so why would those measurements be used in the production line to detect defects?
AES E-Library: Loudspeaker Testing at the Production Line
Maybe you could point me to some other literature that would "dispell this fiction"... It seems there are many others would not agree that THD and IMD are not good indicators, at least at some threshold.
Plastic, like wood, follows the laws of physics. Not sure what your point is here.
Well looking at harmonics is very usefull for finding a speaker that is broken. But it has no uesful purpose otherwise.
See Perception
Plastic has much higher damping that wood, thats the attraction. Wood is stiffer, but the plastic is much stronger and much better damped. I am of the opinion that cabinet vibrations are not a significant source of coloration or anything else, but even if they are I would rather have high damping than stiffness. The stiffness problem can be resolved with the structure - bracing etc. but the damping problem with wood you are stuck with.
I am not a novice at this. No two loudspeakers are exactly the same and even differnt data sets on the same speaker will be slightly different. But the algorithm is most likely the greatest difference. I don't actually know because I have not done the comparison that you did. I do know that everything DID change.
Believe what you want, but the research all indicates otherwise. Read Moore in the AES on "The Audiability of Group Delay". Then read Lidia and my work on the same topic. Then get back to me.
I just looked at the two Abbey data sets and they are not that different. The older one looks as if it has 1/6 octave as opposed to 1/10 octave smoothing, which it could. The newer set found on the actual functioning applet is the better one. The progam itself normalizes to a 0 dB line based on some algorithm and slight changes in this level can make large changes in the polar map. But if you look at the frequency response curves on the top graph for equal angles they are very similar.
I could do what everyone else does and smooth the heck out of the data until it all looked the same.
I could do what everyone else does and smooth the heck out of the data until it all looked the same.
I'd take a monkey tweaked masterpiece over typical fruit of tedious and data intensive optimum way any day
anyway - this is tweaking monkeys area
tedious and data intensive optimum way area is there: Manufacturers - diyAudio
good luck! 🙂
Very interesting paper... I would agree 100% that at some level those measurements would not be useful but I submit that there is a variable threshold whereby something can certainly be declared as operating at a level that many could subjectively consider "significantly less than optimal" and yet still the device is obviously not "broken".
For example: I've heard some systems where I was moved to say, "it's damaged" or "something isn't right" because I could hear audible distortion and yet others would say, "I don't hear that". If we measured it, I'm sure I could prove that it was broken, despite someone else's poor subjective evaluation. Sure enough, upon taking apart the speaker there was clearly separation of the voice coil just starting to take place.
In other cases I'm sure the device was not broken but something was definitely "not right" and certainly a distortion measurement could have proved that subjective evaluation.
Now the above example of a damaged speaker was real and I didn't actually get a chance to measure the level of distortion to prove what level of distortion was audible to me.
In summary, my experience causes me to believe that distortion measurements can be useful beyond only "broken" vs "not broken" evaluations. Further, small changes in these measurements can be indicative of build quality even if the relatively small changes in distortion measurement are not entirely obvious to every listener.
I am a little reluctant to take this thread in the direction of distortion perception, but let me ellaborate. (If the moderators want to shed this to a stand alone thread please feel free.)
I am not saying that it is not possible to have distortion in a system that is audible, of course it is. But it is simply so easy to design a system such that the nonlinear distortion is inaudible that there is simply no excuse for not doing so. The point is that making a THD measurement and getting 1% on one speaker and 2% on another will tell you absolutely noting about which one will sound better. There simply is no relationship there. So, to me, any audible distortion in a loudspeakers syetem means that it is broken. To be audible, a louspeakers distortion would have to be upwards of 15-20% and I'd say that's broken. In a sub this would even be higher.
Now in an amp, .01% distortion can be audible - thats the whole point. There is no realationship by which this number means anything that relates to sound quality - for loudspeakers. Amps are another thing all together, not much more complicated, just completely different.
And then there is group delay. Moore showed that group dealy was not audible under normal listening conditions. But in this same study he did describe how the audibility of group delay would increase with increasing SPL. THD is the exact opposite. At a fixed % THD it is less audible at higher SPLs than lower ones. In a study that Lidia and I did we showed how group delay did in fact have higher audibility at higher SPLs. This is a complex situation. It means that as we turn up our loudspeakers the nonlinear distortion is getting less audible while the group delay is getting more audible. Therefor, one has to surmis that what is often heard at higher SPLs is group delay and not nonlinearity, yet because it scales with level it is often attributed to nonlinearity. It is the ear that is nonlinear and this nonlinearity masks the nonlinearities found in most loudspeakers. There are of course exceptions, but not as many as you might think.
I am not saying that it is not possible to have distortion in a system that is audible, of course it is. But it is simply so easy to design a system such that the nonlinear distortion is inaudible that there is simply no excuse for not doing so. The point is that making a THD measurement and getting 1% on one speaker and 2% on another will tell you absolutely noting about which one will sound better. There simply is no relationship there. So, to me, any audible distortion in a loudspeakers syetem means that it is broken. To be audible, a louspeakers distortion would have to be upwards of 15-20% and I'd say that's broken. In a sub this would even be higher.
Now in an amp, .01% distortion can be audible - thats the whole point. There is no realationship by which this number means anything that relates to sound quality - for loudspeakers. Amps are another thing all together, not much more complicated, just completely different.
And then there is group delay. Moore showed that group dealy was not audible under normal listening conditions. But in this same study he did describe how the audibility of group delay would increase with increasing SPL. THD is the exact opposite. At a fixed % THD it is less audible at higher SPLs than lower ones. In a study that Lidia and I did we showed how group delay did in fact have higher audibility at higher SPLs. This is a complex situation. It means that as we turn up our loudspeakers the nonlinear distortion is getting less audible while the group delay is getting more audible. Therefor, one has to surmis that what is often heard at higher SPLs is group delay and not nonlinearity, yet because it scales with level it is often attributed to nonlinearity. It is the ear that is nonlinear and this nonlinearity masks the nonlinearities found in most loudspeakers. There are of course exceptions, but not as many as you might think.
I've XP sp 3 and downloaded the MS bells and whistles and the Gedlee app works just fine in Firefox.
My ff browser has MS .Net Framework Assistant 1.2.1 in its addons and the .net I got from MS installed without any fuss that I remember. if you're gonna play in in the Windows sandbox then you gotta use their toys even if they are klutzy.
Geddes's speakers don't sound much like anything😀 even when playing insanely loud
. Those "plastic" boards he's now using have to be a million times better than MDF.There is a very long thread, "Geddes on waveguides", where the characteristics of his speakers have been exhaustively discussed and the "Beyond the Ariel" thread has quite a lot of interesting discussion, also. You might search them out.
This thread has been really interesting, especially the last while. Geddes's speakers are very nice but are they that relevant to the subject itself?
no no... that would make your product much less interesting. I would suggest that "no-smoothing" would make the objective data more interesting to a greater number of people here. 😀
I'm sure a "warts and all" presentation would likely result in more criticism.... But if you have an open mind (i.e. not perfect yet) by sharing more data and listening to constructive feedback you might even be able to improve the product.
If the product is already perfect then you have nothing to loose.🙂
As I said, the "no smoothing" is not an option IMO because the FFT data is all constant bandwidth so you have to do some smoothing to go from constant bandwidth to constant % bandwidth. Its how much you do that is the issue. Some makes sense, too much makes everything look good. Toole and JBL use 1/20 octave. Thats difficult for me because I don't have an anechoic chamber like they do, so 1/10 is about as high a frequency resolution as I can get.
Frank - I don't see the relavence to my speakers either, but I didn't bring them up. Data analysis techniques are very relavent IMO however as this goes to the core of the issues at hand.
Frank - I don't see the relavence to my speakers either, but I didn't bring them up. Data analysis techniques are very relavent IMO however as this goes to the core of the issues at hand.
Most of us are stuck in the outdoors and no smoothing seems to work OK if the weather is nice... and it's almost nice enough outside for me to do some testing again. 😀
OK... maybe we should get back on topic.
SL, EG and I presume the thread's author still uses, to some unknown/varying degree, something of a down shelving response. Is there anybody who actually tried this and does proper measurements truly believe that even a small downward tilt in the response sounds bad?
How and how much would those that prefer this direction recommend? At what relative listening level?
I've read on Stig Erik's thread that he likes almost 10db extra at the bottom end. His system looks really impressive as does the sound treatment in his room.
SL, EG and I presume the thread's author still uses, to some unknown/varying degree, something of a down shelving response. Is there anybody who actually tried this and does proper measurements truly believe that even a small downward tilt in the response sounds bad?
How and how much would those that prefer this direction recommend? At what relative listening level?
I've read on Stig Erik's thread that he likes almost 10db extra at the bottom end. His system looks really impressive as does the sound treatment in his room.
I don't tilt any of my speakers down and they all have smooth, declining toward the top octave off axis performance. Their directivity is not controlled as deep or as narrow as Dr. Geddes's designs and I'd bet that's the reason I don't find a tilt necessary. Mine are all recording monitors like studios use. That will unfortunately limit their dynamic range--something I do miss.
It's not hard to make do, 😉
Dan
It's not hard to make do, 😉
Dan
So... flat on-axis then? You aren't using the any high frequency shelving function that is frequently provided with studio monitors?
And what about your average listening level? Do you still find the sound to remain quite balanced as the volume increases?
I think that there is analytical justification for a gradual downward slope at the high end to maybe -2->-3 dB IF the speakers are CD. If they are pistons then the falling DI accounts for the same thing. At the low end the justification for a subtle bass boost comes from the low reverb time in a small room. The ear does a time/ level tradeoff, i.e. in a large room the reverb of the low end is great enough that it actually will add to the loudness. That doesn't happen in a small room so some LF boost (I target about 3 dB at maybe 30 Hz.) seems justified to achieve a "flat" perceived low end response when compared to a larger room.
These are not "shelving" but gradual tilts to the response.
Personaly I think that the bass should be sent to a reverb unit and this increased time of signal duration would be the same effect on the loudness perception as a bass boost, but it should work much better for modal smoothing.
These are not "shelving" but gradual tilts to the response.
Personaly I think that the bass should be sent to a reverb unit and this increased time of signal duration would be the same effect on the loudness perception as a bass boost, but it should work much better for modal smoothing.
Last edited:
My system polar response may not fall as much as the average speaker at the high end because I use 6.5" + 2" + 1" drivers so for most music I also prefer a bit of the downward slope. I have encountered a few recordings where I might prefer to bring it back by 1-2 dB.
I've also tested this on my significant other with similar results (preference towards attenuation at the high end) and she had no knowledge of what was being tested just A vs B.
But doesn't the "cabin gain" add to the response in a smaller room?
Now that's an idea I haven't heard before... I'll have to try it out!
- Status
- Not open for further replies.