HELP! What's important to measure in subwoofers?

[ericguarin]

OK all, this is kinda long, but I could really use your opinions about this!

I write for Car Audio & Electronics, testing some amplifiers so far. The editor would like more subwoofer testing, and I'd be glad to oblige. But, what should we measure?!?! Note: This is for a consumer magazine, not the AES Journal, total testing has gotta be reasonably quick and focus on stuff ordinary folks can understand easily.

SPL:
Output is subwoofer buyer's chief concern, SPL vs. distortion is a key item:
- What frequency(ies)? Again, gotta be quick, and not disturbed by the resonance frequency. 10 Hz? 16 Hz? 25 Hz?
- What THD+N? 10%? 3%? 20%? Don't want something too low, but don't want to damage drivers either.
- I can't justify an expensive laser; is there any cheap way to measure the speaker's excursion while it is playing? It'd be nice to give excursion numbers.

Parameters shift:
For a bunch of reasons, we probably won't measure parameters.
HOWEVER it does seem important to capture how the performance shifts from progressively higher excursion.
- How about showing impedance curves at like 1, 3, 10, 30, 100... watts? (Using 0.1 ohms as a series current detector may be tough with a MLSSA's input sensitivity)
- Or maybe impulse response would be easier to understand? (MUCH easier, can just calculate from the same measurement as the frequency response). Would this show noticeable change at various drive levels?



Let me explain more about setup limitations:
- I'm acquiring a MLSSA system, so want to use that.
- Can use an AP to measure SPL distortion.
- Basically I can measure outdoors, or in the middle of my garage.
Again, I'm just concerned about repeatable measurements, not whether anyone else can duplicate what I'm doing to six decimal places.


Currently I'm thinking of the sub box on the floor, sub facing forward, mic on the floor out in front. Of course this gives a distorted measurement due to the ground plane, but that would scale out adding a vehicle correction. We're not concerned about absolute numbers or comparing to anyone else's tests (which are generally not standard anyway...). Readers can compare to the other tests which will follow and at least that will be apples-to-apples.
- What do you think?

The garage is highly preferable, as the setup can be left in place, and measurement is possible even during bad weather.
- But how about the unavoidable rattling? How can I decide if that is masked enough by the main signal to be irrelevant?
Of course I can clear space around the sub, and rearrange some stuff, but it would be impossible to remove ALL rattles...

THANKS!

[cbdb]

Your room (garage) will always affect your low freq measurements because of the room modes. This is why I would do them out in a field. IMO these kind of measurements are of limited use because as soon as the sub goes into a car (a small room, or a big bass cabinet) many of the sub parameters change, and then change again when you open a window (now its a large ported bass cabinet).

[Brian Steele]

Imitation is the sincerest form of flattery. I was quite a fan of Nousaine's subwoofer tests in Car Stereo Review years ago. Nice mix of objectivity and subjectivity. Might be worth giving them a look.

I'm not a fan at all of most of the current CA&E subwoofer tests. They're usually useless IMO, unless I happen listen to the same music on the same system in the same car at the same level as the tester used for his tests. And even then, who's to say that the tester's idea of what is "good performance" matches my own?

Why don't you want to include parameter measurements BTW? Because my suggestion would be to get the sub Klippel'd, then publish the t/s parameters and the BL and suspension graphs as part of the test report - the t/s parameters will help with custom box design and those graphs will give a pretty good idea of how consistent the subwoofer is within its excursion limits. Pictures can speak a thousand words in this regard . If getting the sub Klippel'd isn't an option, then a graph of the impedance curves at various levels may be useful, but it might be a bit more difficult for the casual reader to translate those curves to how the subwoofer performs at higher levels.

What would help also is a description of what the subwoofer's behaviour is like as it approaches its limits. Does it simply stop getting louder? Does its tone change and by how much? Does it reach its limits very audibly? Does it suffer from motor noise at high output levels? The CA&E report on the subwoofers I now use in my car was decent, but if it had also said that the motor noise was very noticeable at higher levels, I may have chosen another brand / model to fill my needs.

[ericguarin]

Originally Posted by cbdb

Your room (garage) will always affect your low freq measurements because of the room modes. This is why I would do them out in a field. IMO these kind of measurements are of limited use because as soon as the sub goes into a car (a small room, or a big bass cabinet) many of the sub parameters change, and then change again when you open a window (now its a large ported bass cabinet).

Mmm, good point. Actually come to think of it I could run curves at each spot and see how much the room modes affect things. Maybe I can find a sweet spot in the garage

As for "limited use":
If I measure subwoofer A 104 dB at 10Hz at 10% THD+N in my garage, then subwoofer B at 107 dB, that implies each sub can move a certain electromechanically limited maximum amount of air.
That won't change in other situations. In a box in a car sub A might measure 114 dB, but sub B should then measure 117 dB. That does presume the car loads them in the same way, but those differences won't affect the mechanical limits which restrict the maximum SPL.
--> I'm not trying to measure/predict the frequency response in the car, I want to numerically measure the maximum output but in a way readers will relate to.

[raypalmer]

I think it would be worth measuring actual output power versus claimed. Car audio tops the consumer audio pyramid for absurd claims and lies.

Take the "1000w" number on the case and see what you get at an actually tenable distortion ratio into 8 or 4 ohms. I'd be curious to see how far off from reality they are.

[ericguarin]

Originally Posted by Brian Steele

...Nousaine's subwoofer tests...

Good suggestion! Tom knew what he was doing, I liked his reports too. I'll take a look.

Originally Posted by Brian Steele

I'm not a fan at all of most of the current CA&E subwoofer tests...

That's why I want to introduce some numeric measurement

Originally Posted by Brian Steele

Why don't you want to include parameter measurements BTW?...

There are a lot of ways to measure parameters. The best of the "standard" methods is what Vance Dickason does in Voice Coil, constant-voltage sweeps at various power levels from a very low source impedance, creating detailed LEAP5 models. BUT
A) It's hugely time consuming, tough to justify from a money vs. website traffic point of view.
B) Why do that when a lot of folks just shove the numbers into box programs based on electric filter theory? Those are handy, but just a rough approximation, so considering point A, we don't feel inclined to do a LOT of work to get really precise numbers which will help only a very few people.
We'd rather spend the time & budget reviewing more subs, instead.

The Klippel is cool, but expensive and again time consuming. It is really a design tool, and it's measurements overkill for casual readers. It's describing the inner workings of the sub, but what readers really care about is just what is the result-how much SPL can I get?

Originally Posted by Brian Steele

...What would help also is a description of what the subwoofer's behaviour is like as it approaches its limits...Does it suffer from motor noise...?...

Great point! I'll see what I can do.

[Brian Steele]

Originally Posted by ericguarin

There are a lot of ways to measure parameters.

Then use a simple test as a "standard", as it can be repeated by most hobbyists. You can use a WT3 to measure the t/s parameters (use the delta-compliance method to determine Vas). Testing can be completed in under an hour, assuming that you have a test box.

Originally Posted by ericguarin

Why do that when a lot of folks just shove the numbers into box programs based on electric filter theory? Those are handy, but just a rough approximation

A rough approximation is better than nothing at all. As many manufacturers quote t/s parameters, a simple t/s parameter test can also serve to determine how truthful those manufacturers actually are . I've noticed certain manufacturers seem to publish parameters that are quite close to measured ones (JBL, Infinity and even Pyramid (years ago - not sure about now). Others, well, let's just say tend to be a bit "overly optimistic" with their published parameters.

[ericguarin]

Originally Posted by Brian Steele

...Others, well, let's just say tend to be a bit "overly optimistic" with their published parameters.

"Optimistic" is probably not the right word-that would apply to some brand's power ratings

There's just a big difference driving something with through 100 ohms, with millivolts of test signal across the woofer, versus a 0.1 ohm source impedance with several volts of sine wave. Also, you can get variation from sample to sample, carelessness with test box sealing, etc.

[ericguarin]

Ah, missed my other point.

There are just a lot of different ways to measure, and most are not aware that their methods, while "legitimate" are actually poor.
--> Yet, folks have successfully designed boxes from those "poor" numbers for years!
Probably this is because sealed boxes are just not so sensitive, and ported boxes built with some fine tuning. OR the woofers are just shoved into a premade box anyway.

Plus, due to logistics and timing, many measure ONE hand-made sample and probably round the numbers a little because it's silly to publish "Fs = 36.749 Hz" since production will vary.
If parameters I measure don't match, it's not that anyone is lying.
--> So I have NO interest in trying to publish if a manufacturer is "lying" or "wrong"
It's not really so, and it's too arguable.

Another problem with providing parameters would be annoying my neighbors were I to do 552 point LMS sweeps at repeatedly higher voltages.

But I take your point, and will think if I could do something "worthy" with MLSSA.
...If I can get an amp that will respond down to DC

[Brian Steele]

Originally Posted by ericguarin

There are just a lot of different ways to measure, and most are not aware that their methods, while "legitimate" are actually poor.

Actually AFAIK t/s parameter measurement by different systems produces fairly consistent results across the board. If there are gross differences between the results, that's likely due to error in the measurement process, or modification of the process (e.g. trying to measure t/s parameters using large signals).

Originally Posted by ericguarin

If parameters I measure don't match, it's not that anyone is lying.

I'm not interested if Fs is off by a few Hz, or Qts is off by a few points. However, when a manufacturer claims an Fs of 21 Hz and Fs measures 31 Hz, or claims that Qts is 0.31 and measured Qts turns out to the 0.68, that's a whole other matter entirely. And yes, that HAS happened to me, so yes, I would like to know how close the measured specs are to the published ones.

FWIW, every now and then someone brings a driver by me to check it out. As a first step, I usually run a low frequency signal through it for a short while at sufficient amplitude to get the cone really moving and listen for what audibly happens as cone excursion increases and determine a rough idea of the usable Xmax from that (this test has the advantage of automatically performing any "break-in" required for the next series of tests). I then hook up the WT3 and check Fs, Qes, Qms, Qts and Re within a few minutes. If they're close to published specs, I assume the Vas is close to published specs as well and don't bother with measuring that. If they are NOT close to published specs however, I go the extra step and measure Vas using the delta-compliance method. From start to finish this takes about an hour or so (which usually includes explaining to the person why I'm doing specific tests and what the results mean).