Helmuth said:
Helmuth, while I don't agree it is "high end" performance I do agree that the distortion performance of the Summa is likely better than many typicaly speakers.
There are dome tweeters with 0.03-0.05%THD at 90dB and tall/long ribbon tweeters that has 0.03%THD at 110dB.
In the midrange (above 100Hz) there are speakers with aprox. 0.05%THD @ 100dB.
These kinds of speakers use to rock the socks of the typical hifi speaker with 0.5%THD at 90dB.
/Peter
gedlee said:
Is this test covered in any of your articles? I'm very curious about the test method.
Excuse me for being slightly sceptical about that claim unless you back it up. I have good reason to believe that it is a false statement.
/Peter
When you refer to diffraction, are you focusing on diffraction in a horn or a waveguide that produces the HOMs or diffraction in general?
Dave
Pan said:
Peter, you are really too generous!
USD$4800 list.
http://www.soundstagenetwork.com/measurements/speakers/monitor_audio_pl100/
Pan said:
This article was published in AES some years ago.
If you doubt my claim then you have obviously not seen the body of work that we have done in this area.
dlr said:
Hi Dave
All types of diffraction, of which HOM are one type, none of it is good if it is within 10 ms of the direct wave. I design out as much as possible in the speakers themselves and then I place them in the room to avoid room diffraction as much as possible as well. It all has a major effect.
Dr. Geddes,
Before I ask anything or say anything, I want to be clear that in general I agree with the conclusions that you have drawn regarding the thrust of your research in this matter. I think it is very important and explains things (at last) that I and others have been saying for a long time, but now in concrete terms. In terms of the specifics, I admit that my math & theory does not match yours, so I am at a disadvantage in that area. However I do have some questions...
How does one listen to the output of a plane wave tube - and listen at 120dB (safely)??
Did the THD rise quickly as you reached toward that 120dB level (curious).
Another question: if one wanted to use the Gedlee Metric, how would that be possible, short of hiring your firm? If it is not possible, then that creates a barrier (minor or major depending on circumstances) to general acceptance of the "standard"?
Have you had in your hands a Stage Accompany tweeter?
They spec it at <1% THD at 128dB.
I have owned a pair - my experience with them was that WRT their sound, they made every other tweeter (1500Hz and up) sound rather less than "clear" by comparison. (and it is not because of a tipped up HF response nor a peak) (There are other issues integrating them into a system, but that's not one) Everything else sounds like "hash" by comparison - and I do not say such things capriciously or without serious consideration.
Put this in other terms, at least in my personal experience there is/was no way to confuse the SA with other HF reproducers - and not because of the obvious other issues like polar response, freq response, phase response, group delay, etc... so this makes me puzzle over the idea that there are (or may be) some drivers whose THD is relatively high, but will sound (to use this case) indistinguishable with the SA??
Regards,
_-_-bear
PS would you please relate the brand and model of the headphones you used, and what amplifier was used to power them?
Before I ask anything or say anything, I want to be clear that in general I agree with the conclusions that you have drawn regarding the thrust of your research in this matter. I think it is very important and explains things (at last) that I and others have been saying for a long time, but now in concrete terms. In terms of the specifics, I admit that my math & theory does not match yours, so I am at a disadvantage in that area. However I do have some questions...
gedlee said:
How does one listen to the output of a plane wave tube - and listen at 120dB (safely)??
Did the THD rise quickly as you reached toward that 120dB level (curious).
Another question: if one wanted to use the Gedlee Metric, how would that be possible, short of hiring your firm? If it is not possible, then that creates a barrier (minor or major depending on circumstances) to general acceptance of the "standard"?
Have you had in your hands a Stage Accompany tweeter?
They spec it at <1% THD at 128dB.
I have owned a pair - my experience with them was that WRT their sound, they made every other tweeter (1500Hz and up) sound rather less than "clear" by comparison. (and it is not because of a tipped up HF response nor a peak) (There are other issues integrating them into a system, but that's not one) Everything else sounds like "hash" by comparison - and I do not say such things capriciously or without serious consideration.
Put this in other terms, at least in my personal experience there is/was no way to confuse the SA with other HF reproducers - and not because of the obvious other issues like polar response, freq response, phase response, group delay, etc... so this makes me puzzle over the idea that there are (or may be) some drivers whose THD is relatively high, but will sound (to use this case) indistinguishable with the SA??
Regards,
_-_-bear
PS would you please relate the brand and model of the headphones you used, and what amplifier was used to power them?
Peter, please provide examples of the speakers/drivers you have referenced? Especially the "midrange" with 0.05% THD at 100dB??
Curious about this...
_-_-bear
Curious about this...
_-_-bear
bear said:
The test protocal was all in the AES paper, but was done by recording the signal in the PWT and playing it back over Etymotic ER4 Research earplugs. The listening level was always the same, but the driver level varied from 80 dB - about 120 dB in the PWT. The distortion did rise quickly with level. Now if the THD was audible at 25% versus the x% at lower levels then the listeners would have been able to detect this is the samples. Statistically no one could. There was no detection of a sound quality change over this range of playbacks. Criticize the test all you will, but until someone can provide better data, I'll stick with whats available.
Dr. Geddes,
Suggest that you might do well to not start out expecting "criticism"?
In this case I would agree that if the 25%THD was going to be audible, it ought to be audible even with the mic'ing and recording, and playback via whatever means... it's a VERY high level of distortion.
Of course a bit does depend on the order of the THD and nature of the stimulus signal.
I just downloaded your AES papers (again) and will take a look at them time permitting...
What about the other issues I raised?
_-_-bear
Suggest that you might do well to not start out expecting "criticism"?
In this case I would agree that if the 25%THD was going to be audible, it ought to be audible even with the mic'ing and recording, and playback via whatever means... it's a VERY high level of distortion.
Of course a bit does depend on the order of the THD and nature of the stimulus signal.
I just downloaded your AES papers (again) and will take a look at them time permitting...
What about the other issues I raised?
_-_-bear
bear said:
Of course criticism is expected, but that was a per reviewed AES paper and I've been all through the criticisms before. None of them has stood up as yet. People found the paper hard to accept, but no one has refuted it.
I have no data on your other issue so I can't really comment. I don't doubt your observations, but there is so much at play that there are lots of possibilities.
I would say that you might - for your own edification perhaps - borrow or otherwise acquire a pair of the SA ribbon drivers. They are not perfect, but they are sufficiently superior imho to audibly differentiate themselves from the rest of the pack.
The good part is that you can resell them without much difficulty, should you then wish to recover your investment.
Perhaps some of the resistance here and elsewhere is because it is very difficult to grasp your Gedlee Metric without a very strong mathematical background - which is in part both why it is valuable, and why it isn't/wasn't "discovered" until you did it. Most people (non-professionals) would feel much better about it if it could be reduced to a descriptive text and examples that they can mentally correlate back to the "real world."
Regards,
_-_-bear
The good part is that you can resell them without much difficulty, should you then wish to recover your investment.
Perhaps some of the resistance here and elsewhere is because it is very difficult to grasp your Gedlee Metric without a very strong mathematical background - which is in part both why it is valuable, and why it isn't/wasn't "discovered" until you did it. Most people (non-professionals) would feel much better about it if it could be reduced to a descriptive text and examples that they can mentally correlate back to the "real world."
Regards,
_-_-bear
I did that.
1) ignore THD and IMD in loudspeakers - it doesn't matter.
2) look at all electronics as the signal level falls - look for the appearance of higher harmonics.
What could be simpler than that? The metric only helped me to arrive at those conclusions and now that I have I have no more use for it. Thats why I don't push it.
1) ignore THD and IMD in loudspeakers - it doesn't matter.
2) look at all electronics as the signal level falls - look for the appearance of higher harmonics.
What could be simpler than that? The metric only helped me to arrive at those conclusions and now that I have I have no more use for it. Thats why I don't push it.
I just finished reading the paper "A Multiple Regression Model for Predicting Loudspeaker Preference Using Objective Measurements: Part II" from AES 117th 2004.
Basically, the authors were able to develop a model/equation that takes various frequency response and directivity measurements ONLY, and predicts with almost 100% accuracy how well the speakers sounded to a group of listeners!! (pearson R^2 was ~0.99)
Yes, this was blinded. They used 70 different loudspeakers ranging from $100 to $25,000. The used 19 different listening levels.
This basically tells us that loudspeaker sound quality, as determined by listeners, is almost 100% accounted for by measurements that can be derived from frequency response and directivity. Essentially, this is smoothness, flatness, and extension. It also means that DISTORTION is NOT an independent determinant of loudspeaker quality in commercially available loudspeakers. I wish I had seen this paper sooner. Question answered.
SG
Basically, the authors were able to develop a model/equation that takes various frequency response and directivity measurements ONLY, and predicts with almost 100% accuracy how well the speakers sounded to a group of listeners!! (pearson R^2 was ~0.99)
Yes, this was blinded. They used 70 different loudspeakers ranging from $100 to $25,000. The used 19 different listening levels.
This basically tells us that loudspeaker sound quality, as determined by listeners, is almost 100% accounted for by measurements that can be derived from frequency response and directivity. Essentially, this is smoothness, flatness, and extension. It also means that DISTORTION is NOT an independent determinant of loudspeaker quality in commercially available loudspeakers. I wish I had seen this paper sooner. Question answered.
SG
While there are some issues with that study, for the most part I agree with it. I certainly agree that nonlinear distortion IS NOT a factor in subjective preference of loudspeakers. But I do have issues with the range of loudspeakers evaluated as all were piston type sources which are not constant directivity. Since such sources were excluded one cannot say to what extent the results would have been influenced had they been included.
Whoops. I read the paper too fast. With their sample of 70 loudspeakers, their model only achieved an R of 0.86, which means it only accounts for ~75% of the variability in loudspeaker preferences. This model is based on-axis frequency response and predicted in-room frequency response (derived from anechioc measurements). The metric involves:
a) degree of frequency response smoothness over 1/2-octave narrow bands for the on-axis and predicted in-room curves from 100Hz to 12kHz
b) degree of smoothness and flatness for the predicted in-room curve for 100Hz to 16kHz
c) low frequency extension
However, they apparently created another model with an R of 0.94, R^2= 88%! And loh and behold, they DIDN'T print the equation...I assume it's very similar to the above, but who knows.
So basically we haven't accounted for 12-25% of the listener preferences, which means that distortion isn't necessarily excluded. In fact, the authors did find a weak correlation between 2nd/3rd HD and listener preference, wiith R^2 around 20% or so, however, they also found that speakers with more distortion also had worse frequency response, a potential confounder. They easily could have done multivariate analysis to demonstrate that distortion was not an independent predictor of listener preference, but they didn't. Imagine that.
However, I've seen enough expert opinion (Toole, Olive, and Geddes above) and some supportive studies to be convinced that nonlinear distortion in loudspeakers is NOT a significant factor in determining listener preferences. There is ample evidence that linear distortion, that is frequency response aberrations, is the dominant factor. Perhaps it is THE predominant factor, depending on whether you believe Olive can tweak his regression model to have a higher R^2 over more loudspeakers and listeners.
SG
P.S. The article is available for $5-$20 off the AES website. Toole's article is free off the Harman website.
a) degree of frequency response smoothness over 1/2-octave narrow bands for the on-axis and predicted in-room curves from 100Hz to 12kHz
b) degree of smoothness and flatness for the predicted in-room curve for 100Hz to 16kHz
c) low frequency extension
However, they apparently created another model with an R of 0.94, R^2= 88%! And loh and behold, they DIDN'T print the equation...I assume it's very similar to the above, but who knows.
So basically we haven't accounted for 12-25% of the listener preferences, which means that distortion isn't necessarily excluded. In fact, the authors did find a weak correlation between 2nd/3rd HD and listener preference, wiith R^2 around 20% or so, however, they also found that speakers with more distortion also had worse frequency response, a potential confounder. They easily could have done multivariate analysis to demonstrate that distortion was not an independent predictor of listener preference, but they didn't. Imagine that.
However, I've seen enough expert opinion (Toole, Olive, and Geddes above) and some supportive studies to be convinced that nonlinear distortion in loudspeakers is NOT a significant factor in determining listener preferences. There is ample evidence that linear distortion, that is frequency response aberrations, is the dominant factor. Perhaps it is THE predominant factor, depending on whether you believe Olive can tweak his regression model to have a higher R^2 over more loudspeakers and listeners.
SG
P.S. The article is available for $5-$20 off the AES website. Toole's article is free off the Harman website.
bear said:
Hi bear, check post #10 in this thread:
http://www.diyaudio.com/forums/showthread.php?s=&threadid=138461
/Peter
2100 said:
Peter, you are really too generous!
USD$4800 list.
http://www.soundstagenetwork.com/measurements/speakers/monitor_audio_pl100/
Ouch! That was not very funny to see! Surprisingly poor performance.
/Peter
- Status
- Not open for further replies.