• 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!

UMIK-1 or any USB MIC unsuitable for timing measurements?

gnarly

Major Contributor
Joined
Jun 15, 2021
Messages
1,071
Likes
1,516
I have both a USB and 2 channel mic for measurements. I go to the USB one for convenience most of the time.

The issue with USB is you have no reference timing marker for measuring the "flight time" for the impulse, which will give you absolute phase information between different driver measurements providing for reliable phase integration for crossover modeling.

However you can work around this for USB measurements in a multi-driver system via the parallel driver approach and relying on drivers essentially being minimum phase devices (only works when using raw driver measurements as any components in the signal path may alter minimum phase behaviour).

The approach is as follows:
1. Measure tweeter
2. Measure woofer
3. Measure both tweeter and woofer in parallel

Make sure you don't change drive levels nor mic / baffle placement. i.e. ideally you will have both driver terminals exposed and just clip to the next driver to measure

4. Load measurement #3 into your simulator
5. Load both driver responses into your simulator and wire them up in parallel to the source / amp. Ensure you check minimum phase (derived phase) or extract this outside the simulator and load with the driver response
6. Increase woofer "Z" (acoustic offset) until the summed #5 response, aligns to combined measurement response (you loaded into step #4)

You now have a reliable Z offset for crossover simulation.

The downside is this is a single axis technique as Z will change off axis (the amount depends on the drivers / placement).

It would be nice it it worked that way.

Again THE problem is if you are using a USB mic with acoustic timing reference...that again is a 5KHx up sweep...phase and resultant timing delay estimation is bogus for any driver not playing into tweeter-ville.

Try this... take transfer functions of the driver sections with hp or lp in place, above or below the intended pass band, as fits the driver.
Watch how the phase curve changes, and estimated peak IR time changes.

If you can, use linear-phase hp and lp. And let the light shine :)
 

mtg90

Member
Joined
May 24, 2021
Messages
56
Likes
147
Location
Illinois
I have both a USB and 2 channel mic for measurements. I go to the USB one for convenience most of the time.

The issue with USB is you have no reference timing marker for measuring the "flight time" for the impulse, which will give you absolute phase information between different driver measurements providing for reliable phase integration for crossover modeling.

However you can work around this for USB measurements in a multi-driver system via the parallel driver approach and relying on drivers essentially being minimum phase devices (only works when using raw driver measurements as any components in the signal path may alter minimum phase behaviour).

The approach is as follows:
1. Measure tweeter
2. Measure woofer
3. Measure both tweeter and woofer in parallel

Make sure you don't change drive levels nor mic / baffle placement. i.e. ideally you will have both driver terminals exposed and just clip to the next driver to measure

4. Load measurement #3 into your simulator
5. Load both driver responses into your simulator and wire them up in parallel to the source / amp. Ensure you check minimum phase (derived phase) or extract this outside the simulator and load with the driver response
6. Increase woofer "Z" (acoustic offset) until the summed #5 response, aligns to combined measurement response (you loaded into step #4)

You now have a reliable Z offset for crossover simulation.

The downside is this is a single axis technique as Z will change off axis (the amount depends on the drivers / placement).

This is how I did it for many years using my UMIK-1 when working with Xsim and PCD before that. It does works well enough to get by on a single axis and can even be used for 3-ways if you measure each pair separately HF/MF and MF/LF one at a time and implement the right delays in software to match the summed response starting with the midrange first then woofer second. I personally found it was not necessary to use derived phase in my measurements, the measured phase was good enough once offset corrected to use over the crossover region.

That said with the advent of Vituixcad and being able to sim with off axis data I finally switched to an external interface and XLR mic a few years back and let me tell you being able to just import my measurements and start simulation without worrying about manually correcting the timing/Z offset for each new measurement is a huge convenience.

If someone is looking to do just simple room or FR correction measurements of their completed speaker/subs then a USB mic is fine. For actual speaker design work I strongly recommend a 2+ channel interface and calibrated XLR mic, the workarounds with USB mics just aren't worth the hassle.
 
Last edited:

kimmosto

Active Member
Joined
Sep 12, 2019
Messages
226
Likes
537
From what I have seen, am sure others have designed speakers without driver phase data but if you want SOTA simulation, you need it to be accurate.

Since @kimmosto wrote the USB mic warning, am sure he can articulate the issue more accurately than I.
I stopped arguing years ago. People are very good justifying why they have acquired Umik-1 or similar, continue to use it and repeatedly recommend it even to those who have no idea what limitations and difficulties they have to face. Some have had a commercial motive, although they talk about being solution-oriented while writing instructions that seem like a desperate struggle how you can get USB microphone to use time reference if you're lucky and remember to do the same tricks every time and no setting in the system has changed. The price difference to calibrated analog is practically zero, and also USB microphones needs signal output, i.e. a sound card.
In the measurement instructions I wrote, USB microphones are actually prohibited. One version (ARTA) promises that single-channel mode can be used for near field measurements, but that is also a mistake I will correct some day.
 

Rick Sykora

Major Contributor
Forum Donor
Joined
Jan 14, 2020
Messages
3,723
Likes
7,509
Location
Stow, Ohio USA
I stopped arguing years ago. People are very good justifying why they have acquired Umik-1 or similar, continue to use it and repeatedly recommend it even to those who have no idea what limitations and difficulties they have to face. Some have had a commercial motive, although they talk about being solution-oriented while writing instructions that seem like a desperate struggle how you can get USB microphone to use time reference if you're lucky and remember to do the same tricks every time and no setting in the system has changed. The price difference to calibrated analog is practically zero, and also USB microphones needs signal output, i.e. a sound card.
In the measurement instructions I wrote, USB microphones are actually prohibited. One version (ARTA) promises that single-channel mode can be used for near field measurements, but that is also a mistake I will correct some day.

Thanks Kimmo!

Aside from those who refuse to acknowledge the shortcoming, I think @JohnPM is mainly to blame. REW as a tool has evolved from room measuring to measuring speakers to measuring drivers. Shame on John for continuing to enhance REW! ;) Those who bought USB mics for room measurement and perhaps speaker measurement often struggle to understand why they need a different mic setup for design work. I have found some still trying to measure speakers with the mic pointed at the ceiling! Others that did not know they need to change the calibration file based on the mic orientation.

I will echo others that the effort is simpler and VituixCAD makes for powerful simulations using an XLR mic. Thanks again for all your efforts!:)
 

kimmosto

Active Member
Joined
Sep 12, 2019
Messages
226
Likes
537
Few years ago I offered to buy all Umik-1 mics from Finland and trash them to avoid continuous and frustrating discussion about problems related to USB mics in design work with XO simulator. I also proposed miniDSP to upgrade their mics to have chip with 2-channel IO, signal output and internal loopback to feed output signal to input and at the same time with mic signal. But of course, they didn't understand a bit about speaker designing so they proposed that could I add acoustical reference to simulator. Right... My conclusion was that struggling against global stupidity is not my problem anymore. I still hope that e.g. JohnPM would add disclaimer (instead of advertising) that Umik-1 is limited and should not be purchased and used for design work with XO simulator.
 

dcibel

Member
Joined
Feb 8, 2021
Messages
47
Likes
53
I'd actually blame Jeff Bagby and Dayton Omnimic for the USB mic problem more than I would John. In John's defense, REW wasn't purpose build as a loudspeaker measurement suite, but rather aimed at room EQ, it's just evolved over the years. I am glad that John was receptive to including full 2-channel capabilities in REW in recent versions, I am left only with the complaint of how phase is treated always relative to t=0, not the window reference time, but it doesn't seem that will ever change.

In any case, John has recently posted some information relating to the timing capabilities of UMIK-2 at the REW forum, so perhaps there is some hope for USB users yet ;).

Until some proper USB all-in-one solution exists, for loudspeaker measurements for crossover design, 2-channel system should be the only recommendation. As Kimmo mentions, a USB mic with proper timing reference can be easy, costing no more than the existing designs, just with small circuit modification to utilize a loopback on a 2ch USB audio codec.

The J-box is a close all-in-one solution, if it included XLR mic connection with 48V power it could be perfect.

On 1ch vs 2-ch measurement, I have attempted to capture the benefits of 2-channel measurements in this article, with VituixCAD in mind of course:

I have also detailed a simple 2-channel measurement jig here, however for basic speaker measurement you can get started with just a patch cable from output to input on your audio interface or soundcard.
 

dcibel

Member
Joined
Feb 8, 2021
Messages
47
Likes
53
I have both a USB and 2 channel mic for measurements. I go to the USB one for convenience most of the time.

The issue with USB is you have no reference timing marker for measuring the "flight time" for the impulse, which will give you absolute phase information between different driver measurements providing for reliable phase integration for crossover modeling.

However you can work around this for USB measurements in a multi-driver system via the parallel driver approach and relying on drivers essentially being minimum phase devices (only works when using raw driver measurements as any components in the signal path may alter minimum phase behaviour).

The approach is as follows:
1. Measure tweeter
2. Measure woofer
3. Measure both tweeter and woofer in parallel

Make sure you don't change drive levels nor mic / baffle placement. i.e. ideally you will have both driver terminals exposed and just clip to the next driver to measure

4. Load measurement #3 into your simulator
5. Load both driver responses into your simulator and wire them up in parallel to the source / amp. Ensure you check minimum phase (derived phase) or extract this outside the simulator and load with the driver response
6. Increase woofer "Z" (acoustic offset) until the summed #5 response, aligns to combined measurement response (you loaded into step #4)

You now have a reliable Z offset for crossover simulation.

The downside is this is a single axis technique as Z will change off axis (the amount depends on the drivers / placement).
Convenience? All those steps just to capture relative delay and major downside mentioned at the end sounds rather inconvenient to me. Not to mention crossover simulation with this method is locked to measurement distance, not actual listening distance. Maybe okay for small bookshelf speakers, but anything large will be problematic. I admit that I started into this hobby years back with an Omnimic as well following the Jeff Bagby mantra, but have since moved on years ago and never looked back. The Omnimic is just an SPL meter now, mostly collecting dust in the closet.

In any case, if you must stick with USB for convenience...you can skip a few steps. Re-calculating min-phase through HBT process simply isn't necessary, and often the primary source of error in this process. Use the measurements as-is. Just go to VituixCAD Tools Menu -> Auxiliary -> Time Align. Load your 3 measurements, un-check MP (min-phase), solve delay. Enter delay value in driver tab of main window, leave driver coordinates at 0,0,0 in the crossover.

But, with all the work that goes into a DIY speaker, why take a shortcut on arguably the most important part? Take a bit more time and gather the full spatial set of measurement data with 2-channel system properly.
 

Rick Sykora

Major Contributor
Forum Donor
Joined
Jan 14, 2020
Messages
3,723
Likes
7,509
Location
Stow, Ohio USA
Few years ago I offered to buy all Umik-1 mics from Finland and trash them to avoid continuous and frustrating discussion about problems related to USB mics in design work with XO simulator. I also proposed miniDSP to upgrade their mics to have chip with 2-channel IO, signal output and internal loopback to feed output signal to input and at the same time with mic signal. But of course, they didn't understand a bit about speaker designing so they proposed that could I add acoustical reference to simulator. Right... My conclusion was that struggling against global stupidity is not my problem anymore. I still hope that e.g. JohnPM would add disclaimer (instead of advertising) that Umik-1 is limited and should not be purchased and used for design work with XO simulator.

Lol, I get that the repetitive questioning and second guessing can be annoying. By part of @dcibel post above looks as though minidsp took your advice and improved their Umik. Agree some just need to simply follow your advice!

As I mentioned with REW, you continue to improve VituixCAD. For some who do not understand the system aspect, I suspect they think software will simply fix (eventually). Serious speaker design is complex and can understand why some seek aa easy answer. In recent experience, even professional systems can be plagued with reflection, gain and other issues. Even the more experienced engineer may not readily identify when "noise" issues are skewing results versus driver or other design characteristics. Those who spend serious money for Klippel systems encounter can mic and software issues. However, they can turn to Klippel and get support and/or obtain reference baselines.

If hobbyists had a known good baseline, think this would help. Even if just a set of test signals for a fixed test configuration could be used to illustrate more valid versus invalid measurement results. Maybe someone has done this already? Know would not have completed Directiva r1 without @ctrl's experience identifying test fixture issues during the design effort. :)
 
Last edited:

dcibel

Member
Joined
Feb 8, 2021
Messages
47
Likes
53
What sort of baseline measurements do you have in mind? I am writing an article in progress specifically for data validation, perhaps it's a good place to incorporate this information. The document is waiting for my next speaker build so I can populate it with real world examples.

Data validation seems to be an often missed step in DIY efforts, so I am assembling some document to detail all the methods that can be used to verify that your measurements are good, as well to verify that an assembled crossover does exactly as designed. So things like verifying impedance, filter transfer function, driver polarity, port tuning, cabinet resonances, and of course overall speaker response.
 
OP
L

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,456
Likes
3,622
Location
San Diego
Thank you everyone for your in depth comments and insights. I never had an issue with the UMIK-1 for FR measurements but measurements with timing often seemed hard to understand and hard to repeat. I have ordered a regular measurement MIc and I have an interface so look forward to taking measurements I can have confidence in.
 

Rick Sykora

Major Contributor
Forum Donor
Joined
Jan 14, 2020
Messages
3,723
Likes
7,509
Location
Stow, Ohio USA
What sort of baseline measurements do you have in mind? I am writing an article in progress specifically for data validation, perhaps it's a good place to incorporate this information. The document is waiting for my next speaker build so I can populate it with real world examples.

Would need some time to ponder each use case, but the basic approach is to show both a clean measurement and a measurement that has been skewed due a fault in the setup.

Data validation seems to be an often missed step in DIY efforts, so I am assembling some document to detail all the methods that can be used to verify that your measurements are good, as well to verify that an assembled crossover does exactly as designed. So things like verifying impedance, filter transfer function, driver polarity, port tuning, cabinet resonances, and of course overall speaker response.

Some DIYers fail to validate their builds. This would help uncover data errors IF the speaker designer provides solid baseline. The problem is that measurement equipment and conditions often differ so a solid baseline is a challenge too. Too often some designers fail to provide enough measurements to do even rough baselining. The other challenge is the equipment expense to do validation. Many are just looking to do a build and hope they can get a good result. If they are careful and the kit components are high quality, this could happen. Otherwise, miswiring, bad solder joints or simple mechanical shortcomings can readily prevent the ideal outcome. Good luck with your effort but the fix needs to start with more standardized measurements from kit designers AND being able to consistently obtain parts that are within tolerances to match the original design. IF you have the time and energy, cannot hurt to try, but will require a coordinated, focused effort. :cool:

p.s. Any comprehensive speaker design and build validation effort deserves its own thread.
 
Last edited:

dcibel

Member
Joined
Feb 8, 2021
Messages
47
Likes
53
Would need some time to ponder each use case, but the basic approach is to show both a clean measurement and a measurement that has been skewed due a fault in the setup.
I think I can manage a few examples to include, just have to go back through some past conversations with others. Things like simply checking the noise floor of the measuring equipment, SNR of the measurement, what DC offset looking like on a windowed FFT, etc.
Some DIYers fail to validate their builds. This would help uncover data errors IF the speaker designer provides solid baseline. The problem is that measurement equipment and conditions often differ so a solid baseline is a challenge too. Too often some designers fail to provide enough measurements to do even rough baselining. The other challenge is the equipment expense to do validation. Many are just looking to do a build and hope they can get a good result. If they are careful and the kit components are high quality, this could happen. Otherwise, miswiring, bad solder joints or simple mechanical shortcomings can readily prevent the ideal outcome. Good luck with your effort but the fix needs to start with more standardized measurements from kit designers AND being able to consistently obtain parts that are within tolerances to match the original design. IF you have the time and energy, cannot hurt to try, but will require a coordinated, focused effort. :cool:

p.s. Any comprehensive speaker design and build validation effort deserves its own thread.
I will plan to post a thread when I have something worth presenting, it'll be some time yet. Of course, the validation document will be aimed at what can be done using just 2-channel measurement jig and VituixCAD, REW, ARTA. Not all-encompassing, and geared towards design validation.

In the meantime, other documents written for VituixCAD to fill in the gaps and other tips and tricks can be found here:
 

gnarly

Major Contributor
Joined
Jun 15, 2021
Messages
1,071
Likes
1,516
I will plan to post a thread when I have something worth presenting, it'll be some time yet. Of course, the validation document will be aimed at what can be done using just 2-channel measurement jig and VituixCAD, REW, ARTA. Not all-encompassing, and geared towards design validation.

I think a good layman's guide to making dual channel transfer functions is what's probably needed most.
Mag and phase traces of each driver section individually, tell us their levels, polarities, and delays.
And then a transfer of the whole speaker provides verification, as in how well do mag and phase smoothness end up.

I've found sims can be entirely forgone (for active multi-ways),
by just setting up the DIY using transfers.
 
OP
L

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,456
Likes
3,622
Location
San Diego
One other thing I learned when researching microphones is that for the budget Mics, both USB and analog like the UMIK-1, is that the calibration files are by "batch". So while you enter your seriel number to get the cal file that does not mean your Mic was tested, rather a number of Mics from a batch were tested to come with an average calibration file for that batch. I have no idea how much variation exists accross a batch but in order to get an individually tested Mic calibration file you need to spend quite a bit more.
 

staticV3

Master Contributor
Joined
Aug 29, 2019
Messages
8,716
Likes
14,069
One other thing I learned when researching microphones is that for the budget Mics, both USB and analog like the UMIK-1, is that the calibration files are by "batch". So while you enter your seriel number to get the cal file that does not mean your Mic was tested, rather a number of Mics from a batch were tested to come with an average calibration file for that batch. I have no idea how much variation exists accross a batch but in order to get an individually tested Mic calibration file you need to spend quite a bit more.
The UMIK-1 is individually calibrated.

Source:
 
Last edited:

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
21,217
Likes
38,582
One other thing I learned when researching microphones is that for the budget Mics, both USB and analog like the UMIK-1, is that the calibration files are by "batch". So while you enter your seriel number to get the cal file that does not mean your Mic was tested, rather a number of Mics from a batch were tested to come with an average calibration file for that batch. I have no idea how much variation exists accross a batch but in order to get an individually tested Mic calibration file you need to spend quite a bit more.
As StaticV3 said, Umiks are individually calibrated. I believe the Dayton's are too. I've examined several Umik cal files and no two are the same. The 90 degree adjustment from the 0 degree file is the same, but that makes sense.
 
OP
L

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,456
Likes
3,622
Location
San Diego
The UMIK-1 is individually calibrated.

Source:
As StaticV3 said, Umiks are individually calibrated. I believe the Dayton's are too. I've examined several Umik cal files and no two are the same. The 90 degree adjustment from the 0 degree file is the same, but that makes sense.
There is a large amount of conflicting information on this subject and all the "claims" yes or no are made by 3rd parties and the actual manufactures are quite vague. For example here is what the Mini-DSP site says about the UMIK-1 "It is provided with a unique calibration file based on the serial number to insure an accurate measurement."

I have noticed the Mics that are individually measured to make the calibration file make a big deal out of it. For example "Every M23 is individually hand-tuned and tested and is delivered with its own printed frequency response chart. An Electronic Calibration File (ECF) is generated for each microphone during final testing, and can be downloaded free of charge after product registration."

For $79 retail for the UMIK-1 individually calibrated sounds like a tall order but I don't know for sure.
 

Blumlein 88

Grand Contributor
Forum Donor
Joined
Feb 23, 2016
Messages
21,217
Likes
38,582
There is a large amount of conflicting information on this subject and all the "claims" yes or no are made by 3rd parties and the actual manufactures are quite vague. For example here is what the Mini-DSP site says about the UMIK-1 "It is provided with a unique calibration file based on the serial number to insure an accurate measurement."

I have noticed the Mics that are individually measured to make the calibration file make a big deal out of it. For example "Every M23 is individually hand-tuned and tested and is delivered with its own printed frequency response chart. An Electronic Calibration File (ECF) is generated for each microphone during final testing, and can be downloaded free of charge after product registration."

For $79 retail for the UMIK-1 individually calibrated sounds like a tall order but I don't know for sure.
"It is provided with a unique calibration file based on the serial number to insure an accurate measurement."

So what part of the above statement indicates anything other than unique file? Unless they just randomly futz with them to make it look good they must be measuring each one. I've examined several and no two are the same.

I've done the same for the Dayton EMM-6. No two are the same. I've got an IMM-6 which cost me $15, but is a little more now. I've examined several of those cal files and no two are the same.

I just did a quick look checking adjacent serial numbers. If batch tested you would think they might have the same file. They do not have the same file.
 
OP
L

levimax

Major Contributor
Joined
Dec 28, 2018
Messages
2,456
Likes
3,622
Location
San Diego
There is a large amount of conflicting information on this subject and all the "claims" yes or no are made by 3rd parties and the actual manufactures are quite vague. For example here is what the Mini-DSP site says about the UMIK-1 "It is provided with a unique calibration file based on the serial number to insure an accurate measurement."

I have noticed the Mics that are individually measured to make the calibration file make a big deal out of it. For example "Every M23 is individually hand-tuned and tested and is delivered with its own printed frequency response chart. An Electronic Calibration File (ECF) is generated for each microphone during final testing, and can be downloaded free of charge after product registration."

For $79 retail for the UMIK-1 individually calibrated sounds like a tall order but I don't know for sure.
Dayton is pretty clear despite the fact that many online claim it is batch calibrated.

"The EMM-6 microphone is individually hand calibrated using a laboratory standard Brüel & Kjær microphone in conjunction with a DAAS computer measuring system. Each microphone includes stand mount, foam windscreen, transport case, and its own unique printed and downloadable calibration response graph and .txt data file (see below)."

I have no idea now and will probably never know.
 

dcibel

Member
Joined
Feb 8, 2021
Messages
47
Likes
53
Dayton emm-6 calibration is so full of noise you may be better off without it. Maybe with 1/1oct smoothing...

Red trace is the supplied calibration file for an EMM-6 I purchased. Blue is 1/1oct smoothed. Those who know what a mic cal should look like would throw away the garbage supplied with EMM-6. You get what you pay for, so don't expect much from a $60 mic with no accuracy spec.
1711907990255.png


Serial numbers are sequential, so it's also easy to just type some numbers into Dayton's site and download 10 sequential calibration files for comparison. Here's 10 sequential serial numbers:
1711908887810.png


Same 10 serial numbers with 1/2oct smoothing.
1711908958142.png


If I adjust the amplitudes so they align better 1kHz range I think provides a bit better visual.
1711909126018.png
 
Last edited:
Top Bottom