Understanding sub phase/group delay for placement

[jtalden]

Before I put these tests to rest I'd like to share some (probably very uncommon) facts I encountered yesterday.
1. EQ experiment
I tried to boost 63 and 80 Hz at the FR I posted in post #70. I did three tests with Q1 Q2,5 and Q5 on both front subs.
The result was even bigger peak at 45 Hz. I find this very weird.

I don't follow. I recommend that once you decide a favorable 4 SWgroup setting than no further changes should be made to the individual SWs (no EQ or gain changes). Use REW EQ panel on the 4 SWgroup measurement to determine PEQ settings. Then apply those EQs to all 4 SW individually. The new 4 SWgroup measurements will match the predicted response in the REW EQ panel.

2. Location experiment
Now, time for some heavy lifting. I moved front subs to a new place and re-measure them. Turned out their response is quite good - way better than before. I used delay to time align them with mains. To get even nicer align I changed phase on one sub as well. Both needed reversed polarity. IR graph is now perfect.
But when both subs were measured with mains - the FR is very bad.

@jtalden , when you'll have some spare time, would you please care too look and make a comment? This could look as a phase issue ...

I suggest you measure the setups I recommended above to see if they meet you needs. You will get the same SPL response shown above if the settings are correctly applied. PEQ to can be used as needed for fine tuning on the SWgroup.

 

[thexder]

Wow, thank you so much for that. This is really a nice surprise!

I had very little time yesterday so I checked out Option 1 first. What a pleasant surprise.
I didn't see until now you proposed different phase settings as well.
Here are two measurements with phases unchanged. I don't know why second one is quieter. All the settings were the same. I had some visitors and left the gear turned on and then re-measured.
A quick test confirmed sound improvement which really made me happy.
I wonder how the IR graphs look ...

I'm looking forward to more tests today.
Am I understanding you correctly that Option 2 has two options? First, rear distance 3,8 and second, rear distance 1.95?

Now my big curiosity, if you don't mind me asking. How did you figure this out? Manual calculations? Or does REW have prediction options?

Again, I really appreciate your help! I might finally be able to stop spending time setting up gear but rather enjoy it.

 

[thexder]

I don't follow. I recommend that once you decide a favorable 4 SWgroup setting than no further changes should be made to the individual SWs (no EQ or gain changes). Use REW EQ panel on the 4 SWgroup measurement to determine PEQ settings. Then apply those EQs to all 4 SW individually. The new 4 SWgroup measurements will match the predicted response in the REW EQ panel.

Yes, you are right, I was just experimenting and was surprised how EQ was affecting the measurements.
You are recommending I should look at the mains+4subs curve, then define the PEQ parameters and then apply the same parameters to each sub? I kind of global PEQ?

I suggest you measure the setups I recommended above to see if they meet you needs. You will get the same SPL response shown above if the settings are correctly applied. PEQ to can be used as needed for fine tuning on the SWgroup.

I did and I will. This was just a test where I moved front subs to my previous location. The response of an individual looks better and I was wondering if I made the right choice putting front subs in the corner. Now that you helped me with the settings I'm more confident the corners are good despite much less flatter response of each sub. I'm however still intriqued if locations next to speakers would work good as well. What do you think?

 

[jtalden]

I checked out Option 1 first.
I didn't see until now you proposed different phase settings as well.

No, phase was copied from the reference data. I have no way to make SW phase changes. The proposed phase settings are per the reference Post-68 settings. I repeated your nomenclature in column 2 of the table just to highlight them. If there were any other things changed from Post-68 then those also should have been returned as they were. I suspect settings were pretty close as an overlay of my calculated response and your measurements were close other than an improvement at about 46 Hz. That could have been due to a room difference at the time of the measurements or maybe a difference in phase settings.

Am I understanding you correctly that Option 2 has two options? First, rear distance 3,8 and second, rear distance 1.95?

No, just one. The changes are in the last 3 columns. Columns 3, 4, 5 are used by me to record my calculation steps in case there is a discrepancy between my prediction and your subsequent measurements.

Now my big curiosity, if you don't mind me asking. How did you figure this out? Manual calculations? Or does REW have prediction options?

I used the REW alignment tool and spent a lot time finding settings that resulted in favorable response tradeoffs.

 

[jtalden]

You are recommending I should look at the mains+4subs curve, then define the PEQ parameters and then apply the same parameters to each sub? I kind of global PEQ?

Yes, that is the best way. Any differences in EQ between the individual SWs will upset the careful balancing act we made that resulted in a smooth SWgroup SPL response.

I did and I will. This was just a test where I moved front subs to my previous location. The response of an individual looks better and I was wondering if I made the right choice putting front subs in the corner. Now that you helped me with the settings I'm more confident the corners are good despite much less flatter response of each sub. I'm however still intriqued if locations next to speakers would work good as well. What do you think?

I have no interest or time to explore all the various setup options in your room. I helped you find a couple of very favorable setups. You are now on your own to experiment and learn. That is how I did it.

 

[thexder]

Thank you. I'm really thankful for your time. I hoped for an easier way so I could do/learn this on my own. I've spent countless of hours every day to get there and you perfected my work in one day. Hats off to you, sir!
I'm going to check other options you proposed and get back with the results.

I have no interest or time to explore all the various setup options in your room. I helped you find a couple of very favorable setups. You are now on your own to experiment and learn. That is how I did it.

Of course not, I did't mean you should repeat I was just wondering if one can tell the best location just by looking at the response. The second location seem to have great response but yet we were able to get near perfect result with two subs having (IMO) worse response.

Anyway, I'm extremelly happy and if everything stays as per Option 1 - you won't see me complaining.

 

[thexder]

Okay, took me a while to test everything as I wanted to be thorough. I'm sorry for such long response.

You took data from post #68 where phases were set to 75 and 105 but later I changed them in post #70.
I'm glad you took that ones. With delay set to proposed 3.8 this is now it. I tested every phase settings in latest measurements and those two are obviously optimal.
It's great. I guess longer decays at 45 Hz can't be solved but this is still the best sound I have experienced with my gear.
I took a look at impulse plots and none of the subs are now aligned. I guess this isn't so important after all?

I didn't manage to set Option2 as you predicted. The response is far away from yours.

Also tried to run Audyssey. Very good results, a flat FR within +/-3 dB. Even there the 45 Hz has longer decay.

I'm attaching the mdat in case you'd like to see. Many plots should look the same but somehow the 45 hz peak is swinging for approx. 1-2 dB. Can't figure it out why. Nothing in the room was changed.

Again, I would like to thank you for your support!

 

[jtalden]

The phase settings of SW 3 and SW 4 were transposed in my table. That may be the reason the option 2 did not measure as predicted.

Below is the corrected table.

 

[thexder]

Based on my previous attemts phase changes from 75 to 105 didn't make such huge differences. Anyway, I like the first option and now it's time to enjoy listening!
Forgot to mention. Tried EQ too (can be seen in mdat file). Interestingly, it is far away from REW predictions.

However, I'm still curious if impulse aligments are important for a better result?
I mean, my room is a proof they obviously aren't but many people are suggesting IR alignments should be done.

 

[jtalden]

Time alignment is what we have been doing with the phase settings and the distance settings. This is very important as we have seen with the wide scatter of SPL responses obtained using various settings.

The relative impulse positions of the Individual SWs (step 1), or the SWgroup vs the mains (step 2) tells us only that we are not straying too far afield in the sound arrival times that may create a problem. It does not indicate whether the timing is favorable for SPL or not. Favorable timing requires adjustment to the initial settings typically based on impulse locations or distance measurements. The timing adjustments can be done within REW or by trial and error experimentation.

There is more than one favorable time alignment for smooth SPL in most setups, including yours. They have tradeoffs of other factors such as group delay and sound character. The charts will not show us which one a listener will prefer.

 

[thexder]

I see, thank you for the explanation.
Yes, listening is important, I've spent too much time looking and relying on plots.
Did you use alignment tool along with impulse plots?
I tried to use alignment tool with average plots (two subs pairs) but looks like this can only be done for single sub measurements. More subs, more work!
Theoretically, is it possible to get a good SPL and aligned impulse positions as well? Seems like a tough job with multiple bass sources on different locations.
Sorry for asking so many questions, this is something worth learning.

 

[jtalden]

Did you use alignment tool along with impulse plots?

No??
The alignment tool shows the impact of changing 3 parameters; delay time, polarity and gain of the 2 chosen impulses. The original 2 Impulses are not changed. If favorable settings are found within the tool, copies of the 2 impulses including the parameter changes can be saved as new impulses and the resulting sum impulse can also be saved. If the tool parameter changes are actually implemented in the setup then new measurements will exactly reflect the prediction made by the alignment tool. So, the alignment tool is indeed considering the original impulse location gain and polarity and providing the resulting impact of changes to those parameters. There is no need to look at the impulse chart separately during this process. It is helpful, but not necessary, to first time align the initial rise of the impulses using the AVR or MiniDSP so that it is more obvious to us if timing changes considered within the tool are moving too far from the starting point alignment.

I tried to use alignment tool with average plots (two subs pairs) but looks like this can only be done for single sub measurements. More subs, more work!

It can be done on 2 pairs of SWs. We just need to first use the REW trace arithmetic tool to combine each pair using "A + B". We can then use the alignment tool on the resulting 2 Impulses. This is what I did to find the find the favorable SW timing alignment.

Theoretically, is it possible to get a good SPL and aligned impulse positions as well? Seems like a tough job with multiple bass sources on different locations.

It is very unlikely that aligning any defined feature of 4 impulses representing distributed individual SWs will always result in a favorable timing alignment. We need to be careful however as to what exactly is being proposed. It may well be possible to use a narrow sweep within the primary SW frequency range and align those resulting impulse peaks for instance. There are lots of methods to find favorable settings and it is not wise to suggest a method will not work without appropriate study/investigation.

 

[thexder]

Great, this is really helping me learn very useful functions of REW.
I tried to play with 2 sub pairs but didn't succeed. I included mains of course (I used trace arithmetic A+B for left+right speaker).
I shall practice more, I guess.

What is the difference in trace arithmetic tool between A+B, A-B, A*B, etc?

 

[jtalden]

The only operators you need to know are summation, A+B and average, (A+B)/2. These are vector operators so it is important the time alignment is maintained. The calculation includes the impact of phase as well as the SPL. It is best if full-range measurements are used for all measurements rather than the truncated frequency range used in you measurement. See REW help for other notes.
In your case we are summing 4 SWs and averaging the L,R mains. In your file you summed the mains instead.
Attached is my mdat for Option1.
Measurements;
1-6: Your measurements
7: Avg of L, R mains
8-10 Summing the SWs with no timing change.
11-13: Summing the SW pairs with alignment tool changes
14-16: Summing 7 and 13 with alignment tool changes

 

[thexder]

OK, much more clear now, thank you. This is really helpful and can save a lot of time instead of doing trial and error experiments.
Couple of questions though:
1. I see A+B, (A+B)/2 and average look pretty similar except A+B is louder. You used (A+B)/2 for all actions, I assume this is the best option? Probably it's important to use the same algorithm for all plots.
2. Delay slider can slide left and right. Sliding it to left can be set up in the AVR. What about sliding it to the right? Does any AVR offers negative delay setting?

Makes a lot more sense now why miniDSP can be very handy. It's easier to time-align subs with delay instead with phase changes. However, that's no shortcut for determining optimal HP/LP filters.

 

[jtalden]

1. I see A+B, (A+B)/2 and average look pretty similar except A+B is louder. You used (A+B)/2 for all actions, I assume this is the best option? Probably it's important to use the same algorithm for all plots.

My mistake. I should have summed the 4 SWs instead of averaging them. The good news is that this does not impact Step-1; the favorable integration of the 4 SWs. It does impact the gain choice on Step-2; the handoff from SWs to mains. The SWs SPL level in my SPL charts is about 12 dB lower than the summation would have been. For Step-2 my table offered a gain change of the SWs LFE vs mains gain level of 0 dB in as a reasonable starting place. That should have been be a 12 dB reduction of gain level on the LFE output (assuming I now have this correctly in mind). In the end, the gain level between LFE output and mains level is chosen based on the listeners preference. Just adjust the LFE level as needed.

2. Delay slider can slide left and right. Sliding it to left can be set up in the AVR. What about sliding it to the right? Does any AVR offers negative delay setting?

It's relative timing we found that counts. Adding a positive delay to A is the same as adding a negative delay to B. No matter the number of the speakers and subs in a system there is sound from one of them that arrives at the LP first. That speaker becomes our reference and can be set as low as 0 ms delay if needed. All the others then have more delay, but all still must retain the proper relative timing between them.

 

[thexder]

Not sure if I follow. The way I see it you used A+B average for plot SX4 (looking at mdat from post 89).
If you would have used just A+B it would not be marked with the sign "Trace Arithmetic A,B average", right?
I repetead thisprocedure today and got the same result as you did. *
Or are you talking about older mdat from post 68?

* I followed this procedure:
1. trace arithmetic (A+B)/2 for L+R
2. trace arithmetic (A+B)/2 for SUB1 + SUB2
3. trace arithmetic (A+B)/2 for SUB3 + SUB4
4. align tool: SUBS 1+2 vs SUBS 3+4 (set gain +4,6 and +3,8 m delay)
5. trace arithmetic (A+B)/2 for SUBS1+2 and SUBS3+4
6. trace arithmetic (A+B)/2 for SUBS1+2+3+4 + mains (joined in step 1)

 

[jtalden]

Yes, your steps are equivalent to the steps I took and you obtained the same (wrong) answer in your Step-6 as a result. As stated, I should have summed the 4 SWs instead of averaging them. That mistake did not impact the SPL shape of the 4 SWgroup. It only impacted their overall gain level. It impacts your Step-6 result; the relative gain level of the SWgroup (LFE gain) Vs the mains.
You can easily confirm this just by making new measurements. Use L+R channel signal for the measurements. Measure the 4SWgroup alone, then measure the L+R mains alone, then measure the SWs and L+R mains all together. That's only 3 measurements. Measurement 3 (all together) should equal the A+B trace arithmetic of measurement 1 + measurement 2. If not, there is an error someplace. If you then overlay the high frequencies of my predicted Option-1 response you should find that the measured bass range is 12 dB higher than my prediction.

Again, just adjust the LFE gain to the level that you prefer. There is no correct level. If you listen to music at modest levels you will probably like the LFE gain higner than if you listen at reference levels.

 

[thexder]

So the proper thing would be to use trace artihmetic A+B (instead of (A+B)/2 for step 6?
I compared the A+B and (A+B)/2 and I see the sum only 6dB hotter.
Did you made the same mistake for mdat in post #68 or are we talking about post #89?

Because I liked the sound immediately when your changes were introduced. And the response looks great too which confirms your suggestions were spot on. I came pretty close to my target curve.
12dB seems a lot and I doubt I would get better sound with subs running 12dB lower.

 

[jtalden]

So the proper thing would be to use trace artihmetic A+B (instead of (A+B)/2 for step 6?

I suggest you not spend so much time on recreating my mistaken analysis process. It would probably be more useful to understand the correct process for your situation (see options below). Since different setup situations lead to different analysis steps we also must understand how to select the process steps needed for different situations - it takes while. You can see, I still get confused.
I now see my mistake was only in averaging SWs 1,2 and 3,4 instead of summing them. I correctly summed the 2 groups using the alignment tool. The net impact of this mistaken analysis was then to under estimate the gain level of the SWgroup by 6 dB instead of the 12 dB that I recently stated above. Again, this only impacts the plotted SPL level of the predicted LFE output. That's a level that normally finally adjusted to listener preference. It also impacts any comparison of predicted Option-1 SPL plots to new measurements of Option-1 settings as previously indicated.

I compared the A+B and (A+B)/2 and I see the sum only 6dB hotter.

That is correct summing 2 identical sounds results in a 6dB gain. The alignment tool uses vector summation. Summing 4 identical sounds results in a 12 dB gain. SWs measurements are not identical sounds so the actual vector summation of the 4 will probably average less than the max 12 dB possible.

Did you made the same mistake for mdat in post #68 or are we talking about post #89?

Post-68 was your data that I analyzed in Post-75 and posted the actual mdat of that analysis in Post-89. Both contain the error I reported. The 4 SWs should have been summed.

Because I liked the sound immediately when your changes were introduced. And the response looks great too which confirms your suggestions were spot on. I came pretty close to my target curve.
12dB seems a lot and I doubt I would get better sound with subs running 12dB lower.

As I noted it's the relative LFE level compared to the mains level that is impacted by this mistake. The actual level LFE level you prefer may be very different than the one I might prefer in your room and I have no way to know what level I might prefer. Adjust the LFE to you preference is the message. I only pointed this out as a mistake to so that if you measure the option 1 settings and notice that bass level is significantly higher than the mains than I charted you would understand why. There was nothing special about the level I plotted. I just looked at the response with no LFE level change and decided It look reasonably good without change.

Option-1:
I can run through the Post-68 analysis again correcting the mistake. You will then have a clean mdat of the correct process used to find Option-1.

Option-2:
I recommend new measurements for that purpose. It is difficult to really see the actual average mains levels in your Post-68 data since you did not take full range measurements for all 6 measurements. We only have a hint as to where the mains SPL level actually is.

Setup

• Post 68 Settings for all speakers, AVR, Room layout and other conditions
• Measurement channel = L
• Reference channel = L
• Acoustic timing = Active
• Mic Location = LP
• Measurement Sweep = 15 - 20k Hz

Measure

1. L (alone)
2. R (alone)
3. SW1 (alone)
4. SW2 (alone)
5. SW3 (alone)
6. SW4 (alone)
7. Optionally, set all Option 1 settings and measure L+R+SWgroup (i.e., measurement channel L+R)

With measurement 7 we can confirm that the prediction matches the measurement. This is how we find if mistakes were made in the analysis.

Option-3: Adjust the LFE to your preference and enjoy the music.

 

[thexder]

Oh, I'm not analyzing. I'm making notes so I'll be able to repeat this process some day. Hopefully all by myself.
I believe there could be an universal rule. Measure L, R and each sub together.
Then:
1. (A+B)/2 for mains
2. time align subs with alignment tool and sum pair if necessary
3. sum all subs with (A+B)/2
4. sum SW group + mains (here I was asking you which algorith should be used)
Here I've been wondering: when you say summing - you mean "Aligned sum" in Alignment tool? And when averaging you mean trace arithmetic (A+B)/2?

But than again, gain can be set according to taste so I guess summing algorithms are't so important?
Plus, all this can be checked/confirmed with actual measurements.

So I vote for Option 3, thank you very much.
I'm happy with the result. As I posted earlier, I got pretty close to the target curve.
I added a little EQ at 80 Hz area (I gained -3dB on a graph but had to add -10dB on the subs. Looks like subs are not that active at 80 Hz).
Based on many measurements I don't think I can tame that 80 Hz peak.

I think the 120 Hz area could bother me but until I buy an AVR with PEQ there's not much I can do. Perhaps I could try locate the loudest area in my room and try some absorption.

 

[FargateOne]

Dear Jtalden, thank you for this on line class! I am not sute that I would successfully pass the exam!

 

[thexder]

We may not pass but we are sure much closer than before!
I'm really glad there are many good people who are willing to help.

I focused on multichannel sound lately. Not sure if I need some time to adjust but sometimes I feel like something isn't as it should be at low frequencies.
I suppose any critical abnormality would be seen on a measurement? I can't see anything in particular.
Or are some movie mixes above average and my room is now too good to hide this?

 

[thexder]

Many months have passed and this is still a superb setup. I did nothing but enjoyed since this awesome learning lessons. I guess I needed some time to adopt because I see now that low frequencies are more detailed when it comes to movies.

Thank you again, @jtalden!

 

[FargateOne]

@jtalden Hello, many months later and covid confinement obliged (so bored) I would like to begin a new chapter of "Understanding how to use phase and alignment tool to smooth XO in a 5.1 system" I started this thread with "one xo to rule them all' (to paraphrase famous Lord of the ring moment). See post #37 et #38.
Since then, I made some experiments.
My receiver give the possibilty to put one XO for the sub (say 100Hz) and different xo for fronts (xo70), for CC (xo80) and for surrounds (xo100). I ask Rotel and they are LR-24 shape filter. A LPF for the the sub and HPF filter for the satellites. The smoothest response in my system is when I put those xos with the same delay we had found: sub at 240cm, CC at 165c, and fronts at 270cm.

But mixing the xos should change the phase isn't it?

If I want to check if I messed up I must take some measurements.

Here what I want to do: does it mak sense?:

Measurement Setup:

• AVR XO active
• Full range sweeps
• REW acoustic timing activated (in my cas it is the right surround)
• Mic at LP
• EQ filters actives (is it a good idea or whould put EQ filters into the receiver to off?)

Measurements:

1. FL (alone - no SW) xo at 70
2. FR (alone - no SW) xo at 70
3. CC (alone - no SW) xo at 80
4. SL (alone - no SW) xo at 100
5. SR (alone - no SW) xo at 100
6. SW xo100 [Measure this using the Surround right channel with the right surround speaker disabled)

What do you think? Does it worth the effort?