TOPIC: Re: Low quality design or faulty mic ?

Hi all.
I have received my UMIK-1 two weeks ago only to discover unacceptable high noise and spurious peaks. I have tried to contact minidsp support (2 e-mails and 3 web-form posts) with no luck, thus this posting.

I have documented my problem and I present it bellow. I would like to know if all UMIK-1s are like this or I am the unlucky one. In the second case I'm sure that at some point tech support will contact me and find a solution. In the first case (all mics are like this) I'm afraid that a recall is in line.

The issue is that the sound data coming from the microphone contain high levels of noise, 1Khz (and multiples) spurious peaks and inter-modulation distortion products when a signal is present. Overall my microphone is far from a measurement instrument.

To demonstrate I set up a comparison test with the UMIK-1 and the Behringer ECM8000. The key points are:
- Connect the UMIK-1 and the ECM8000 to the same computer.
- For the ECM8000 I used a Soundblaster Audigy 2 NX USB sound card. This card has a microphone input and rotary knob for mic gain adjustment. The phantom power was provided by a quick and dirty homemade circuit.
- Windows recording levels were set to 100% for both devices
- The two microphones were placed next to each other and in from of a loudspeaker.
- The REW software was used for initial adjustment of matched gains and measurements
- A 800 hz tone was produced through the loudspeaker and the gain of the ECM8000 through the soundblaster was adjusted to match the level produced by the UMIK-1.
- The spectral content of both microphones was recorded with REW's RTA.
- The 800 Hz tone was stopped and the ambient sound spectrum was recorded
- A short recording of some music playing through the speaker was recorded at three different playback levels simultaneously from the two mics. No compression or further amplification was done to the audio file. The audio is indicative of the level of distortion and noise the UMIK-1 produces.

I hope that miniDSP will be decent enough to acknowledge this issue and provide a fix. If this is by design and I had kown, I would had never bought this microphone.
See the screenshots bellow and most important download and compare the two sound files; you will be surprised by the UMIK-1 output.

Microphone - speaker setup:


Soundblaster and ECM8000 connection and gain adjustment:


Gain matching the two microphones:


Ambient sound. Notice how UMIK-1 is producing false readings as REW measures it's noise and artifacts that exceed ambient level:


800 hz as heard from Behringer ECM8000:


Same level 800 HZ through UMIK-1:


Ambient spectrum from both mikes overlapped for easy comparison:


Recording with playback through the speaker at three different sound levels:


And finally, let your ears be the judge; listen to the two recordings of the same sound at the same time from the two microphones. Can you tell which is better ?
ECM8000
UMIK-1

It would be nice if you can test your mic and post the results.
Thanks and sorry for the long post and big pictures.
George

I have made a test with ARTA with behringer ecm8000 and UMIK1 (got this on 2013/04/04). With the UMIK1 I have the same problems like you.

1) I can reach a maximum of only -40dB (normal mesurement speakers loudness)--> the sensitivity is too low, and I don´t know how to change this.

2)the Noise of the mic is about -55db --> not enough for measurement to the signal of -40db.

3) I have also in spectrumanalysis with a 800Hz-tone: the peaks in 1KHz, 2kHz, 3kHz.......


sorry I don´t know how to put pictures into this Forum.



Joachim

Hi Joachim and thanks for the post.

I ordered mine end of February and I got an order confirmed e-mail on February 26. Then a "sorry for the delay" e-mail on March 9 and a "items posted" on March 15. I got the item on March 26 so the dates indicate that we had units from the same batch (and a bad one by the looks of it).

I have noticed that the noise and spurious tones are affected by the gain adjustment in windows's recording devices tab. I do not know the design of the electronics, but typically the gain adjustment should be implemented by a programmable analogue gain block at the input amplifier (before the Analogue to Digital converter). Since this gain affects the problem, it is already present in this block. So as a designer you would have to limit the maximum gain as more input gain would make the spikes really climb up the dbFS scale...

Having limited input gain means we need to make up for the lost input signal to the A to D by louder sound. That of course creates an array of other problems coming into play from speaker distortion and various items buzzing in your room with the loud sound to unhappy neighbors and sore ears...

I would like to know :
- If previous batches had this problem (high noise and spurious tones)
- Sensitivity of older vs newer batches. If sensitivity has gone down, the designers probably saw the problem and tried to limit it to acceptable levels by compromising the input gain.
- Is miniDSP guys aware of this and if so how long before they reply to my messages ? I have contacted them with the best of intentions and as polite as you can be.

I'd appreciate if more people could help us here.

Cheers
George

My understanding is they have been on a 4-day holiday since Thursday and depending upon where in the world you reside (time zone) that could have started as early as Wednesday morning. Hopefully they will be reading emails in just a few hours and begin to respond. I too have sent them an email concerning the UMIK... however on another issue. All the calibration data below 990Hz in the newly released calibration files is bogus.

As far as I know there has only been two batches of UMIKs produced... seems few owners of the first batch have complained about anything but this second batch has be a different story. If you read the threads here and over at hometheatershack you can get caught up. Yes, others have complained about abnormally high noise floors and that 1kHz spike (and its multiples), also there has been some confusion with the sensitivity factor in the calibration file... its been changed for various reasons not entirely clear anymore.

The most I can offer at this point is to tell you to read all the threads about this second batch.

Hi monomer, thanks for the answer.

I have sent miniDSP the first e-mail on 26th of March, that makes it 12 days ago. I think an answer is due...

Thanks for the hometheatershack tip, I'll look it up over there.

Regards

George

Hello George,

Sorry for missing your email. We did receive them in the This email address is being protected from spambots. You need JavaScript enabled to view it. which as explained for email, is indeed a place where automated emails come from our shopping cart so not always the best/easiest way to get in touch with us. Please use "This email address is being protected from spambots. You need JavaScript enabled to view it." in the future. With quite a big part of our team was away last week for National holidays we have delays in answering all emails (yours included). With our sincere apologies for not getting back to you in a timely manner, let's get down to solving your question:

We can indeed see that the second batch of microphone had an issue with the noise floor. The microphone was rated at -74dBFS noise floor but it sounds like a higher noise floor than usual was triggered by an internal modification. We've fixed this issue in the latest batch of microphone as we investigated on the latest batch in the past few days.

The previous version of the microphone was rated for higher SPL rating. These noise being close to the acoustic noise floor, they actually wouldn't be that much of an issue for a measurement of a speaker (i.e. need at least 30dB above noise floor). But you'd indeed need to play the tone at a louder level to be above the noise floor. For a lot of end users, it's actually not an issue since typically you'd want to be away from the acoustic noise floor by quite some margin.

We now increased in the latest batch the sensitivity (circa -9dBFS for the sensitivity) by using the analog gain and new front end. It's a good tradeoff to get a more sensitive microphone. Just lower Max SPL but I guess that it's not an issue here for the type of speaker you're measuring?

Once we have a supply of microphone (waiting for a batch of 400pcs) we can organize returns back to us for a replacement for those of you who don't need the high SPL but rather use the microphone with a low level like yourself.

Sounds good? Let us know if you have any questions.

DevTeam

Hi Devteam !!

Thanks for the answer. There is no problem with the delay; I guess it happens. I hope you all had a nice time.

Also thanks for the replacement offer.
As I mention in my e-mail, I think that the issue at hand is with the "soundcard" portion of the device. I would like to see a clean input from it with no spikes. It's a shame that your electronics are limited to 74dbFS. 16 bits nowdays give you easily the theoretical 96db. And for the claimed 24 bits of UMIK-1 with input gain all the way down we should see at least that much.
Please post at some time a no signal (ambient sound level at maximum gain) against dbFS spectrum analysis.
If the improvement is not too much I will rather modify the mic myself.

On a general note, I think more people will benefit from a higher gain and lower maximum sound pressure level, but that's just my opinion.

And as a footnote, it's nice to see that your team is not afraid to acknowledge an issue and do something about it.

Regards

George

Hi again.
I have opened up the mic and had a go improving it. I must say that things are looking pretty good. Without the microphone transducer connected and a 2.2 kohm loading resistor in it's place, I got from the original -63db down to -90db (27db improvement). See the screenshots bellow.

With the microphone transducer in place, I get -88db noise/spike floor (it was about -68db for me).
It's not a very easy mod but still I did not have to use any active components (as was my initial intention).
If the guys from miniDSP are OK with it, I can post more details here; but I think that they have already solved it so there is no point. I just couldn't wait

Cheers
George

Original spectrum with 2.2kohm instead of microphone transducer at the input :


Moded with 2.2kohm resistor instead of microphone transducer at the input:


Final result with ambient noise showing. -88.4 db for the nasty spikes :

Heheh.. indeed you couldn't wait.. We were preparing some instructions for you as promised but we realize that you were eager from the point we mentioned that you could fix it yourself.. It's great seeing that you figured it out. Great job George!

Do feel free to show your mods by any means.. We have no problems with it.
As a disclaimer, we just want to say that the tricky step here is to "open" the microphone. It's not very easy if you don't have a tool that has pin able to go inside the little holes. But with a bit of jiggling, it will move and open the mics. Inside you can have access to the internal DIP jumper to modify the analog gain.

We can document all this later for those of you wanting to do a modification. Modifying these resistors will mean modifying the internal input gain but it's a "known" gain difference. Just need to update manually the calibration file to account for that change. We can document all this later on.

DevTeam

Hi again !

Please send your instructions as well, I will be glad to test them.
You certainly have inside info (ic's involved, pcb layout etc) which I had to partly and up to a point reverse engineer. So your solution should be optimum.
I will contact you through e-mail with a full report.

Regards
George

Good stuff- please keep the info coming-!

I've opened the mic too. With two nails and a flat pliers it's not a difficult task. Just cut the head of the two small nails, put them in the pliers and rotate the ring with the two holes. No scratches at all...
Dear George, what is your mod? I think that you've maybe modded the supply of the mic capsule... ? Can you explain you mod please?
Dear minidsp, I'll post the photos of the mic so other users can adjust the gain themselves.

Regards,

Michele

Hi Michele, all.
Sorry for the delay, I was going to do a proper post with pictures and all, but got hooked up. And it seems that I won't make it today either.

So a quick description of two solutions, an easy one and the advanced that with the mic closed I got -90db to the top of the noise peaks.
I will make a post maybe tomorrow with more details.

Simple mod:
It will give you about 17 db of improvement and it only takes one capacitor:


More advanced mod:
The one I ended up with needs the following steps :
- Remove the resistor that gives power to the microphone feed from the 3.3 Volts.
- Use a LM4040 3V reference in parallel to the filtering capacitors of the mic power to get 3V
- Power it from USB 5 Volts through a choke (600 Ohm at 1Khz, Murata) and a 180 ohms resistor
- Added 100uF Very low ESR ceramic capacitor in parallel to the caps and LM4040
- Added 47uF very LESR ceramic cap just before the 180ohms and after the murata choke.

I also moved an 680ohm resistor from it's position on the pcb to a better place to improve noise pickup and let the differential input of the A2D work better (I do not know if this helped much though).
This is now giving me -90db with the microphone assembled.

I hope to find the time to make a full post tomorrow with more details and tips on how to make an initial measurement to keep for reference, open the mic, mod details etc.

Good luck

George

Good Stuff George! Thanks again for your time sharing..

- Here are few info worth for those of you trying to do a modification on a RevB microphone and trying to remove these spurious peaks close to the noise floor by basically switching the power to the capsule back to 3.3V.
NOTE that these changes are to be made by end users knowing how to rework an SMD. If you need help, feel free to get in touch with us. We are planning to do a new batch of SMD board that could be replaced with a simple soldering job. We'll soon give an update to the whole community. It just takes time to work a manufacturing batch last minute.

Please see the diagram below as a reference for location of pads.


1) R5 from 680 ohm change to 1K ohm
2) R6 from 680 ohm change to 1k ohm


3) R17 from 330 ohm change to 0 ohm
4) R18 from 330 ohm change to 0 ohm

5) R11 from nc change to 0 ohm
6) R12 from 0 ohm change to nc

- The trick recommended by George of adding a larger decoupling cap on the Mic power + LDO is also a great idea. That's something done already on revC.

Hoping this info helps.

Devteam

My mic has serial number 0565, is it affected? Thanks.