Using two microphones to remove room modes from bass measurements

[Keith_W]

How about down to 60-70Hz ?

Probably not, unless your room is the size of a basketball stadium and you are sitting very far away. Calculate the Schroder frequency of your room, that will tell you where your modal zone is. All speakers within the modal zone should be corrected together with the room, preferably using spatial averaging.

 

[3ll3d00d]

All speakers within the modal zone should be corrected together with the room, preferably using spatial averaging.

it's still technically correct to separate speaker design from room correction which means there is potential for some filter designed to lift/tame the low end of a woofer for the same reasons that you would do the same to a sub. However shifting the frequency up an octave or two doesn't change the fact that you should be able to do this via your model.

The other thing to bear in mind is that where you place the speaker/sub in room can have a big impact on it's native response even ignoring the listening position response and that that can be v hard to model e.g. if you position a sub such that it fires into a solid surface in such a way as that the volume it fires into is largely enclosed then you can turn a sealed sub into something that behaves like a bandpass albeit one where one "enclosure" is leaky, modelling that a priori is not generally possible & anechoic is a waste of time so all you have left is in room. Similarly designing a filter for anechoic response when you plan to mount in or on wall would not be too sensible.

 

[DavidR]

I've made a lot of measurements of a Peerless 10" woofer in a closed box using the Mode Compensator (MC) and the Umik2 with very interesting results. One is that I've been able to determine why the MC Mode output (combined Mic1 and Mic2) doesn't match the Umik2 towards low frequencies. Another one is the MC Mode results for various Mic1/Mic2 positions with Mic1 fixed and increasing Mic2 separation. A third one is that I think that the MC may be used for NF measurements that will elmiinate room mode and other reflections that may be nearly equal to boh mic positions, but I still have a question related to baffle diffraction influence of a small baffle. The Mode output seems always to remove a signfiicant number of irregularities that I think are reflections due to limits of my basement (low ceiling, steam pipes). I'll be posting graphs to demonstrate. It may be that it is most appropriate on a quasi-anechoic baffle that I have yet to set up for testing, but that will be done at some point. There is still a question related to result one above that I will detail. This is all about a function of using two mics, not specifically the Mode Compensator.

It seemed to be a very straight forward scheme, but has turned into an interesting exercise of discovery for me. I thought I could just stick the mics in place, make simple meaurements and be done with it.

First I want to thank NTK again for the Python script for piston sound pressure calculations. The Python library for graphing also has controls that allow for zooming in/out and panning. That helped in choosing specific Mic1/Mic2 placements when interpreting the curves.

Before getting into measurements, I want to cover what I learned from the Python code. I modified it in order to see more details related to ka (frequencies) for a given driver starting with the Peerless 850146 10". I also went to back to the 15W/4531G for some comparison. There are simliarities, of course, but the differences between a small midwoofer and large woofer are enlightening.

To start let me say that I'm still pleased to have the Mode Compensator because it provides two microphones calibrated within the provided preamp. I still have plans to do other testing using the pair together.

The biggest issue in using two mics is due to the fact that near-field (NF) measurements at different locations do not follow the far-field (FF) SPL drop, separate from any diffraction or other influence. That should be apparent from the ka curves shown in the graph curves. More importantly, even when a second microphone is positioned at a point that is 6dB down, the MC Mode output has a slightly increase with decreasing frequency as compared to a direct 10mm measurement. The measurements will come in later posts. The Peerless 10" curves are below.

In order to see more clearly I removed some ka curves to eliminate those that are not decreasing monotonically. Also, keep in mind that these curves are calculations for a flat, pistonic circular radiator. Drivers are not close to those characteristics, but these act as general guidelines. I suspect that real drivers are much better, but have no data to support that idea. Yet.

One could assume that this driver enters FF -6dB/oct around 20cm+. That's close, but they are all asymptotes.

I needed more detail, so I expanded it.

I decided to use the ka=1.0 533Hz curve to set the mics. All frequencies below that would be at or more than 6dB down from Mic1 to Mic2. In fact, that is the explanation for the Mode output of the MC diverging with decreasing frequency. There is no constant 6dB/oct range in the NF.

As can be seen in all graphs, the curves show that at any spacing of two mics in the NF there will not be an equal amount of SPL change with distance among frequencies. The lower the frequency, there is increasing divergence from single-mic results and vice-versa. This was disappointing. I had thought that there might have been an REW calibration issue with how I did that. I'll put measurements in other posts. Despite the issues raised here the MC Mode output has interesting results.

 

[DavidR]

As I was deciding what to post next I noticed in old measurements I maintained from 2004 an interesting similarity. I used REW to adjust the SPL of a Mode Compensator Mode output that conflicts with Umik2 measurements. This is an overlay of the REW adjusted SPL with my LAUD (calibrated to 2.83V @1m) MLS measurement that takes the diapragm diameter into account. This was made on my 2m x 2m baffle, so no diffraction or room mode influence. I was actuall astonished to see this. I expected to see the LAUD measurement more like the Umik2 10mm on-box baffle response. Note the file name, 8x3. LAUD was originally a DOS 3.1 app. I still use it in a Windows 7 PC window. LAUD NF measurements were always limited to 500Hz. I still have the LAUD measurement impulse file as I kept absolutely everything I measured, so I could load it up later for any post-processing. Now I need to do some more investigation of current measurements.

One other note. I found the Peerless specs online for this popular woofer. The sensitivity of my old NF measurement is maybe 2db higher that I attribute partly to the level measured with small mic placements. I used 1/4", tested one time. Moved 1mm made a full dB level change, one reason that I stopped using absolute measured NF results and made a best-guess NF level adjustment to try to match frequency limited FF measurements.

The Mode Compensator measurement is fairly similar to the Peerless datasheet up to 1kHz.

 

[DavidR]

Before I get to the MC measured 6dB mic separation results, this graph shows what happens with Mic1 at a recommended 5cm with multiple measurements made moving Mic2 to increase separation. What surprised my is how the Mode output remains the same except for the level below 1kHz. Above that the swings are expected. However, the MC has the lowpass filter to alter the Mic2 phase that is said to provide for accurate response up to 1kHz. I expexted to see "noise" below 1kH with differences with Mic2 changes. But that didnt' happen. The only thing I can attribute that to is the the MC will remove any and l room modes and common reflections, no matter the mic2 location. Those levels may be nearly constant at both mic locations, the basis for the MC, while combining the Mic1 and Mic2 responses. That seems to be case even if Mic2 is not at the point 6dB down from Mic1. The noise is removed, but the MC Mode output absolute level changes due to change in the Mic2 level relative to the source driver output. Seems counter-intuitive at first, but it's hard to argue with measurements.

 

[DavidR]

Here are the MC measurements, Mic1, Mic2 and Mode outputs. This mode output is the shown in the comparison to LAUD NF measurement above.

For comparison, here are the MC Mode and Umik2 NF measurements. These are the puzzling measurements with unexplained difference. I had attributed that to the Mic separation that still makes sense, but in seeing the nearly perfect overlay of the MC Mode with LAUD 1/4" measurement from 22 years ago...I'm still unsure about the reason.

 

[DavidR]

Given the close match of the MC Mode output to my measurement from 22 years ago in Post #64 above I'm starting to think that maybe the Umik2 calibration at the low end is in question. Earlier overlay of the MC Mic1 to Umik2 shows deviation from about 40Hz down in a 0.5m measurement, about the same in any comparison between MC and Umik, yet it's nearly a perfect overly of the old LAUD measurement. I guess I'll need to get my LAUD system up and test with it again.

Mean time, I was curious about the results of measurements for this woofer when positioned close to the floor. This makes in-box measurements problematic in a different way, not just influence of diffraction. In my range usage in a 3-way this puts the woofer essentially in 2pi space at the low end. I'm posting the Umik2 measurements here. I need to re-do the MC measurement, seemed very odd. The AtFloor location is more problematic with a 2-mic scheme.

In this case, the transition from NF to FF changes from what I see. My thought it that the NF is extended given that the 2pi space, in this case, boosts the low end as much as 3dB @20Hz with a tapered reduction as the wavelength decreases. Note that I adjusted the level of one to overlay the other to my best guess.

Since I can't edit me earlier post, that LAUD NF measurement was, I think, in-box not on my large baffle. The FF measurement was definitely on the large baffle. There couldn't be such a close overlay of in-box to quasi-anechoic baffle.

 

[john61ct]

Might be worth seeing if an independent service could do a re- (check) calibration of the UMIK-2 for a reasonable fee?

Ideally local...

 

[DavidR]

Might be worth seeing if an independent service could do a re- (check) calibration of the UMIK-2 for a reasonable fee?

Ideally local...

I don't know if there is a local service, I may send it in. First I want to try to verify that it is the problem and not an error on my part. If needed, I'll contact PartsExpress and/or MiniDSP about it.

 

[DavidR]

The sound card in my old LAUD system is not responding, so I can't make a new measurement with the that system (I have a backup I'll get out), but I do have a stand-alone mic preamp for measurements as well. I connected the mic/preamp into the Scarlett 2i2 and made a 10mm measurement. Even without loading the mic cal file there is a close result as can be seen in the overlay below. The LAUD mic is nearly the same as the Mode Compensator output. This puts the Umik2 calibration into question. The Mode output (Mic1 + Mic2) of the Mode Compensator does deviate from the others above about 200Hz, but it may be that it's closer to true up to 1kHz as claimed since the single mic measurements will all start to deviate due to the limits of a close-mic position that is typical and expected.

I will probably contact miniDSP about this. For the cost of the mike the calibration should be much better than this. The mic capsule I used came with a calibration file and was excellent and not expensive. This was my second mic capsule due to an accident on my part in moving into a new home, so I had to build a second mic wand. Re-measuring drivers showed a perfect match between the two mics.

This result makes me think that the Mode Compensator is more accurate than it seemed in earlier comparisons to the Umik2. This is very disappointing. Above 200Hz the Umik2 and LAUD mic are surprising close up toe 20kHz. Without a calibration file for the LAUD mic. My previous idea that the deviation had to do with the issue of NF vs FF has less to do with the disparity in the two measurement systems (Umik2 vs MC).

 

[DavidR]

Three months on and I've continued to pursue this. In May added to a thread at the miniDSP community board on Umik-2 started by another member with an issue. Ultimately miniDSP authorized Deer Creek to verify mine. After some back-and-forth on side issues they said it verified against their reference mic and they returned it. I measured again, no change. I was close to pressing for a replacement, but continued to look into everything I could. I had asked the REW author if anything could alter the measurement, he said no, recommended what I had already looked into. So I kept digging into everything.

I was able to resurrect my old Liberty Audiosuite (LAUD) system. The Fiji card was fine, Win98 settings was the issue after repairing the motherboard some time ago. I measured the woofer again with it. It uses scripts to make the close-mic measurement using stepped sine after using an automated optimization of the mic and feedback inputs. It then uses the woofer diameter coupled with the calibrated microphone response to produce a calibrated measurement that is then modified to be represent SPL for 2.83V @1m. This is a DOS 3 program from 1998, has functionality I never even tried to use.

LAUD measurement. I set the markers to make it easy to read the slope. This confirmed that the old microphone used in it's low end matches that when used in REW in Windows 10 using the m-Audio 1010LT card. More on that to come.

For reference I ran loopback calibration for all of the unbalanced outputs of the 1010LT card used for the Panasonic and ModeCompensator measurements. Six channels almost all identical. For a nearly 30 year old card design. Purchased used on ebay some years ago.

When I started using REW I left all of the defaults untouched, but tried some change in preferences. I re-ran measurements with all mics, still found nothing until I looked closely at the most recent Panasonic REW measurement. It also had a downward tilt not in earlier measurements. Shortly after that I found the IR Windows option in REW (I was close to asking the author for similar). What I found was that the default window in REW is set for...measuring room response that needs a long window. Default is 500msec using a Tukey 0.25 window. That is how I left REW when making sound card and amplifier distortion measurements.

With some experimenting in REW I found that a close-mic measurement must have a short window that totally eliminates any influence outside of the driver and that the correct window is a Blackman-Harris. I experimented with the windowing for all three microphone measurements. Most important is the window length, second is the window type. All windows must, of course, go to zero. Several can be used such as Hann, but the Blackman-Harris seems to be the optimum for side-lobe control with a small amount of SPL magnitude.

The optimum window width for this 10" woofer with a 10mm mic position is around 100msec. Each measurement may benefit from a small change from this, +/-, more often -. The window width affects the low end extension as well. Other driver sizes may be slightly different due to mic position needed for the driver diameter, but 100mm will probably be a good starting point. The problem is that I can't find a way to determine an absolute window width that would be appropriate for every case. As an aside, however, the two-mic measurement scheme is much less affected with regard to slope.

This is the result of setting all three microphone measurements to 100mm.

The Umik-2 slope now matches the other mics. However, it is still down 1db across the band from 20-60Hz merging about 85Hz. The difference at 10dB originally was about 5dB down, but in setting the window the 10dB value will swing a lot, but this is for a woofer close-mic. Using it for room EQ will something else entirely, but I am not interested in that myself so I haven't experimented with that. It does appear, however, that the window type is far less important for room EQ.

To demonstrate I made a video uploaded to youtube to demonstrate the impact of the impulse window on the SPL response. REW is amazing at the level of detail easily manipulated with real time display of the SPL frequency response.

REW close-mic measurement:

 

[john61ct]

Good stuff!! Thank you on behalf of all of us...