Acoustic mesurements in room - a question.

[onununo]

I noticed that when measuring the frequency response of my stereo loudspeakers in room, if I make them play together pink noise in mono, and measure using an RTA, I get + 6dB bass as opppsed to approx. + 3dB in the mid-high range, wrt the measurement of a single louspeaker playing the same pink noise. I suppose this is due to the fact that low frequencies, having long wavelength, sum coherently and therefore we get double the SPL (i.e., +6dB), whereas the mid-high frequencies having shorter wavelength and due to the multiple reflections of the room, sum incoherently, and hence the resulting measures amplitude is the sum of the rms amplitudes, resulting in something like +3 dB.
This means that the MEASURED response of both speakers playing together has approx 3dB higher bass with respect to the response of a single speaker.
Now my question is: is this bass "boosting" effect of both speakers playing together also perceived acoustically, or is it only something that appears in the measurements? In case it is, am I right in taking the measurement of both speakers playing together the pink noise as target for balancing the overall system response with EQ, or should I instead take the response of a single speaker? Something tells me that I should make reference to the response of one speaker only, or even better do an rms average of the two single-speakers playing alone since the response of the single speaker is the one I shoild rely on, rather than making reference to the response of both speakers playing together, but something else tells me that I should make reference to the response of both speakers playing together, since when listening to the stereo system this is what they normally do. If to the listener, the perceived system response changes like in the measurements, then only one of the two hypotheses must be correct, but which one?

 

[Keith_W]

1. Read Book 2 of the REW eBook. In particular, pay attention to what it says about bass measurements.

2. What you measure isn't necessarily what you hear. Your measurement depends on whether it was taken CORRECTLY and represents real-life listening conditions. For example, if your measurement was taken with the listening chair removed, or with the microphone not in the listening position, or you point the mic in the wrong direction, or fail to load the calibration file, or a half dozen other variables that results in a poor quality or unrepresentative measurement, then your measurement is lying to you. But if it was a well-taken measurement, what you measure is what you will hear.

3. Since you listen to both speakers, the correct measurement for studying the bass result is to play all speakers and subwoofers together. You can choose to measure them all at once, or you can measure each speaker/sub individually with a timing reference, and then vector sum all of them.

 

[onununo]

1. Read Book 2 of the REW eBook. In particular, pay attention to what it says about bass measurements.

2. What you measure isn't necessarily what you hear. Your measurement depends on whether it was taken CORRECTLY and represents real-life listening conditions. For example, if your measurement was taken with the listening chair removed, or with the microphone not in the listening position, or you point the mic in the wrong direction, or fail to load the calibration file, or a half dozen other variables that results in a poor quality or unrepresentative measurement, then your measurement is lying to you. But if it was a well-taken measurement, what you measure is what you will hear.

3. Since you listen to both speakers, the correct measurement for studying the bass result is to play all speakers and subwoofers together. You can choose to measure them all at once, or you can measure each speaker/sub individually with a timing reference, and then vector sum all of them.

Dear Keith, thank you so much for answering.

1. Great to have an advanced guide to REW, I really felt such a thing was necessary so thank you for your work. While I read your guide, may you give me 2 words of anticipation about what shuold I know about bass measurement that is clear I don't?

2 and 3: I am pretty aware of the fact that measurement do not necessarily represent what we hear, this is also why I was asking if the MEASURED 3dB boost in the bass region is PERCEIVED or somehow psychoacoustically cancelled. As for the measurements, I am measuring with full band periodic pink noise L+R, with careful setting of the FFT resolution, overlapping, windowing, etc, averaging 320 samples acquired while moving the calibrated measurement microphone held in perfect vertical position as recommended for diffused field measurement by the manufacturer, moving it around the position of the listener's head while sampling, with everything (chair, furniture, etc.) in place. It is in short the so-called RTA/MMM technique. if well executed, this measure should result thebsame as in multiple (a lot!) sweep-measurements taken and vectorially summed toghether, shouldn't it?

It all begs to the question of which measureement should be considered as base for balancing the spectral response of the system, whether the single speaker measurement or the 2 speakers playing together, given the 3dB difference in the bass region. Bear in mind that I don't want to equalize the system to flat starting from that measurement (also because we all know that eq-ing to a flat line in room would not result in a flat perceived response), but I happen to have ruler flat speakers on anechoic conditions, along with all their mesurements in such conditions and in an ideal room, so what I am aiming at, is to obtain by calculus the target curve in my room, but in order to get it I need to measure the response of the speakers in my room, and I'm not sure which one of the two I should choose.

 

[Keith_W]

Dear Keith, thank you so much for answering.

1. Great to have an advanced guide to REW, I really felt such a thing" was necessary so thank you for your work. While I read your guide, may you give me 2 words of anticipation about what shuold I know about bass measurement that is clear I don't?

In a nutshell, all rooms will have peaks and dips in the bass response. I divide the causes into 3: (1) loudspeaker to loudspeaker phase cancellation issues, (2) "local" issues, and (3) "room" issues - room modes, SBIR, etc. The book describes a way to tell the difference between them, and what to do about it.

2 and 3: I am pretty aware of the fact that measurement do not necessarily represent what we hear, this is also why I was asking if the MEASURED 3dB boost in the bass region is PERCEIVED or somehow psychoacoustically cancelled. As for the measurements, I am measuring with full band periodic pink noise L+R, with careful setting of the FFT resolution, overlapping, windowing, etc, averaging 320 samples acquired while moving the calibrated measurement microphone held in perfect vertical position as recommended for diffused field measurement by the manufacturer, moving it around the position of the listener's head while sampling, with everything (chair, furniture, etc.) in place. It is in short the so-called RTA/MMM technique. if well executed, this measure should result thebsame as in multiple (a lot!) sweep-measurements taken and vectorially summed toghether, shouldn't it?

If you want to determine the audibility of a peak or dip, use psychoacoustic smoothing, or ERB smoothing. There is a newer version of that eBook that I am working on that talks about different smoothing methods and what they mean. Lately, I have been a bit too busy to write. But then nobody is paying me to write that eBook so I figure you guys can wait

The RTA/MMM technique is superior to the sum of individual sweeps in that it is reliable, has excellent repeatability, and it's easy to do. However, I am not a fan. That eBook tells you why.

It all begs to the question of which measureement should be considered as base for balancing the spectral response of the system, whether the single speaker measurement or the 2 speakers playing together, given the 3dB difference in the bass region.

This is the process when equalizing speakers:

1. Measure each subwoofer and speaker individually with a time reference.
2. Time and phase align each subwoofer to the speakers. The goal is to obtain a flat summation, or as flat as possible.
3. Measure all the subwoofers and speakers together and chop off bass peaks with DSP. You can use a target curve if you want.
4. Decide how much the remaining dips bother you. It may involve more diagnostic measurements, repositioning subs, more advanced DSP techniques, or even purchasing more subs.

 

[onununo]

I am reading your ebook just now, very useful in that it clearly summarizes the process of measuring through REW. Yes, as I am reading, the RTA/MMM method has the great plus of letting me take a measurement in a bunch of seconds, already spatially averaged, where it would take me literally hours of careful microphone positioning to get it using the sweep technique, although I reckon I would also get the time information, whithout which I can't obtain the IR, waterfall, phase plot, etc. But For the kind of equalization I am doing, I am not interested to the absolute phase of each speaker, but rather to the effects that the phase relationship among the speakers makes to the overall amplitude, and this is captured by the MM technique (in fact this is the very specific reason why I am measuring L, R and L+R separately). I may do a sweep measurement from time to time, to check the time domain, but I am not into room acoustics treatment, nor I EQ using linear phase filters (I use Equalizer APO), so knowing the absolute phase is irrelevant for my purposes.
Having said that, please have a look at tthe attached plot, that shows the frequency response of my speakers. They are already pre-equalized by automatic linear phase equalization (they're studio monitors and have their own DSP). The blue and red lines represent L and R, and the red line is R+L. I have applied 1/3rd octave smootihing to ease viewing, and lowered the L+R curve by 3dB to align the high frequency portion of it to the curves of the L and R channels measured individually. You will notice that the bass region of the L+R curve is approx. 3dB higher than that of each speaker measured individually. As mentioned, I believe this must be due to the fact that due to the wavelength, the symmetry of the setup and the relatively short listening distance (2.3m), the L and R bass sums in-phase, whereas as the frequency raises, this phase reletionship gets fuzzed and so higher frequencies sums randomly together, hence L+R measures +6dB in the bass region but only +3dB in the mid-high region when compared to the measurement of R or L alone.
It is difficult to say if this +3dB in the bass are really audible, because the only way to tell is to switch from mono with both channels active to mono with one channel only active, and it's difficult to concentrate only on the spectral balance without getting fooled by the different spatial configuration from single to both channels. Nevertheless, after having made a number of trials, it looks like the higher SPL in the bass region with both channels active is indeed audible.

I have set up a fairly complex procedure to obtain the so-called "House Curve" specific for my room, based on the linearity of my speakers and the availability of all kind of measurements, but it starts with the measured frequency response of the speakers in my room (not the curves attached, but the un-equalized response, which by the way is affected by the same "problem" affecting the curves attached). I believe I should use the average of the R and L curves (specifically the "rms average" provided by REW), because otherwise using the combined R+L measurement would result in a wrong +3dB higher bass. But I am not sure...

 

[bmc0]

I suppose this is due to the fact that low frequencies, having long wavelength, sum coherently and therefore we get double the SPL (i.e., +6dB), whereas the mid-high frequencies having shorter wavelength and due to the multiple reflections of the room, sum incoherently, and hence the resulting measures amplitude is the sum of the rms amplitudes, resulting in something like +3 dB.

Yes, this is correct. Multiple high-frequency sources should be measured separately[1] and averages should be RMS (not vector). The auditory system can "hear through" the room to a substantial degree at high frequencies, in part due to the directional characteristics of the physical hearing apparatus (head shadowing, etc).

At low frequencies, the speaker and room are not so separable and what you measure is closer to what you hear than it is at high frequencies. Keep in mind that the correlation between channels varies considerably with music. Ideally, the low-frequency response should be similar for both monophonic and uncorrelated inputs (levels adjusted for coherent/incoherent summing).

In short, the RMS average of L and R is a better choice in this case.

[1]: Or all together using a stimulus which is uncorrelated in all channels (I don't believe REW has an option for this).

 

[onununo]

Very well, and thanks a lot for giving me the confirmation.
And now that I think about it, this is especially true for my system, considering that my subs are one close to the other and wired in mono (L and R channels get summed to mono in both the subs).