G3P Elevated BM & LFE Issue

[ttocs]

Aug 3, 2021 7:43:31 GMT -5 marcl said:

Aug 1, 2021 22:16:29 GMT -5 ttocs said:

5.1 with two subwoofers
Left Sub Mono
Center Sub LFE
Large LCR
Small SL and SR
Surround sound field

After running Dirac in a 5.1 config then leveling using XMC-2 Levels, I ran REW to check. The first plot is that result which shows that XMC-2 Levels don't work, so this showed me how much adjustment was still needed.

XMC-2 Levels checked with REW
View Attachment

After using REW to determine proper levels adjustments.
View Attachment

Am I understanding correctly?

After running Dirac, you used the pink noise source from the XMC-2 to check levels and adjusted them with an SPL meter (REW?).  Yes.
This produced the incorrect levels.  Yes.
After that you used REW as the pink noise source and adjusted levels using an SPL meter, producing the correct levels.  Yes.
Dirac calibration was done with the default Dirac target curve?  Yes. 

This seems to indicate that maybe the pink noise source in Levels in the XMC-2 is played without any Dirac level processing and so it reflects the levels without the level alignment that Dirac does, but clearly with Dirac filter correction.  When you use REW as the pink noise source, Dirac level adjustment is definitely applied and you add final level adjustments on top of that.

Yes to the first 4 points, but the last section is questionable. I find the same behavior whether using the User EQ or Dirac EQ. I pointed this out when bringing the BMbug to light and using only User EQ.

Bottom line, Levels pink noise in the G3P are not accurate. The inaccuracy seems to be in Subwoofer channels and Small channels, so my belief at the moment is that it's a BMbug artifact.

I can use Levels to adjust all the channels to be the same SPL. Then when I run REW the levels of Small channels above XO are not equal to Large channels in the same frequency range, so I adjust them down. Then you add the BMbug on top of this knowing that it's elevated above the Small channels by at least 4dB, so this needs to adjust down -4dB and then the amount that the Small channels were lowered by.

After running Dirac and using the XMC-2 Levels pink noise resulted in the following settings:
Large Front Left: -1.0dB
Large Front Right: 0.0dB
Large Center: 0.0dB
Small Left Surround: +4dB
Small Right Surround: +3dB
Left Sub Mono: -4dB
Center Sub LFE: -4dB
and results in the following REW plot:

After using REW sweeps measurements for levels using Surround mode, the settings are now:
Large Front Left: -1.0dB
Large Front Right: 0.0dB
Large Center: 0.0dB
Small Left Surround: -1dB
Small Right Surround: -2dB
Left Sub Mono: -9dB
Center Sub LFE: -12dB
and results in the following REW plot:

edit: See corrected posts below. Found corruption in XMC-2 and did Factory Reset and manually entered settings with new Dirac run.

[ttocs]

Aug 3, 2021 8:51:10 GMT -5 markc said:

Is it the same result when a Blu-ray calibration disc pink noise is played through the G3P level matching adjustments instead of the internally generated ones?

I just used the Dolby ATMOS bluray disc and it's a weird result. I used REW's SPL meter with a UMIK-1 mic, and the Dirac EQ is engaged.

With the system setup for 5.1, I ran the test for 5.1.2. All the channels are really close to each other in level except the Right Surround, which is about 5-6dB low. Weird. I ran the track lots of times looking for settings which might be different, but it was the same each time. 

Then I changed the system setup to 5.1.4 and ran the 5.1.4 test. This resulted in the Surround Left test to run on both the SL and Front Left together, same for the Right Surround test running on both Right channels together. I've run these Dolby Disc tests on previous firmwares and they used to be correctly interpreted for each channel and also pretty close to the XMC-2 Levels settings, mostly. Now, the Left Front is one level, the Right Front is +2dB vs LF, and the Center is +4dB vs LF. Makes no sense.

I then tried the Studio Six Digital ATMOS app for TV4K and ran the 5.1 test which the XMC-2 shows as Surround. This ran all the speakers and sub within about a 3dB range, so this seems pretty good I guess, margin of error sort of thing. So this pretty much agrees with REW.

I'm too perplexed to continue right now, gotta think. I'll run this again later with User EQ.

edit: I decided to try User EQ now, and guess what? It's what I would've expected.

With the Dolby Disc running 5.1.2, the system setup for 5.1, all the channels are pretty close to what XMC-2 Levels is. Up 1dB, down -1dB here and there, so within tolerance I'd say.

See next post for my theory.

edit: Still postulating, but see following post.

[ttocs]

FW2.3
User EQ vs Dirac EQ

I just discovered that Levels doesn't act the way it used to prior to FW2.3. It used to be that Levels could be adjusted for User EQ and it would result in all the channels being about the same level, except for this BMbug thing. And the same could be said for Dirac EQ, in that after running Dirac we might still need to adjust the levels a tiny bit. Now that's all gone.

User EQ still works the same.
Dirac EQ DOES NOT.

With the system setup as 5.1.4, I re-ran XMC-2 Levels and got all channels equal when Dirac EQ is active, but the Levels settings are waaaay weird with a 13.5dB range from low to high.
Then ran the Dolby Disc again using the 5.1.4 test and the LCR are kinda close, but again, Left is one level, Right is +1dB, and Center is another +1.5dB, weird, but then the SL and SR are sharing the pink noise with the Front Left or Front Right. So this is broken when using Dirac EQ.

To recap, if that's even possible, User EQ is actually more stable in Levels being more universal amongst discs and apps for checking levels, but Dirac EQ throws anarchy into the mix.

So now a Factory Reset will be done to clear everything and start fresh as a check. There's got to be something not right.

edit: After Factory Reset, manually inputted speaker setup, and a new Dirac run, things seem pretty normal when using Dolby Disc. Speaker channels are pretty equal, within reasonable tolerance, subwoofer is +5dB, but that's normal for this firmware. But things are more normal.

When I last applied FW2.3 I used Saved Settings. Is this what got things wacky? With this fresh Factory Reset I manually entered everything, like I usually do, but like I said, previously I used saved settings.

Something went off the rails and called an audible, which was audible. This has never happened to me before.

I'll run REW next.

edit2: REW shows things back to normal, well, with the BMbug still present, but at least it's showing what I've been used to with 2.3. The XMC-2 must've been corrupted. First time for this. But also the first time I've used Saved Settings. Not saying that's what did it, but it's coincidental for sure.

[p4t]

Aug 3, 2021 8:51:15 GMT -5 ttocs said:

Aug 3, 2021 7:43:31 GMT -5 marcl said:

Am I understanding correctly?

After running Dirac, you used the pink noise source from the XMC-2 to check levels and adjusted them with an SPL meter (REW?).  Yes.
This produced the incorrect levels.  Yes.
After that you used REW as the pink noise source and adjusted levels using an SPL meter, producing the correct levels.  Yes.
Dirac calibration was done with the default Dirac target curve?  Yes. 

This seems to indicate that maybe the pink noise source in Levels in the XMC-2 is played without any Dirac level processing and so it reflects the levels without the level alignment that Dirac does, but clearly with Dirac filter correction.  When you use REW as the pink noise source, Dirac level adjustment is definitely applied and you add final level adjustments on top of that.

Yes to the first 4 points, but the last section is questionable. I find the same behavior whether using the User EQ or Dirac EQ. I pointed this out when bringing the BMbug to light and using only User EQ.

Bottom line, Levels pink noise in the G3P are not accurate. The inaccuracy seems to be in Subwoofer channels and Small channels, so my belief at the moment is that it's a BMbug artifact.

I can use Levels to adjust all the channels to be the same SPL. Then when I run REW the levels of Small channels above XO are not equal to Large channels in the same frequency range, so I adjust them down. Then you add the BMbug on top of this knowing that it's elevated above the Small channels by at least 4dB, so this needs to adjust down -4dB and then the amount that the Small channels were lowered by.

After running Dirac and using the XMC-2 Levels pink noise resulted in the following settings:
Large Front Left: -1.0dB
Large Front Right: 0.0dB
Large Center: 0.0dB
Small Left Surround: +4dB
Small Right Surround: +3dB
Left Sub Mono: -4dB
Center Sub LFE: -4dB
and results in the following REW plot:


After using REW sweeps measurements for levels using Surround mode, the settings are now:
Large Front Left: -1.0dB
Large Front Right: 0.0dB
Large Center: 0.0dB
Small Left Surround: -1dB
Small Right Surround: -2dB
Left Sub Mono: -9dB
Center Sub LFE: -12dB
and results in the following REW plot:

This result get me a little confused. I thought your previous result all speakers set as small below the crossover point will get elevated. But this one seems to be okay other than the volume levels a few db louder across the whole freq.

[ttocs]

Aug 4, 2021 12:12:39 GMT -5 p4t said:

This result get me a little confused. I thought your previous result all speakers set as small below the crossover point will get elevated. But this one seems to be okay other than the volume levels a few db louder across the whole freq.

I'm confused also. Please see my edits above.

The XMC-2 must've been corrupted somehow and caused those anomalies in REW and also when using the Dolby Disc and Studio Six Digital ATMOS generator. So it's good to check multiple ways because the Dolby Disc was the most confused of the bunch, causing pairs of channels to operate together and very strange sequence of raised level - channel by channel. Made no sense at all.

Factory Reset, manually input all new settings, new Dirac run, now we're back to just the basic BMbug.

The key thing here is that only the Dirac EQ slots were affected, not User. So the only guess I have right now is that the saved settings I used were not totally accepted by the processor, even though the Dirac Slots were brand new runs on top of Restored Saved Settings.

So please disregard those posts. I'll maybe strike-through those but leave them for context.

[markc]

My take is that I am in a similar camp to you.

My XMC-2 is a Forrest Gump box o' chocolates.
I am never sure what I am getting or going to get at any time.
I should not need to want or need to measure with REW to check if what I am hearing is my over-active paranoid "this processor is crap" imagination or an actual problem.

On Tuesdays Thursdays and three out of 5 Fridays and/or Saturday plus/minus Sundays, a bit-streamed HMDI DTS or Dolby feed is played exactly how it should be.

Oh wait. OR is it played with an error? I can't be quite sure, but I just want to watch my movie / TV show and put up with it sounding like it is and try to blame the source.

[ttocs]

Saved and Restored settings might be a problem with Dirac, at least it is for me.

I Restored the Saved Settings from firmware previous to FW2.3.
There was no Dirac EQ Slot in the Saved Settings so I ran a new Dirac session with these Restored Settings.
After running Dirac, I used Levels to adjust the channels.
Then I ran REW.
The REW results are exactly like the ones I called "corrupted" in previous posts. The Small channels show elevated BM about 4-5dB above the frequencies above the XO, and the frequencies above the XO are elevated about 6dB above the Large channels. So it's saved and restored settings that are causing this to happen.

Can saved settings from any firmware version be restored successfully? It seems to be yes if User EQ is used, but no if new Dirac filters are created.

As I mentioned in a previous post, I have always manually entered all settings after a firmware update except for almost two weeks ago when I reverted back to FW2.0, then progressed to 2.1, 2.2, and 2.3. I don't recall if these saved settings are from 2.1 or 2.2, but it's one of them.

[p4t]

Aug 4, 2021 20:14:27 GMT -5 ttocs said:

Saved and Restored settings might be a problem with Dirac, at least it is for me.

I Restored the Saved Settings from firmware previous to FW2.3.
There was no Dirac EQ Slot in the Saved Settings so I ran a new Dirac session with these Restored Settings.
After running Dirac, I used Levels to adjust the channels.
Then I ran REW.
The REW results are exactly like the ones I called "corrupted" in previous posts. The Small channels show elevated BM about 4-5dB above the frequencies above the XO, and the frequencies above the XO are elevated about 6dB above the Large channels. So it's saved and restored settings that are causing this to happen.

Can saved settings from any firmware version be restored successfully? It seems to be yes if User EQ is used, but no if new Dirac filters are created.

As I mentioned in a previous post, I have always manually entered all settings after a firmware update except for almost two weeks ago when I reverted back to FW2.0, then progressed to 2.1, 2.2, and 2.3. I don't recall if these saved settings are from 2.1 or 2.2, but it's one of them.

Just a suggestion. Have you try start from fresh. Factory reset XMC2 then taking a new measurement again using umik1 for dirac. And save the result. After that test the result with REW. 

[ttocs]

Aug 4, 2021 21:14:06 GMT -5 p4t said:

Aug 4, 2021 20:14:27 GMT -5 ttocs said:

Saved and Restored settings might be a problem with Dirac, at least it is for me.

I Restored the Saved Settings from firmware previous to FW2.3.
There was no Dirac EQ Slot in the Saved Settings so I ran a new Dirac session with these Restored Settings.
After running Dirac, I used Levels to adjust the channels.
Then I ran REW.
The REW results are exactly like the ones I called "corrupted" in previous posts. The Small channels show elevated BM about 4-5dB above the frequencies above the XO, and the frequencies above the XO are elevated about 6dB above the Large channels. So it's saved and restored settings that are causing this to happen.

Can saved settings from any firmware version be restored successfully? It seems to be yes if User EQ is used, but no if new Dirac filters are created.

As I mentioned in a previous post, I have always manually entered all settings after a firmware update except for almost two weeks ago when I reverted back to FW2.0, then progressed to 2.1, 2.2, and 2.3. I don't recall if these saved settings are from 2.1 or 2.2, but it's one of them.

Just a suggestion. Have you try start from fresh. Factory reset XMC2 then taking a new measurement again using umik1 for dirac. And save the result. After that test the result with REW. 

Haven't gotten there yet, but it's on the agenda. Maybe tomorrow.

[ttocs]

Strange anomalous Small Channels elevated full frequency spectrum issue.

I'm not going to chase this "anomaly" much further, because, in my opinion it's something fairly unique. I just happened to have reverted back to FW2.0, and then updated to each firmware since in a progression, and I used saved settings for each update to speed up the process. Somehow, these saved settings, which are saved internally and to a USB stick and both provide identical results, cause Small channels while using a Dirac filter to be totally elevated for the entire frequency spectrum. The key things that are in play here are that the settings I think are from FW2.0 or 2.1, don't recall which, and Preset2 which is what I use for User EQ have PEQ settings for the LCR and SL channels, and Preset1 is what I use for Dirac but these saved settings don't have any Dirac filters nor does Preset1 have any PEQ settings. So the issue is that Preset1 has this elevated Small Channels issue whether Dirac is run anew or a Dirac Project is loaded into Preset1, but if I remove the PEQ settings from Preset2 this problem goes away. I've done this probably a dozen times in the last couple days and every time I remove the PEQ settings from Preset2 the Preset1 Dirac problems go away.

This is NOT a Dirac problem. It's somehow a strange issue with whatever FW these saved settings are from, and the fact that PEQ is present on the User Preset that somehow affect the Preset that normally has Dirac on it but Dirac isn't saved on these settings, so Dirac is added after the Restore Settings and it doesn't matter whether Dirac is a new run or loaded from a saved project.

Factory Reset, manually entered new settings, Dirac, no problemo.
Factory Reset, manually entered new settings, manually entered PEQ on one Preset and Dirac on the other Preset, no problemo.
Factory Reset, Restore These Saved Settings with PEQ on the User Preset and no Dirac on either, then load a Dirac filter on the other Preset, problem.
=======================================================================

Aug 4, 2021 21:14:06 GMT -5 p4t said:

Just a suggestion. Have you try start from fresh. Factory reset XMC2 then taking a new measurement again using umik1 for dirac. And save the result. After that test the result with REW. 

Dirac was never saved in the settings that caused this anomaly, so I have amended your suggestion to include only what was done before except that this test won't have PEQ on Preset2. This test did not cause a problem, but it also doesn't have any PEQ and wasn't saved from a previous firmware. So this operates as it should except for the BMbug.
=======================================================================

I have saved settings for whatever the current firmware had been and used those when doing Factory Resets, but I've not used settings from a previous firmware after updating, until last week when doing all that testing and using one set of saved settings saved time. But again, it wasn't until I added the PEQ to the saved settings that the strange elevated Small channel thing happened. So I guess, since I'm nailing coffins shut, I'll add some PEQ to this set of FW2.3 saved settings on Preset2 and see what happens.

[ttocs]

Like Tommy in Goodfellas, I didn't see this one coming. This is all within Firmware 2.3. No other firmware involved.

This PEQ Saved/Restored Settings bug is a current problem in FW2.3. After experimenting with PEQ the week before last I decided I want to use PEQ for the User EQ that I setup Preset2 with. But can't if I expect to Restore Settings ever. Fresh PEQ doesn't seem to cause a problem, it's only when Restoring Saved Settings that this happens.

I just now did the following:
Factory Reset
LPS reboot
Manually entered Speaker Size and XO settings on Preset1, did not Level
Manually entered Speaker Size and XO settings on Preset2
Manually entered PEQ settings for LCR on Preset2 on Filters 1-4
Leveled Preset2
Assigned "Surround" to HDMI5 for 5.1
Saved Settings to USB
Factory Reset
LPS reboot

Restore Settings from USB with PEQ on Preset2, no Dirac at all,
Switched HDMI inputs away and back

On Preset1 loaded a saved Project into Dirac Slot1
Switched HDMI inputs away and back
Leveled Preset1
Switched HDMI inputs away and back
Ran REW sweeps on Large Front Left and Small Surround Left

The Small SL is elevated all across the frequency spectrum with additionally elevated BM.


After writing this I reverified Levels, no change, and measured again, same results.

[JKCashin]

So... if you have created new settings but still have the old ones, you can do a comparison.... the files are stored in XML format so are easy to compare.

[JKCashin]

If you want to send me the files I can see what's different about them.

[ttocs]

Aug 6, 2021 22:51:25 GMT -5 JKCashin said:

So... if you have created new settings but still have the old ones, you can do a comparison.... the files are stored in XML format so are easy to compare.

Thanks for the offer. There's really nothing of interest in the Settings files that I've compared. The issue is really that one preset with PEQ is disturbing the other preset with no PEQ but is running through a Dirac filter. As soon as each PEQ filter level is set back to 0.0dB the issue disappears. This issue doesn't happen if PEQ is setup from scratch, it only happens when saved settings with PEQ are restored and then a Dirac filter is loaded and used. It's a very specific sequence.

Making Things Work (with some extra effort) Restoring Settings shouldn't require a cold boot, but that's what brings things back to working.
As I was writing this I did a couple minute cold boot. I then went back to Levels and immediately saw that the Small channels and Subwoofer channels were different from the Large channels and re-Leveled them. Now they are back in line as shown in the plot attached. So after Restoring Settings a cold boot is needed apparently to make sure everything is "set", including PEQ.

I was reminded earlier today that when Dirac was first enabled last year, the XMC-2 didn't load a new Dirac filter until switching to the other Preset and back again. I tried this today but it didn't work. I also did LPS reboots, and switching HDMI inputs away and back, but nothing would allow Levels to work like the cold boot did.

The Red trace is the new one after a cold boot and re-leveling.

[ttocs]

Finally found the culprit. PEQ is not needed to show a problem, however, it does enhance the problem, but today I'm not using PEQ.
Today is all about how the problem interacts with Dirac.

Factory Reset
Cold Boot
Manually entered settings for speaker size, XO, etc for both Presets.
Can't do a Cold Boot without losing the newly entered settings, but if they were saved first, then they can be restored, or instead use LPS first which makes them stick. Either way it needs a reboot for the settings to stick.
After successful reboot while retaining settings I Leveled Preset2 for User EQ, and saved settings, then LPS reboot, then Leveled Preset2.
Saved the settings again with the new Levels for Preset2.
And another LPS reboot. Any time I use the Menu to change something while testing I always reboot to be sure.

Next, loaded a saved Dirac project and loaded that into Preset1 Slot1.
LPS reboot.
Level Preset1 using REW SPL meter.
Switch HDMI inputs away and back.
Measure with REW sweeps for Large Front Left channel and Small Surround Left channel. The Small channel bass is very elevated, higher than the usual BMbug.
LPS reboot.
Measure again with REW, same very elevated results.
Cold boot for at least a couple minutes.
Measure with REW and find that the Small Surround Left has now fallen in line with the usual BMbug amount of elevated bass.

This is to show that it's not PEQ's fault, but PEQ can make the problem be extend beyond the XO if there is PEQ in upper frequencies in some channels of Preset2. But again, this is a short term problem. After a Cold Boot the problem doesn't exist.

This problem comes up after loading a Dirac filter. The only way to "clear" the levels imbalance is to perform a cold boot. Last year when the first firmware for Dirac became available we noticed that weird levels were happening after loading a Dirac filter, but that a Preset swap - change to other Preset and back - would clear the issue. Now, this looks similar in that it needs a cold boot to clear the issue.

It is not fun to need to cold boot after using Dirac. But without it, the actual volume of each channel might not be correct, even though the settings in Levels "are" correct.


Here is where I need to question how pervasive the BMbug is with regard to when a Dirac filter is active. It may not be as bad as some of the plots have shown, only because maybe some of those would show lower levels of bass if a cold boot had been performed. Don't get me wrong, there is a big problem with BMbug bass, but "some" tests might have shown more exaggerated levels of bass when LPS was used vs a cold boot.

Every time a Dirac filter is loaded, I'll need to do a cold boot.

[ElectricKoolAid]

Just wanted to say thanks ttocs for all the investigation that you've done on this. I've been watching this thread every day hoping for an update or any further information. Has Emotiva said anything further on their end as to if they may be able to fix this?

[ttocs]

I'm optimistic.

[muslhead]

Aug 19, 2021 13:30:18 GMT -5 @j*bleep*hrbc said:

Just wanted to say thanks ttocs for all the investigation that you've done on this. I've been watching this thread every day hoping for an update or any further information. Has Emotiva said anything further on their end as to if they may be able to fix this?

As do i.
Disappointing but not unexpected Emotiva tech support is MIA. This is their MO and why people are looking elsewhere for answers to their audio/video solutions.

[marcl]

Aug 20, 2021 7:50:14 GMT -5 muslhead said:

Aug 19, 2021 13:30:18 GMT -5 @j*bleep*hrbc said:

Just wanted to say thanks ttocs for all the investigation that you've done on this. I've been watching this thread every day hoping for an update or any further information. Has Emotiva said anything further on their end as to if they may be able to fix this?

As do i.
Disappointing but not unexpected Emotiva tech support is MIA. This is their MO and why people are looking elsewhere for answers to their audio/video solutions.

Reliable sources confirm they are making progress ... patience!  They heard us loud and clear.

[okjazz]

Does anyone here have tested or know the impact of the Dirac 3.1.1 to the BM bug? Improve? Same? Worsen?

I would be on side until the BM bug gets resolved.