Stupid speaker test question...please help a n00b


Why aren't speakers tested by measuring the output sound waves vs the input wave signals? Would this not be the easiest way of testing distortion introduced by the speaker? Assuming you control all the other parameters of the test of course...

Thanks for the help!
spartanmorning
I know that some independent testers (HTShack.com) use a parking lot for anechoic style testing. The Canadian Nat'l Lab has a large chamber that some Canadian manufacturers use and Stereophile uses a "quasi-anechoic" measurement technique that I've never seen thoroughly explained. Either way, the FR numbers that result from any of these tests never particularly reflect what I get in my own room.

Marty
>02-23-12: Unsound
>I'd be leery of putting too much stock in Sterophiles testing of first order speakers. Unlike many other speakers manufactures, first order speaker manufacturers usually suggest testing speakers at the preferred listening positions, not 1 or 2 meters, something that seems to have been challenging for Stereophile.

While I wouldn't read anything into small peaks and dips resulting from being more off-axis to one of the drivers, the narrowing and broadening horizontal dispersion are endemic to the design and captured accurately enough to show why first order acoustic cross-overs are not the best idea.
Unsound... Everyone has access to a perfect anechoic test chamber. It is called "outdoors". I think it was KLH when it was a startup who used this.
I agree with a lot of your observations here. However, (whether true or "quasi" - and, as you note, it's often hard to say how the numbers came to be) the anechoic data that I've seen published for speakers has not been very useful for my main FR concern - determining that a speaker has an acceptable octave to octave balance in my room.

The published data has had little (if any) correlation to what I hear (and measure, for that matter) in my own listening room. Obviously, I can't speak to your experience applying this data. If you find it useful in "apples to apples" comparisons, that's great. I just can't say that it's been useful FOR ME in that regard.

Marty
I think products like the Audyssey just might well be the game changer in the near future. Still, knowing what a speakers low frequency output is in an anechoic chamber gives one advantages when making an initial purchase. Even room correction can't make a speaker do what it's not capable of. Truth be told, I doubt most manufacturers have access to anechoic chambers that are accurate below 200 Hz. I suspect most use computer modeling to predict anechoic response that low down in frequency. As compromised as that might be, it still allows for apple to apple comparisons, that can be user modified as needed.
Unsound,

While I completely agree with your observation re: vast majority of music lies >150hz, the octave to octave balance of a speaker is priority one for me. If the 3 octaves (or so) below 150hz aren't well balanced with the remainder of the speaker's bandwidth above, I won't be able to live with that speaker. To me, this concern is far greater than midband accuracy, imaging, dynamics, etc.

Anechoic response is often used to demonstrate how well balanced a speaker is from bottom to top. IME, once you get that speaker into a listening room, the story is wildly different.

Marty

PS I now use Audyssey on my 2.2 channel main system for this reason. It eliminates the guesswork.
The vast majority of music is above 150 Hz. Even with the diversity that exists in most real listening environments, having a baseline at least lets one try to correct for it.
Where should they play, in a water polo pool?

I'm tempted to say that that would be roughly as useful as an anechoic chamber, but that would be a (slight) overstatement.

I never said that I had a better place to measure than an anechoic chamber (other than the "right", but impractical answer: In your own listening room), I just said that an anechoic chamber is a poor environment for simulating real world performance (per Johnnyb53). If you want to make the case that any baseline is better than no baseline, I won't argue, but....

Hopefully, my point was understood: Anechoic measurements may be the most logical way to measure a speaker, but IME, they're not very useful for predicting in-room performance. By the way, "not very useful" isn't the same as "useless". Above 250hz or so, smooth anechoic FR usually translates to smooth in-room response in that region. Below 150hz, IME, anechoic response is pretty much useless. Again, IME, for predicting critically important (to me) octave to octave balance, anechoic FR only weeds out the really bad desgns.

This thread has morphed into the question of which measurements might be useful in evaluating speakers - and my conclusion is that, generally, anechoic FR is of sharply limited utility.

Marty
Where should they play, in a water polo pool? Without anechoic measurements we wouldn't have a baseline. Without a baseline it would all become a crap shoot. With reasonable speaker/ placement we can differentiate between direct and reflected sound surprisingly well. Appropriate room treatment can go a long way towards attaining a fairly neutral environment. As you have pointed out, digital room correction is starting to further help in that regard too.
Johnnyb53 wrote:

Basically, if you create a loudspeaker to measure well in an anechoic chamber (the only way you can accurately evaluate its output), it will sound unnatural in a number of ways when placed in a room.

I might have said "different" rather than "unnatural" to describe the in-room performance, but his point should be taken. An anechoic chamber is (for measurement purposes) infinitely large - with no contributions to the measured performance from reflected sounds. To achieve this in your listening room (for instance, to eliminate quarter wave reflections), the speakers would need to be +/- 5 meters from the nearest wall (including the floor, to be a purist). I thinks Johnny's generalization is pretty safe: few rooms will allow such speaker positioning.

Marty

PS The contribution of reflected energy is usually VERY audible. Audyssey is - broadly speaking - a system that attempts to adjust your system's in-room response to something more akin to anechoic response. A simple "before" and "after" test with Audyssey will quickly demonstrate just how far from "anechoic-ish" response you will get in your listening room.

PPS I'm pretty confident that anechoic testing arose to "level the playing field" for comparing speaker test results. Unfortunately, the level field is IMHO also the wrong field. If you want to determine which is the better of two football teams, by all means schedule a head to head game. Just don't play it on a basketball court.

Marty
I'd suggest listening to speakers at the appropriate listening position. Caveat, first order speakers might not be the best choice for dancers. I'd be leery of putting too much stock in Sterophiles testing of first order speakers. Unlike many other speakers manufactures, first order speaker manufacturers usually suggest testing speakers at the preferred listening positions, not 1 or 2 meters, something that seems to have been challenging for Stereophile.

02-22-12: Unsound
Please define "typical room"?

Of course there are many listening room configurations, some on a slab, some suspended floor, some with 8' ceilings, some with high or vaulted ceilings, some with enclosed rooms, some with open architecture, etc., etc. One could say, however, that a typical room will be rectangular with painted sheetrock walls, carpeted or with area rugs, with a mix of hard and soft furniture, pictures, wall hangings, and shelves, some holding record and/or cd collections. And the astute buyer will match the speaker's dynamic range, bass extension, and system power to the room size.

One thing is sure--a speaker designed for a uniform power response in the "average room" is going to sound more natural in a wide variety of room configurations than one that is voiced only for nearfield or anechoic use.
Basically, if you create a loudspeaker to measure well in an anechoic chamber (the only way you can accurately evaluate its output), it will sound unnatural in a number of ways when placed in a room. Some loudspeakers are designed (e.g., James, Mirage, MBL, Gallo, Ohm) to interact with the room and are the better for it when it comes to natural-sounding listening. But in an anechoic chamber these speakers don't measure so well.

For nearfield monitors, anechoic measurements are valid, but for regular home audio speakers, the *power response* is more important. That is, how linear is the response when it is in a typical listening room, where walls reinforce certain bass frequencies and hard surfaces can make the treble sound unlistenably bright. If you have a narrowed dispersion at the crossover point, this won't be apparent in nearfield listening or anechoic measurement, but will sound thin and hollow at that frequency in a typical room because the narrowed dispersion at that frequency results in less reflection--and therefore less energy--there.
>quite in disagreement with you on this. you clearly state you are speculating - have you heard any true 1st-order x-over speakers? $#

I've heard Dunlavies + Thiels and they subjectively don't sound as correct as good designs with higher order cross-overs.

I'm just making a high probability (90%?) educated guess on the reason where it's most likely the polar response problem that makes most consumer two-way speakers sound wrong too (I don't like those either) unless they're built to clock radio dimensions which precludes bass extension and more realistic listening levels.

Timbre perception results from a weighted combination of what our brains identify as the direct sound and its reflections which are the direct result of polar response. Where the response spectra vary too much (we seem to accommodate high frequency roll-off which would be consistent with evolution in natural surroundings that become more diffusive and absorbing with higher frequency) it doesn't sound like it would have live.

Subjectively polar response uniformity predicts speaker preference regardless of a listener's country of origin, preferred musical genre, sex, and other factors. Sean Olive has actually produced a formula using weighted values of amplitude response over a listening window and first reflection angles for a variety of rooms but AFAIK the coefficients remain a Harman Group trade secret.

To avoid running into higher frequency drivers' mechanical limits first order designs generally use cross-over points where the lower frequency drivers are becoming acoustically large which makes directivity non-monotonic.

Note the peaks and dips in off-axis response as the Dunlavy SC-IV/a transitions from woofer to midrange and midrange to tweeter as the lower frequency driver gets acoustically large:

http://www.stereophile.com/images/archivesart/D4afig06.jpg

http://www.stereophile.com/content/dunlavy-audio-labs-sc-iva-loudspeaker-measurements-part-2

(The vertical polar response should be interesting too with all the driver overlap but Stereophile doesn't measure far enough off-axis to capture what you're hearing from floor and ceiling bounce).

which is not natural and not preferred (although you might object less to the polar response problems than the stored energy issues that go with metal drivers which are avoided in first order designs).

Note speakers I like or don't object to just don't do that

http://www.linkwitzlab.com/Pluto/resp1.gif

http://www.stereophile.com/images/archivesart/N33fig4.jpg

http://www.stereophile.com/content/nht-33-loudspeaker-measurements

There are lots of others that should have similar polar response which I haven't verified.

Wave guides mated to large mid-bass drivers, dynamic driver dipoles that limit dispersion through acoustic cancellation, and wide dispersion designs with acoustically small baffles all work well.

>Merely having a 1st-order x-over ckt does not make a speaker 1st order, just FYI.

Right. Drivers have a pair of high-pass poles and the voice coil inductance creates a low-pass filter so first order electrical can yield up to third order acoustic.
02-21-12: Drew_eckhardt
.....I speculate that these reasons are why first order designs don't sound natural.
quite in disagreement with you on this. you clearly state you are speculating - have you heard any true 1st-order x-over speakers? Merely having a 1st-order x-over ckt does not make a speaker 1st order, just FYI.
Well Drew, I suspect that many of those who bought the relatively large percentage of high end speakers that have first order designs (most of which have garnered very favorable reviews along the way) considering that few companies (less than 10?) bother with the difficult task of making them, would disagree with your opinion that they "...don't sound natural."
>Why aren't speakers tested by measuring the output sound waves vs the input wave signals?

Because it's less relevant due to how our hearing works than other measurable parameters like intermodulation distortion and stored energy.

The phase distortion (reproduced via electronic all-pass filters so that the effects can be tested without the polar response differences that go with multiple drivers and varying cross-over functions) by cross overs through fourth order at typical frequencies is not detectable in blind comparisons with musical signals (clicks at long intervals can be differentiated).

While first order acoustic cross-overs avoid that they also allow excursion to double at a given input signal level for each octave below the cross-over point until their slope gets steeper. To avoid audible distortion as you reach the speakers' mechanical limits you need a higher cross-over point which means the lower frequency driver is becoming acoustically large with increasing directivity so there's a bigger directivity mismatch switching between drivers. You also have a +3dB peak off-axis someplace which isn't something our brains encounter in nature. I speculate that these reasons are why first order designs don't sound natural.

>Assuming you control all the other parameters of the test of course...

You'd do well to read _Sound Reproduction: Loudspeakers and Rooms_ by Floyd Toole. It does a great job summarizing what we know about hearing and sound reproduction.
02-21-12: Spartanmorning
lol thanks guys. i have to say i can't think of another field that is operates in this much of a grey area regarding performance. imagine if cars, computers, machines, or any other engineering devices operated like this. Quite frankly, I don't like this side of hi-fi.
Spartanmorning, you do not need to get so frustrated or disillusioned with this hobby (errrr.....obsession) of hi-end audio. You are new to this field so you do not know the various ins & outs but the it's not hard - just requires some common sense & extensive listening with your ears. Whenever you get an opportunity to listen - audio gear in a friend's house, audio gear a A/V show, audio gear at a dealer, live concert performances - take the opportunity & GO!

Look you cannot specify a speaker's performance by it's output sound waves & there are several reasons:
* the output of a speaker is very dependent on the quality of audio gear driving it. So, which brand or brands of audio gear should be considered the "gold" standard in driving a speaker? Who decides this "gold" standard? Is the decision unbiased or is there collusion in the decision? Who decides what the deciding factors should be to settle upon the gold standard to drive loudspeakers?
* output of a loudspeaker is dependent impedance matching of the amplifier & speaker & cable. Who decides which exact components to select? And, worse, the amplifier-cable-speaker match varies with every speaker! So, if you get such a spec (i.e. measurement of the output sound waves) from one manuf can you make an apples-apples comparison with a similar spec from another speaker manuf? NO! 'cuz 99.999% the 2nd speaker manuf used very diff. audio gear to drive his loudspeaker (to get favourable results).
* What should be the amplitude of the signal driving the loudspeaker? As many have stated distortion is a very heavy function of amplitude (& frequency).
* We have the audio spectrum of 20Hz-20KHz: at which frequencies should the output sound waves be measured? Who/which body decides that? Why is this body entrusted to make this decision?
* What should be the driving signal into the loudspeaker? Should it be a sine wave? Should be it be a square wave? Should it be music? If music, what kind of music? If you say Beethoven, I'll say Mozart! You are not wrong & neither am I! So, now what?

imagine if cars, computers, machines, or any other engineering devices operated like this.
Now, let's look at this part of your statement - computers, cars, machines, engineering devices ALL give you stardardized specifications that hold true across planet Earth whether you are in Tibet or in Washington D.C. or anywhere in between. If you have standard measurement equipment (which can be bought from standard measurement equipment manuf) anybody with the skill in that area can verify the performance. The test conditions require that the device be put into a known state and this can be done by the experts at, say, Consumer Reports, and then the test can be run. When you get the results you can compare apples-to-apples.

Loudspeaker manuf in some sense give you the same -
* they tell you speaker efficiency in dB/W/m or dB/2.83V/m(which lets you estimate amplifier power requirements)
* impedance & phase plots (which again give you an estimation of how easy/hard the speaker is to drive)
* max SPL level (which lets you estimate if it reaches your listening levels + if it'll work in your room)
* some manuf like Green Mtn Audio even give you distortion specs (which give you an idea of how loud you can play the speaker)
* Cross-over point frequencies (which give you an estimation if you'll hear artifacts in the sensitive presence region)

you have all the info/specs you need to make an informed decision of whether the loudspeaker needs to be put on your short-list or not. Thereafter, you NEED to listen to it & see if you like it. No way around this last part. Short-ciruiting this last part often means short-ciruting your listening pleasure!!

Hope that this helps to make you less disillusioned.....
FWIW.
lol thanks guys. i have to say i can't think of another field that is operates in this much of a grey area regarding performance. imagine if cars, computers, machines, or any other engineering devices operated like this. Quite frankly, I don't like this side of hi-fi.
See, I'm convinced I'm an audio neanderthal. I've been into this hobby for over 30 yrs and have NEVER checked a single spec or chart! Always skipped right over any and all tech info in reviews. Did it all by ear baby! Must be why I've got a tube-based system!!;)
I'm not so sure about squarewave testing being meaningful. It will reward first order (analog) x-over designs, but the audible benefits are debateable. I won't take sides in that debate, but Siegfried Linkwitz has written extensively on the subject, and he comes down squarely (pardon the pun) on the side of higher order designs that won't fare as well on squarewave tests.

Linkwitz certainly qualifies as an authority, since the Linkwitz-Riley crossover solution (even order only for "true" L-R, odd order "quasi" L-R is possible for 3rd order and above) is pretty widely considered a classic bit of design theory. Obviously, there are great designers (Dunleavy, Vandersteen, et al) out there using first order crossovers so.....

While L-R is NOT universally accepted as an optimal solution, Linkwitz's analysis does cast some doubt on the real world benefits of goood performance on a squarewave test.

Marty

PS For those pointing out that THD tests, etc. are of limited utility in judging a speaker's performance, I'd say that this is one more example where test results for speakers may raise more questions than they answer.
One limiting issue in hifi is measurement vs perception.
Apparently, everything that can be measured doesn't matter and some things which matter are not measured...or perhaps measurable (yet). Add in literally hundreds of tweaks ranging from Schumann frequency generators to having a Seance before listening doesn't help much, either.

Toss in a good mix of those in favor of DBT, SBT and whathaveyou and you've got a real mess.

Contentions that 'amplifiers run in there linear, non-clipping area all sound alike' without reference to what speaker is being driven, don't necessarily help.

'Ya can't take it TOO seriously because 'that way lies madness'.

Some of what would seem to be basic advice like....listen before buying or find a good sales guy to trust and work with.....may not help because a person new to this 'hobby' doesn't necessarily know who to listen TO or what to listen FOR.

Advanced HIFI does a poor job of selling itself, but that's another thread.
Square wave reproduction (as noted by Unsound) is a darned good objective test. Decades ago Ohm cited this in their promotional material. I have been told that a few other speakers pass this test.
IMHO, John Dunlavy built some of the finest loudspeakers, and he depended on tests more than most. He seemed to think that step and pulse response were particularly important.
"I have to say I find this field of 'hi-fi' quite frustrating"

You're not alone. We are here to learn.
We are well beyond my knowledge level (obviously) but I have to say I find this field of 'hi-fi' quite frustrating. There seems to be so much misinformation and disinformation that obtaining any semblance of understanding would require a formal acoustics degree of some variety.
Spartanmorning, One reason is that few manufacturers who use more expensive underhanged speakers would have clear advantage. No incentive for everybody else who is using overhang drivers to publish distortion. The other reason is that speaker distortion is not only performance at particular frequency and sound level but also at complex signal (music). Non-linear membrane action and membrane bending at higher frequencies creates intermodulation distortions. Speaker also has different phase shifts at different frequencies and presents complex load to amplifier that is difficult to measure. Specifying distortion at particular sound level and particular frequency driven by some perfect lab gear would not make much sense. There are few people on this forum who know speakers much better (design them for living) and I hope they will comment on this interesting thread.
Ears are less sensitive to distortion at lower frequencies. 5%? 10% I know it would look awful on a chart.
At higher frequencies ears are much better.

As for sensitivity being the sole measure needed? Well, I'd like to know if I were getting a speaker with high phase angles before popping for that good tube amp.
The dyns and va should be compared in optimum setups in the same room.
It is easy to have that much variation so that the speaker with better bass spec actually sounds thinner. Maybe the speaker is built to a higher 'Q' or there is an amp issue with speaker / amp not getting along at low frequencies where a high phase angle may play havoc with an amps ability to produce enough power. All sorts of other possibliities exist.

Now, I dont' know....how easy IS it to measure speaker distorion and how many flavors does it come in? Is time / phase coherence part of distortion?
Doing such measurements sounds like a legitmate use of an anachoic chamber and mega bucks worth of amp.
oh and PS...I would go by my ear but it is so difficult to tell A/B type comparisons. also hearing one place vs hearing it at home...different sources...etc...
Thanks guys but as a noob that really boggles my mind. you mean with all of the resources we have online in this day in age and no one is doing distortion comparisons amongst speakers for given decibel levels??? WHY????? cynical marketing reasons?
THD below 50hz at highish SPL (90 db +/anechoic or quasi anechoic) is off the charts for most speakers. Per Sarcher's post, this test is sometimes run by HT websites, especially for subwoofer testing. Subwoofers are purpose designed for clean response in this range and most of them still suck.

No one wants to see that # for a typical full range speaker.

Marty
Green Mountain Audio is the only manufacturer I can recall consistantly giving speaker distortion specs.
Ah Specs. I learned a while ago especially with speakers to only look at the sensitivity spec. The most current reason is my Dyn C1's are rated 45-22khz. My brother in law has Vienna Acoustic (forget which model) that are rated 32-25k and we both agree my Dyn's go lower. It could be due to the fact the VA's distort more at the lower frequencies (just a guess). Besides if I were a manufacturer I would look for a frequency band where the distortion is the lowest to publish. For me the same holds true for amps regarding watts (but that is a long story).

Bottom line is let your ears decide what sounds best ;-)
Exactly, we kill ourselves to get our electronics purer than Ivory soap, all the while our speakers are distorting much more. It is sort of like the dirty little secret of hi-fi. No one speaks of it much. It seems the emperor has new clothes. BTW woofers can distort up to 10%.
Because most speakers have high distortion compared to electronic equipment. Over 1% thd is normal for most speakers. It would be nice if manufactures would publish that spec. The only place I have seen it measured was on some reviews by Secrets Of Home Theater online.