That gets back to the larger question.
Assuming a speaker has a flat 1m on axis response, Is it better to only EQ below the Schroeder frequency, or is there value in EQ above that? There are strong proponents on both sides.
I have had much better results using EQ over the entire range. I found it very difficult to settle on the target curve that worked best for my situation though.
If I knew my speaker was truly flat on an anechoic basis, then I would stick with EQ-ing the lower frequency response effects. Not necessarily just through the Schroeder frequencies, but up through the midrange floor bounce. Above that the typical room curve tends to be what you get from speaker directivity and room absorption, i.e. if you adjust most speakers to a room curve you will be flattening the axial response in a roundabout way.
David S.
That's good to hear as it's essentially where I ended up after lots of experimentation with various EQ schemes.
Hi gentlemen!
I'm testing MMM with RTA REW.
My microphone is a ECM8000 of crossspectrum with calibration file. The measurements are made at 90 °.
My speakers are Yamaha HS7 and my listening point is just 0.6meters.
See photo.
When I IR measurements in listening point I do not fall in the upper frequency. But when I use Pink Noise PN I have a brutal fall.
We can see in the following screen.
This is the setting of the parameters of RTA in REW.
Filters calculations with DRC-FIR Denis Sobreing, but for this I need a measure IMPULSE with phase information.
RTA I have only magnitude and I can not export as impulse, to be treated in DRC-FIR.
Excuse my bad English. Thank you all for this wonderful post.
I'm testing MMM with RTA REW.
My microphone is a ECM8000 of crossspectrum with calibration file. The measurements are made at 90 °.
My speakers are Yamaha HS7 and my listening point is just 0.6meters.
See photo.

When I IR measurements in listening point I do not fall in the upper frequency. But when I use Pink Noise PN I have a brutal fall.
We can see in the following screen.
This is the setting of the parameters of RTA in REW.


Filters calculations with DRC-FIR Denis Sobreing, but for this I need a measure IMPULSE with phase information.
RTA I have only magnitude and I can not export as impulse, to be treated in DRC-FIR.
Excuse my bad English. Thank you all for this wonderful post.
> Measure 1 speaker at a time.
> "PinkPN" Signal is correct
> "RTA 1/48 Octave" Mode
[more smoothing can be applied if desired after the response is saved.]
> "Forever" Averages
> Use the same FFT length for both the RTA and the signal generator.
There may remain a slight additional roll-off of the HF compared to the sweep, but it should be very small.
> "PinkPN" Signal is correct
> "RTA 1/48 Octave" Mode
[more smoothing can be applied if desired after the response is saved.]
> "Forever" Averages
> Use the same FFT length for both the RTA and the signal generator.
There may remain a slight additional roll-off of the HF compared to the sweep, but it should be very small.
> Measure 1 speaker at a time.
Its oK
The measurement was performed with a single speaker.
> "PinkPN" Signal is correct
Its ok
> "RTA 1/48 Octave" Mode
OK I change spectrum to 1/48 RTA
[more smoothing can be applied if desired after the response is saved.]
> "Forever" Averages
Its Ok
> Use the same FFT length for both the RTA and the signal generator.
Its the same 32768
There may remain a slight additional roll-off of the HF compared to the sweep, but it should be very small.
We'll try again. Thank you for the suggestions.
You could also select "Adjust RTA Levels" to match the same reported level as the sweep measurement to make it easier to compare the 2 measurement options.
Hello again Jtalden, it's dark now and I bedded children. But I have been measured in 1/48 RTA loop mode, and have flat response. Thank you!
Now I have export txt file response to PCM 32bits floart to pass by DRC-FIR. I guess I have to generate minimum phase to do so.
Now I have export txt file response to PCM 32bits floart to pass by DRC-FIR. I guess I have to generate minimum phase to do so.
When I IR measurements in listening point I do not fall in the upper frequency. But when I use Pink Noise PN I have a brutal fall.
That is to be expected. The steady-state response should display a rolled-off treble (and a rising low-end) compared to the a flat direct response (short window). See for example here:
http://www.aes.org/tmpFiles/elib/20151002/17839.pdf
For the link to work, you need to amend the date portion of it to match today's date.
This is the result of MMM averages around 160 0.3meters around the listening point.
Compared to a single Sweep boost the listening point.
In my case I think are very similar and the resulting EQ, It could be indistinguishable.
Compared to a single Sweep boost the listening point.
In my case I think are very similar and the resulting EQ, It could be indistinguishable.

Yes, very close. I am surprised that the MMM didn't smooth out the bass more. Perhaps because your listening point is very near the speakers?
@pano yes I would say the proximity is an issue. When I did my MMM the up and down and left to right would have been more than isabido's listening distance 🙂 I suspect over a much smaller area there is much less spatial averaging occurring.
edit: I wonder whether stepping back to a couple of meters and performing an MMM in a much larger arc would be usable for the eqing at the listening position... Might be worth experimenting with.
Tony.
edit: I wonder whether stepping back to a couple of meters and performing an MMM in a much larger arc would be usable for the eqing at the listening position... Might be worth experimenting with.
Tony.
Last edited:
60cm is super close. Interesting to see how similar the two measurements are.
Wonder what you would get with 2 measurements - 1 at each ear placement? 🙂
Wonder what you would get with 2 measurements - 1 at each ear placement? 🙂
^^I think that this gives best possible view of room response, recommended!
i have to agree because sweeping all over my listening area after some time you see the response and it does not change.ie you can sweep for an hour and the average will stay the same. that response fully encompasses the space.
then, when i eq with rephase, the predicted response is EXACTLY what will be measured when i do the sweep again.
before i used to average 9-12 sweeps to do my correction but i like the method much better as it is much quicker.
A curious possibility crossed my mind, that with MMM the effect of the observer is naturally included in the results, and with greater similarity between measurements. How far away from the mic do most people stand when measuring using a single point measurement? Clearly the human body can have tremendous absorptive and reflective characteristics - would this not naturally be one of the hidden benefits of MMM, that the effect of the observer is more clearly (and naturally) represented?
Brainditch
Brainditch
^There will always be some variance and dispersion of measured response, no matter how we do them. Prehaps a person should sit at listening position when doing measurements. I do MMM sweeps (or RTA) while sitting at spot, holding the mic in my straight arm. Averaging will do the job.
I don't think it it is wise to try to corrrect high Q (sharp) deviations. Use 1/1 - 1/3 octave smoothing for measurements and for eq! Only room modes would benefit from higher Q corrections, but then only 1/3 of excess dB! I personally don't like any room mode EQ, speaker and listening positioning are far more better ways to cope with them. I have minidsp 2x4HD and 4x10HD in my two systems. I don't use rephase or dirac, only parametric EQ/XO and delay.
I don't think it it is wise to try to corrrect high Q (sharp) deviations. Use 1/1 - 1/3 octave smoothing for measurements and for eq! Only room modes would benefit from higher Q corrections, but then only 1/3 of excess dB! I personally don't like any room mode EQ, speaker and listening positioning are far more better ways to cope with them. I have minidsp 2x4HD and 4x10HD in my two systems. I don't use rephase or dirac, only parametric EQ/XO and delay.
I'm a simple guy...
I like the simplicity of the approach below (REW in RTA mode)
I haven't waded through the 30+ pages to confirm if this method is still regarded as being reasonably accurate compared to others?
I like the simplicity of the approach below (REW in RTA mode)
I haven't waded through the 30+ pages to confirm if this method is still regarded as being reasonably accurate compared to others?
Room measurements should never be done at only one mic location. My friends have learned to me an easy way to do multiple measurement.
We use REW in RTA mode, Forever averaging. Play pink noise or pulsating pink noise with one or more speakers at the time, start RTA recording and move the mic slowly, 3-dimensionally around your head at your "sweet spot". The radius of movement should be 0.5m (2 feet). Continue recording for say 15 seconds. Then after stopping you can toggle for varius smoothing to see how the response looks.
The measurement reflects true room power response and audible spectral balance well. For learning measure several different speakers and compare measurements to your audition (how you hear them to sound).
Extract from PDF says:
With regards to mean-squared sound pressure averages, is there a setting for this in REW?
Im using a miniDSP mic. As far as I know it is not calibrated for random incidence diffuse field?> Calibration : the pink noise has to be checked for its frequency response and microphone should
be calibrated for random incidence diffuse field.
> Mic shall not be pointed directly to loudspeaker : due to semi-diffuse field, microphone can be
pointed randomly during scanning or kept at 90° from source (grazing incidence), both techniques
give near results with small mics (1/4" or less) but allways keep microphone in direct view of
loudspeaker.
> Averaging is to be done on mean-squared sound pressures and on a long enough time : no short
average or moving average. An average on a whole sequence of 30s seems to give good
resolution and reproductibility (ISO/WD 16283-1 recommends 60s for frequencies under
100Hz*11)
With regards to mean-squared sound pressure averages, is there a setting for this in REW?
Last edited:
- Home
- Loudspeakers
- Multi-Way
- Moving Mic Measurement