• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

inconsistent results from two Umik-2

Gruesome

Active Member
Forum Donor
Joined
Apr 21, 2023
Messages
177
Likes
181
Location
California
If anybody else has access to two Umik-2's, it would be great if they could do a quick comparison measurement, i.e. setting up REW, a microphone stand and a speaker, and just swapping out the Umik-2 and the calibration. I'm getting 2 to 3 dB discrepancies, with one mic having a tilted response with respect to the other, i.e. 2dB higher below 200Hz and 3dB lower above 2kHz than the other. Not so good for room equalization. I posted on the miniDSP forum, but so far nobody seems to be able to do the equivalent test.
I'm not really expecting anything better than 0.5dB or thereabouts, but 2-3dB discrepancy seems a lot. The results (SPL vs frequency curves) are very reproducible for each mic itself between removal and remounting on the microphone stand, the microphones (with calibrations) just don't agree with each other. 108 and 111 are one Umik-2, 109 and 110 the other:
comparison 4180 2dB higher below 200Hz and 3dB lower above 2kHz.jpg



The agreement between one of the Umik-2's and the microphones built into my Surface 7 tablet seems almost better (after aligning overall sensitivity, i.e. matching SPL's at 1kHz) than the agreement between the two Umik-2's - ok, not quite, but still:
comparison Surface 7 mic (top trace) vs Umik-2 (bottom trace) - different mic positions same r...jpg

(The Umik-2 and the Surface 7 tablet were in different positions, so one should probably ignore the low frequency room mode details).
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
That looks like one or the other is wrong. Or both (yikes). Did you accidentally do something like swap the cal files?

I've used some recording omnis creating a cal file from published mic response and gotten closer correspondence between that and a Umik 1.
 
OP
Gruesome

Gruesome

Active Member
Forum Donor
Joined
Apr 21, 2023
Messages
177
Likes
181
Location
California
Yes and no: when I do swap calibration files intentionally, it makes things a bit worse, since the lower response Umik-2 has a slightly larger correction (lower sensitivity) from its calibration file. I did quite a few sanity checks, position in room, distance from speaker, opening a door, but I don't want to repost everything from the miniDSP forum.
I was just hoping somebody here might have access to a second Umik-2. It shouldn't take more than 5 minutes. It would be really helpful.
 

ROOSKIE

Major Contributor
Joined
Feb 27, 2020
Messages
1,915
Likes
3,394
Location
Minneapolis
I have two Umik-1's
They are quite close.
One is also custom calibrated by Cross spectrum labs.
All three 'versions' are pretty close to each other, though the cross spectrum is different from miniDSP's own values by a almost a couple DB in some areas(mainly very high and very low)
Try one of theirs.

They also can do a 2 and I believe you can send yours in for $60 if that makes more sense.
 
OP
Gruesome

Gruesome

Active Member
Forum Donor
Joined
Apr 21, 2023
Messages
177
Likes
181
Location
California
Thanks, Rooskie! Some deviation (say, of order 1dB) near the high and low frequency limits, where the mic calibration corrections increase to 4-5dB I would understand and accept.
But that overall 5dB slant, and 3dB systematic uncertainty pretty much from 2kHz upwards is not acceptable.

I didn't know I could send Umik-2's bought elsewhere to Cross Spectrum.
I think though I would like miniDSP to sort their calibration out first, before throwing more money at the problem.
I'm hoping they find there is some user error on my part (I sent them my REW .mdat files), or some misunderstanding between REW and the mics.
 

MikeP

Member
Joined
Apr 7, 2022
Messages
11
Likes
16
Don't count on MiniDSP to sort this thing out. I have a similar issue with Umik1 and Umik2 microphones having 3-4dB difference at the low end of the spectrum (besides other inconsistencies). I've sent all the relevant data to them (incl. REW files) and ask for support. They replied eventually, but no concrete solution or pertinent explanation was given.

It might be better to get one of your mics properly calibrated by a third party.
 

ROOSKIE

Major Contributor
Joined
Feb 27, 2020
Messages
1,915
Likes
3,394
Location
Minneapolis
Thanks, Rooskie! Some deviation (say, of order 1dB) near the high and low frequency limits, where the mic calibration corrections increase to 4-5dB I would understand and accept.
But that overall 5dB slant, and 3dB systematic uncertainty pretty much from 2kHz upwards is not acceptable.

I didn't know I could send Umik-2's bought elsewhere to Cross Spectrum.
I think though I would like miniDSP to sort their calibration out first, before throwing more money at the problem.
I'm hoping they find there is some user error on my part (I sent them my REW .mdat files), or some misunderstanding between REW and the mics.
You will save so much time just getting a pro calibration.
The standard deviation of the UMIK-1 and similar mics is +-4db and miniDSP calibration is often off the mark.
I'd suspect the 2 has the same performance.
In the end they are $80 & $195 mics that include power.
You could spend far, far, far more on more accurate oftb mics.
Once CSL calibrated you should have the equivalent of a $$ nice mic.

See post #469

See post #10

Honestly just go to Cross Spectrum...
 
OP
Gruesome

Gruesome

Active Member
Forum Donor
Joined
Apr 21, 2023
Messages
177
Likes
181
Location
California
Cross Spectrum is out of Umik-2's, does not know when they get them back in, and does not answer my email.

After finally hearing back from miniDSP support, I did some more tests, and now both microphones behave similar, and both match the old data for one of them.
I can not reproduce the old sloping shape of the frequency response in REW measurements with the other one at all. The mics are within about 1 dB, the main difference is which calibration curve gets applied.

I tried restoring the default analysis settings in REW, but I must have changed something, since the old data are not reproduceable for one of the two microphones.
So, I'm still not 100% happy, since it would be nice to know what went wrong, but I can now get reproducible data, and I hope I will be able to use my measured KEF R3 near field response as a 'calibrated source' for future measurements.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
20,524
Likes
37,057
Cross Spectrum is out of Umik-2's, does not know when they get them back in, and does not answer my email.

After finally hearing back from miniDSP support, I did some more tests, and now both microphones behave similar, and both match the old data for one of them.
I can not reproduce the old sloping shape of the frequency response in REW measurements with the other one at all. The mics are within about 1 dB, the main difference is which calibration curve gets applied.

I tried restoring the default analysis settings in REW, but I must have changed something, since the old data are not reproduceable for one of the two microphones.
So, I'm still not 100% happy, since it would be nice to know what went wrong, but I can now get reproducible data, and I hope I will be able to use my measured KEF R3 near field response as a 'calibrated source' for future measurements.
Well I've been there. Some setting or something and couldn't figure it out, but repeated later testings gave expected results. Either that or you can now claim Umik 2's have a break in/burn in time? :)
 

sam_adams

Addicted to Fun and Learning
Joined
Dec 24, 2019
Messages
976
Likes
2,368
Pics of measurements are nice and all, but, having access to the actual data usually produces better results. Can you make the .mdat file(s) available?
 
OP
Gruesome

Gruesome

Active Member
Forum Donor
Joined
Apr 21, 2023
Messages
177
Likes
181
Location
California
Pics of measurements are nice and all, but, having access to the actual data usually produces better results. Can you make the .mdat file(s) available?
Of course! I'd actually appreciate if somebody could tell me whether anything looks wrong or not recommended with the settings!
I'll attach one from 5/18, and then another new one with the previously odd microphone that now seems to behave just like the other.
The later file is I think after I tried to get the analysis settings in REW back to how they were initially (truncation of saved data after a few seconds, limit on cal file dB boost, etc.), but didn't make a difference.
You have to rename them from .pdf to .mdat after downloading (.mdat is not an allowed extension here). Unfortunately the files without data truncation, and also the one with all four measurements from 5/18, are too big.
 

Attachments

  • 4180 3.5in 130.pdf
    1.9 MB · Views: 55
  • KEF R3 5_18 111 4180 48kHz windows eff off.pdf
    1.4 MB · Views: 60

sam_adams

Addicted to Fun and Learning
Joined
Dec 24, 2019
Messages
976
Likes
2,368
Well, just a quick look, one measurement was made from 3.5in. and the other from 1m. The closer one would emphasize the tweeter measurement while the further measurement would have more contribution from the low frequency driver. Other than the fact that the measurements were of different sweep lengths and different distances, there doesn't seem to be anything inherently wrong with them—assuming that they were made with the same mic using the correct .cal files.
 
OP
Gruesome

Gruesome

Active Member
Forum Donor
Joined
Apr 21, 2023
Messages
177
Likes
181
Location
California
Sorry, that was a typo, or rather leftover from earlier measurements in the notes. All measurements are from 3.5". Now of course you have me think that that (room effect) is the only thing that immediately makes sense. But I did measurements in the order 108 -109 - 110 - 111, and the difference is between the two groups 108/111 and 109/110, and I certainly did not move the microphone stand back and forth. I'll reconstruct from my measurement notes when I actually moved from 1m to 3.5". Back in a second...
Double checked: the older file (sequence number 111) does say 3.5" in the measurement notes inside the file, and today's measurements were all from 3.5", so I must have inherited some notes and forgot to overwrite them before saving today's measurements. In any case, the 111 one is the one with the bass-heavier tilt to the response curve, which I can no longer reproduce.
 
Last edited:

JohnPM

Senior Member
Technical Expert
Joined
Apr 9, 2018
Messages
340
Likes
901
Location
UK
Very large clock adjustments in those measurements, a bit unusual.
 
OP
Gruesome

Gruesome

Active Member
Forum Donor
Joined
Apr 21, 2023
Messages
177
Likes
181
Location
California
Very large clock adjustments in those measurements, a bit unusual.
Thanks, that gives me a hint where to look further! Could that lead to incorrect window settings that mismeasure power differently at different frequencies? Because something like that would be needed to change the frequency response.
Or maybe the sampling frequency was wrong, and REW looked at a much shorter time interval than it was thinking? (I do not know how REW works in detail. I thought/think it would measure the time (number of samples) between the two timing pulses at the beginning and end of the sweep, and complain if the microphone time difference deviates significantly from what REW sent to the the output. A wrong sampling rate should make this comparison fail completely.)
 

JohnPM

Senior Member
Technical Expert
Joined
Apr 9, 2018
Messages
340
Likes
901
Location
UK
It shouldn't have any effect on frequency response or impulse response since REW has corrected it, but it suggests the source clock is less accurate than usual.
 
OP
Gruesome

Gruesome

Active Member
Forum Donor
Joined
Apr 21, 2023
Messages
177
Likes
181
Location
California
It shouldn't have any effect on frequency response or impulse response since REW has corrected it, but it suggests the source clock is less accurate than usual.
Thanks, John. So as long as I see a clock correction in the automatic comments for the measurement, I know that REW found the start and stop chirps roughly where it expected, and recorded the full time window.
It could also be my PC (tablet) clock that is less accurate, right?
 
Top Bottom