SPL meter specs for speaker testing

[diydidi]

Hi All
I've lately been getting a lot of speaker repair jobs with the strangest problems.
Just the other day I received a pair of powered EV's. The complaint was that the one was louder than the other. After busting my balls for hours, working without a schematic, probing the preamp and amp boards, only to realise it is a damn BTL arrangement, I figured out that one of the tweeters was softer than the other.
If I had an SPL meter, I could have figured this out quicker. Does anyone here use these things in their shop?
I had a look at some specs of units for sale online. Most of them have good resolution, but only go to 8khz... Is this normal?
Will it work for my application?
What will the actual procedure be with regards to the actual test setup etc. I know that distance to speaker can mess-up my findings.

[Mark Black]

I never had an SPL meter in my shop, but I think you could accomplish 'quick & dirty' comparison measurements by use of any reasonable microphone through a preamp monitored by scope. Perhaps a chopstick or some such taped to the microphone to ensure distance stays the same, and maybe some foam on the end of the stick to minimize direct vibration transfer to the mic.

Purely conjecture on my part since I never tried it, and the variables which can be induced by even very small positioning differences of the mic might swamp the results but it should be relatively easy to try.

[km6xz]

Accurate sound measurements are hard to achieve and not worth doing for repair shop that is only interested if something is defective, using relative measurements. A small cheap electret element is best for minimizing mic induced variations in polar response. By comparing two speakers fairly close, above bass frequencies you can isolate the speaker response from the room response so on axis measurements with identical mic and speaker placement you can get a good idea of what speaker is doing relative to another speaker. The closer the more accurate in that room acoustics plays less of a role in the mid or high frequencies. Low frequencies are more complex in their interaction with the room.
Real speaker response measurements are done in rooms with controlled reflections(or none) and averaged over a number of sweeps from different positions or mathematically smoothed. Response curves tell us remarkable little about a sound system but it is one of the things customers want to see. Microphone curves are even less telling of what it sounds like. Polar plots give more actionable information but still do not tell much about what it sounds like. For that, real sound sources and real listeners are the only way to judge. In your case, just testing for major differences between two speakers at some close measuring point, say 10 cm, sure can tell you if one is way down, or where the cross-over point is. As mentioned above, a cheap 1/4in electret mike element will have a very flat response, and be sensitive enough. High quality low noise floor large capsule mics will usually not measure as flat, but a quality mic is not valuable because of its flatness of response. If it was, we could spend $1.25 instead of $5,000 and get the job done.

[Mark Black]

Juan - you're still here - I'm so glad, and Merry Christmas/Happy Holidays as well! Same to you Stan over in Russia too!

[J M Fahey]

The BIG problem is the room, you move your microphone 1 inch and can get a > +/-10dB peak or dip.

Published speaker curves are "avereged" or "smoothed" , I'll search and post a real, unsmoothed one and you will be horrified, there's n ot a slighter adjective for that.

So only way you can get a useful bench test at the servicing level is, as Stan said, a broad comparison one:

1) minimize room influence.

Personally I have 2 SPL meters, a cheaper one similar to the common Radio Shack one, which is guaranteed to 8kHz (more on that later) and a very good and expensive Japanese one, with tons of functions I will never use, including digitized data output for later processing on a PC, 10 or 12 memories, lots of measurement modes, etc.

In any case I always use it on the "Flat" setting, slow/integrating response, and plain read what the display shows so I can do the same just with an electret capsule, a simple preamp, a diode and some RC to smooth the output which can simply be measured in DCV , not DB (which would require serious calibration) .

I stick the mike as close to the dustcap/voice coil as possible, basically at the edge of the speaker baffle hole if backmounted or the front gasket or edge if frontmounted.

If I can't pull the cloth out (say, a Marshall 4x12") I "normalize" by measuring, say, 5" away from the cloth.

Signal source?
One and only: some pink noise MP3 ... unfiltered, let the crossover do its thing.

In your case I would feed some 5 or 10W of pink noise into the cabinet (it's loud and won't make you popular) and measure woofer(s), then mid(s) , then tweeters, note values down and then pull it from where it's sitting, put the other one in exact same place and orientation and measure again.

Room standing waves are terrible and even measuring close up they have some influence.

This is a "polite" , "publishable" frequency response the kind you are used to see, "smoothed 1/6 " which means they take a band 1/6 octave wide and average all the mess inside it into a single number :


this is the real truth (mind you, measured inside an anechoic chamber) , not to be shown to kids and impressionable people:


that's why sinewave measurements are accurate (in theory) but USELESS.
Way too much data for our brains.

That's why for these comparisons I prefer humble pink noise, which is fine to get just "one" number for SPL comparison.

Here's a 5 seconds sample, set your player to loop so it's continuous (with short silences between end and restart).
http://www.mediacollege.com/audio/_a...pink-noise.mp3

If that annoys you, search for longer 5 or 30 minute samples, which of course will be heavier.
Get MP3, no need for "better" WAV.