Comparing CALIBRATED microphones

[bjs]

Hi,

Has anyone compared CALIBRATED microphones or know of websites that have?

There are a number of places who will calibrate microphones or sell calibrated ones but I've never seen any tests comparing the results.

I've often wondered what our hard earned calibration money is really buying us...

[brucek]

I've often wondered what our hard earned calibration money is really buying us

If you have your microphone calibrated by an accredited lab such as West Caldwell, then you don't need a comparison to know it's within the specs they claim. This is the lab that calibrated the ECM8000 from our cal file on the download page.

You can check their scope and accreditation on their site.

brucek

[bjs]

I'm aware of West Caldwell. Exactly what +/- dB variance will two mics have after calibration with them between 20-20kHz? I couldn't find anything definitive on their site.

Regardless the point is whether anyone actually tested that assumption. We seem to be great at comparing uncalibrated mics...but no-one has the courage to compare calibrated mics!

[tenzip]

The difference between two mics, both calibrated, will be in the tolerance specs of the equipment used to calibrate them.

If West Caldwell doesn't give those specs on their website, I'm sure they will tell you if you send them an e-mail and ask for the tolerances of their equipment. I have no doubt their equipment is all calibrated using equipment/standards that are traceable, and they will be able to give you a +/- tolerance for each piece of equipment they use.

[brucek]

I couldn't find anything definitive on their site

Just click the scope button.....

brucek

[bjs]

brucek wrote:

Just click the scope button.....

brucek

Yes, I saw that. But it didn't answer question as I said earlier.

The specs show the error is about +/- 0.5dB from 20-20kHz as a statistical quantity and is the BEST CASE if both mics met IDEAL test conditions when they were tested. So in the best case, two mics could differ by 1dB. In the typical or worst case it is indeterminate.

Hence the question about comparing calibrated mics in the real world.

[brucek]

Hence the question about comparing calibrated mics in the real world.

Unfortunately, testing them would prove problematics, and you would never be able to do so within a dB.

You need to be cautious when comparing microphones, since small mic element positional differences can account for some large changes in response that aren't due to the actual microphones response differences.

In addition, where you're not testing in a controlled environment, rapid changes in system response can occur, and even the smallest positional difference in a mic element can exaggerate the outcome.

brucek

[bjs]

brucek wrote:

Unfortunately, testing them would prove problematics, and you would never be able to do so within a dB.

Yes, one does have to know what they are doing...but it is wrong to say it is impossible or, for that matter, even very hard.

Here's an example. Three separate measurements of a tweeter where the entire test setup is completely dismantled and re-setup in between each test. I measured the mic location with a tape measure but didn't obsess about it so the mic location error circle is about 1/2 inch. Note the vertical scale is 1dB/division and, while I windowed out the first major reflection, I didn't worry about early small reflections. Peak errors are a small fraction of a dB and one could easily do better with more attention to detail.

[tenzip]

I guess I don't understand what you're driving at. What exactly are you trying to determine about these microphones?

As I said earlier, the difference between two microphones will be within the tolerance of the equipment used to calibrate them. So if West Caldwell's calibration chain has a total range of +/- .3dB, then the most your microphones will differ is .6dB. Assuming you got everything else perfect.

Just for the record, I have absolutely no idea what kind of tolerances West Caldwell specifies, the numbers above were just picked out of thin air.

[bjs]

tenzip wrote:

I guess I don't understand what you're driving at.

No worries mate. The question is probably only relevant to loudspeaker designers anyway. If people had such test comparisons I'm sure they would have posted by now.

[ludwignew]

I have a question concerning calibration.

When you talk about "mic calibration", what do you mean? That you calibrate the microphone of a measuring system like the radioshack or the phonic or whatever?

Or you really get into the mic and modify its parameters (because that would be reaaaaally hard to do)?

[brucek]

When you talk about "mic calibration", what do you mean?

The calibration is carried out at a company that will supply you with a text file. The file lists the deviation from a flat response over its bandwidth. You apply those values after a measurement.

brucek

[JMB]

Is there a problem with using a calibrated microphone to make a measurement and then using an uncalibrated one to make the same measurement and subsequently subtracting the corrected calibrated file from the uncalibrated one to develop a calibration file for the uncalibrated mic?

To simplify:
Take a measurement with a calibrated mic
Apply the calibration curve for correction
Take same measurement with no changes in position or environment with the uncalibrated mic
Subtract the corrected measurement from th calibrated mic from the uncalibrated mic reading
Invert the result and use this as your calibration file for the uncalibrated mic.

Thanks for your feedback,

Jay

[brucek]

That is indeed the method we used here before we found a lab that would calibrate SPL meters for us.

It's a bit of a pain, but it can be done.

It's best to do the measuring with as little room influence as possible, so outdoors or near-field is best, as any spiky changes from comb filtering or reflection cancellation render the comparison invalid at those points.

You will have to choose a frequency to normalize the plots to. Overlap the two measurements at the chosen normalize frequency, and then use the REW "add offset to data" feature.

Then convert the plots to text with REW and then use Excel to normalize them and to create the differences file.

brucek

[JMB]

Thanks for the quick response and explanation, Jay