How to measure smooth, nice curves for microphone datasheets

[zoom]

hi,
I am just evaluating my results from a few microphone measurements I have done last week. I had the occasion to use a Rohde&Schwarz UPV and some B&K stuff for acoustical measurements in an anechoic chamber. Probably a few results are of interest, my motivation was to learn more about microphone responses when the sound comes from different angles to the mic. This is an important thing you need to consider for recording. Usually this is an issue which is normally not shown by polar plots and in frequency response diagrams of manufacturer datasheets.

ok, introduction: probably you are interested to know how a nice smooth curve is generated, although this doen's show the reality? It's not a secret, but everyone have to be aware what was going on if you have a look on response curves. I have done this with Excel with 256 normalized (1kHz=0dB) measurement points for a sweep. Here we go...

Graph [1]: The raw curve 'as it is'
Graph [2]: The same with a scale which is much more suitable. +10 to -30 dB is the usual range.
Graph [3] und [4]: Smoothing by making averages. Averaging of three or five measurement points. But it can be improved...
Graph [5] und [6]: Using a thicker lines... and on picture [6] we have an averaging of 13 points. That is what the markting wants! ...technicans would prefer graph [2]




But you still see that this is a real measurement and not a line made by a drawing program.

btw. this was a capsule of a well known German manufacturer. guess which one? ;-)

Cheers Stefan

[Ethan Winer]

Excellent Stefan!

Yes, microphone (and loudspeaker) makers have been using averaging for years to make their products look better than they really are.

--Ethan

[1954U1]

Nice post!

But, about the frequency steps.. maybe I dont understand..
why 9 logarithmic intervals between 10-100, 100-1000, 1000-10000 ?

[Chrisc_o]

Very cool, much appreciated.

[zoom]

Originally Posted by 1954U1

But, about the frequency steps.. maybe I dont understand..
why 9 logarithmic intervals between 10-100, 100-1000, 1000-10000 ?

this should be correct like it is. Write all the frequency values on each line of the logarithmic scale and you will see it fits.
Unfortunately Excel doesn't allow me to cut the graph above 20kHz, so I have to show the unused range up to 100kHz.

btw: it is a Schoeps MK8 which is shown here.

[1954U1]

Idiot me..

Whish I have an anechoic chamber at home,
the off-axis freq measurements are sooo useful..
Its a shame the vendors datasheets dont report it.

[boxymoron]

Even the raw freq response is impressive compared to most speaker measurements. Could you explain the testing setup? I'm interested in knowing how you excite the mic, and calibrating procedures.

[zoom]

Having a studio with good acoustics at home would be probably more useful than an anechoic chamber ;-)

I use this chamber sometimes for work (my company rents it for measuring hearing aids in freefield conditions) and last week I had some time to do some private measurements in there...
Used equipment:
1/4" B&K measurement microphone and power supply (for the 200V polarisazion voltage),
B&K Calibrator for the reference 94dBSPL (1Pa) at 1kHz,
B&K turntable (for the polar-diagrams) - which is controlled by an old DOS computer with a selfmade Turbo-Pascal Program. ...stone aged - but it works :-)
Rohde&Schwarz UPV Audio Analyzer
Power amplifier and measurement loudspeaker (coaxial 2 way - similar to a Tannoy system)

Calibrating procedure for the setup:
1. after warming up the equipment I calibrated the measurement mic with the B&K calibrator to 94 dBSPL at 1kHz and adjusted the analyzer-section of the R&S UPV to it.
2. Then I mounted the meas.mic to the reference point in the room and adjusted the Generator at 1kHz so that the speaker produces 74 dBSPL at this point. 20dB less than 94dBSPL. Reason: I wanted to measure with a reduced SPL to avoid unneccessary higher THD which is produced by the loudspeaker.
3. Reference point in the room was in 2.20m distance to the sound source. With the B&K microphone the frequency response of the loudspeaker from 50-2000Hz was measured. To equalize it, this measurement was stored as an inverted curve. So the Generator uses that to calculate a correction for each frequency.
4. Then I've replaced the B&K measurement mic with my microphone I want to investigate. Results were measured in dBV.

Evaluation of the measurement results: The dBV values had to be normalized to 0dB at 1kHz.
For polar diagrams a Matlab script was used to generate the Graph. I've done only one measurement with the turntable, because you will see more characteristical details of a microphone if whole frequency sweeps under different angles are measured.

here an animated polar diagram of my MK4 cardioid with some steps in between (22 frequencies 1/3 oct):


Sometimes the 180° doesn't look typical for a cardioid. The reason is a scaling limitation I had with the Matlab tool. Interesting is everything above 8kHz: you see (compared to the manufacturer data) a high frequency boost for sound which comes from the side.

I hope this is of interest here ;-)

cheers
Stefan

[Francis Vaughan]

Any chance you can tell us that analyser parameters were?

[zoom]

Originally Posted by Francis Vaughan

Any chance you can tell us that analyser parameters were?

I don't know exactly what you mean... The analyzer reference is depending on the sensitivity of the capsule of the measurement microphone @ 1kHz. Usually this is a value around -132 dBV when I put the 94dBSPL calibrator onto the mic.
During a measurement session I do the calibration up to 3 times a day (in the morning, afternoon and after everything is finished - to check that nothing went wrong what i've measured before)

[zoom]

Some measurement details of my little measurement orgy will be posted in the Geekslutz Forum.

here is the first: Some special acoustical measurements on a MK8 capsule regarding its symmetry.

Have fun
Cheers
Stefan