[Ubbe2]

As the speed of sound are 343 meters per 1000mS it is a delay of 1mS for each 0,34 meter of distance. If a person stands 1 meter away and speak to you it will be a 3mS delay until the sound reach your ears. You don't react to any lip sync being out of order even if a person are 3 meters away, 10mS. So if your speakers are calibrated to use a 3 meter distance, 10mS, or 0,875 compensated 2,65 meters, 9mS, doesn't matter to lip sync. I think it is said that 50mS may be where you could see a lip sync problem.

Audessey measure and present the timing and distance to Denon. Denon use the incorrect formula to convert back to time and applies that value to their DSP. Audessey know about the problem and have let Denon know about it but Denon do nothing about it. Denon perhaps reason that a couple of milli seconds doesn't matter, and maybe rigthfully so. But then it pulls the carpet out from under the feet of those who make a living of selling applications and microphones who claim that delay has to be within micro seconds between speakers to sound perfect. But 1mS are the difference between your left and right ear and between the front of your face and the back of your head. So how relevant can it be if you need to sit still like a statue and in the exact same position every time you want to listen to perfect sound. The sound engineer that did the mix surely didn't sit like that, with his head in a wise, while mixing the sound we are listening to.

 

[aron7awol]

So that discrepancy between the 0.5ms achieved and 0.1ms that you'd expect from the UI is quite blatant when you look for it, hence those of us with REW can play around and make a thread like this and start digging, cos it's an interesting intellectual exercise, plus we get to call out Denon for incompetence, and easily produce a measurably better result, but I'm not convinced that improving the accuracy to 0.1ms is audible.

Forget 0.5ms or 0.1ms, a relative delay difference between two speakers is audible at 0.01ms. I have top and bottom centers and I just tested it again before posting (I can adjust relative delay between them on-the-fly using a miniDSP). 0.01ms is just barely audible, but 0.02ms is actually quite easy to hear. 0.1ms either way (0.03ms or 0.23ms vs my optimal 0.13ms) is a huge difference, and enough to effectively collapse the entire phantom image and anchor to an individual speaker. BTW, it's not just me, either. I had a non-audiophile friend over to check out my system and I played pink noise through the centers and we honed in on his optimal delay, and he was easily hearing differences of 0.02ms as well. I would almost say that 0.01ms is a tiny enough difference not to matter, but interestingly when watching real dialogue content there are specific voices which reveal my current 0.13ms is definitively better than 0.12ms and 0.14ms.

While my case is somewhat unusual in that it is constantly involving dialogue, which we are very sensitive to, the fact that 0.1ms makes such a huge difference proves to me that the differences we're talking about here are absolutely having significant and audible effects. I would encourage anyone who doesn't believe me to test this using something like Equalizer APO if the normal signal chain does not allow delay adjustment on-the-fly. Play speaker cal pink noise using REW through two adjacent speakers and adjust delay while it's playing.

 

[Soulburner]

While I've never done an on-the-fly test, my center image during stereo music doesn't sound different compared to before I did the tweak.

 

[bluesky636]

Releasing a new tool to hopefully make this process easier for everyone. The Denon/Marantz Distance Correction Calculator calculates the correction factor required, as well as provides feedback on the improved time alignment.

Looking for some people to help test it. I ran it this morning against a new calibration, and it worked as expected for me. The alignment feedback pretty well matched my actual measured.

To use it either copy it to your own Google Drive with File | Make a copy. Or File | Download to copy locally.

Denon/Marantz Distance Correction Calculator

docs.google.com

View attachment 3285420

Not interested in re-reading the entire thread

Is the spread sheet calculator still valid or have any changes been made to it?

 

[KMO]

The statement was made several times early in this thread by various people that the absolute timing reference does not matter and the delays can be adjusted to an arbitrary reference channel in REW. This statement is incorrect. The lip sync function adds an absolute delay compensation to the audio path, not a relative delay compensation. Even though the lip sync compensation can be manually adjusted by eye, any automatic lip sync compensation that the receiver applies of its own accord will be out of calibration.

You're wrong here. (And your next paragraph as a consequence heads off into space, and I can't really make sense of it).

The choice of timing reference in REW, which is what was being talked about, has no bearing on anything. Firstly, we never use the absolute numbers from REW to give anything to the receiver - the choice only determines which channel determines "0" on the graphs we're looking at. And that is irrelevant because we're only trying to see how close together the impulses are.

Secondly, the lip sync is not significantly affected by the speaker distance settings we're messing with, because the receiver has a separate control for overall delay. The actual delay added to any channel is

Master delay setting + ((distance of furthest channel - distance of this channel) / speed of sound)

So adding any constant distance offset to all channels changes nothing in the receiver's behaviour. The only way to properly change overall lip sync is via the master delay setting.

Doing what we're doing by multiplying all channels by a constant 0.875 will have a small effect. Say the subwoofer's at 4m due to DSP delay - the furthest speaker - and the centre's at 2m, then we end up changing relative distance between sub and centre from 2m to 1.75m. The reduction of speaker delta by 25cm means 0.8ms less delay on the centre.

This is a pretty small difference, compared to the master delay which is adjusted in 1ms steps. It's below what I'd claim to be able to perceive (and I think I'm fairly picky).

And I suspect many people won't have such a large difference between furthest speaker and centre anyway.

D&M is ignoring the issue because they don't think the difference is noticeable to most consumers and it's too much expense/risk to acknowledge it and publish a fix, so they are just going to let their receivers that are widely regarded as some of the best out there keep calibrating themselves incorrectly forever.

That's speculation at this point. It's basically impossible to communicate with D+M and get meaningful responses on anything other than catastrophic bugs like the HDMI not working with an Xbox. I wouldn't expect them to bother discussing something this trivial, or attempt to backport a fix given the potential compatibility issues like messing up MultEQ-X's existing compensation.

I'd wait to see whether anything changes in new AVRs.

If the instructions that come with the AVR are followed by the consumer, the presence of this bug essentially negates all of the advantage of having increased resolution in the distance setting registers,

No, you still gain from the increased resolution, but maybe only by a factor of 2 rather than 10.

How'd I do?

Mostly correct, but a bit waffly On the other hand, still clearer than reading the thread. So well done! No complaints about the bits I didn't quote.

I hate having to do all this research and make all these tradeoffs.

Yep, choosing AVRs is pretty hellish - you have to choose which bugs/glitches/UI things will annoy you least. Never going to be perfect.

This one is way down my list of possible annoyances. At least you can correct for it once, then forget about it. (Until next calibration - and that can be an issue. Relearning all the tricks after not touching it for a year. "Why did I do that...?")

 

[KMO]

Forget 0.5ms or 0.1ms, a relative delay difference between two speakers is audible at 0.01ms. I have top and bottom centers and I just tested it again before posting (I can adjust relative delay between them on-the-fly using a miniDSP). 0.01ms is just barely audible, but 0.02ms is actually quite easy to hear. 0.1ms either way (0.03ms or 0.23ms vs my optimal 0.13ms) is a huge difference, and enough to effectively collapse the entire phantom image and anchor to an individual speaker.

I see no reason to assume your tests of vertical phantom imaging tell us much about horizontal phantom imaging. The human auditory system is massively asymmetric across axes!

Can you run the same test with your heads turned sideways? Or with the L+R speakers?

And quoting someone else:

While I've never done an on-the-fly test, my center image during stereo music doesn't sound different compared to before I did the tweak.

Well, no it wouldn't affect that, because presumably the left and right speakers are also symmetrical, and hence have equal distance, and hence are not affected by this tweak.

I agree that this change is most likely to be A/B noticeable as improved phantom stereo imaging.

But it only significantly changes the timing between two speakers who are at significantly different distances - a non-symmetrical pair.

And then that's between a couple of surrounds or a surround and a front L/R.

But in such cases (a) you're not facing them to be probably appreciating the imaging anyway, and (b) I've never found such asymmetric pairs to be capable of great imaging ever, even if you turn to face them. Smooth pans, maybe, but not the sort of stereo imaging you get from a symmetric pair.

This fix can only improve non-optimal asymmetric setups, so I think there's a limit to what can be achieved.

The ideal is fully symmetric speaker layout, at which point this adjustment does nothing.

 

[Soulburner]

The ideal is fully symmetric speaker layout, at which point this adjustment does nothing.

I'm trying to wrap my head around this.

If two speakers are symmetric, as in, each the same distance from you (and you faced the space between them), then timing errors in either direction would de-center a sound image. The image would pull towards the speaker that shot first.

What am I missing?

 

[KMO]

I'm trying to wrap my head around this.

If two speakers are symmetric, as in, each the same distance from you (and you faced the space between them), then timing errors in either direction would de-center a sound image. The image would pull towards the speaker that shot first.

What am I missing?

Yes, that's correct. But when both speakers are at the same distance, and have equal distance in the AVR, then multiplying both the distances in the AVR by 0.875 keeps both distances equal, and they continue to play sounds simultaneously.

We're not changing that pair of speakers timing. Their impulses were always overlapping in the REW graph - there was no error to be corrected.

We're only changing the correction for speakers at different distances. Speakers with equal distance set in the AVR always play simultaneously.

 

[KMO]

Another concrete visualisation.

5 speakers, all at 2.00m, 2.00m set in the AVR.

AVR will introduce no delay. From the formula above "(distance of furthest channel - distance of this channel)" is always zero.

If you multiply all of those by 0.875 to get 1.75m, nothing changes. Delay is still zero on all channels.

But if the surrounds were at 3.00m, then the AVR will delay all 3 fronts by (3.00m - 2.00m) / 300m/s = 3.3ms.

Surrounds play first together, then 3 fronts play together 3.3ms later.

Multiplying everything by 0.875 then gives you 2.625m surrounds and 1.75m fronts.

Front delay is now (2.625m - 1.75m) / 300m/s = 2.9ms.

Surrounds play first together, then 3 fronts play together 2.9ms later. Which is correct for a 1 metre difference. (3.00m - 2.00m) / 343m/s = 2.9ms.

At no point were the fronts ever doing anything other than playing simultaneously. L/R stereo phantom imaging was never faulty, and did not change.

 

[Soulburner]

I got you.

 

[Pixelatto]

The additional adjustment range comes at a price. The lower distance settings that get entered into the registers have slightly less resolution because of the fixed bit width of the registers.

Not sure if I understand this one - would appreciate if anyone could clarify?

 

[KMO]

Not sure if I understand this one - would appreciate if anyone could clarify?

Yes, it's a bit circuitous.

Basically the steps you're adjusting in are not quite as fine as claimed. You actually are getting 3.43cm = 0.1125ft = 0.1ms steps. Not 3cm = 0.1ft = 0.0875ms steps.

(Although the UI falsely suggests 1cm steps anyway)

 

[Ubbe2]

The resolution of any distance/time are 0.1mS and the min distance that can be set are 0.46 meters 1.3mS and max are 18 meters 52mS, more or less 512 steps using a 9 bit register size. At 17 meters the 3cm step size has a 0,2% resolution and at a 1 meter distance it will be a 3,4% resolution.

 

[Pixelatto]

Many thanks for clarifying guys, much appreciated.

 

[aron7awol]

I see no reason to assume your tests of vertical phantom imaging tell us much about horizontal phantom imaging. The human auditory system is massively asymmetric across axes!

Can you run the same test with your heads turned sideways? Or with the L+R speakers?

Sure, we localize specific locations of sounds above us more poorly than those along the horizontal plane at ear level (in front of us, at least) and while that does come into play here somewhat, it's really separate from the precedence effect, which is having the far greater effect here.

The impact of the poorer above-ear-level localization comes into play with the actual perceived location of the top center sound, which ends up a bit below the actual physical location. It's why, after thorough testing of top only, bottom only, and phantom top-bottom center, I would recommend a top only center over a bottom only center, in a case where there must be only one, because the perceived lower-than-actual location is a benefit for dialogue anchoring on-screen. And then back to this context, the perceived phantom image location falls in between the perceived locations of the two speakers rather than their actual physical locations, so the perceived phantom image location with impulse responses perfectly aligned and equal levels is also a bit below vertical center, which makes sense.

What we're really dealing with here is the precedence effect, where the apparent location of the perceived sound is dominated by the (perceived) location of the sound that reaches the ears first. And it's important to note that the precedence effect is so strong that it has been shown to overcome level differences of ~10-15dB before breaking down. And this Haas effect perfectly explains why having relative delay discrepancies causes a panning phantom image (created by varying relative levels over time) to stick and then jump (when the level difference finally becomes enough to overcome the delay difference) rather than panning smoothly like it does when there is no relative delay discrepancy, which is the exact experience I described earlier in this thread based on my testing of this stuff previously.

I just did the same test with my head sideways and the effect is essentially identical. Maybe it's technically the tiniest bit easier to perceive horizontally, but I'll just call it effectively the same.

 

[bluesky636]

Not interested in re-reading the entire thread

Is the spread sheet calculator still valid or have any changes been made to it?

Nobody knows?

 

[primetimeguy]

Nobody knows?

Just multiply by .875

 

[bluesky636]

Just multiply by .875

I know that. My question was is that still valid, so I guess it is.

Here is what I got using the spreadsheet:

My next question is, I used to add 2 ms delay to the center channel for music using Dolby Surround per the implementation of Dolby ProLogic IIx (per the Meridian recommendation). Has anyone done that and have you done it to the original baseline center channel distance or the corrected distance? If you do, should 2ms also be aded to the subwoofer? I don't have REW and have no interest in it.

Note that I use a modified version of cascading crossovers for the LFE LPF along with Audyssey Flat. Surrounds are Polk bipolar speakers.

 

[SuperD00ps]

Nobody knows?

It works as before.

 

[bluesky636]

It works as before.

Doesn't answer my other questions.