent and therefore is not an error. What happens when we "fix" the delay by bringing the amplitude peaks together in time?

When we shift an entire channel by say 5ms, we are shifting every single frequency in it by a different portion of its phase. A 200 Hz signal has a period of 5ms. When we shift those amplitude peaks by 5ms we shift a 200 Hz signal one cycle over (no audible change), but a 100Hz signal moves 1/2 a cycle over (phase inversion) and a 300 Hz signal also moves 1/2 cycle over (phase inversion). This time-shifted channel may now be causing a perceived loss of signal in the very low bass range because the LFE is now combining with the other channels to cancel out the low bass frequencies from about 50Hz - 120 Hz.

The shifting would appear to be increasing the amplitude of the signal or "cleaning it up", but over the frequency spectrum the results are a mixed bag.

If we're looking just at the LFE band, we cause phase inversion at different frequencies depending on the number of milliseconds shifted:

5 ms = 100Hz
6 ms = 83Hz
7 ms = 71Hz
8 ms = 63Hz

The graph below shows phase shift for different time shifts vs. frequency for 5ms to 8ms shifts. To read, 0 is no phase change and -1 to +1 indicate -180 to 180 degrees of phase change. Both -180 and +180 degrees are 100% out of phase. You can clearly see that the meat of the LFE band is now completely out of phase so tones in this range will be subdued rather than enhanced.

I think this explains the apparent paradox I experienced with Yoshimi in my post above that "fixing" an LFE delay caused a loss of low bass in the end result that I heard.

And now I leave this to the friendly crowd in here and hope that my admittedly rusty signal analysis skills did not fail me.

 

[svtmike]

I thought I would test my theory using Track 1 of Yoshimi, "Fight Test"

What I did is mix the LFE at 1/4 power (-6dB) with the front L/R channels and measure the spectrum of the mix.
Then I mixed the LFE with the 5ms delay taken out with the front L/R channels and measured the spectrum of that mix.

What I thought I might see is a decrease in the content in the low bass, but instead I see an increase.

L/R mixed with LFE -6dB followed by L/r mixed with LFE -6dB with 5ms delay removed:

You can see the very low end is increased about 3dB and everything above 200 Hz looks pretty similar.

I think I might try doing a "bass managed" version mixing the sync'd LFE into the front L/R and silencing the LFE track.

Or I'll wait for the Atmos version to arrive and just enjoy that!

 

[Cheezmo]

I thought I would test my theory using Track 1 of Yoshimi, "Fight Test"

What I did is mix the LFE at 1/4 power (-6dB) with the front L/R channels and measure the spectrum of the mix.
Then I mixed the LFE with the 5ms delay taken out with the front L/R channels and measured the spectrum of that mix.

What I thought I might see is a decrease in the content in the low bass, but instead I see an increase.

L/R mixed with LFE -6dB followed by L/r mixed with LFE -6dB with 5ms delay removed:

View attachment 110715View attachment 110716

You can see the very low end is increased about 3dB and everything above 200 Hz looks pretty similar.

I think I might try doing a "bass managed" version mixing the sync'd LFE into the front L/R and silencing the LFE track.

Or I'll wait for the Atmos version to arrive and just enjoy that!

Easier to enjoy because more difficult to analyze and find that the same bass issues may exist. Ignorance is bliss. Seriously.

 

[svtmike]

Oh 100% agree that enjoying is easier.

But I enjoy learning ( and sometimes relearning! ), so the analyzing is fun too. And in the case of egregious flaws like channel inversion, analysis and fixing leads to more enjoyment.

 

[mandrix]

Nothing wrong with analyzing. I just don't have time so adapt to whatever is presented. So much music, so little time.
I have a stack of new releases unheard as yet.
Seem to be stuck on PT - FOABP. But that's OK. Love it.

 

[Cheezmo]

Oh 100% agree that enjoying is easier.

But I enjoy learning ( and sometimes relearning! ), so the analyzing is fun too. And in the case of egregious flaws like channel inversion, analysis and fixing leads to more enjoyment.

I know, I spent months analyzing and fixing my 12,000 multichannel music files. I'm kind of burned out on it.

 

[MidiMagic]

More incompatibilities between products.

 

[Cheezmo]

More incompatibilities between products.

What product incompatibilities are you referring to?

 

[quicksrt]

I know, I spent months analyzing and fixing my 12,000 multichannel music files. I'm kind of burned out on it.

12,000 of them had errors needing corrections? Wow, yeah that would burn me out.

I’m interested in serious issues on the big albums. I was surprised to find out Tull’s Aqualung quad has issues with channels on several tracks. Things like that I need to know. Polarity issues like on Blow By Blow SACD is important to know about. An important album for Jeff Beck.

 

[Cheezmo]

12,000 of them had errors needing corrections? Wow, yeah that would burn me out.

I’m interested in serious issues on the big albums. I was surprised to find out Tull’s Aqualung quad has issues with channels on several tracks. Things like that I need to know. Polarity issues like on Blow By Blow SACD is important to know about. An important album for Jeff Beck.

12,000 were checked, I fixed 3,126.

 

[ssully]

Artist	Title	Formats	Release Year	LFE Issue?	Adjustment (ms)	Notes / Other Issues
America	Homecoming	DVD-A/SACD	2001	NO
Beck	Sea Change	DVD-A/SACD/BDA	2002	YES	8ms​	SACD & BDA: FL & FR are out of phase (DVD-A is OK)
Beck	Guero	DVD-A	2006	YES	0ms​	Track 1-12 LFE is out of phase / Track 13 is OK
Beyonce	Beyonce: The Visual Album	BD-V/DVD-V	2014	NO
Derek and The Dominos	Layla and Other Assorted Love Songs	DTS DVD-V	2011	NO
Doobie Bros.	The Captain and Me	DVD-A/SACD	2001	NO
Dylan, Bob	Blood on the Tracks	SACD	2003	YES	6ms​
Eagles	Hell Freezes Over	DTS CD/DVD-V	1997	MAYBE		LFE either out of phase or misaligned
Eagles	Hotel California	DVD-A/SACD	2001	NO
Eric Clapton	461 Ocean Blvd.	BDA	2013	NO		Give Me Strength' Box Set
Fagen, Donald	The Nightfly	DVD-A/SACD	2002	YES	4ms​
Fagen, Donald	Kamakiriad	DVD-A	2003	YES	4ms​
Fagen, Donald	Morph the Cat	DVD-A	2006	NO
Flaming Lips	Yoshimi Battles the Pink Robots	DVD-A	2003	YES	5ms​
Guns 'N' Roses	Appetite For Destruction	BDA	2018	NO
Foo Fighters	In Your Honor	DVD-A	2005	NO
Hill, Faith	Cry	DVD-A	2002	YES	4ms​	(different delays on different tracks)
Lukather, Steve	Candyman	DTS CD	2002	YES	TBD​	(per HaikuBass)
Lynyrd Skynyrd	Southern Surroundings	DVD-A/BDA	2012	NO
Mallet, Alain	Mutt Slang	DTS DVD-V	2016
Mallet, Alain	Mutt Slang II: A Wake of Sorrows	FLAC D/L	2020	NO
Morrison, Van	Moondance	BDA	2013	NO
Neville, Aaron	Nature Boy: The Standards Album	DVD-A/SACD	2003	YES	5ms​
New York Voices	Let It Snow	SACD	2014	YES	TBD​
Orbison, Roy	Black & White Night	DVD-A/SACD	2004	NO
Porcupine Tree	In Absentia	DVD-A	2003	YES	9ms​	LFE needs 7.6dB boost / Center channel is out of phase
Porcupine Tree	Deadwing	DVD-A	2005	NO
Queen	A Night at the Opera	DVD-A	2002	YES	6ms​	Withdrawn 1st mix
R.E.M.	Automatic for the People	DVD-A	2003	YES	7ms​	(per MrSmithers)
R.E.M.	In Time: The Best of R.E.M. 1988-2003	DVD-A	2003	YES	TBD​	(per AYanguas, approx 6ms)
R.E.M.	Document	DVD-A	2003	NO
R.E.M.	Green	DVD-A	2005	YES	7ms​	(per MrSmithers)
R.E.M.	Out of Time	DVD-A	2005	YES	7ms​	(per MrSmithers)
R.E.M.	Reveal	DVD-A	2002	YES	4ms​	(per MrSmithers)
R.E.M.	Monster	DVD-A	2005	YES	7ms​	(per MrSmithers)
R.E.M.	Up	DVD-A	2005	YES	7ms​	(per MrSmithers)
R.E.M.	New Adventures in Hi-Fi	DVD-A	2005	YES	2ms​	(per MrSmithers)
R.E.M.	Around the Sun	DVD-A	2005	YES	3ms​	(per MrSmithers)
Scaggs, Boz	Dig	DVD-A	2001	YES	TBD​	(per MrSmithers)
Steely Dan	Everything Must Go	DVD-A	2003	YES	7ms​
Steely Dan	Gaucho (1998 Mix)	DTS CD	1998	NO
Steely Dan	Gaucho (2003 Mix)	DVD-A/SACD	2003	NO
Steely Dan	Two Against Nature	DVD-A	2002	NO
Sting	Brand New Day	DTS CD/DVD-A/SACD	1999	NO
Toto	Toto IV	SACD/BDA	2003	YES	4ms [1-2,4-9] 8ms [3, 10]​
Toy Matinee	Toy Matinee	DTS CD/DVD-A	2001	NO
Washington Jr., Grover	Winelight	DVD-A/SACD	2002	NO

[Delay time in bold means the results have been verified for all tracks.]

(If anyone can help with the missing albums, that would be great - a screenshot of foobar's 6 channel oscilloscope (or similar) at a moment where there's a big bass hit, like in the images I posted above, will allow me to discern which albums are good immediately, and which ones are bad and need LFE delay figured out.)

Has anyone discerned a pattern in this? Sorting by year doesn't show one. Did he use different facilities for different projects?

 

[ssully]

Any chance QQ could be compiling a crowdsource master chart like this for *all* known surround releases?

 

[Paulosophy]

I'm not going to be the one to tell Elliot Scheiner that he made a dozen mistakes.

​

 

[jimfisheye]

The offset business...
I still want to know what software got away from someone with a PDC bug! Could it be the infamous Protools? Some plugin popular at a couple mastering houses with incorrect PDC reporting? Perhaps one of the highly latent Waves plugs? (All low hanging fruit there.)

The Lfe with other than just sub bass...
To be fair, the old-school sub use and mixing for it went like this:
Your mains speakers roll off on the bottom wherever they do. You add a subwoofer to the array with an integrated low pass crossover filter built into the speaker cabinet itself.
When you want to "extend" the sub bass for a mix element, you send some of that track to the subwoofer channel in addition to the mains. The sub crossover only lets the sub bass through. You push up that level until the mix sounds right between the random point your mains speakers roll off the sub bass and the added bass from the subwoofer. And if there's a frequency range where both mains and sub respond... that just is what it is. You mix to taste and things WILL be technically skewed.

Crude like that!

A more modern approach is to calibrate a speaker array (including subwoofer) to where a frequency sweep put to the array as a whole transitions between mains and sub woofer balanced and calibrated. The Lfe channel ends up only getting already band restricted program with precision. The crossover transition is calibrated with no doubling up and no holes.

FYI, Protools DAW still has channel strips set up for the old-school method. Double sending a track to the sub in addition to the main output. You have to dial up your own routing and workflow to do the modern method. (Which isn't hard or restricted or anything.)

Some people just consider the older method the official way to mix for a subwoofer and you'll get Lfe tracks with full program. Yes, learn your bass guitar and kick drum parts! They expect you to have a sub woofer with a built in low pass crossover. They never imagined a speaker managed system with big fronts and no sub and speaker managing that Lfe track into the front L/R mains! Probably tell you not to do that and to get a sub.

Obviously I think the modern method is more correct than the old-school random response method! Just my opinion. But the old method is established and we're going to keep seeing mixes like that. Adjusting your sub volume to taste for those kind of mixes is expected rather than expecting a calibrated system.

 

[ssully]

The offset business...
I still want to know what software got away from someone with a PDC bug! Could it be the infamous Protools? Some plugin popular at a couple mastering houses with incorrect PDC reporting? Perhaps one of the highly latent Waves plugs? (All low hanging fruit there.)

seen this?

There, the culprit is clearly an EQ plug-in.

A more modern approach is to calibrate a speaker array (including subwoofer) to where a frequency sweep put to the array as a whole transitions between mains and sub woofer balanced and calibrated. The Lfe channel ends up only getting already band restricted program with precision. The crossover transition is calibrated with no doubling up and no holes.

This again makes the result depend on the mixer's monitoring setup.
More ideally, an 80 Hz 'transition' would be standard and required.

More ideally still: no LFE unless absolutely required by the original definition of LFE

 

[svtmike]

seen this?

There, the culprit is clearly an EQ plug-in.

Because I am a nerd, I did an experiment this morning in Audacity to see if I could make it easier to compare full range channels to the LFE. So, I did this:

- Duplicated the FL channel, then applied a Low Pass Filter to it.
- Duplicated the LFE channel

I then amplified both of these duplicate channels so the peaks would be obvious. Interestingly, I was seeing longer delay (by about 2ms) than by comparing the unfiltered FL channel to the LFE channel.

So I applied the same low-pass filter to the duplicated LFE channel and compared it to the original. It was 2ms delayed relative to the original.

So yes, absolutely software filters can have the exact same behavior as their hardware counterparts (probably because they are written to simulate the circuitry of the filters).

So my process to make the channels easy to compare is now:

- Duplicate the FL channel, and apply the Low Pass Filter
- Duplicate the LFE channel, and apply the identical Low Pass Filter

Amplify both duplicated channels, and measure the delay between the peaks.

 

[jimfisheye]

seen this?

There, the culprit is clearly an EQ plug-in.


This again makes the result depend on the mixer's monitoring setup.
More ideally, an 80 Hz 'transition' would be standard and required.

More ideally still: no LFE unless absolutely required by the original definition of LFE

Interesting. Not a Fab Filter user. Thought I heard good things about their plugins? I use the stock ReaEQ most of the time unless I need linear phase. EQ is more about how you use it than the plugin.

I'm not sure off the top of my head if that would be a PDC bug situation or an operator error example where someone should have used linear phase. I'd have to demo it.

I can say that I'm one of those who is addicted to parallel processing though! I'm usually always working some mix elements with multiple parallel track instances and working in "layers" like that. That means using linear phase eqs often enough to facilitate that.

Is this just easy to stumble into with standard eq phase shifting at those frequencies? Might have to do a little study! I've been using a linear phase crossover style eq to split bass elements for Lfe routing. Vetted the eq by nulling the split and recombined against the source and it's a perfect null. Waves linear phase multiband compressor plugin, FYI. Just using the crossover eq function in this example. Probably some newer lighter weight plugin available but I'm stuck in old habits as usual.

 

[jimfisheye]

The question listeners are interested in is if the mix was dialed up with the Lfe error in place or if this was a mastering mistake that altered it. If it was an unintentional move on the mixing board... The mix was still dialed up with intentional balances listening like that. It may have been an unintentional move but balances were then dialed up with that in place. At the end of the day, listening as is is how the mixer heard it and intended it.

If this was introduced in mastering though, then it's an alteration of the mix from how it was heard during mixdown. Then it should be corrected.

All you can do is guess based on conventions and familiarity with the artist's mixes.

 

[Mickb]

Hi,
Could use some help here. I've got flac copy of Beck - Sea Change (2002 SACD), & trying to determine delay. Looking at track 04 (Lonesome Tears), it appears to be closer to 15ms out on LFE, & not 8, unless I'm not comparing correctly?

 

[svtmike]

On this one the FL and FR are out of phase with C, LFE, RL and RR channels.

Fix that by either inverting FL and FR, or inverting C, LFE, RL, and RR first. Then measure the delay. (My personal opinion is that FL and FR are polarity-correct and the remaining 4 channels are inverted in the final mastering, based on the polarity of the initial transient attacks.)

The phase inversion is going to cause much more audible change than the LFE delay.