Inconsistent measurements with Umik-2

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
Hi,

I have purchased a Umik-2. I also own a Umik-1 (which is defect), and made 100's, if not 1000's of measurements on the speakers I am developing with the Umik-1.

With the Umik-2, I have not been able to produce a consistent and correct measurement. I run on a Mac (REW V5.20.3 - JRE 1.8.0_302 64-bit on OS X 10.12.6). When selecting 48khz or 96khz, each measurement looks different (as in all over the place), the impulse reponse is a mess and my measurement (on the same, unchanged speaker) is different every time, even if I don't change any settings.

I have replaced pretty much everything in my setup to exclude any possible causes, new cables, other laptop etc etc. I cannot create a single consistent measurement.

Are there known issues with the UMIK-2? Can anyone provide any help or insight? I am pretty stuck at the moment, not being able to use my favorite measurement tool.

Any pointers or help to troubleshoot this is appreciated.

Stefan
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
5,023
What are the UMIK-2 settings in Audio Midi Setup? Sample rate should be set the same as you are using in REW.
 

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
Yes. I tried 48, 96 and 192 khz and ensured settings match with Midi setup.
 

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
I have had the Mic checked by the supplier. It seemed to measure fine on their setup. (their setup is REW on windows, I run REW on my Mac).

I did some more tests this morning and I think i found the root cause. It looks like there is a correlation with the length (end frequentie) of the sweep.

Here is what i did:

1. Setup my MIDI settings (both Umik-2 and Output device) for 48KHZ.
2. Opened up REW and made a 48 kHz measurement where the sweep was 200hz to 20khz. (I am only measuring my tweeter here):

47578


3. Changed the sweep range from 200hz to 24khz and measured again:
47579


As you can see the measurement is wrong.

4. Closed REW (saved measurements of course)
5. Changed MIDI settings to 96khz
6. Measurement with sweep from 200hz to 25khz:
47580


7. Changed Sweep 200hz to 30khz:
47581


Response incorrect from 22khz.


It seems the sweep window has an impact on how REW processes the responses. Hopefully this helps to narrow down the problem? I added the MDAT file to this post.

Cheers,

Stefan
 

Attachments

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
5,023
That looks like an effect of variable latency in the chain. You could try changing (probably reducing) the buffer size settings in the REW soundcard preferences, but the most reliable solution would be to enable the acoustic timing reference. It is also best to avoid using 0 dBFS sweeps as that doesn't leave any headroom in the digital path.
 

sm52

Member
Joined
Mar 14, 2019
Messages
381
If you change the polarity of the impulse at 96 kHz, they will become similar to the measurements at 48 kHz. Graph 2 is different from everyone else. Maybe something went wrong during the measurement.
 

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
Hello,

i have reduced the buffer-sizes, set measurement to -3 dBFS and used the acoustic ref option. There is no difference. I was not able to produce consistent measurements. See MDAT file attached and example below. A Umik-1 measures fine btw.

47590


Stefan
 

Attachments

sm52

Member
Joined
Mar 14, 2019
Messages
381
The second episode is completely unsuccessful. I took your mdat from post # 8, set t = 0 correctly. Of the six graphs, four are the same. Those. normal. You can use them. Disable graph 2 and 6.
 

Attachments

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
The second episode is completely unsuccessful. I took your mdat from post # 8, set t = 0 correctly. Of the six graphs, four are the same. Those. normal. You can use them. Disable graph 2 and 6.
Thank you, and yes the 1st episode was better. But I cannot trust my measurement system and that is a big problem of course. I use REW to analyse the design of my speaker. I design these in FreeCAD and get them professionally printed. This is very costly. I need to ensure that the measurements to my dummies are 100% accurate before i have them professionally printed.

I appreciated your help and support!

Stefan
 

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
Then you need to spend some money(~200$) and get an external sound card with two inputs and two outputs.
It works just fine with the Umik-1. Just not with the Umik-2. I am happy to spend the money needed - that is not the issue.
 

sam_adams

Member
Joined
May 23, 2017
Messages
164
Can you post screenshots of your Audio MIDI Setup.app settings for the output device and the Umik-2 input?
 

John Mulcahy

REW Author
Joined
Apr 3, 2017
Messages
5,023
I'd say you are getting dropouts in the data from the mic, though I can't say why that might be, whether some software issue or something hardware related, like an intermittent connection. It isn't easy to see the dropouts in your plots as you have a very small left window width, increasing the left window to 10 ms shows this plot, where the glitches from the dropouts are easier to see:

47762
 

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
Thank you John - i'll do some tests with cables etc. The cable is obvisouly different from the Umik-1 as it's a USB-C connection cable. I purchased an extra long one, as the one that comes with the Umik-2 is too short for decent measurement of my speakers in the setup i use.

I think i tried another one the other day, but i'll doublecheck. Thanks for your support.

Stefan
 

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
I just made 2 audio recordings using the Umik-2. One with the original USBC cable that comes with the Umik-2. Another one with a longer USBC cable i use for measurements. The recording made with the longer USBC cable has static & tics. It is very audible when I play back the recording. The shorter cable does not have that.

The random results in the measurements might well be down to a cable issue. I will be experimenting a bit more and make some measurements with the shorter cable later today.

It also could explain why the Umik-1 works fine. It has a different USB connector, hence I use a different cable for it.

I certainly hope to put this issue to bed soon.

Stefan
 

satefan

Registered
Thread Starter
Joined
Nov 24, 2019
Messages
21
Okay - i made an experimental setup so i could measure with the UMIK-2 and the original short cable. It compromises my setup a bit, but good enough for testing.

With the short USBC cable, i get consistent measurements without the drop. The measurement below is without any filters active. The little wrinke you see at 27khz is the membrane of the Bliesma Berillium tweeter breaking up just a bit. This is also visible in the datasheet, so not a problem at all.
47803



My longer USBC cable was brand new, but still crap. I'll be getting a good one in the next few days and this problem should be over with.

Thanks for all help and support.

Satefan
 
Last edited:

skid00

Member
Joined
Mar 20, 2020
Messages
57
My AV System  
Main Amp
2 x Adcom GFA 555
Front Speakers
Carver Amazing Platinum Mark IV
USB standard IIRC, is 15 foot long cables, max. I had a 20 foot long cable not drive the DAC at all on a new laptop, but would work on an older laptop.
 

jschwender

Member
Joined
Mar 16, 2021
Messages
101
My AV System  
Front Speakers
Nubert digital pro
Other Speakers or Equipment
Philipd dss940
How long is your working cable, please? Brand?
 
Top Bottom