Topic: Some mic measurements, now with graphs

[notlance]

I just finished performing measurements on most of my microphones, and I thought y’all might be interested on what I found.  I was surprised by some of the results.

First of all a disclaimer:  Since in most cases I only had one mic to measure, my sample size is far too small to be statistically significant.  So it is not possible to say that all mics of that model behave the same as my one sample

I measured these mics by comparing them to a mic with a “known” flat response.  In fact I used the average of 4 mics: the front and back capsules of two Sennheiser MKH800 Twin mics.  These were my standard mics because bought these mics new, their published frequency response is quite flat from 20 to 20kHz, and each capsule comes with its own frequency response measurement.  After analyzing the measurements I made on the MKH800 mics I became even more confident that they are a good standard.

Each mic I tested was 1m from the right speaker in my listening room, set on axis with the tweeter (which happened to be 1.16m high).  I fed pink noise to the speaker, and measured the response of each mic.  I averaged the response of the 4 MKH800 Twin capsules, and subtracted this average from the measured response of each mic resulting in the mic’s approximate frequency response.  My listening room is about 7m x 4m x 2.2m and it is somewhat dead acoustically but certainly not anechoic.  I have treated it to reduce its most egregious acoustical problems.  I used a Rane MS 1 pre-amp and a Cakewalk UA-1G USB audio interface.  The measurements were done by TrueRTA software running on my laptop.

Here are the mics I tested:
Condensers:
2 each Sennheiser MKH800 Twin
1 Sennheiser MKH800, set to cardioid
2 each Milab DC-196, set to cardioid
1 Beyerdynamic MC740, set to cardioid
1 Beyerdynamic MC742 stereo mic, tested both upper and lower mics, set to cardioid
1 Sennheiser MKH835T cardioid
1 Nakamichi CM300 cardioid (CP-1)
1 Nakamichi CM300 omni (CP-2)
4 each Milab LSR2000 cardioid
4 each MXL603S cardioid, modified by Michael Joly
1 MXL991 cardioid unmodified
2 each Busman SDC cardioid
2 each Behringer ECM8000 omni measurement mics

Ribbons:
1 Busman Stereo ribbon, (Blumlein) tested both upper and lower
1 Beyerdynamic M160 hyper-cardioid
3 each Beyerdynamic M130 figure-8
1 Peluso R14 figure-8

Dynamics:
1 Beyerdynamic M101 omni
2 each Beyerdynamic M201 hyper-cardioid

My comments will be in the next post.

[notlance]

Mic measurement comments, Condensers.

Others have said it before, but now I understand better why some mics are so bloody expensive and why others are not.  For example, the Sennheiser MKH800 mics are in the neighborhood of $3000 each.  So what do you get for that kinda money?  For one thing, consistency.  The MKH800 Twin mics are not factory matched, nor are they consecutive serial numbers (10016 and 10138), but any one of the four capsules (front and back on each mic) are within +- 1dB of any other capsule from 20 to 20kHz, and within +- 0.5 dB from 70 to 20kHz.  These kinds of differences could easily be the result of my measurement errors rather than the mic themselves, so for all intents and purposes these mics are perfectly matched.  The MKH800 also matches closely to either of the MKH800 Twin mics.  The Milab DC-196 mics are a close match.  (My frequency response measurement of the Milab DC-196 mics agrees with the measurements supplied by the factory, so that give me assurance than my measurement results are reasonably accurate.)  The upper and lower capsules of the Beyerdynamic MC742 are almost identical.  Even the 4 Milab LSR2000s (which are used vocal performance mics) are very close to each other.

With the Chinese mics it’s a different story.  For example, out of the 4 MXL603S mics, two of them are close enough to each other that I would consider using them as a stereo pair.  The Busman SDC mics are well matched, however.

On an absolute basis the European mics again have an advantage.  Their responses are smoother and more extended than any of the Chinese mics I measured.  I’d even have no problems recording with the Milab LSR2000, which as I said, is a vocal performance mic.

As for the Chinese mics, the MXL991 was particularly bad.  No low end, a peak at 4 kHz, a dip at 6 kHz, and another peak at 9 kHz.  However, my sample size was just one, so maybe I got a bad mic.  Good thing it was free.

I would recommend the Michael Joly mod that he does to the MXL603S, MXL991 and other similar Chinese SDCs.  His mod lowers the 5 kHz bump these mics almost always have, and turns a barely useable mic into something pretty good.  His mod does not change the low end response, which varies quite a bit for these Chinese SDCs, so if you’re going to spend the money to get mic modded, pick one with a good low end.

I tested the Behringer ECM8000 measurement mics just to make sure they close enough to flat to be considered a measurement mic.  I’m happy to say that they both measured flat enough to be useable as a measurement mic.  One was better than the other, but both were acceptable.  I’d never use them for recording since they are too noisy, but not bad for $50.

Next I’ll comment on the ribbons and the dynamics, but it might be a while before I get it posted.

[Roger Gustavsson]

Will you show us any measurements?

Roger

[boojum]

NL -  great work!  I would sure like to see the details of what you have done.  Even though it is only your mics it would be very inteteresting to see your figures.

Cheers

[Church-Audio]

I just finished performing measurements on most of my microphones, and I thought y’all might be interested on what I found.  I was surprised by some of the results.

First of all a disclaimer:  Since in most cases I only had one mic to measure, my sample size is far too small to be statistically significant.  So it is not possible to say that all mics of that model behave the same as my one sample

I measured these mics by comparing them to a mic with a “known” flat response.  In fact I used the average of 4 mics: the front and back capsules of two Sennheiser MKH800 Twin mics.  These were my standard mics because bought these mics new, their published frequency response is quite flat from 20 to 20kHz, and each capsule comes with its own frequency response measurement.  After analyzing the measurements I made on the MKH800 mics I became even more confident that they are a good standard.

Each mic I tested was 1m from the right speaker in my listening room, set on axis with the tweeter (which happened to be 1.16m high).  I fed pink noise to the speaker, and measured the response of each mic.  I averaged the response of the 4 MKH800 Twin capsules, and subtracted this average from the measured response of each mic resulting in the mic’s approximate frequency response.  My listening room is about 7m x 4m x 2.2m and it is somewhat dead acoustically but certainly not anechoic.  I have treated it to reduce its most egregious acoustical problems.  I used a Rane MS 1 pre-amp and a Cakewalk UA-1G USB audio interface.  The measurements were done by TrueRTA software running on my laptop.

Here are the mics I tested:
Condensers:
2 each Sennheiser MKH800 Twin
1 Sennheiser MKH800, set to cardioid
2 each Milab DC-196, set to cardioid
1 Beyerdynamic MC740, set to cardioid
1 Beyerdynamic MC742 stereo mic, tested both upper and lower mics, set to cardioid
1 Sennheiser MKH835T cardioid
1 Nakamichi CM300 cardioid (CP-1)
1 Nakamichi CM300 omni (CP-2)
4 each Milab LSR2000 cardioid
4 each MXL603S cardioid, modified by Michael Joly
1 MXL991 cardioid unmodified
2 each Busman SDC cardioid
2 each Behringer ECM8000 omni measurement mics

Ribbons:
1 Busman Stereo ribbon, (Blumlein) tested both upper and lower
1 Beyerdynamic M160 hyper-cardioid
3 each Beyerdynamic M130 figure-8
1 Peluso R14 figure-8

Dynamics:
1 Beyerdynamic M101 omni
2 each Beyerdynamic M201 hyper-cardioid

My comments will be in the next post.

What are you using as a reference microphone?

Chris

[flipp]

I just finished performing measurements on most of my microphones, and I thought y’all might be interested on what I found.  I was surprised by some of the results.

First of all a disclaimer:  Since in most cases I only had one mic to measure, my sample size is far too small to be statistically significant.  So it is not possible to say that all mics of that model behave the same as my one sample

I measured these mics by comparing them to a mic with a “known” flat response.  In fact I used the average of 4 mics: the front and back capsules of two Sennheiser MKH800 Twin mics.  These were my standard mics because bought these mics new, their published frequency response is quite flat from 20 to 20kHz, and each capsule comes with its own frequency response measurement.  After analyzing the measurements I made on the MKH800 mics I became even more confident that they are a good standard.

Each mic I tested was 1m from the right speaker in my listening room, set on axis with the tweeter (which happened to be 1.16m high).  I fed pink noise to the speaker, and measured the response of each mic.  I averaged the response of the 4 MKH800 Twin capsules, and subtracted this average from the measured response of each mic resulting in the mic’s approximate frequency response.  My listening room is about 7m x 4m x 2.2m and it is somewhat dead acoustically but certainly not anechoic.  I have treated it to reduce its most egregious acoustical problems.  I used a Rane MS 1 pre-amp and a Cakewalk UA-1G USB audio interface.  The measurements were done by TrueRTA software running on my laptop.

Here are the mics I tested:
Condensers:
2 each Sennheiser MKH800 Twin
1 Sennheiser MKH800, set to cardioid
2 each Milab DC-196, set to cardioid
1 Beyerdynamic MC740, set to cardioid
1 Beyerdynamic MC742 stereo mic, tested both upper and lower mics, set to cardioid
1 Sennheiser MKH835T cardioid
1 Nakamichi CM300 cardioid (CP-1)
1 Nakamichi CM300 omni (CP-2)
4 each Milab LSR2000 cardioid
4 each MXL603S cardioid, modified by Michael Joly
1 MXL991 cardioid unmodified
2 each Busman SDC cardioid
2 each Behringer ECM8000 omni measurement mics

Ribbons:
1 Busman Stereo ribbon, (Blumlein) tested both upper and lower
1 Beyerdynamic M160 hyper-cardioid
3 each Beyerdynamic M130 figure-8
1 Peluso R14 figure-8

Dynamics:
1 Beyerdynamic M101 omni
2 each Beyerdynamic M201 hyper-cardioid

My comments will be in the next post.

What are you using as a reference microphone?

Chris

pretty sure he used what I bolded for his reference mics

[notlance]

Yes, as I said in my original post and was pointed out by flipp, I used 2 Sennheiser MKH800 Twin mics as my reference mics.

I will try to post my measurement results but I'll have to figure out how to do it.  Stay tuned.

[illconditioned]

Checking in to thread... waiting for the figures/graphs...

  Richard

[notlance]

OK, here is a test to see if I can get this graph to display.

This first graph is the difference between the front capsules of my two MKH800 Twin mics that I used as reference.  Since these mics are my reference mics, I cannot display an "absolute" frequency response for them.  But this graph shows how consistent these mics are, and it shows my test method is reasonable since two mics that should have about the same response do have the same response.  These mics have SN 10016 and 10138 and are “unmatched”.

[notlance]

That seemed to work OK.

Here is the data for two Milab DC-196 mics, set to cardioid.
First graph:      Milab DC-196 SN 33468
Second graph: Milab DC-196 SN 33469
Third graph:     Difference between the two Milab mics

[notlance]

One more set then I'm going to bed.

Graph 1: Beyerdynamic MC740, set to cardioid

Note that the Beyerdynamic MC742 is a stereo mic.
Graph 2: Beyerdynamic MC742 upper capsule, set to cardioid
Graph 3: Beyerdynamic MC742 lower capsule, set to cardioid
Graph 4: Difference between MC742 upper and lower capsules

More tomorrow.

[notlance]

More mics.  Note: For mics that don't have a serial number, I give them a color code.  For example the mics below are coded "red" and "black".

Graph 1: Behringer ECM8000 measurement mic "red"
Graph 2: Behringer ECM8000 measurement mic "black"

These are omnidirectional mics

[notlance]

Graph 1: Sennheiser MKH435T (cardioid)
Graph 2: Nakamichi CM300 CP-1 (cardioid)
Graph 2: Nakamichi CM300 CP-2 (omni)

[notlance]

This next group are all cheap Chinese SDC cardioids.  The first 4 are MXL603S that have been modified by Michael Joly, the 5th one is a stock MXL991.  The MXL603S and MXL991 are basically the same mic.

Graph 1: MXL603S Joly mod "black"
Graph 2: MXL603S Joly mod "blue"
Graph 3: MXL603S Joly mod "green"
Graph 4: MXL603S Joly mod "red"
Graph 5: MXL991

I do have data on the unmodified MXL603S mics, but it was taken using a different test setup.  I think there are enough variables between the tests performed on the unmodded mics and the modded mics that I do not feel comfortable publishing the unmodded mic data.

[notlance]

Graph1: Busman SDC 041 cardioid
Graph2: Busman SDC 042 cardioid

[notlance]

These mics are all Milab LSR2000 cardioids, which are often used as vocal performance mics.

Graph 1: Milab LSR2000 4097 cardioid
Graph 2: Milab LSR2000 6054 cardioid
Graph 3: Milab LSR2000 6056 cardioid
Graph 4: Milab LSR2000 6137 cardioid

[notlance]

Up to now, every mic has been a condenser.  These next three mics are dynamics.  The first two are Beyerdynamic M201 mics, a fairly common hypercardioid often used on drums.  The third mics, the M101, is a little less common and it’s an omni version of the M201.

Graph 1: Beyerdynamic M201 35019
Graph 2: Beyerdynamic M201 043072
Graph 3: Beyerdynamic M101

[notlance]

The next few post are all ribbons.  This first group are Beyerdynamic ribbons, the M160 and the M130.

I was surprised at the lack of low end on the M160; is the ribbon damaged?  Since I do not have another M160 to compare it to, I can’t be sure.  The mic sound OK, and quite frankly the M160’s LF rolloff has not been an issue in use, but I tend to use it to mic instruments without much bass.

The M130s are quite variable, with better bass than my M160 sample.  The M160 and M130 are often used as an MS pair, and from my experience they work quite well in that configuration.  One thing I find interesting is how my M160 and the M130 SN 2652 are almost a perfect complement to each other in an MS configuration.  They seem to have built-in stereo “shuffler” equalization where the Mid LF are cut and the Side LF are boosted.  I’ve used my M160 and one of my M130s (don’t know which one) in MS to record a choir and it did sound quite good, IMHO.  But it’s hard to say exactly why.
OK I’ll shut up already and give you the data.

Graph 1: Beyerdynamic M160 hypercardioid
Graph 2: Beyerdynamic M130 2629 figure-8
Graph 3: Beyerdynamic M130 2652 figure-8
Graph 4: Beyerdynamic M130 3675 figure-8

[notlance]

Next is a Peluso R14 ribbon, figure-8.  Just bought this mic and have not had a chance to use it yet.

Graph 1: Peluso R14 00012 figure-8

[notlance]

This next mic is perhaps the surprise of the lot.  It’s a Busman Stereo Ribbon in a Blumlein configuration.  I was surprised how uniform it is (for a ribbon) and how well matched the channels are.  Now its high end above 10 kHz has some problems, and it’s Blumlein so you have to live with the limitation of that configuration, and it’s a ribbon with not much output, but I’ve made some nice recordings with it IMNSHO.  This is a 300 buck mic.

Graph 1: Busman Stereo Ribbon left channel
Graph 2: Busman Stereo Ribbon right channel
Graph 3: Busman Stereo Ribbon difference between left & right channels

This is all the data I have for now.  I do have a few more mics I plan on testing, in particular a Studio Projects LSD2 which I forgot about, but I need to attend to the rest of my life for a while.  It will get done when it gets done.

[Roger Gustavsson]

Thank you, for sharing this with us!

Roger

[page]

Thank you, for sharing this with us!

[Church-Audio]

Unfortunately you cant really use a cardioid mic as a measurement source.. And get reliable data from it when measuring omnidirectional mics. Also the mics capsule size will also have a huge effect on the disturbance of the sound field making it also less reliable for a measurement standard. You can get an "idea"
 but you can not base anything from the data you are collecting because of the issues mentioned above.. Measuring Microphone frequency response is one of the hardest things to do in acoustics. Its easy to measure a speaker.. But not a microphone. The substitution method simply does not work with microphones of different polar patterns and types. The other huge part of the puzzle that you dont have is a db calibration for the Sennheiser mic.. You cant have that with out using a calibrator on the mic then using the same calibrator on all the other mics so you have at least matched the sensitivity. I can see by your graphs that there are "similarities" in all the plots.. This indicates a flaw in your testing. You must be able to subtract the measurement errors from the graphs.. In other words you must be able to subtract the flaws in your sound source from your graphs by using the Mic calibration file and zeroing out your reference curve. But you cant do that because you dont have calibration data about your measurement mic. That's why I spend $150 a year getting my reference mics calibrated. So I have a reliable reference. I think its great that you did these tests... i think its also important to point out that they are in no way accurate. Very small shifts in placement between mics and the sound source will also give you bad results that are 100% inconsistent.

Chris

[kbergend]

Thanks for sharing all this, very informative!

[notlance]

Reply to Church-Audio:

I’m not sure my measurements are all that bad, and I’ll try to show why I believe they are valid.  Let me do this by answering your concerns in order.

“Unfortunately you cant really use a cardioid mic as a measurement source.. And get reliable data from it when measuring omnidirectional mics.”
I agree.  In fact, as I was setting up one of my omnidirectional mics for measurement I was thinking “This does not make sense since my standard mic is a cardioid, but let’s see what I get.”  Measuring figure-8 mics with a cardioid standard is probably suspect also.  That is why I listed the polar pattern of all the mics I tested, so the observer could view the graphs with whatever size grain of salt they deemed appropriate.

“Also the mics capsule size will also have a huge effect on the disturbance of the sound field making it also less reliable for a measurement standard.”
My reference mic has a capsule size that is just about the same size as most of the microphones I measured, so the disturbance to the sound field of the standard mic and the measured mic should be similar.

“You can get an "idea" but you can not base anything from the data you are collecting because of the issues mentioned above.”
OK, let’s toss out all the data taken on the omnidirectional mics, and consider the cardioids only.

I believe my data on those cardioid mics is valid because it is consistent with all the independently measured data I have available to me.  For example, I have 6 factory supplied frequency response graphs for the Sennheiser MKH800 and the Milab DC196 mics and my measurements are consistant with the factory measurements to a surprising degree, within about +-1 dB.  In some cases I do not have the frequency response graphs of my individual mics, but I can reference the microphone model’s frequency response spec.  My data for the Beyerdynamic MC742 and MC740 agree closely with the specs for these models, for example.  Did I just get lucky with my measurements, or are they at least reasonable?

In addition, what my ears tell me matches my measurements.  For example, the recordings I have made with the Busman SDC sound like they have a presence bump around about 5 kHz.  I can hear the high-frequency rolloff of the ribbons.  To my ears the Sennheiser MKH800 and the Milab DC-196 sound a lot alike, and they also measure a lot alike.  My ears are not calibrated in dB, but what I hear correlates to what I measure, so that is at least a reasonableness check.

“Measuring Microphone frequency response is one of the hardest things to do in acoustics. Its easy to measure a speaker.. But not a microphone.”
I agree it is difficult to measure microphone frequency response, but if the test is carefully set up and one considers all possible sources of error, it is not impossible. Apparently you are able to do it, so why can’t I do it also?  (And speakers are not that easy to measure either, BTW.)

“The substitution method simply does not work with microphones of different polar patterns and types.”
Yes, I agree and I have already acknowledged my error in comparing omni and cardioid mics.

“The other huge part of the puzzle that you dont have is a db calibration for the Sennheiser mic.. You cant have that with out using a calibrator on the mic then using the same calibrator on all the other mics so you have at least matched the sensitivity.”
I don’t care about sensitivity, and I don’t need to measure it for my frequency response measurements to be valid.  The only thing a sensitivity measurement would do is shift the frequency response curves either up or down relative to each other without changing the shape of the curves.

“ I can see by your graphs that there are "similarities" in all the plots.. This indicates a flaw in your testing.”
What are these similarities?  Many of the mics show a presence boost, but that is by design.  I guess I have not poured over these graphs long enough to see the common errors, although I have spent too much time on this little experiment already it seems to me.

“You must be able to subtract the measurement errors from the graphs.. In other words you must be able to subtract the flaws in your sound source from your graphs by using the Mic calibration file and zeroing out your reference curve. But you cant do that because you dont have calibration data about your measurement mic. That's why I spend $150 a year getting my reference mics calibrated. So I have a reliable reference.”
I understand I need to subtract the non-random measurement errors from the graphs.  I do not have a calibration file, but I do have some factory supplied frequency response graphs for the Sennheiser mics.  These graphs are shown in the attached files called “Sennheiser 10015” and “Sennheiser 10138”.  The mics are “flat” enough to use as a reference, certainly from about 100 Hz to 10 kHz.  I could create a calibration file from these graphs, but that would be tedious and applying that correction would not significantly alter the shape of the frequency response graphs I measured.  The Sennheiser mics have their worst deviation of about +1 to + 2 dB between 10 kHz and 20 kHz.  So would adding 1 to 2 dB from 10 to 20 kHz make a big difference in any of my measurements?  Would that change the plots from unusable to acceptable?

What if the Sennheiser plots are in error, and my reference mic are much worse than I am assuming?  Well, I can compare the Sennheiser mics to each other, and I find the comparisons are consistent with the plots supplied by Sennheiser.  For example, the Sennheiser plots show at 50 Hz the front capsule of SN 10138 is about 2 dB lower than the front capsule of SN 10015.  My measurements confirm this.  In fact, my measurements are in agreement with the Sennheiser plots across the spectrum.  I can look at the plots I get from the Milab DC-196 mics and see they agree with the factory supplied plots.  If the Sennheiser factory plots are wrong, then the Milab factory plots would have to error in a complementary way in order for my measurements to match the Milab factory data.  This is not very likely.

Nevertheless, I would love to have an appropriate calibrated reference mic, if for nothing else my own peace of mind.  Can you recommend an appropriate microphone?  And where should I get it calibrated?  And how do you know the calibration data you have is accurate?

This post continues below.

[notlance]

“I think its great that you did these tests... i think its also important to point out that they are in no way accurate.”
Actually, if I thought my tests were “in no way accurate” I would be doing a disservice to everyone here by publishing bogus data.  Since I thought my measurements were reasonable, I made them public.

“Very small shifts in placement between mics and the sound source will also give you bad results that are 100% inconsistent.”
So how much error is introduced by shifting mic placement?  And how small is a “very small” shift?  Is +- 5 mm a very small shift?  To find out, I performed the following test:  I placed a mic 1.16 m high and 1 m from a speaker.  For this test it does not matter what mic I used, but it happened to be a Sennheiser MKH800 Twin.  With a pink noise source, I plotted and saved the frequency response at that location.  Then I moved the mic 5 mm higher, and ran another frequency response plot.  I repeated the plots with the mic 5 mm lower, 5 mm closer, and 5 mm farther back from the center location.  The only variable in this test is the location of the mic.  So now if I subtract the response of the mic in the center location from the responses at the other locations I can plot the error introduced by a shift in placement.

These plots are shown in the attached file “Placement Error”.  This file contains four plots offset from each other in order to be readable.  From top to bottom they are 5 mm up, 5 mm down, 5 mm forward, and 5 mm back.  These plots should NOT be compared to each other, rather the deviation within a plot should be observed.  Note the entire scale of this plot is 10 dB.  Each plot is +-0.5 dB, and some are +-0.25 dB which indicates the magnitude of the error introduced by a shift in placement.  Now some of this error is random error introduced largely, I would suspect, by the pink noise source, particularly at low frequencies.  I ran some tests to confirm this random error, but I’m not going to go into it at this time because this is turning into a thesis as it is.

My conclusion is I can vary the mic placement by +-5 mm without significantly affecting the frequency response, at least using a pink noise source.  I chose +-5 mm because I know I can place a microphone within that error margin, and NOT have “bad results that are 100% inconsistent.”

I have host of information that tells me my data is valid.  My measurements are consistent with outside sources of information, it is constant with itself (i.e. one measurement does not contradict another), and it is consistent with what I hear.  I am sure you know what you doing and I know you have a good reputation on this forum.  But with all due respect the only thing telling me my data is bogus is you.  You have raised some valid concerns and I hope to decrease my measurement error based on your suggestions, but I’ll stand by this data.

[Church-Audio]

“I think its great that you did these tests... i think its also important to point out that they are in no way accurate.”
Actually, if I thought my tests were “in no way accurate” I would be doing a disservice to everyone here by publishing bogus data.  Since I thought my measurements were reasonable, I made them public.

“Very small shifts in placement between mics and the sound source will also give you bad results that are 100% inconsistent.”
So how much error is introduced by shifting mic placement?  And how small is a “very small” shift?  Is +- 5 mm a very small shift?  To find out, I performed the following test:  I placed a mic 1.16 m high and 1 m from a speaker.  For this test it does not matter what mic I used, but it happened to be a Sennheiser MKH800 Twin.  With a pink noise source, I plotted and saved the frequency response at that location.  Then I moved the mic 5 mm higher, and ran another frequency response plot.  I repeated the plots with the mic 5 mm lower, 5 mm closer, and 5 mm farther back from the center location.  The only variable in this test is the location of the mic.  So now if I subtract the response of the mic in the center location from the responses at the other locations I can plot the error introduced by a shift in placement.

These plots are shown in the attached file “Placement Error”.  This file contains four plots offset from each other in order to be readable.  From top to bottom they are 5 mm up, 5 mm down, 5 mm forward, and 5 mm back.  These plots should NOT be compared to each other, rather the deviation within a plot should be observed.  Note the entire scale of this plot is 10 dB.  Each plot is +-0.5 dB, and some are +-0.25 dB which indicates the magnitude of the error introduced by a shift in placement.  Now some of this error is random error introduced largely, I would suspect, by the pink noise source, particularly at low frequencies.  I ran some tests to confirm this random error, but I’m not going to go into it at this time because this is turning into a thesis as it is.

My conclusion is I can vary the mic placement by +-5 mm without significantly affecting the frequency response, at least using a pink noise source.  I chose +-5 mm because I know I can place a microphone within that error margin, and NOT have “bad results that are 100% inconsistent.”

I have host of information that tells me my data is valid.  My measurements are consistent with outside sources of information, it is constant with itself (i.e. one measurement does not contradict another), and it is consistent with what I hear.  I am sure you know what you doing and I know you have a good reputation on this forum.  But with all due respect the only thing telling me my data is bogus is you.  You have raised some valid concerns and I hope to decrease my measurement error based on your suggestions, but I’ll stand by this data.

You have your opinion I have mine. I just wanted to point out your tests are highly flawed. And they are highly flawed, I dont want people to look at graphs and make decisions on purchases based on false data that is not a service people need around here. I am sure you have convinced your self that you are doing the tests in a correct manor but I can assure you there is problems with your methods.. I have pointed them out. I did not want to get into a pissing contest. Its very easy to measure a speaker compared to a mic.. Also you are looking at Freq plots from Sennheiser mics how did they get the data they are printing out? How flat are the mics? did someone have there finger on the plotter? was smoothing used?? You don't know so you don't have any point of real reference that's why I spend $1000+ for a proper nist traceable analyzer mic and spend $150 a year to get it calibrated not to mention the $500 calibrator that I use to calibrate the sensitivity to my preamp / program. If it was just a matter of looking at  a plot and saying hey that's flat I would have done that This is what I do for a living after all. But go ahead and do your tests this is how you will learn Its how I learned. But be sure to point out that the tests are NOT perfect so people don't go thinking they are.

Chris Church

[chris319]

My listening room is about 7m x 4m x 2.2m and it is somewhat dead acoustically but certainly not anechoic.

Not to be harsh, but you have no idea what effect room acoustics are having on your tests.

[Roger Gustavsson]

It is a pity that Sennheiser have no comments on how their microphones are measured. Neumann have their microphones measured in free-field conditions, the IEC 60268-4 standard.

I would have expected the omnis to have better responses (flatter bass) than shown here. Even low price omnis use to be very flat at lower frequencies.


Roger

[Church-Audio]

Nevertheless, I would love to have an appropriate calibrated reference mic, if for nothing else my own peace of mind.  Can you recommend an appropriate microphone?

Shoot me a PM and I'll lend you a reference microphone. It will be a second gen calibration, so that will limit your accuracy to around +/-0.3dB.  Accuracy above 10kHz will be limited due to placement error, but that's not terribly important for many mics as their response tends to drop above that point.  So +/-2dB at 20kHz is fine.

As for room acoustics, use the poor man's anechoic chamber:  go outside.  You can do one test outside to study the effects of your indoor tests.

John you have the free field calibration data for your mic? If not then there is not much point.. Also just going "outside" is not enough. I only wish it was that easy.

Chris

[Church-Audio]

Of course I do.  And this discussion shouldn't progress in the manner of "it's impossible".  It's not a matter of possibility, it's a matter of margin of error.  So first we have to define what the acceptable margin of error is, then ascertain whether or not we can achieve that margin with the test methodology.

The difficulties of attempting an on-axis, free-field measure in a non-reflection-free environment aren't that hard to understand.  It starts with the basic inverse square law of sound intensity.  So we know that the closer we get to a sound source relative to reflective surfaces, reflections become less prominent relative to the source signal.  If we can stay 10cm from our source, we probably don't have to worry much about the room's acoustics at all.  This can be easily demonstrated in any room with a speaker.

Two problems: first, it's not easy to find a perfect full-range driver.  That isn't really so terrible though, so long as we start with a calibrated microphone and a reasonably full-range driver, we can compare response via the subtraction method.

That works great for omnidirectional microphones, not too well for cardioid microphones.  This is of course due to proximity effect.  That's the second problem.  So if we want to get an idea of the far-field low-frequency response of the microphone, we need to get at least 1m away.

Now room acoustics become troubling, because the reflections are no longer sufficiently small compared with the source signal.  That can lead to some different inaccuracies.  So we probably can't get an accurate low-frequency measurement of a cardioid microphone indoors . . .

Or can we?  Again, if we start with a reference microphone at the same location, we will know the room's response once we adjust for the reference microphone's response.  And so we can derive the microphone under test's low-frequency response.

I am curious why you think the outdoors does not very closely approximate a reflection-free environment.  What is the margin of error there?  If the nearest reflection is say 20m away?  This can be calculated mathematically without too much trouble.  And verified experimentally, again with the reference mic at increasing distances from the speaker.  That result can be compared with the same test in the room, and the effect of the room can be derived.

Some of the other comments . . . the error in mic positioning should be a frequency-dependent effect.  I don't get the objection based on capsule size.  That is relevant if you are trying to measure an acoustic environment, because the microphone will impede upon that environment by its presence.  But if you are measuring the microphone itself, the acoustic effects of its capsule and housing are an integral part of the microphone's response.

I never actually said outdoors was not a good environment.. I simply said it was not as simple as placing a speaker outside and measuring the sound with it.. as a test source.. I am sure you can explain why.

Yes its pretty difficult to measure a microphone accurately, just ask AKG, Sennheiser, Shure.. How many $100's of thousands of dollars they have spent trying to do that "simple task" you speak of. Its not simple it never was and it takes real gear to get it right. Not just placing a mic in front of a speaker and calling a day. The single most important part of the equation has not even been talked about at least not be anyone but me. figure out what that is then tell me how easy it is Its not if it was there would be a lot of mic companies out there with "flat mics" selling them for lots of $$$.  Because the real truth is most mics made today are pretty flat. But flat in and of it self does not tell the whole story.



Chris

[Church-Audio]

You know what, I'm not really into mystical gnosticism, so you can either share what you know or not, I am not particularly interested in guessing what you feel like you have to hide.  As I recall you haven't published frequency response specifications for any of your microphones.

Strict laboratory techniques for microphone measurement are well-described in the relevant standards, what I am discussing is what margin of error can be expected for the home "lab" environment.  When you want to know if microphone X is relatively flat or has a 10dB peak at 6kHz, that degree of accuracy is not hard to achieve.

I would challenge the assertion that most microphones are flat.  Clearly most are not, especially directional microphones, which have the lion's share of the market.

  No secrets just basic facts its not easy to measure a microphone with any kind of meaningful accuracy with out matching the raw sensitivity of the mic and with out proper placement of a sound source even outside. I still stand behind that statement. 

[illconditioned]

Let me just step in for a bit.

I agree, instead of negative statements, please try to provide some suggestions as well.

One method I have found to be quite good is to put two *small* mics side by side.  Hold them in front of a speaker and play pink noise at approx 80dBA or so.  (I wear earmuffs when doing this, otherwise it sounds like I'm behind a waterfall.)

I record with a stereo rig, Edirol or whatever.  Then I play back and analyse the spectrum with say two to five seconds of averaging.

This is a "quick and dirty" sanity check, that the mics are both outputting a good signal, and that they have similar responses.  I've done this with: small omnis (Countryman B3, Sennheiser MKE2, etc) and with slightly larger mics (AT853, CA11, Sennheiser MKE40, etc).

Enhancements.  Get a calibrated mic.  I could also do one mic right after another, being careful to put both mics in exactly the same position.

I think the biggest problem is the room response, which can vary quite a bit with position.  The second is the speaker response, which again can vary quite a bit.

Anyway, this kind of "sanity check" is good enough for small (5mm) omni mics.  It would be a good way to match a dozen WM61a capsules, for example.

  Richard

[illconditioned]

Enhancements.  Get a calibrated mic.  I could also do one mic right after another, being careful to put both mics in exactly the same position.

I think the biggest problem is the room response, which can vary quite a bit with position.  The second is the speaker response, which again can vary quite a bit.

Anyway, this kind of "sanity check" is good enough for small (5mm) omni mics.  It would be a good way to match a dozen WM61a capsules, for example.

  Richard

That's an entirely valid way to compare small omni capsules.  The speaker and room response aren't too important, since they will be the same for all units under test, although it does help to be diligent about identical placement.

Treating the room and doing a few helpful things with the drivers can make the charts look prettier insofar as they will look flatter, but the spiky kind of graphs you get without controls are valid for subtraction.

Yeah, I'm just suggesting this as I discovered it by chance.  My *guess* is that a lot of people (like Coresound, etc) are using this technique.  I don't imagine all of these guys are as obsessive about testing as Chris Church .

  Richard

[notlance]

My listening room is about 7m x 4m x 2.2m and it is somewhat dead acoustically but certainly not anechoic.

Not to be harsh, but you have no idea what effect room acoustics are having on your tests.

The effects of room acoustics should be negated because the reference mic and the test mic are measured in the same room at the same positions with the same sound source.  Therefore, the room acoustics affect both mics equally and are canceled out when the reference mic curve is subtracted from the test mic curve.

[Church-Audio]

Enhancements.  Get a calibrated mic.  I could also do one mic right after another, being careful to put both mics in exactly the same position.

I think the biggest problem is the room response, which can vary quite a bit with position.  The second is the speaker response, which again can vary quite a bit.

Anyway, this kind of "sanity check" is good enough for small (5mm) omni mics.  It would be a good way to match a dozen WM61a capsules, for example.

  Richard

That's an entirely valid way to compare small omni capsules.  The speaker and room response aren't too important, since they will be the same for all units under test, although it does help to be diligent about identical placement.

Treating the room and doing a few helpful things with the drivers can make the charts look prettier insofar as they will look flatter, but the spiky kind of graphs you get without controls are valid for subtraction.

Yeah, I'm just suggesting this as I discovered it by chance.  My *guess* is that a lot of people (like Coresound, etc) are using this technique.  I don't imagine all of these guys are as obsessive about testing as Chris Church .

  Richard

Who is obsessive ? 

[Church-Audio]

My listening room is about 7m x 4m x 2.2m and it is somewhat dead acoustically but certainly not anechoic.

Not to be harsh, but you have no idea what effect room acoustics are having on your tests.

The effects of room acoustics should be negated because the reference mic and the test mic are measured in the same room at the same positions with the same sound source.  Therefore, the room acoustics affect both mics equally and are canceled out when the reference mic curve is subtracted from the test mic curve.

How can you guarantee the same position? when the capsule sizes are all different? And they are all different.

[Church-Audio]

No secrets just basic facts its not easy to measure a microphone with any kind of meaningful accuracy with out matching the raw sensitivity of the mic and with out proper placement of a sound source even outside. I still stand behind that statement.

Well I offered to lend OP a reference mic for which I know its sensitivity, on-axis response, polar response, and so forth.  The same lab reports you probably have.  And of course placement is important with respect to the sound source and the acoustic environment.  I am trying to help OP mitigate those factors by explaining some useful techniques.  You apparently just want him to give up.  I don't accept that.  You also haven't defined "meaningful accuracy".  How would you define that, in quantitative terms if you please?

I will try to demonstrate some of these techniques in the next few days . . .

I don't think you have a nist traceable mic.. You may think that it is some labs are saying that have nist traceable gear but unless you get a nist number with your calibration data its not nist. Also sensitivity is relative to the input the mic is connected to and the signal chain. You need a calibrator to "calibrate" the signal / sensitivity to any new source you connect your mic to and you then need to calibrate all the other mics under test to that same calibrator in order to make the tests anywhere near accurate. And eyeballing the curve is not enough. But again everyone has there own opinion.

[page]

My listening room is about 7m x 4m x 2.2m and it is somewhat dead acoustically but certainly not anechoic.

Not to be harsh, but you have no idea what effect room acoustics are having on your tests.

The effects of room acoustics should be negated because the reference mic and the test mic are measured in the same room at the same positions with the same sound source.  Therefore, the room acoustics affect both mics equally and are canceled out when the reference mic curve is subtracted from the test mic curve.

How can you guarantee the same position? when the capsule sizes are all different? And they are all different.

Ok, I've watched this quietly long enough; Ok, so you are exactly in the *same* space, you're off as much as 1cm in any direction (so your field of testing expands). Fair, it's not the exact same spot, but will it make that much of a difference? I think that depends on what you're going to use the results to infer.

I still don't think you have addressed Jon's bit of what meaningful accuracy and tolerances are, but maybe I just missed that.

edit: jon beat me to the post.

[Church-Audio]

My listening room is about 7m x 4m x 2.2m and it is somewhat dead acoustically but certainly not anechoic.

Not to be harsh, but you have no idea what effect room acoustics are having on your tests.

The effects of room acoustics should be negated because the reference mic and the test mic are measured in the same room at the same positions with the same sound source.  Therefore, the room acoustics affect both mics equally and are canceled out when the reference mic curve is subtracted from the test mic curve.

How can you guarantee the same position? when the capsule sizes are all different? And they are all different.

Ok, I've watched this quietly long enough; Ok, so you are exactly in the *same* space, you're off as much as 1cm in any direction (so your field of testing expands). Fair, it's not the exact same spot, but will it make that much of a difference? I think that depends on what you're going to use the results to infer.

I still don't think you have addressed Jon's bit of what meaningful accuracy and tolerances are, but maybe I just missed that.

edit: jon beat me to the post.

The bottom line is when you have different size capsules you have a huge problem with measuring frequency response because you don't really have a meaningful reference point. That's the really big part of the problem. How accurate are these tests its impossible to say since I did not do them but I can tell you that there will be huge problems because of the method being used. + - 40% Maybe more maybe less. The problem gets further complicated by the fact that this is not a true point source. And yes moving the mic 1 cm  can make a huge difference. When you are talking about measurement. I just wanted to make that point.. I did not intend for this to get all blown up like it has.

[runonce]

I didn't say I did, I said it was second gen.  That means it was a test done in comparison to a NIST (caps please, it's an acronym) calibrated mic.  Therefore, the margin of error increases the farther you get from that calibrated mic.  All of this I said up front.  Strictly speaking, I don't need a NIST-calibrated mic because I don't calibrate my own reference mic, I send it off to a lab.  Thus, I only really need a calibrated mic to validate my own test methods.

You still refuse to estimate the margin of error of my tests or notlance's.  Your position seems to be only you can measure frequency response because you say so.  Here is your problem:  if I can't measure frequency response, then neither can you.  If you can't, why do you have a calibrated microphone?  Instead, you need to send your mics to the test lab rather than doing your own test.  This I have done.

Why don't you call up Herb Singleton at Cross-Spectrum Labs (not a NIST lab, yes), and explain to him why his methods are complete bunk, because you know better?  Funny, but I think they were good enough for Crowley & Tripp.  Have you sold a mic patent to Shure?  I shure  haven't.

You also haven't raised a serious objection to the outdoor test.  That tells me you have probably never tried it.  It wasn't my idea, I got it (along with most of my other good ideas) from Harvey Gerst.  He's designed a microphone and speaker or two.  You could ask Mr. Satz for a character reference there, I believe they are acquainted.

Again, margin of error.  No, notlance won't hit +/-0.1dB.  Who cares?  What accuracy did he achieve?  What exact are his errors?  Specifics, please, not innuendo.

Your point about sensitivity is absurd.  The reference mic has a measured sensitivity . . . but even if it didn't, who cares?  So long as the chain does not change between the reference mic and the mic under test, you only need to compare sensitivities and normalize 1kHz.  The exception would be for mics with unreasonably high output impedance or complex output impedance.  Or mics that aren't phantom powered (or just professional-style XLR output mics).  You will have that issue because the sensitivity of plug-in power mics is a function of the supply resistance.  Even so, it doesn't really matter, just normalize response at 1kHz and state the test conditions.

I've been doing some simple tests tonight, results to follow shortly.

I was going to go through each point line by line but I give up! I don't have enough spare time to explain / educate you. I will say one thing you are taking this very personally. I was just making a statement that these tests were not 100% accurate and should not be taken as such. Talk to anyone that measures microphones for a living they will tell you the same thing there are way to many variables with the way these mics are being placed / tested. I don't consider Cross spectrum labs to be a good source of info.. Try talking to the guys as Nist or NRC like I have. And maybe you will get a better idea of whats involved.

It seems like you are trying to mischaracterize his tests on your own terms to discredit them, rather than accept them for what they are. I cant see why/how anyone would read his post and methodology - and think the goal was "100% accuracy"

Pretty sure this was not set up to be about "accuracy"...but rather a loose comparison.

My only observation is - for this loose comparison, one could use any mic for "the standard" and make all comparisons to that mic...and still garner some useful info. I think the OP just picked his best documented, flattest "known" mic...

His tests seem to emphasize the differences between the mics...not make reference-quality measurements. Those differences will always be the same, regardless of the comparison mic.

And despite your claims to the contrary, you sound more like the person, taking something personally, and making it a pissing contest...

Try talking to the guys as Nist or NRC like I have. And maybe you will get a better idea of whats involved...That's why I spend $150 a year getting my reference mics calibrated...You don't know so you don't have any point of real reference that's why I spend $1000+ for a proper nist traceable analyzer mic and spend $150 a year to get it calibrated not to mention the $500 calibrator that I use to calibrate the sensitivity to my preamp / program.

sounds like a pissing contest, Chris...

[page]

But 1cm is really pretty poor placement accuracy.  I think anyone could do better than that if they tried.

I agree it is. My intention was primarily to provoke the response of "how much difference does it make", not that it's difficult to achieve mic placement tollerances that are tighter.

Next, there is an even easier way to measure placement error:  measure a mic, remove it from the stand, put it back in the stand, and measure it again.  Without moving it, measure it a third time.  Now we know our placement error and also our overall resolution.

precisely (no pun intended). 

[Massive Dynamic]

This is all the data I have for now.  I do have a few more mics I plan on testing, in particular a Studio Projects LSD2 which I forgot about, but I need to attend to the rest of my life for a while.  It will get done when it gets done.

I'd be interested in seeing graphs of all the LSD2 possibilities when you find the time. Thanks.

[illconditioned]

Enhancements.  Get a calibrated mic.  I could also do one mic right after another, being careful to put both mics in exactly the same position.

I think the biggest problem is the room response, which can vary quite a bit with position.  The second is the speaker response, which again can vary quite a bit.

Anyway, this kind of "sanity check" is good enough for small (5mm) omni mics.  It would be a good way to match a dozen WM61a capsules, for example.

  Richard

That's an entirely valid way to compare small omni capsules.  The speaker and room response aren't too important, since they will be the same for all units under test, although it does help to be diligent about identical placement.

Treating the room and doing a few helpful things with the drivers can make the charts look prettier insofar as they will look flatter, but the spiky kind of graphs you get without controls are valid for subtraction.

Yeah, I'm just suggesting this as I discovered it by chance.  My *guess* is that a lot of people (like Coresound, etc) are using this technique.  I don't imagine all of these guys are as obsessive about testing as Chris Church .

  Richard

Who is obsessive ? 

OK, you got me there!!!

  Richard

[Church-Audio]

I didn't say I did, I said it was second gen.  That means it was a test done in comparison to a NIST (caps please, it's an acronym) calibrated mic.  Therefore, the margin of error increases the farther you get from that calibrated mic.  All of this I said up front.  Strictly speaking, I don't need a NIST-calibrated mic because I don't calibrate my own reference mic, I send it off to a lab.  Thus, I only really need a calibrated mic to validate my own test methods.

You still refuse to estimate the margin of error of my tests or notlance's.  Your position seems to be only you can measure frequency response because you say so.  Here is your problem:  if I can't measure frequency response, then neither can you.  If you can't, why do you have a calibrated microphone?  Instead, you need to send your mics to the test lab rather than doing your own test.  This I have done.

Why don't you call up Herb Singleton at Cross-Spectrum Labs (not a NIST lab, yes), and explain to him why his methods are complete bunk, because you know better?  Funny, but I think they were good enough for Crowley & Tripp.  Have you sold a mic patent to Shure?  I shure  haven't.

You also haven't raised a serious objection to the outdoor test.  That tells me you have probably never tried it.  It wasn't my idea, I got it (along with most of my other good ideas) from Harvey Gerst.  He's designed a microphone and speaker or two.  You could ask Mr. Satz for a character reference there, I believe they are acquainted.

Again, margin of error.  No, notlance won't hit +/-0.1dB.  Who cares?  What accuracy did he achieve?  What exact are his errors?  Specifics, please, not innuendo.

Your point about sensitivity is absurd.  The reference mic has a measured sensitivity . . . but even if it didn't, who cares?  So long as the chain does not change between the reference mic and the mic under test, you only need to compare sensitivities and normalize 1kHz.  The exception would be for mics with unreasonably high output impedance or complex output impedance.  Or mics that aren't phantom powered (or just professional-style XLR output mics).  You will have that issue because the sensitivity of plug-in power mics is a function of the supply resistance.  Even so, it doesn't really matter, just normalize response at 1kHz and state the test conditions.

I've been doing some simple tests tonight, results to follow shortly.

I was going to go through each point line by line but I give up! I don't have enough spare time to explain / educate you. I will say one thing you are taking this very personally. I was just making a statement that these tests were not 100% accurate and should not be taken as such. Talk to anyone that measures microphones for a living they will tell you the same thing there are way to many variables with the way these mics are being placed / tested. I don't consider Cross spectrum labs to be a good source of info.. Try talking to the guys as Nist or NRC like I have. And maybe you will get a better idea of whats involved.

It seems like you are trying to mischaracterize his tests on your own terms to discredit them, rather than accept them for what they are. I cant see why/how anyone would read his post and methodology - and think the goal was "100% accuracy"

Pretty sure this was not set up to be about "accuracy"...but rather a loose comparison.

My only observation is - for this loose comparison, one could use any mic for "the standard" and make all comparisons to that mic...and still garner some useful info. I think the OP just picked his best documented, flattest "known" mic...

His tests seem to emphasize the differences between the mics...not make reference-quality measurements. Those differences will always be the same, regardless of the comparison mic.

And despite your claims to the contrary, you sound more like the person, taking something personally, and making it a pissing contest...

Try talking to the guys as Nist or NRC like I have. And maybe you will get a better idea of whats involved...That's why I spend $150 a year getting my reference mics calibrated...You don't know so you don't have any point of real reference that's why I spend $1000+ for a proper nist traceable analyzer mic and spend $150 a year to get it calibrated not to mention the $500 calibrator that I use to calibrate the sensitivity to my preamp / program.

sounds like a pissing contest, Chris...

And that's why I removed the post.. I realized that it was a total waste of my time and a bit to harsh... And I did not want to really get into it with anyone but again I can see things got out of hand. I can see the post police are watching my every move Anyway I will leave you all to it.

[page]

I can see the post police are watching my every move

[Church-Audio]

I want to apologize to everyone that I have offended. I really should not have said anything, I am sorry. Its not a big deal if the tests are not 100% perfect very few tests in reality are. Anyway that's all I wanted to say. I would still like to be in the circle of trust.   

[page]

I want to apologize to everyone that I have offended. I really should not have said anything, I am sorry. Its not a big deal if the tests are not 100% perfect very few tests in reality are. Anyway that's all I wanted to say. I would still like to be in the circle of trust.   

haha, things like this don't bother me as long as people explain things and are civil. My only point to try and take you to ask was on the metric issue and you finally spoke on that point so my checklist was resolved.

no worries (to me at any rate), we've done worse here...

[kirk97132]

Chris is there any chance you will publish the graphs for your mics?  Ya know now that we are all in the circle of trust and all.

[Church-Audio]

Chris is there any chance you will publish the graphs for your mics?  Ya know now that we are all in the circle of trust and all.

lol I am really flattered, I am in the circle of trust. I am actually still working on publishing graphs of my mics.. Because I do want to get it right I have posted a few graphs in the past. I don't want to mislead anyone by publishing things that are not 100% correct. I am actually going to build a test chamber that will be good down to 100hz. I am also going to be taking my mics to an independent lab the NRC. To get individual mic response tests done but its very expensive almost $3,000 per session.

[illconditioned]

I seem to remember an article where the person ran the mics not only outside, but also on a tower they constructed high over the ground to get near perfect acoustics!  I can just see some geeks like us climbing a tower somewhere late at night...

  Richard

[Church-Audio]

I seem to remember an article where the person ran the mics not only outside, but also on a tower they constructed high over the ground to get near perfect acoustics!  I can just see some geeks like us climbing a tower somewhere late at night...

  Richard

The best way to test is with the speaker on the ground facing up.. The mic can be suspended at least 10 feet from the ground.. The speaker needs to be baffled. So that there are no wrap around reflections allowed to hit the ground.. Or a tower that is at least 10 feet off the ground and as small of a foot print as possible to avoid reelections. When I worked for a speaker company here in Canada as a lab technician measuring loudspeakers. We had an arm that was attached to the side of the building and it was about 15 feet off the ground x 40 feet long to avoid reflections.. The huge problem with outside tests is WIND.... There is no getting around the wind issue. It effects your measurements big time.. If you look on the earthworks website they have an interesting paper on how they measure the mics they sell.. Pretty interesting.. They are using the spark gap method and speakers.