Regarding multi-sub alignment, sub to main alignment and EQ set up

[Niklasmagnus]

Also, how are you analysing to not recommend 90 or 100hz XO from my data?

 

[jtalden]

I completed the 60 Hz alignment. Two favorable alignments were found:
1 - SWs delayed by an additional 3 ms (distance setting in AVR reduced by 1.0 m)
2 - SWs inverted and delayed by an additional 8.5 ms (distance setting in AVR reduced by 2.9 m)
These alignments are measurement 6 and 8 as shown in the chart and file attached.

 

[jtalden]

I completed the 80 Hz alignment. Two favorable alignments were found:
1 - SWs delayed by an additional 5.5 ms (distance setting in AVR reduced by 1.9 m)
2 - SWs inverted and delay unchanged
These alignments are measurement 6 and 8 as shown in the chart and file attached.

 

[jtalden]

Also, how are you analysing to not recommend 90 or 100hz XO from my data?

I am expecting that the closer the XO gets to the SWs null around 115 Hz less SPL reinforcement and EQ ability there will be in the XO range. If we compare 60 to 80 Hz alignments above there is not much difference in overall SPL, but there is less SPL reinforcement across the XO range using the 80 Hz XO.

 

[Niklasmagnus]

Thanks for you input. I will now test the 4 different setting and check seat-to-seat and sound quality. I might add 1-2 EQ filterns to reduce the strong 43hz peak to get better understanding of the sound quality.

 

[jtalden]

Please let know if differences are heard between some or all of them.

 

[Niklasmagnus]

I have done verification measurement on your input. I guess you would like to see the results of your good work.
- The XO60 verification looks good for both suggestions.

- The first recommendation (SWs delayed by an additional 5.5 ms) (called _a_ in file) for XO80 did not align with the simulation. I took new measurement with _a_ settings and did tried to get the sum to align your suggestion, measurement 13-15. I used the alignement tool and got quite good match (16 Aligned sum_c_) with delay -1,37ms. Verified with 17 and 18.
(open first "...file 1 of 2" and then "...file 2 of 2 " to get measurements number in correct order.

will run seat to seat now and apply EQ before start doing listening test.

 

[jtalden]

- The first recommendation (SWs delayed by an additional 5.5 ms) (called _a_ in file) for XO80 did not align with the simulation. I took new measurement with _a_ settings and did tried to get the sum to align your suggestion, measurement 13-15. I used the alignement tool and got quite good match (16 Aligned sum_c_) with delay -1,37ms. Verified with 17 and 18.
(open first "...file 1 of 2" and then "...file 2 of 2 " to get measurements number in correct order.

It appears a difference in mic location between the measurements used for targeting vs those now used for verification may be responsible.

The SPL support in the 80 Hz range for the 80-a target is slightly better than the 80-a verification (Traces 6 vs 9+10) as you indicated.

The step response of the 80-a verification shows the SWs to be delayed a little more than the 80-a target (Traces 6 vs 9+10).

The step response of the 80-c verification shows the SWs to be delayed slightly less than the 80-a target (Traces 6 vs 17+18). That may explain why the 80-c SPL verification better matches the 80-a SPL target.

You indicated a delay of -1.37 ms. I don't understand how that relates to all this this data. The verifications of 80-a, 80-b and 80-c all have about the same 5.5±1 ms delay that was targeted. The 80-a verification delay is slightly more than the 80-a target delay. The 80-c verification delay is slightly less than the 80-a target delay. It's very possible that the 80-a alignment SPL is more sensitive to small mic positional changes than the 80-b setup and that is why there was no SPL problem with 80-b.

 

[Niklasmagnus]

You indicated a delay of -1.37 ms. I don't understand how that relates to all this this data.

I used the "Align phase at cursor" ad the new line match you blue quite good. Maybe this is all wrong, but I was playing around at tried this feature.

It's very possible that the 80-a alignment SPL is more sensitive to small mic positional changes than the 80-b setup and that is why there was no SPL problem with 80-b.

Anyway, I´m quite sure the 80-a are vary sensitive to mic location since its no more then +-1cm from when I took the original measurements. AND the verification for all other are in same mic location as for 80-a. So maybe I should go with your suggestion for setting 80-a instead of my 80-c

 

[jtalden]

Okay, I understand now. You are using the alignment tool properly. From your screen shot, you took the L the SWs (13 and 15) measurements to find a new alignment. The SWs were set at 5.5 ms delay and your adjustment was -1.37 so an overall delay of 4.13 instead of 5.5 we initially targeted. That gave a good alignment for the L main. I would have suggested that you vector average L and R (L+R)/2) as we did for the targeting. The delay adjustment may then have been a little closer to the original 5.5 ms unless there was another small change that caused the shift. Whatever the cause, it is within the range we were targeting. With some setups and some alignments there indeed can be high sensitivity to exact test conditions if you want the results to overlap all is best done in one session. The minor SPL deviation you found does not necessary indicate that the alignment is not favorable.
Don't feel obligated to test 2 or more the favorable alignments I recommended. I was just asking if you do test 2 or more we are interested in your feedback as to how they compare. I also forgot to reinforce your comment about applying some EQ to smooth the bass range first. I wouldn't bother to do listening tests without doing that.

 

[Niklasmagnus]

When adding EQ to the subs, should there be an increase in group delay? And what is acceptable GD below 100Hz?

 

[jtalden]

When adding EQ to the subs, should there be an increase in group delay?

There is no expected detriment to group delay when applying PEQ. There may be improvements.

And what is acceptable GD below 100Hz?

Web browse for 'audibility of group delay'. You will find some studies and a raft of opinions. Test results vary widely depending on the conditions chosen. Most of the arguments don't deal with all the interactions that occur in a SW setup situation.

My Thoughts
Group delay for SW setup is often difficult to measure accurately at the LP due to the room effects.
There is no direct relationship between the delay setting and the SPL support or the sound quality.
Attention to group delay is a misguided focus for a SW setup.

So:
The theoretical target group delay for a SW XO is the one that is associated with the delay that provides the closest phase tracking of the direct sound through the XO range. However, room effects often disrupt the direct sound at the LP adversely impacting the SPL support. Since SPL is the more influential factor it is often better to depart from the theoretical timing as needed to find a timing that provides more favorable SPL support. In those cases I usually suggest that the delay timing is not changed more than -1 to +0.5 XO wavelengths. It is probably safe to extend that to -1.5 to +1 wavelengths if significantly better SPL support is found.

Favorable SPL support is normally found at roughly 1/2 XO wavelength intervals of delay timing. These do sometimes result in a noticeable difference in sound quality. Since the SPL is similar in these cases I suspect the difference in sound is due the differing interaction of the later arriving XO range bass waves as they travel around the room.

 

[Niklasmagnus]

The GD a XO is quite good, I think. But at around 35 hz the GD gets increased due to EQ. is this normal?

Yellow: SUbs and L with EQ
Blue: SUbs and L no EQ
Purple: Only subs no EQ
Red: Main L no EQ


EQ:

 

[jtalden]

I cant' analyze those charts. Please provide a direct comparison of 2 measurements.
Post an .mdat with using one of the channels containing, with and without EQ applied, I will then take a look. The EQ gain levels applied are much larger than I have ever applied, but even so I suspect the apparent difference is very misleading.

 

[jtalden]

Below is a way to look at the impact of the EQ alone. I tried to keep similar scaling for the charts to help compare to your measurements. Note that the greatest impact to GD of the EQ alone is about ±10 ms at different frequencies than your chart indicates. Plotted GD of measurements at the LP is pretty meaningless in nonminimum-phase situations.

 

[jtalden]

Try again -
If your using a PEQ device that provides a proportional Q filter type instead of a constant Q filter type then the results are as posted below. The MiniDSP uses a proportional Q type filter. This type of PEQ filter has a bigger impact on SPL and GD. It is still not directly proportional to the LP measurements however.

 

[Niklasmagnus]

Hi, sorry for late reply. I had problem to find time from family to take the measurements. Attached are mdat with and without EQ. I included both output L and R since there are some differences. Why, I have no idea of?? I have connected my external soundcard to the CD rcas and running stereo mode.
Also included the txt file for minidsp2x4.
BR Niklas

 

[jtalden]

I summed the L and R channels; with and without EQ.

• The first chart is the GD without FDW applied so we are looking at the apparent GD using the entire window for all frequencies. This allows the room to influence the result for the entire 500 ms window. These late arriving room reflections and modes cause comb filtering of sound and thus chaotic phase changes. Since GD is the slope of the phase at any given frequency it also appear chaotic. This chart is not representative of the direct sound.
• The second GD chart has a 3 cycle FDW applied. This trace includes the first 3 cycles at each frequency so the later arriving room effects are filtered out. Note that there are still some strong room effects, but we can now easily see the trend of the direct sound. The large spikes in the GD trace indicate where room effects create nonminimum-phase conditions. EQ in these areas is very inefficient and not normally recommended. I find some EQ there to help the surrounding frequencies helps so long as it is done sparingly. Direct sound nulls cannot be removed, but some late arriving sound at those frequencies can be helpful to smooth the SPL. The spectrogram can show this better.
• The third GD chart has a 1 cycle FDW applied. In this case it shows the GD of the first cycle of each frequency. Because this is a very tight window it can sometime under estimate the phase rotation and thus GD a little. It is probably better to use a 3 or 4 cycle FDW and ignore any frequencies where there are spikes.

The conclusion then is that the GD of the direct sound was not significantly impacted by the addition of the EQ. The EQ boost filter at about 32 Hz removed the large dip with later arriving sound that was reflected around the room. We can point out that this is GD as it is late arriving sound so the first chart is indeed correct in that sense. Our choice is little or no SPL in those areas or more reasonable SPL albeit late arriving. We are much more sensitive to SPL irregularity than we are to group delay, but too much boost or too narrow filters can also cause sound quality problems also. I suggest moderation with EQ, but it is an individual preference.

 

[Niklasmagnus]

This was very interesting but a bit hard to understand. So I must add a few follow up question just so I know I understand it correctly.

I summed the L and R channels; with and without EQ.

• The first chart is the GD without FDW applied so we are looking at the apparent GD using the entire window for all frequencies. This allows the room to influence the result for the entire 500 ms window. These late arriving room reflections and modes cause comb filtering of sound and thus chaotic phase changes. Since GD is the slope of the phase at any given frequency it also appear chaotic. This chart is not representative of the direct sound.

So what we measure here is the combined sound during all 500 ms, okey. Is our hearing also working in same way, that we sum the direct and up to 500 ms late sound? I´m fooling myself with a nice looking graph or what?

• The second GD chart has a 3 cycle FDW applied. This trace includes the first 3 cycles at each frequency so the later arriving room effects are filtered out. Note that there are still some strong room effects, but we can now easily see the trend of the direct sound. The large spikes in the GD trace indicate where room effects create nonminimum-phase conditions. EQ in these areas is very inefficient and not normally recommended. I find some EQ there to help the surrounding frequencies helps so long as it is done sparingly. Direct sound nulls cannot be removed, but some late arriving sound at those frequencies can be helpful to smooth the SPL. The spectrogram can show this better.

The spikes you are referring to; is it both positive and negative? 32hz neg, 100hz and ~200hz pos.
Is this meaning that its unwise to boost around ~32 hz and reduce around 100 hz?
Do I have a null at 32 hz that cannot be boost, but instead I have boosted the late arriving sound to fill up the gap measuring-wise? i.e. I get fooled by the measurement since its not representing our hearing? In spectrogram we see the late energy peak with EQ but also the longer decay time.

Spectrogram:
With EQ

Without EQ:

• The third GD chart has a 1 cycle FDW applied. In this case it shows the GD of the first cycle of each frequency. Because this is a very tight window it can sometime under estimate the phase rotation and thus GD a little. It is probably better to use a 3 or 4 cycle FDW and ignore any frequencies where there are spikes.

The conclusion then is that the GD of the direct sound was not significantly impacted by the addition of the EQ. The EQ boost filter at about 32 Hz removed the large dip with later arriving sound that was reflected around the room. We can point out that this is GD as it is late arriving sound so the first chart is indeed correct in that sense. Our choice is little or no SPL in those areas or more reasonable SPL albeit late arriving. We are much more sensitive to SPL irregularity than we are to group delay, but too much boost or too narrow filters can also cause sound quality problems also. I suggest moderation with EQ, but it is an individual preference.

Would it be wise to increase the subs overall volume and then negative eq all above the target for minor boost in the 32 hz region? or what is the correct action in your opinion to handle the dip at 32-33 hz and spike at 43 hz? also spike at ~100 hz?

 

[jtalden]

Yes, it is confusing to draw any firm position on what EQ settings are best technically given our hearing mechanisms. There is even less confidence in doing that for a particular setup and listener preference. I'm no expert on the hearing mechanism, have little info on your setup, and don't know your listening habits and preferences, but I shared my overall thoughts/conclusions in the last paragraph above.

More to some of your additional questions:
The spikes are indicative of a strong room influence. The positive or negative direction is not an issue. It is more an artifact of the chosen chart settings (I think). Cutting peaks in the direct sound is pretty effective. A strong deep null in the direct sound cannot be removed with EQ. There will always be the null in the direct sound, but placing more energy there with EQ may provide enough late arriving, out of phase, energy such that the measured (and heard) SPL at the LP is present. The spectrograms you posted show this well. The question is whether the that late arriving SPL and extended decay tail is helpful to your overall sound quality or more detrimental. I cautioned against going too aggressively trying to fill-in that SPL completely. Only your own experience can find the best answer for you.

I do favor to cut more than boost with EQ. All these EQ choices however have tradeoffs. All you can do is try to understand some of them and experiment with a couple EQ schemes to see what you prefer. Program material also is a factor. There may not be much music that will be impaired by a SPL dip at 32 Hz and possibly most movie sound effects will be more impactful and will not be degraded significantly by the longer decay tail there? Higher frequency issues are likely to be more sensitive to aggressive EQ. SPL is the most important characteristic, but it is not the only factor that defines our enjoyment.