Microphone matching in large quantities: how ?

FWIW was wondering how microphone manufacturers select 'matched pairs' when they have a big pile of mics.
They obviously won't take a mic and try to find the best next mic from
the population (say by summing their outputs and polarity-flipping one
in response to broadband noise) and then continue in this fashion for the next pair.

And then, do they match more on freq.response or on sensitivity ? Can become a compromise fairly soon, unless it's an A-brand manufacturer with tight fabrication tolerances.

Best way I could think of is that they make a freq-graph of each and
make pairs based on that...
but if they make these anyway, why don't you see them more often included ?

The matching for A-brands seems less fuzzy to me, but considering say a cheap matched pair of SDC-mics with actually not too bad matching I wonder how they do it.
Graphing each mic might be too expensive there I guess, so maybe they manage to maintain tight tolerances as well ? (be it at more crappy frequency-responses, hence indeed no plot supplied :wink: )

Bye,

Peter

E-V once (>25 years back) published a few pixs of their production test. It was a doghouse size anechoic chamber, the tech held the mike through a hole, and looked at a 'scope. I bet it was a swept-tone speaker, and crayon-marks on the 'scope indicating the allowed tolerance. If the trace fell between the marks, "good"; if not, "reject".

And guessing how production errors fall, 90% of mikes might lie inside of +/-2.5dB 100Hz-10KHz and 30% might fall inside +/-1dB. IIRC the article did imply that the same mike might get sold as different models based on this test.

Was it RE14 cheap and RE15 tight-spec? Sumthin like that. All RE14s were good, but two RE15s were prefered when used as a Pair. Or so they said. I never found a lot of diff amongst my RE14s. If finding an RE15 was tough, then the RE14s should scatter high/low of RE15 spec. But if the gals were on a roll, and making all mikes the same, then E-V might have more RE15-grade mikes than RE15 orders, and be selling-down the excess RE15s with RE14 badge and price.

Some of your total production slop is minimized if you snag consecutive units. Your film or spacer-stock may be +/-10% thick, but from one inch to the next is is likely to be much closer than 10%. The 10% covers slop from roll to roll, or from beginning to end of a roll.

When we did newspapers with color overlay, the first 1,000 non-rejects still had the color all over the place while Joe dialed-in the rollers. It would be "right on", Joe would go in back for a snort, the color would drift, he'd trim it again.... long runs where relative unit-to-unit error was "zero", with jumps as accumulated drift got corrected.

Actually, before Joe went for his snort, he'd call "500 samples!". We'd put the next 10 stacks on a different skid and set it by the door. The guy who sold the ads which supported the paper would stack them in his Rotary Wagon and go call on the advertisers, show them how great their ad looked. Unit-to-unit match was not the issue, but when Joe had it trimmed good, he could take a snort with assurance that the next 500 would be equally good. (After "Samples!" we'd let the perfection fall off. You've seen your Weekly Shopper with the red 0.15" off alignment.... Joe sometimes used a linen magnifier to set-up the Samples to impress the ad-buyers.)

Film thickness is a different variable than alignment of red on black on two presses, but I suspect it drifts similarly.

So the tech pulling "matches" has a real advantage over you. He can decline units made right after a film-roll change, or a shift-change, or on Monday Morning. When the line is running smooth, he can pull sequential units, and have a higher probability that #13 and #14 etc are "matched".

Beyer does the frequencyplot comparison. They store all frequency plots of every mike they make and select the matched pairs by comparing the plots.
Their cheapo range mikes don´t get delivered with a plot (though they did record the frequency plot to check if the mike is within the specs) but the more expensive ones always include the plot for the customer

Thanks both for the response :thumb:

As in PRRs story, can understand a certain degree of matching is guaranteed by the fact that subsequent mics in a reasonably controlled process don't have a reason to be really different (apart from goofs and those changes in supplies etc).

Apart from that it there's of course the discussion how well matched we actually need, but I was mainly wondering about 'how they do it', thanks Jensenmann for the Beyer-info.
Can't imagine that Beh. C-2 pair I couldn't resist for $40 took an alike route though :wink: but that's another story... c'mon Uli, how do you do it ? Skip it ?
I checked my non-subsequent-ser# pair of Beyers (nothing to be desired there), should repeat for those C-2's.

I recall some info in SOS-mag that we shouldn't be too obsessed w.r.t. matching though, but again, this was about the how of matching. Better won't hurt of course.


Regards,

Peter