Combining measurement methods to get a more accurate EQ - MMM + Gated Nearfield + Klippel Measurements

[johnp98]

... You have made many references to the Schroeder crossover frequency, but what you actually mean is the transition frequency (or transition region). The Schroeder crossover frequency doesn’t apply to small rooms like home hifi, and if you try to calculate it from measurements in a small room, you get low values that are too low to describe what you actually wish to convey, namely, the frequency (region) below which room modes dominate and above which reflected sounds dominate. This is referred to as the transition frequency or transition region.

Thanks for the clarification! I thought I heard them being used interchangeably so did not know that nuance and difference. Will keep in mind for future posts.

 

[johnp98]

Looking at your step responses in post 23 we can see a peak at around 7 to 7.5 ms that is greater in amplitude than the direct sound. This is usually due to non-minimum phase behaviour as described in John Mulcahy's explanation of minimum phase. See section, "Minimum Phase and Ivertibility." You may want to check your step response out to 100ms to see if there are any other peaks greater than the direct sound...

It does look like there are numerous peaks occurring later in time that are larger amplitude than the initial signal. I wonder if this is primarily the subwoofers?

@ernestcarl was kind enough to spontaneously make and send a quick and dirty FIR filter in rePhase and so I loaded that up with Equalizer APO and here are the results.

Left HTM Combo EQ, No phase correction (narrow window to 10ms)

(wider window to 100ms)

Left HTM Combo EQ, Post rePhase (narrow window to 10ms)

(wider window to ~100ms)

Left HTM Combo EQ, No phase correction

Post rePhase:

Listening tests did seem improved with more 'punchiness' and 'tight' sounding bass for sure.
Das Spiegel by Chemical Brothers seemed particular good at revealing increased attack for the bass.

So thank you @ernestcarl for spontaneously making those filters and sending them to me!

Here is the link to my raw measurements pre and post if that is of interest.

Defiantly has me interested in actually pursuing FIR / phase correction.

 

[johnp98]

You might want to give http://drc-fir.sourceforge.net/ a try. User gmad from diyAudio has written some useful scripts including a default starting place to try to make using it simpler. https://www.diyaudio.com/forums/ful...ectrical-loudspeaker-correction-networks.html There is also a GUI that goes with it if you want to use that instead: https://www.diyaudio.com/forums/full-range/242171-towers-25-driver-range-line-array-73.html

There are good explanations in the thread including how it is different than rePhase. The big deal is the addition of psychoacoustic filtering, adjustable frequency dependent window (FDW) and excess phase correction. Denis DRC documentation also does a great job of explaining this and showing what is going on. If you have the time it is worth a read as excess phase correction at low frequencies is where it is at. This allows you to clean up the step response overall but especially the low end.

Sweet! Thanks for all the information (and information that you have contributed on numerous forms elsewhere).
I am quite busy with work this week but certainly will want to do a lot more reading and tinkering around and try and learn Denis DRC.

I am wondering if there is anything specific that I will have to consider or measure or keep in mind with my setup when doing DRC.
My current setup is: PC --> Receiver --> L&R output to HTM 12s (bass management has them crossed over at 100hz), Subwoofer output (mono) to Behringer amp which has its own EQ filters and own phase and delays on each of my two 18" subwoofers (one positioned front and center, the other positioned directly behind the MLP in a semi nearfield location).

I have tried to time/phase align the speakers and subs but I don't know if the complexities of having the AVR set the crossover from speaker to subs, and then the subs having their own delays and phase will make DRC more complicated? I guess it is measuring the output / phase of the whole system at its current state and then correcting it, so maybe it does not complicate it, just for some reason I am wondering if I should phase correct each subwoofer independently and pair one sub with the left speaker and the other with the right.

Anyways, just wanting to make sure I have optimized all the upstream factors as I get the sense that DRC is the final icing on the cake and certainly would need to be re-done if one started messing with the delay and phase and position of speakers and subs.

Good luck on your journey!

I just got into learning about sound reproduction at the start of covid and it is wild to think of all that I have learned.
1 year ago I was rocking some value village speakers set to "large" and subs improperly positioned. Now I have read numerous books, watched countless videos, assembled two speakers and two subs, got a UMIK-1, done tons of measurements and optimization and even remained married throughout the process!

But most importantly I am enjoying my system and listening to music more than ever!
I have even helped setup systems for my siblings and made numerous people jaw drop as they listen to an above average system (and no I have no desire to go toe to toe with any $100,000 system, just want to have mine hit way above the money invested).

Ok, rant over

 

[Kvalsvoll]

It does look like there are numerous peaks occurring later in time that are larger amplitude than the initial signal. I wonder if this is primarily the subwoofers?

@ernestcarl was kind enough to spontaneously make and send a quick and dirty FIR filter in rePhase and so I loaded that up with Equalizer APO and here are the results.

Left HTM Combo EQ, No phase correction (narrow window to 10ms)
View attachment 138129
(wider window to 100ms)
View attachment 138131


Left HTM Combo EQ, Post rePhase (narrow window to 10ms)
View attachment 138130

(wider window to ~100ms)
View attachment 138132


Left HTM Combo EQ, No phase correction
View attachment 138133
View attachment 138136

Post rePhase:
View attachment 138134
View attachment 138135

Listening tests did seem improved with more 'punchiness' and 'tight' sounding bass for sure.
Das Spiegel by Chemical Brothers seemed particular good at revealing increased attack for the bass.

So thank you @ernestcarl for spontaneously making those filters and sending them to me!

Here is the link to my raw measurements pre and post if that is of interest.

Defiantly has me interested in actually pursuing FIR / phase correction.

Excellent results, especially if it sound better as well. The changes you have done - using eq on the speakers and then additional FIR room correction - are measurable, as you have confirmed, and both the speaker eq and this last FIR correction shows a measured response closer to the ideal situation.

Since your speakers are born with quite excellent off-axis response, they are likely to respond well to eq. But does a more smooth and flat frequency response sound better? In my experience, it does. And we can find support in science for this. What I have observed, it that speakers where the response above say 1k is difficult to place within around 5dB range, sound quite colored - there is a distinct sound character added by the speakers, to all recordings and instruments. Getting the response flat gives a more neutral and natural sound.

 

[HoberM]

Hi Johnp98 !

I use a really similar method with very good results too!

I use the "Jeff B." quasi anechoid measurements method, to get quasi-full polar reponse, and apply EQ as you do. Separing above and below schroeder FR. Targeting peaks (visible on multiples measurements around listenning position) for bass, and smooth EQ on hights to handle some RT and FR related stuff...

So. I'm do you mind to share which phase correction did you apply ? the wavelet capture of phase corrected measurement is impressive with a nice room/speaker coupling. I'm jaleous !

 

[johnp98]

I use a really similar method with very good results too!

I use the "Jeff B." quasi anechoid measurements method, to get quasi-full polar reponse, and apply EQ as you do. Separing above and below schroeder FR. Targeting peaks (visible on multiples measurements around listenning position) for bass, and smooth EQ on hights to handle some RT and FR related stuff...

Nice!

So. I'm do you mind to share which phase correction did you apply ? the wavelet capture of phase corrected measurement is impressive with a nice room/speaker coupling. I'm jaleous !

Attached is what @ernestcarl so kindly generated and sent to me.

With the phase correction it now has me interested in more advanced digital room correction so I am looking into DRC-FIR.
Phase correction has always seemed intimidating for some reason, but I think I should be able to manage it

Anyways, when work settles down I will probably dig into DRC-FIR and try it out and then post the results.

 

[ernestcarl]

With the phase correction it now has me interested in more advanced digital room correction so I am looking into DRC-FIR.
Phase correction has always seemed intimidating for some reason, but I think I should be able to manage it

Anyways, when work settles down I will probably dig into DRC-FIR and try it out and then post the results.

There are lots of people out there who know more about DRC-FIR software/program than me -- unfortunately, I'm definitely not one of these guys!

I am a little more familiar with rePhase and REW.

Ideally, one corrects and optimizes each driver to get a perfect linear-phase xo IR summation from the very beginning.

But, to be honest, even a crude type of after-the-fact / post correction at the main listening position may be adequate enough for one's own needs.

----


The method I used is really kind of simple and 'basic' once you get the hang of it:

1. Take speaker measurements at the listening position (preferably multiple points, time-corrected, and vector averaged).

2. Apply FDW (around 3-10 cycles).

3. Extract the excess phase version of the vector averaged IR and export as text.

4. Import excess phase text file in rePhase.

5. Use the 'Filters Linearization' tab and apply filters -- doesn't have to to be the exact crossover of your drivers as long as it flattens the excess phase curve the most.

6. Use the 'Paragraphic Phase EQ' tab' and use sliders to flatten the phase curve.

7. Generate your initial impulse correction, and then re-import or drag this wave file back into REW.

8. Convolve the original (none phase corrected response) with the impulse correction wave file by performing the simple 'Trace arithmetic' function 'A times B'.

9. Apply some dB offset to get the magnitude level exactly the same as the original. Let REW generate/calculate minimum phase and excess phase curves and individually check them -- including the group delay curves, step response, ETC, and spectrograms.

10. Iteratively adjust excess phase correction curve in rePhase as desired -- again, re-import generated file into REW and convolve this with the original IR.


THE VERY LAST STEP IS TO CONFIRM THE IMPULSE CORRECTION WITH ACTUAL MEASUREMENTS AT THE LISTENING POSITION; AND, YEAH, AS WELL AS DOING SOME LISTENING TESTS!


---


There's a massive thread at diyAudio on rePhase:
rePhase, a loudspeaker phase linearization, EQ and FIR filtering tool


*Some speaker and room combos are 'easier' to 'correct' this way than others, and so it just happens that @johnp98's measurements turned out quite well here.

 

[johnp98]

The method I used is really kind of simple and 'basic' once you get the hang of it:
----

Yeah I just might have to give rePhase a try! As stated earlier I don't know why I am intimidated by rePhase and DRC but I am, so I just got to commit to learning them. I think it is the lack of concise guides and thus thanks for the outline of what you do, that is certainly helpful!

*Some speaker and room combos are 'easier' to 'correct' this way than others, and so it just happens that @johnp98's measurements turned out quite well here.

Hummm.... I wonder.... If I learn DRC and you know rePhase, maybe it would be interesting to compare the two results (DRC vs rePhase).

Would your convolution / adjustments be any better if I took more measurements "preferably multiple points, time-corrected, and vector averaged" or took them pre vs post my own EQ? I certainly don't want to sign you up for more work, as the files you already gave me are getting daily use, but I think I might give DRC a try this next week and then compare to the rePhase results. I could even activate a Dirac trial and then toss it into the mix as well. Could be something interesting to post (might even deserve a new thread of its own). Thoughts?

 

[HoberM]

Thanks you for the methodoloy.
It's look like what I've read from pd0 et Bear. I'll try that soon I hope

Did you checked the pre-ringing ?

 

[ernestcarl]

I think I might give DRC a try this next week and then compare to the rePhase results. I could even activate a Dirac trial and then toss it into the mix as well. Could be something interesting to post (might even deserve a new thread of its own). Thoughts?

I could go take a look and see with you what differences might be between the more automatic correction and the more 'manual' method I described. Though, I'm not sure I would necessarily be able to tell which one is actually better. It may be worth a try just to see the actual differences.

Did you checked the pre-ringing ?

I suppose a tiny bit is indicated in the graphs, but very low in level so nothing to be concerned about:

John's:

*actual measurement (not predicted)

It's just the bass driver starting a little earlier.

*actual measurement (not predicted)

And I don't really think there is anything too radical about the phase correction wave file:

 

[ernestcarl]

Would your convolution / adjustments be any better if I took more measurements "preferably multiple points, time-corrected, and vector averaged" or took them pre vs post my own EQ?

Perhaps not much difference in the bass -- but if we are doing a full-range correction up to the higher frequencies it also helps screen out anomalies by comparing a more wider set of measurements rather than over-relying from a single-point information.