Improving measurements with a new mic

2234rew

Member
Joined
Nov 24, 2020
Messages
222
There are two separate issues when talking measurement microphones. Was the mic properly calibrated? And if using a USB mic, potential timing issues.

I made a comparison a while back between analog and USB mic. And in the case of the USB mic, one with and without Audiolense's "clock drift correction" using my triamp system and Lynx Hilo.

Analog mic:
View attachment 47849
USB mic with clock drift correction:
View attachment 47850
USB mic without clock drift correction:
View attachment 47851

Bernt's clock drift correction works very well.

There are (very) small differences in the timing between the analog mic and USB with clock drift correction. There are also frequency response differences at the extremes between all three measurements.

The analog mic shows a perfect channel match in the low frequencies and natural rolloff in the top octave. Whereas the USB mic shows a bit of low frequency mismatch between channels and a bit of a rising top end. Now whether any of that is audible...

Of course, the USB mic measurement without clock drift correction is unusable timing wise.

While the USB mic is "convenient" I sold mine and stuck with the isecmcon analog mic.
Thanks mitch

I definitely has clock drift correct enabled with UMIK-2

The measurements on the screen looked fine (step response)

It was music playback that sounded very off. Vocals that should be dead centre were heard left and right , from the speakers

I use Harry Belafonte "Dayo" live at Carnegie Hall as my centre vocal test.

With 2 different UMIK-2's it sounded so wrong in 2 different systems. After multiple days trouble shooting and repeating test in slightly different measurement positions.

M23 on first try nailed it. All done in 15 minutes
 

2234rew

Member
Joined
Nov 24, 2020
Messages
222
And I forgot to mention something really important - with my M23 I am using the small Focusurite Scarlett Solo Gen3 for ADC.

So I am using USB clock drift with M23 + Focusrite ADC , just like with UMIK-2

I had same "vocals coming from left and right speakers, not center" with 2 different UMIK-2s, in 2 different systems.

My system is very humble but the other is a $100k Magico system in a dedicated $50k room seperate to his house. Seperate building really, not just seperate room. He hardly even needs room correction based on REW measurements.

And even in that room, I ran Audiolense XO with UMIK-2 measurements, multiple times, slightly different positions (moving an inch or so each time) , 2 different UMIK-2s and each time I went to load my playlist on Qobuz for testing, track 1 is The Harry Belafonte song and every time, vocals were coming from left and right speaker.

Turn correction off, and vocals are dead center.

Run measurement and correction with M23 - vocals dead center first time.

UMIK-2 really drove me nuts for multiple weekends.

I suspect the forum comments I've seen around, that UMIK-2 is better for speaker correction than room correction (smaller capsule than UMIK-1) may have something true.

@dsnyder0cnn if Harry's vocals during verse of Dayo, Banana Boat Song (Live from Carnegie Hall version !) is dead center, then you're UMIK-2 from Cross Spectrum Labs is great and no need to change.

If vocals during verse is left+right , time to get M23 or @Mitchco's recommended isecmcon
 
Last edited:

dsnyder0cnn

Member
Joined
Nov 28, 2017
Messages
48
And I forgot to mention something really important - with my M23 I am using the small Focusurite Scarlett Solo Gen3 for ADC.

@dsnyder0cnn if Harry's vocals during verse of Dayo, Banana Boat Song (Live from Carnegie Hall version !) is dead center, then you're UMIK-2 from Cross Spectrum Labs is great and no need to change.

If vocals during verse is left+right , time to get M23 or @Mitchco's recommended isecmcon
Your setup sounds similar to mine (but much nicer mic), I I'm using a Focusrite Scarlett 2i2 gen2 with my Cross•Spectrum Labs EMM-6.

I finally had some time to take measurements with both mics this evening. Here's what I'm seeing in the amplitude domain (SPL) with psychoacoustic smoothing:

47852


47853


I'm satisfied with his match, especially since my EMM-6 is 5+ years older than the UMIK-2. Distortion levels were significantly lower with the UMIK-2 sweeps:

47854


In the time-domain, there are some differences. I'm using a Gustard X16 DAC (L-FAST filter) to play the sweeps that I recorded with the UMIK-2. I used REW's Acoustic Timing Reference for both, but alignment was not perfect…the start is off by about 590 us with visible differences in group delay over the first 20 ms.

47855


47856


Scaling out to almost 120ms, the amount of drift seems to increase (note the difference in scale). The UMIK-2 starts out 590 us ahead but ends up falling behind as time progresses:

47857


47858


I mean, they are not radically different. I did my best to place both mics in exactly the same location, but the differences could be, at least in part, due to less than perfect mic placement consistency.

I've not attempted to record any sweeps using the UMIK-2 in Acourate. I've heard that it's possible, but I'm not interested in messing around with ASIO4ALL or other nonsense. Now that I see that my two Cross•Spectrum Labs mics agree within reasonable margin of error, I'll stick with @Mitchco 's advice. It's the EMM-6 + Focusrite for sweeps that I use for building filters and the UMIK-2 for analysis and verification.
 

2234rew

Member
Joined
Nov 24, 2020
Messages
222
Your setup sounds similar to mine (but much nicer mic), I I'm using a Focusrite Scarlett 2i2 gen2 with my Cross•Spectrum Labs EMM-6.

I finally had some time to take measurements with both mics this evening. Here's what I'm seeing in the amplitude domain (SPL) with psychoacoustic smoothing:

View attachment 47852

View attachment 47853

I'm satisfied with his match, especially since my EMM-6 is 5+ years older than the UMIK-2. Distortion levels were significantly lower with the UMIK-2 sweeps:

View attachment 47854

In the time-domain, there are some differences. I'm using a Gustard X16 DAC (L-FAST filter) to play the sweeps that I recorded with the UMIK-2. I used REW's Acoustic Timing Reference for both, but alignment was not perfect…the start is off by about 590 us with visible differences in group delay over the first 20 ms.

View attachment 47855

View attachment 47856

Scaling out to almost 120ms, the amount of drift seems to increase (note the difference in scale). The UMIK-2 starts out 590 us ahead but ends up falling behind as time progresses:

View attachment 47857

View attachment 47858

I mean, they are not radically different. I did my best to place both mics in exactly the same location, but the differences could be, at least in part, due to less than perfect mic placement consistency.

I've not attempted to record any sweeps using the UMIK-2 in Acourate. I've heard that it's possible, but I'm not interested in messing around with ASIO4ALL or other nonsense. Now that I see that my two Cross•Spectrum Labs mics agree within reasonable margin of error, I'll stick with @Mitchco 's advice. It's the EMM-6 + Focusrite for sweeps that I use for building filters and the UMIK-2 for analysis and verification.
Thanks for sharing.

As noted, nothing looked out of place in my simulations with Audiolense XO

I have to admit I didn't do verification measurements with REW with the UMIK-2

The AL XO simulations looked good. It was the listening that sounded horrible - as mentioned, the vocal was coming from left and right speaker, instead of dead center.

It's possible that REW verification measurements and phase plots would show a problem with my UMIK-2 (maybe, maybe not) but I have trusted AL XO simulations.

Everything sounded correct with my M23 + Focusrite Solo Gen3 (on first attempt) so UMIK-2s were sent packin :-)

With M23 I have done REW verification measurements and been very happy. It's just that I didn't do any with the UMIK-2s . They caused me too much lost time and heartache. I was happy to send them away

How do the 2 mic measurements actually sound @dsnyder0cnn ? Playing the live track I mentioned, they both sound the same? With the same correction? Vocals center?
 

dsnyder0cnn

Member
Joined
Nov 28, 2017
Messages
48
Thanks for sharing.

As noted, nothing looked out of place in my simulations with Audiolense XO

I have to admit I didn't do verification measurements with REW with the UMIK-2

The AL XO simulations looked good. It was the listening that sounded horrible - as mentioned, the vocal was coming from left and right speaker, instead of dead center.

It's possible that REW verification measurements and phase plots would show a problem with my UMIK-2 (maybe, maybe not) but I have trusted AL XO simulations.

Everything sounded correct with my M23 + Focusrite Solo Gen3 (on first attempt) so UMIK-2s were sent packin :-)

With M23 I have done REW verification measurements and been very happy. It's just that I didn't do any with the UMIK-2s . They caused me too much lost time and heartache. I was happy to send them away

How do the 2 mic measurements actually sound @dsnyder0cnn ? Playing the live track I mentioned, they both sound the same? With the same correction? Vocals center?
As you've found, the quality of the final results depends on not just your filter design work but also the quality of the recorded sweeps. That's why I always do end-to-end measurements in REW to verify the results. I'm a Roon subscriber, so for me, that means playing pre-recorded left and right sweep files in Roon using the same DAC that I use for playing music and measuring the output using REW and my best mic, which now it seems is the UMIK-2.

I've not attempted to build filters using measurements from the UMIK-2. If I decide to give it a go, I'll probably just import the measurements I took using REW into Acourate rather than attempting to record in Acourate due to lack of ASIO drivers. Thanks for sharing your reference track for testing. I'll let you know if vocals fail to be centered with either one. :)

I'll test with this track: https://tidal.com/browse/track/25807180
 

2234rew

Member
Joined
Nov 24, 2020
Messages
222
Thanks for sharing your reference track for testing. I'll let you know if vocals fail to be centered with either one. :)

I'll test with this track: https://tidal.com/browse/track/25807180
Thanks, this is the correct Tidal album link (Track 7):


Whenever I do room correction now, this is the first thing I listen to, to make sure vocal is centre.

If that is wrong to start with (vocals coming from the left and right speakers separately) I don't both listening anymore. Back to the drawing board. Luckily haven't had that issue since getting M23.

47874
 
Top Bottom