Equalizing loudspeakers based on anechoic measurements (community project)

[QMuse]

At any of those frequencies moving my head a bit to side to side or back and forward causes audible SPL shifts in both channels, all local interference effects. In general - I'm not sure how valid this kind of test is. In any case, the higher the frequency the more annoying it seems to get

Yep, not easy to do, but it can still be done.

2kHz is -3dB according to ER/MMM and is perfectly in between with ON/LW. Such difference should be audible so you'll be able to detect if you're hearing ER/MMM or ON/LW at your listening position.

 

[TimVG]

Yep, not easy to do, but it can still be done.

2kHz is -3dB according to ER/MMM and is perfectly in between with ON/LW. Such difference should be audible so you'll be able to detect if you're hearing ER/MMM or ON/LW at your listening position.

In that case I'm not detecting the dip as presented in the MMM. I've tested with sine waves and tone bursts (100 cycles seems to be the limit). Again I'm not sure this is very valid due to the variations when moving side/side - back/forth and also because I'm 100% aware of what is going on.

 

[QMuse]

In that case I'm not detecting the dip as presented in the MMM. I've tested with sine waves and tone bursts (100 cycles seems to be the limit). Again I'm not sure this is very valid due to the variations when moving side/side - back/forth and also because I'm 100% aware of what is going on.

You should use low SPL so that you can barely hear 2kHz tone. That way you can more effectively compare it to 1kHz and 3khz tones. It would also be good that somebody switches the tones for you so you can sit as still as possible.

 

[TimVG]

You should use low SPL so that you can barely hear 2kHz tone. That way you can more effectively compare it to 1kHz and 3khz tones. It would also be good that somebody switches the tones for you so you can sit as still as possible.

That'll have to wait, home alone! In the meanwhile I've been trying to actually find under what circumstances the dip becomes audible. I've concluded it is the result of vertical interference between the tweeter and midrange, and it simply doesn't happen in the horizontal plane on the listening-axis. If I move closer to either of the speakers it becomes quite clear under certain angles I'm guessing at 30-40° above and below the intended axis.

 

[QMuse]

In the meanwhile I've been trying to actually find under what circumstances the dip becomes audible. I've concluded it is the result of vertical interference between the tweeter and midrange, and it simply doesn't happen in the horizontal plane on the listening-axis. If I move closer to either of the speakers it becomes quite clear under certain angles I'm guessing at 30-40° above and below the intended axis.

Probably so, but the main thing that is of interest here is if you can hear that dip from the same location where MMM was taken (presumably your LP).

That'll have to wait, home alone!

Oh, I see. Stay calm and watch for these guys!

 

[TimVG]

Probably so, but the main thing that is of interest here is if you can hear that dip from the same location where MMM was taken (presumably your LP).

I'd have to say no. Listening on the intended axis and moving slighter further back and/or side to side does not change the relative SPL of 1, 2 and 3kHz as much* as when listening on a severe cancellation point (I estimate it's most off around 30° up and down).

*there is still always some local interference stuff going on which is alleviated moving sometimes no more than 10cm in a certain direction.

 

[QMuse]

I'd have to say no. Listening on the intended axis and moving slighter further back and/or side to side does not change the relative SPL of 1, 2 and 3kHz as much* as when listening on a severe cancellation point (I estimate it's most off around 30° up and down).

I was referring to my discussion with @edechamps if flat ON/LW matters more than smooth in-room response/PIR. If you can't hear that dip than it wouldn't make sense to EQ it, but if you can than it should be corrected as in that case smooth in-room response does matter.
That dilemma matters from your LP and you probably wouldn't be listening F206 towers from 30° up.

*there is still always some local interference stuff going on which is alleviated moving sometimes no more than 10cm in a certain direction.

You can test that with making 2 sweeps that are 10cm apart. It would be interesting to see what happens to the dip with those 2 measurements.

 

[TimVG]

I was referring to my discussion with @edechamps if flat ON/LW matters more than smooth in-room response/PIR. If you can't hear that dip than it wouldn't make sense to EQ it, but if you can than it should be corrected as in that case smooth in-room response does matter.
That dilemma matters from your LP and you probably wouldn't be listening F206 towers from 30° up.



You can test that with making 2 sweeps that are 10cm apart. It would be interesting to see what happens to the dip with those 2 measurements.

Let's also consider that these are steady-state measurements and that actual program material (and real life sounds) is (are) transient in nature - It's a complex sound field and these steady-state measurements aren't an actual representation of what we hear - Not just the dip in the ER/SP, but also the decline of high frequencies (or should we say build up of low frequencies?) Anyhow, here's 4 measurements of the left speaker around where my head would approximately be.

 

[jlo]

Maybe it would be interesting if you can generate 3 single tone files with 1kHz, 2kHz and 3kHz and play them to see if you will hear 2kHz quiter than other 2 or not.

This won't work : it is too dependant to head position and you will only perceive kind of comb filtering. You need pink noise pass-band filtered with 1/3oct width centered to 1, 2 or 3kHz.

 

[TimVG]

This won't work : it is too dependant to head position and you will only perceive kind of comb filtering. You need pink noise pass-band filtered with 1/3oct width centered to 1, 2 or 3kHz.

I will try this tomorrow!

 

[jlo]

I will try this tomorrow!

It will work better but I don't think it is a good method to detect if the dip has to be EQ'd or not.

 

[TimVG]

It will work better but I don't think it is a good method to detect if the dip has to be EQ'd or not.

I don't believe it needs to be.. according to the anechoic data the direct sound is more or less flat. It is Qmuse who wonders if it is the direct sound I hear or the sum of direct and reflected. Since real sounds are transient in nature and complex as opposed to sine waves I'm not sure this is a good method to evaluate this. What is your take on the matter jlo?

 

[Pio2001]

I think that in order to investigate the ON/LW vs ER/SP question, we should experiment with an extreme case.
What if we put two identical speakers side to side and play the same signal in both ? Wouldn't we get strong comb filtering in the sound power while having a correct on-axis response ?

We could listen to how it sounds. Then, it would be interesting to look for musical content that is sensitive to ON/LW (drums ?) and musical content sensitive to ER/SP (organ, choir ?).

If it works, we could then test the hypothesis that some music would sound better if the speaker is equalized according to the LW curve, and some other style of music would sound better if the speaker is equalized according to the PIR curve.

 

[Chromatischism]

Wouldn't we get strong comb filtering in the sound power while having a correct on-axis response ?

Why would the on-axis sound not be affected?

 

[Pio2001]

It would not be affected as long as we sit exactly in the symmetry plane that separates the two speakers.

But I see the problem now : even if our head is perfectly positionned, each of our ears are a few cm away from this plane.
We should evaluate the frequency range affected. The idea might not work after all.

 

[QMuse]

I don't believe it needs to be.. according to the anechoic data the direct sound is more or less flat. It is Qmuse who wonders if it is the direct sound I hear or the sum of direct and reflected. Since real sounds are transient in nature and complex as opposed to sine waves I'm not sure this is a good method to evaluate this. What is your take on the matter jlo?

Passband limited pink noise is a complex sound. If you make short bursts out of it you can even make it transient, but I don't really see the need for that.

@jlo, My logic here is quite simple: if there would be a difference in level between those 3 tone samples (2khz is lower) than the dip at 2kHz should be EQ-ed. Otherwise it need not to be.

Btw, I have no idea what would be the outcome of such listening test but I do consider it extremely interesting and valuable for the topic discussed here as it may help resolve the dilemma whether to EQ to flat ON/LW or to smooth PIR.

 

[Pio2001]

even if our head is perfectly positionned, each of our ears are a few cm away from this plane.
We should evaluate the frequency range affected.

Done.
If I put two KH-120 speakers side to side, their acoustic axis will be 18 cm apart.
My ears are roughly 16 cm apart from each other.
If I stand 2 meters in front of the pair of speakers, the difference between the two speakers, seen from each ear, will be about 1,3 cm.
That is about 0.04 ms.
Here is the spectrum of a white noise replayed with a delay of 0.045 ms with itself (2 samples at 44100 Hz) :

The frequency response is flat within 1 dB up to 4 kHz. After that, it decreases rapidly and reaches -5 dB around 6 kHz.

 

[Chromatischism]

I think that in order to investigate the ON/LW vs ER/SP question, we should experiment with an extreme case.
What if we put two identical speakers side to side and play the same signal in both ? Wouldn't we get strong comb filtering in the sound power while having a correct on-axis response ?

We could listen to how it sounds. Then, it would be interesting to look for musical content that is sensitive to ON/LW (drums ?) and musical content sensitive to ER/SP (organ, choir ?).

If it works, we could then test the hypothesis that some music would sound better if the speaker is equalized according to the LW curve, and some other style of music would sound better if the speaker is equalized according to the PIR curve.

Couldn't you just use a speaker that measures great on-axis and poorly off-axis and play with toe angles in a small room?

 

[Lbstyling]

Tim, it came from a paper by Tapio Lokki, which I mentioned here: https://www.audiosciencereview.com/...-audiophile-journey.14547/page-22#post-467125. As @Duke pointed out, the tonality didn't come up as a preference class because of the sources in the study, but it did in the underlying class of attributes.

Broadly speaking, I think it's pretty clear that Toole prioritizes spaciousness, which I believe likely favors wide-dispersion speakers to facilitate lateral reflections, as well as listening setups that reduce IACC. I speculate that prioritizing clarity/definition may favor narrower-dispersion speakers and listening setups that reduce early reflections (e.g. those who bring up the ISD gap).

These aren't necessarily mutually exclusive, but probably a matter of priorities, and there will always be tradeoffs.

Anyway, just armchair theorizing...

This is my experience from running a horn with strongly narrowing (but very smooth) directivity.

Far be it for me suggest that Toole is wrong (its much more like I am a statistical outlier or something else is at play, BUT running a large JMLC horn from 20khz to 500hz sounds sublime.

Even when you are in a small, narrow room with lots of hard surfaces, it sounds like your in a studio. You subjectively get space between notes, clarity and detail. Almost too much. It sounds comparitively dry. Orchestral music sounds big, but pop songs often sound very 'simple', like they need to add more instruments.

Just some thoughts.

 

[Pio2001]

Couldn't you just use a speaker that measures great on-axis and poorly off-axis and play with toe angles in a small room?

I'm not sure that I've got that at home.
The KH-120 exhibits such a difference but it is not very big. The same order of magnitude of the typical error when a target curve is chosen for equalization.
The 305P have a very good off-axis performance. So it is of no use for this experiment.
I also have a pair of Dynaudio Gemini and Boston A26, but I don't know how they measure, and sending parcels from France to US and back is a bit complicated.