Paul’s first and perhaps most important comment on this topic is a simple one:  When one is making loudspeaker measurements, especially for dissemination to the public, it is important to have a way to corraborate one’s measurements.  There are a lot of things that can go wrong in acoustic / speaker measurements.  Modestly good test equipment, with a careful setup, and a way to cross-check results, can work quite well.  On the other hand, many thousands (or more) of dollars worth of test gear, with the wrong setup, a small glitch, or just simply no check by another method to confirm all is well, can lead one right down the garden path out of sonic “Eden”. 

Now, quite often, speaker ratings are published by people who either don’t know what they are talking about, or who exaggerate the data simply to make it look better.  We are going to be posting a perfect example of the latter, soon…  But, it is hard to tell, when a vendor specifies a speaker at, say, 93 dB efficiency, and when you get it, you find out it is really, oh, say 84 dB efficiency — well, did they just not know that sensitivity is normally referenced to a 1 watt, 1 meter standard, and that if you test at 1/2 meter, at 1 watt, you have to knock off 6 dB?  Do they know that if you test a 4 ohm speaker at 2.83volts, that is TWO watts, so knock off another 3 dB to get the 1watt, one meter rating?  We suspect they know, but such errors turn up disturbingly often even on certain well respected vendors’ websites.  Even more often, we suspect, other, less ethical sellers knowingly “cheat”, and then feign ignorance if caught.

That’s one area of concern, and, as noted, we’ll address that further, soon.  But, in this discussion, we are more concerned with actual measurement errors, and in particular, those that crop up in Thiel-Small testing.

Let’s say someone has run T-S numbers on a driver and is ready to do their box modelling.  But, are those numbers accurate?  One way to cross check, if one is properly set up for it, in a lab or quasi-lab setting, with a largish test baffle to work with, is to also run a response curve on the driver in question, with equipment known to be SPL calibrated.   This gets very tricky when using a driver with a large inductance or some other anomaly that causes it to roll off or have a droopy response above 200 Hz or so.  (You also need to be able to eliminate room effects in the curve, or know what they are and compensate.)  Nonetheless, if room effects are not a big problem, and a woofer is fairly flat up to several hundred Hz or higher, you should get pretty good agreement between the level reported by, say, a LMS setup, and your predicted T-S efficiency (or sensitivity — be sure not to mix up the 2!)

However, when Paul worked at Credence, we had many woofers for which this did not work well (the measured response did not go flat, high enough, etc.)  Credence was lucky*, though:  Paul had many years before found a method to fairly quickly and (very important!) accurately measure the BL of almost any woofer, and some mids, to within 1 or 2 percent.  We call this our “Direct BL measurement” (BL.d) and it is an extraordinary way to confirm or question a Thiel-Small parameter set.  The reason why, aside from the accuracy of the BL.D measurement, is that normally, when Thiel-Small parameters are calculated all the way down to the “mechanical” parameters, which include BL, the BL is one of the last things that pops out of the math.  It turns out to be quite sensitive to errors in the measurements that one would never suspect, if looking only at Fs, Qms, Qes, Qts, Vas, and Re.)  You might say it’s like a sort of twist on “Murphy’s Law”:  If anything has gone wrong in the T-S measurement, it will probably show up as a discrepancy between the “direct” BL.d, and the calculated T-S BL number.

There IS another way to cross-check, used by some reviewers:  They cut the moving parts out of the speaker, guessing somewhat “where to cut” to trying to approximate the moving mass of the working speaker.  (There can be some problems with this method, particularly when progressive suspensions are used in a speaker, but, generally it will reveal any big problems with a measurement.  However, if one has no samples of a unit to spare, or simply doesn’t want to chop up what may be an expensive driver, it becomes an impractical way to check.)  One also has to remember that putting a vivisected moving parts assembly onto a scale does not include the air load on the driver – which again must be calculated / approximated.  At any rate, Paul has used this “direct mass measurement” technique on various occaisions, through the years, just to make sure both the T-S measurement/calculation of BL, and the “Direct” BL.d measurement were not agreeing in error.  No really significant error of that type has turned up, yet.  Basically, we have a lot of confidence in the BL.d measurement.  Note that a “Klippel” type BL measurement will also work well for this kind of BL cross-check, if, like almost any other speaker measurement, it is carefully done.

*This saved us a lot of grief and cut up prototypes / samples customers sent in, at Credence.

Finally, another way to cross-check woofer data is to build a simple sealed box and run a curve on your woofer in that box:  Assuming either the driver is a low inductance transducer, or that inductance effects are taken into account, and typical box losses are also minimized and / or taken into account*, below 100 Hz, your curve should look reasonably close to the curve your box modelling software projects. 

*Ok, yes, things can go wrong with this approach, too.  But, it may be the easiest cross check for most home builders with modest test equipment to do. 

When it comes to buying drivers, one simple check before purchasing is to see if the efficiency / sensitivity rating matches up with what the Thiel-Small numbers predict, if possible.  Unfortunately, many sellers fail to provide an “effective piston area” value (Sd, in Thiel-Small terminology), and without that, you can only guess at calculated T-S efficiency / sensitivity.  We sometimes wonder if sellers omit Sd simply so buyers can’t really “do the math”.  (Or have their modelling program, such as Pro-Response, do it for them!)  Another check, if a woofer goes up into the upper bass region with fairly flat response, is to compare the curve level in the 200-400 Hz range with the predicted T-S efficiency, and / or with the stated efficiency of the driver.  Be careful to keep sensitivity and efficiency sorted out, as always, though.  And, don’t worry too much if you find a discrepancy of 1 dB or less.  Unit to unit variations, and a whole host of other factors can creep into even a good company’s data.

To sum up, we repeat our initial statement:  When one is making loudspeaker measurements,  it is important to have a way to corraborate one’s measurements.  And we’ll add, if you are depending on someone else’s measurements, you should try to ascertain if they cross-checked them.  If not, you may need to do so, yourself.   Our friends at SL Audio are a good place to start…

Coming soon!!!  Laughable(?) speaker ratings we found on eBay.