Hi doug20, I think you are missing my points (or I'm missing yours), but I don't know how to say it so that you don't misinterpret.
The Toole paper you posted is very interesting. I haven't read it all yet but will do so.
I certainly have not stated that I can hear things that measurements do not show. I hypothesized that there *MAY* be certain aspects of how a speaker performs that are not quantified by the normal measurements that are done.
I agree 100% that "imaging" is an illusion that our brain creates for us. However this was not specifically what I was talking about, I was talking about the ability for the speaker to reproduce complex waveforms, and whether or not any existing measurements can accurately predict how well one speaker can do this compared to another. This is in line with the OP's original questions.
Rather than just dismissing my supposition, actually pointing out WHY it is wrong or false, and how to do measurements that do show this would be constructive, and meet the goals of the OP's first post
You seem to think that I as an "Unknown" in some way was attacking Zaph and Dr Geddes, by daring to suppose that maybe there could be some things that current measurements can't or don't show (a supposition that seems to have caused you to place me in the subjectivists camp). I would put it to you that one who doesn't question, will not learn, and to blindly accept everything that someone with more knowledge than oneself says is foolish. If this was how all people who have done research in audio thought, then we would never have advanced at all!! Ironic isn't it
Tony.
The Toole paper you posted is very interesting. I haven't read it all yet but will do so.
I certainly have not stated that I can hear things that measurements do not show. I hypothesized that there *MAY* be certain aspects of how a speaker performs that are not quantified by the normal measurements that are done.
I agree 100% that "imaging" is an illusion that our brain creates for us. However this was not specifically what I was talking about, I was talking about the ability for the speaker to reproduce complex waveforms, and whether or not any existing measurements can accurately predict how well one speaker can do this compared to another. This is in line with the OP's original questions.
Rather than just dismissing my supposition, actually pointing out WHY it is wrong or false, and how to do measurements that do show this would be constructive, and meet the goals of the OP's first post
You seem to think that I as an "Unknown" in some way was attacking Zaph and Dr Geddes, by daring to suppose that maybe there could be some things that current measurements can't or don't show (a supposition that seems to have caused you to place me in the subjectivists camp). I would put it to you that one who doesn't question, will not learn, and to blindly accept everything that someone with more knowledge than oneself says is foolish. If this was how all people who have done research in audio thought, then we would never have advanced at all!! Ironic isn't it
Tony.
No problem, its the internet its hard to sometimes understand the POV. I apologize to if my POV isnt coming out right.
Speakers do not reproduce anything more then a soundwave (CSD on and off axis shows lots about it). The content constantly sent through them creates the complex in room waveform.
We can measure how a speaker handles any soundwave and what the in-room response is at any point during a song. Spectral analysis gives us a great deal of info.
We can use all these measurements to compare speakers. Some of the comparison will not be audible some of the comparison will be audible.
Im far from being an expert on this but for years I have ask for people to post what isnt measureable.
Rather than just dismissing my supposition, actually pointing out WHY it is wrong or false, and how to do measurements that do show this would be constructive, and meet the goals of the OP's first post
You seem to think that I as an "Unknown" in some way was attacking Zaph and Dr Geddes, by daring to suppose that maybe there could be some things that current measurements can't or don't show (a supposition that seems to have caused you to place me in the subjectivists camp). I would put it to you that one who doesn't question, will not learn, and to blindly accept everything that someone with more knowledge than oneself says is foolish. If this was how all people who have done research in audio thought, then we would never have advanced at all!! Ironic isn't it
Tony.
Again, I apologize but I didn't post saying no one should question Zaph or Geddes. I posted that no one should question the data without posting some data to back up that questioning.
I do not blinding accept anything. People like Zaph and Geddes offer raw data and opinion that we can choose to use as much as we like. Geddes does hard core research, Zaph is just a audio hobbies that loved measuring speakers. You have to take it all with a grain of salt. I do not question Zaph measurements, CSDs, Distortion measurements matter to me EVEN if they are not audible. That is my choice alone. I take all those and buy the drivers that fit all my needs, I then listen to see if I like the drivers. The data comes first the subjective listening comes second, I would never make a conclusion on the value of a driver without both.
Last edited:
This is not a study, but an actual case:
I contracted a company to do some relayout of an amplifier. The main criteria was it cannot be of inferior sound quality compared with my prototype. Well, when they finished a sample, we hooked it up for listening comparison first, and it was clearly inferior compared to my prototype. To avoid any controversy, we just did distortion measurements with the real load connected; sure enough, the new layout had worse distortion figures, and they were obligated to redo the layout. At this point, I made a few suggestions on criterias to follow. The modified layout was then compared in listening tests, and sounded very similar to the prototype. We also did distortion measurements as before, and sure enough, the distortion figures were comparible.
I'm not sure whether we can say there is a correlation or not, but it sure was useful in this particular case.
Distortion is distortion. If you 'dumb it down' to a single THD number, I'll agree it's pretty meaningless. But if you measure high levels of high-order harmonics, whether it's in the driver or in the amp, most people are going to hear that. Olson showed it 50 years ago and Earl's distortion formula shows it by weighting the high orders more heavily.
Now you can say most decent drivers are 'good enough' but that's not the same as claiming distortion doesn't matter. Earl is using very good drivers in his systems so of course it's not such a big deal. B&C costs a lot more than Pyle.
I was talking about "typical" distortion in loudspeakers, not electronics - that's different. Electronics can easiliy generate "types" of nonlinearities that are audible, loudspeakers cannot. And, as pointed out, I am careful not to say that distortion in a loudspeaker cannot be audible, it can. I am saying two things 1) THD and IMD are meaningless numbers even if the distortion is audible and 2) in a decent loudspeaker system design the nonlinearities are such that audibility is not an issue. It CAN BE if the system is poorly designed, and lest face it a whole lot of them are. But does a plot of THD versus frequency of a driver tell us anything? - No, not much at all. Individual orders above fourth or fifth "might be" meaningful (this has never been quantified), but a simple test will show that these orders are difficult to impossible to measure reliably because they are generall so low. 2nd and third order are irrelavent, and fourth and maybe fifth not very important. The haigher orders are very hard for a loudspeaker to generate because it is a massive mechanical system and these orders require a lot of force. Electrons are pretty light and generating the higher orders is not hard at all - in fact it is essentially inevitable - but no one looks at these orders. The high order content as the signal is LOWERED is highly correlated with an amplifiers sound quality - otherwise known as crossover distortion.
There are distortions in time, in magnitude, and in frequency. I always concern myself with time as the primary and then dynamic matching of magnitude and last is frequency because frequency is by far the easiest to fix with the tools available.
If you can find it you should read the "log" of when 5 (maybe less or more) test equipment manufacturers showed up at an ALMA meeting with new equipment to measure phase in speakers. Vance Dickinson got all to agree to measure the same speaker in the same position with a microphone all agreed upon in a singular position also. All manufacturers were certain their system was the one that was right. The test results of all five were quite different. This caused a big problem and several equipment manufacturers left ALMA because of this. Check this out with Vance Dickinson, Lion Audio/Video Consultants .
My point- there are one heck of a lot of assumptions when measuring loudspeakers.
If you can find it you should read the "log" of when 5 (maybe less or more) test equipment manufacturers showed up at an ALMA meeting with new equipment to measure phase in speakers. Vance Dickinson got all to agree to measure the same speaker in the same position with a microphone all agreed upon in a singular position also. All manufacturers were certain their system was the one that was right. The test results of all five were quite different. This caused a big problem and several equipment manufacturers left ALMA because of this. Check this out with Vance Dickinson, Lion Audio/Video Consultants .
My point- there are one heck of a lot of assumptions when measuring loudspeakers.
Distortion is distortion. If you 'dumb it down' to a single THD number, I'll agree it's pretty meaningless. But if you measure high levels of high-order harmonics, whether it's in the driver or in the amp, most people are going to hear that. Olson showed it 50 years ago and Earl's distortion formula shows it by weighting the high orders more heavily.
Now you can say most decent drivers are 'good enough' but that's not the same as claiming distortion doesn't matter. Earl is using very good drivers in his systems so of course it's not such a big deal. B&C costs a lot more than Pyle.
I posted the question above because I didnt know. I do know he does separate the two in all discussions. His post above confirms that he keeps them separate.
YES!!!
Floyd Toole's Sound Reproduction, 2008
As far as I know, this is the only published body of work performed on a large scale, with many people participating, and performed in a rigorous way (aka, scientific method). IMO, this should be required reading before posting.
The issue is basic: there are many many variables when it comes to sound reproduction - trying to explore every variable individually (which is what is required to understand the relationships between things) would take any one of us a lifetime. And even if that were possible, it would only be one person's interpretation. Thats why it is very very important to have lots of different people involved.
Only large companies have these sort of resources, and I have tons of appreciation that someone like Toole (and Harman) would share all this valuable information.
Interestingly, I have also done some distortion measurements that could not be repeated after I used a different test control box with some changes in the interface. Driver distortion measurements are also very hard to coorelate unless the test setup is designed to look into a specific problem. If the driver is not operating in a piston mode, results will vary enough to be misinterpreted.
There's the rub. If you design the system so distortion isn't a problem, then distortion isn't a problem.
An example might be running a metal cone high enough that the cone breakup isn't suppressed enough. Anyone can hear that (fatiguing) sound. The driver may sound fine if you cross it low enough and not so good if you cross it too high. Zaph's measurements of the otherwise excellent Seas W18 show the problem. The 5K breakup has huge spikes in all the harmonics at 5K. There's also a smaller spike in the 3rd at 1.7K (5K/3) and the 5th at 1K (5K/5). If he had plotted the 7th and 9th, they'd probably show similar behavior at 5K/7 and 5K/9.
Attachments
Last edited:
Subs are a different thing, but the basic ideas still hold. The thing is that I did some tests a few months ago on some subs and found that many commercial ones had distortion that was extremely high. I mean totally modified and often clipped waveform and ac tually a lot of correlated noise. These subs would be audible - but then I'd really just call them "broken". My subs were a lot cleaner, but I could still track some port noise and so I had to redesign them. The redesign worked very well. This "port noise" would be highly audible, but would not show up as THD in many tests. I think that subs have to be evaluated differently.
There's the rub. If you design the system so distortion isn't a problem, then distortion isn't a problem.
An example might be running a metal cone high enough that the cone breakup isn't suppressed enough. Anyone can hear that (fatiguing) sound. The driver may sound fine if you cross it low enough and not so good if you cross it too high. Zaph's measurements of the otherwise excellent Seas W18 show the problem. The 5K breakup has huge spikes in all the harmonics at 5K. There's also a smaller spike in the 3rd at 1.7K (5K/3) and the 5th at 1K (5K/5). If he had plotted the 7th and 9th, they'd probably show similar behavior at 5K/7 and 5K/9.
That data has obvious problems. It just doesn't look right. (Look at all the orders right at 5 k - they all jump up discontinuosly - not possible. ) But anyways, you are pointing to an extreme case, finding a unique problem and calling that proof of the general case. That doesn't work.
What I see done all the time is people comparing two curves nonlinearities and saying that one is better because the distortion is lower when there is no obvious problem. Its the marginal cases that become completely unreliable, not the extr'eme one. To call THD representaive of what we hear, it has to be consistant in all cases and its not. That makes it an unreliable indicator of perception. It can find the agregious problems, but it cannot be used to improve a loudspeaker that isn't broke.
This is not true in the least. These days anyone can do a superb job, IF they know WHAT they are doing. I do all my measurements in my living room on a PC.
See my commerical posts if you want to know if this works out OK or not.
Last edited:
Yes, that was my opinion as well, or something like that. So I don't accept his data, but I do accept his point that disortion measurements CAN find a "broken" speaker.
I think probably more a case of me honing in on one bit and then reading the rest with coloured glasses on
Yes I guess doing spectral analysis in room, and comparing to the spectral analisis of the original recording (probably after it has been through the amp) could be quite telling. The interpretation of the differences I guess is going to be the tricky part, something that only experience will bring I suspect. I guess when I first posted I was thinking along the lines of the measurements that we typically see posted, and taking the quotes at face value (especially zaph's one, less so Dr Geddes) that *those* typical measurements are the only things that are necessary to understand how a speaker will perform.
But the problem with that is if we knew then we would be able to work out how to measure it
It is the proposition that there may be things that we don't even know we should be measuring or that people haven't thought about that I find interesting No problem but we will have to agree to disagree on this (to a point). I agree 100% if someone is simply stating that someone elses position (especially one that is backed up by years of research) is wrong, but don't have anything whatsoever to back them up. But If someone is not saying that someone is wrong, but merely postulating that there might be something more, in the interests of maybe getting some discussion going, then I feel that is a different matter
Absolutely the point I was trying to make!! that sums it up very well!!
I think we are on the same page. cheers,
Tony.
Oh and I just thought I'd clarify why I said that the difference of opinion on the importance of distortion was amusing. It wasn't the fact that the two have a difference of opinion, it was more to do with context of it in regards to this thread, and the fact that I felt that these two people were being held up as sources of truth, when on at least one aspect there was a very big difference of opinion
I was actually very pleased when I first read about Dr Geddes findings with regards to distortion, because *my* particular drivers were ones that Zaph had pretty much dismissed as being no good because of too high distortion (among other things), me not really being in a position to test it started to have some doubts about my purchase, however a bit later it became clear (through the thread I posted near the begining) that the measurements Zaph had for T/S params for the drivers in question were very suspect indeed, probably due to damaged or very out of spec drivers). This is a good case in point about how we can pick and choose which research we think is important, ie the stuff that validates our own beliefs or prejudices!
I had another possibly unpopular thought this morning as well (and maybe I should wait till I've fully read the Toole paper, but here it is anyway)
In order to be able to properly interpret the results of measurements one has to surely have a baseline to work off. How does one get that baseline? I imagine experience is the key, ie testing, and listening, and then correlating things in the measurement to things that one hears, and training oneself to be able to spot the things that matter. However, doesn't this then completely remove the objectivity, as the baseline would be arrived at from a subjective viewpoint? Even if we can agree on a purely objective reference (as Toole appears to be attempting to do), how much subjectivity is in the interpretation of the results for any abitrary set of measurements? After all it is a human brain processing the data that is presented in visual form.
Just another thought...
Tony.
I was actually very pleased when I first read about Dr Geddes findings with regards to distortion, because *my* particular drivers were ones that Zaph had pretty much dismissed as being no good because of too high distortion (among other things), me not really being in a position to test it started to have some doubts about my purchase, however a bit later it became clear (through the thread I posted near the begining) that the measurements Zaph had for T/S params for the drivers in question were very suspect indeed, probably due to damaged or very out of spec drivers). This is a good case in point about how we can pick and choose which research we think is important, ie the stuff that validates our own beliefs or prejudices!
I had another possibly unpopular thought this morning as well (and maybe I should wait till I've fully read the Toole paper, but here it is anyway)
In order to be able to properly interpret the results of measurements one has to surely have a baseline to work off. How does one get that baseline? I imagine experience is the key, ie testing, and listening, and then correlating things in the measurement to things that one hears, and training oneself to be able to spot the things that matter. However, doesn't this then completely remove the objectivity, as the baseline would be arrived at from a subjective viewpoint? Even if we can agree on a purely objective reference (as Toole appears to be attempting to do), how much subjectivity is in the interpretation of the results for any abitrary set of measurements? After all it is a human brain processing the data that is presented in visual form.
Just another thought...
Tony.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.