Lot's of the odd shaped speakers will almost surely show a poor on-axis response due to closeby diffraction causing artifacts due to certain designs. It does not mean it's going to be a bad speaker, it just means that you will hear less focus in sound imaging. With these type of speakers, it's still possible to have good sonic balance. If someone is only trying to use on-axis response to separate groups, I think it's like putting the Summa's in a group of not so flat on axis response. That would be the way to do things back in the days when measuring capability was much more limited and things were simplified. I don't know how far the paper you mention dates back, but we really should learn what we can from the past, but also move forward.
Last edited:
Have you guys read Sean Olive's blog on listening training?
Audio Musings by Sean Olive: Harman's "How to Listen" - A New Computer-based Listener Training Program
Audio Musings by Sean Olive: Harman's "How to Listen" - A New Computer-based Listener Training Program
Of course they should admit it like Prada, Gucci, etc. Instead of lying about improved SQ just sell products based real subjective WOW, I want that because others can not afford it.
Why can there be no jumps? Is there some accepted rate of change that is considered the highest slope for individual distortion product orders? What is it in the specifics of the physics of a dynamic driver that limits rate of change of distortion products that would invalidate measurements that show such?
Dave
I don't think I've seen any high priced audio brand claim total SQ improvement, some might focus on certain design aspects that might benefit certain portions of improvement, while avoiding addressing other areas.
And rightly so, perhaps; for some it's a "deal breaker," and Earl is working toward mitigating the issue....
Good questions, tough answers. One finds that the audiblity of distortion is more source dependent than listener. Very low levels of harmonics are detectable with sine wave, but on complex tones the levels are much higher and on complex music even higher still. So the real question is what do you use as the test case. I have always used music, I see no point in finding audibility thresholds of sine waves.
I have seen cases where 25% second and third harmonic THD was not audible so I would not consider 5% to be a problem. Hearing "something" and proving that its what you are measureing is a real problem - just try it sometime. We want to always jump to the cause-effect relationship, but thats not proof. Even correlation does not show cause-effect, but without correlation there cannot be cause-effect. Its all much more complex than people want it to be which is why so many jump at the simple answer. We as humans do not like complexity and we will simplfy things down until we understand them even nif that means that our simplfication is incorrect.
Parsimony is a great concept and I buy into it completely, but we have to always be careful not to jump to any answers because it suites our concepts of what the answer should be.
So yes, your speaker has an audible problem. You find that the distortion is large at arround 700 Hz. Is there any reason to believe that the two things are related? or that one causes the other? No, not really.
Yes, there would be a maximum rate of change. I don't know what it is but I know that its not instantaneous. (it would be more a function of the signal processing that the speakers mechanics. Even if the speaker were able to change it characteristics that fast, and it couldn't, the signal processing would not be able to track it. Hence instantaneous changes like that are just not physically possible.)
What happened in that test is what someone else alluded to. When the signal hit that peak frequency the signal path clipped somewhere and wham, the distortion jumps way up. When the cliping goes away, it drops back down.
IOW, if I'm reading this correctly, it's possible in the electronic devices such as amplifiers or processors but not the speaker d/t to its mass. Correct?
Dan
Tough answers indeed but insightful! 🙂 The final sentence may save me a lot of heartache trying to look in a specific area when it may not in fact be the right place to look at all. The measurements With and without the crossover seem very similar, however I started to notice the problem AFTER putting in the crossover, I'd figured this was probably just because the levels of the mid bass were now higher, but it could be the notch filters I've put on the midbass drivers. Perhaps time to get that 100uF polyprop to replace the cheap BP electro I was using for prototyping! The figure of 25% really surprises me, but if that's what the research shows then I can see that 5% really shouldn't be a problem. I wonder whether it will sound any different now that I have read this response! I never rule out psychoacoustics as a possibility!
Thanks Earl 🙂
Cheers,
Tony.
Thats 1/2 of it, the other half is the signal processing. A pure sine wave will not be a single bin in an FFT, its spead across several bins because of uncertainty (its called leakage). The harmonics will have the same problem. But if all of a sudden the signal clips, then the meter jumps up instantly. Leakage doesn't apply.
That sort of makes sense to me, but the details are indeed over my head and I'm quite sure that's simplified. Looks like I've got homework.
Thanks,
Dan
Hi Dan, If you have Joe D'appolito's "testing Loudspeakers" make sure you read the section on analog to digital conversion (and sampling), I'm at work so don't have it handy to give a page reference. It has a very good coverage of how it works and also the limitations. He covers the subject of leakage that Earl mentioned. It was a real eye opener for me with respect to digital signal processing in general, and just how approximate it is!
Tony.
Tony.
So far what I've gathered from designing my second loudspeaker(so take this in that context), the 3 most important things to measure are the polar response of the woofer, tweeter and of the completed project. It is enlightening as to what the final sound of the speaker will be. Once that's right, any other measurable imperfections are up to you wether it is worth it to fix them, but before that's right you are wasting your time fixing anything else. Just my 2 cents. Don't expect anything useful or accurate from the graphs posted by manufacturers in the cone break up region, so picking your woofer is a guess. To quote Dr. Geddes on manufacturers graphs, "they are optimistic."
Dan
Dan
To split a hair, that is the case if the sine wave period is not an integral multiple of the fft length, and you need to apply a window.
But if you can set the fft to be synchronously with the sine freq, there's no need for a window and hence no leakage. You get a single spike in a single bin.
But at any rate, clipping will make the meter jump as you say and the above doesn't really make a big difference.
jd
yes skeptic I would agree with you there, it is another tool in the toolbox though, and certainly helps with understanding of what you are measuring and how to get as good a result as you can.
I think that this thread can hopefully fill in some of the missing info on how to interpret the results.
It's all very well to say the goal is accuracy (as has been said multiple times here), but obviously no speaker measures perfectly, so what do we hone in on to work on. For example I'm targeting broad peaks in freq response, because I read somewhere a long time ago that they are more noticable than sharp dips (which are a lot harder to do anything about anyway).
I think it was Loren42 made a comment (maybe not in this thread) about the behaviour of the speaker off axis in the crossover region being constant (or the drivers behaving similarly).
I guess what us less experienced people need to know is not that we want the results to be accurate, but what accurate actually is! 🙂 Again I probably should finish reading the toole article 😉
Tony.
Last edited:
That is spliting a hair. I was well aware oif this fact and I have used this technique before by locking the FFT window to an exact sine wave such that there is no leakage. Hence there will not be any leakage on the harmonics either! Using synchoronous averaging, you can measure harmonics down below the noise floor - useful when the level of the signal falls in say an amp and its down in the noise, but yet this is precisely where you want to know what the distortion is. This technique works quite well.
Dr. Geddes,
What is your definition of a "decent loudspeaker system" in terms of design? I'm curious because the most common design, even among audiophiles is based on a 6-7" woofer, which is not really capable of generating the displacement to reproduce meaningful bass without easily exceeding xmax, and the average distortion "envelope" to a more complex signal may be worse than -20dB in these kinds of two ways?
- Status
- Not open for further replies.