no in mentioning amplifier performance it reminded me of the first time i heard a 4-way martin PA system with jbl tops that was powered with H/H Mosfet V series amps for the lows(v 800) and mid's(v 400) but used Bryston 3B's for the horns. that was the moment i knew higher fidelity could be expected from PA gear even horns.
for any other soundudes the x-over was BSS fds360 if memory serves.
for any other soundudes the x-over was BSS fds360 if memory serves.
Yes I understood and I know you quoted another person's statement.
I have forgotten to mention load impedance. The more "wild" load impedance (vs. frequency) with dips as low as like 1 - 2 ohm, the more will be the crossover distortion pronounced.
anyone have anything they would like to say on the topic of audibility of distortion in horns....
Amplifier distortion is not really my interest, but I did a little work on it. When I say "two amplifiers that measure the same" I was referring to the standard THD + Noise measures that are used and not the ability of one to measure differences in more advanced tests.
In more advanced tests, like I developed, it is possible to measure differences in very low level NLD. To what extent this is audible I really can't say because I did not do those tests. Most of the AVRs that I tested were fine - the one that I used was fine, but there were a few others that weren't so good. The ones with chip amps seem to be the best. So maybe not "any" - there will always be excepts - but "most".
Clearly a casual statement: which should read:
"There could be audible differences between amplifiers that measure the same on standard tests."
The question is what is a "standard test". How is it defined? Is it a set of measurements that John Atkinson performs in Stereophile, e.g.?
Parasound Halo Integrated integrated amplifier Measurements | Stereophile.com
Parasound Halo Integrated integrated amplifier Measurements | Stereophile.com
or everything or a recent APx555 analyzer menu, documented in AP white papers?
which produces a few more graphs than could be fit in an entire issue of Stereophile
which produces a few more graphs than could be fit in an entire issue of Stereophile
Regarding DDF post, it reminds me of the situation that I found when I started to study NLD. There were lots of hypothesis about the thresholds of perception of THD, etc. for sine tones and they seemed all over the map. And pure tones are such an idealistic test signal - not very close to reality - that I had my doubts about its validity. What was clear is that there is a threshold below which no NLD could be heard, but there was no metric that could be used to say that "A is better than B" regarding NLD. One might be able to say that both had inaudible NLD but one could not say that if they were both above that level which one is preferred. To study the perceptual effects of NLD one had to have a valid metric of the underlying problem.
This is where things started to come apart because it was clear very early on that THD and IMD did not meet the requirements, because they lacked any correlation to perception. One could not say if a system had audible or inaudible NLD and if it did was it only a "little bad" or "terrible"? Basically a situation where no one could really say much of anything about the audibility of NLD with any certainty.
We showed that metrics could be developed that were accurate, but alas, that was a lot of work and no one wanted to go through all that. "We all know that THD is not accurate, but it is what we have always used." It amazes me how even today people will report THD numbers as if they had some meaning.
This is where things started to come apart because it was clear very early on that THD and IMD did not meet the requirements, because they lacked any correlation to perception. One could not say if a system had audible or inaudible NLD and if it did was it only a "little bad" or "terrible"? Basically a situation where no one could really say much of anything about the audibility of NLD with any certainty.
We showed that metrics could be developed that were accurate, but alas, that was a lot of work and no one wanted to go through all that. "We all know that THD is not accurate, but it is what we have always used." It amazes me how even today people will report THD numbers as if they had some meaning.
THD vs level, frequency actually gives similar information as the GedLee Metric - just not a single number
if the THD vs level, frequency is further shown as discrete harmonics vs level then you do in principle have all the information needed for calculating the GedLee Metric from the graphs within the limitation of applicability of the static nonlinearity assumption built into the GedLee Metric
measuring distortions below the noise may have some merit
needed resolution clearly depending on all of "system gain structure", source and listening room noise floors
but even 10 second averaging should be adequate considering human hearing seems to "average" more on the order of 100 ms time scale
if the THD vs level, frequency is further shown as discrete harmonics vs level then you do in principle have all the information needed for calculating the GedLee Metric from the graphs within the limitation of applicability of the static nonlinearity assumption built into the GedLee Metric
measuring distortions below the noise may have some merit
needed resolution clearly depending on all of "system gain structure", source and listening room noise floors
but even 10 second averaging should be adequate considering human hearing seems to "average" more on the order of 100 ms time scale
There is no static linearity assumption in the metric. We used only static nonlinearities in the tests, but that was for simplicity. The metric does not assume this.
As some British author pointed out, one has to consider the phases of the harmonics if a valid measure of the metric is to be worked out. But, yes, if you had the complex magnitudes of all the harmonics at every frequency and level, then you have all the information that you need. You just need to figure out how to deal with all that information and how to simplify it down to something usable - like a single number - that can be shown to be correlated to perception. Sounds simple enough.
As some British author pointed out, one has to consider the phases of the harmonics if a valid measure of the metric is to be worked out. But, yes, if you had the complex magnitudes of all the harmonics at every frequency and level, then you have all the information that you need. You just need to figure out how to deal with all that information and how to simplify it down to something usable - like a single number - that can be shown to be correlated to perception. Sounds simple enough.
Sure.
By the late 1970's I had decided to use EV cone drivers and JBL HF drivers. The EV cone drivers had less power compression than the JBL cones of the time, but the JBL HF drivers were superior in sound quality to any of the EV drivers I was familiar with.
By around 1983 we (Southern Thunder Sound, Inc.) were using JBL 2445s (4" diaphragm, 2" exit) for hi mid and 2425 (1.75" diaphragm, 1" exit) for high frequency. The 2445 had too much HF distortion to be acceptable running from 1200 Hz to past 5000 Hz for mains use, though we did use them without tweeters in stage monitors on 2385 horns.
The bi-radial 2385 horns use a pinched throat, the narrow aperture allows for wide HF dispersion, but the diffraction effects also makes HF distortion sound more objectionable than the round throat exponential horns used in the main PA system.
In 1992 I did a series of tests comparing 2445s to the recently introduced EVDH1 drivers (3"diaphragm, 2" exit) and found that they had far less HF distortion than the 2445, and could be used all the way up from 1200 Hz, and sounded cleaner than the combination of 2445 and 2425 covering the same range.
My experience does lead me to agree with Earl Geddes that diffraction, HOMs, and other forms of linear distortion certainly can have an undesirable effect on the sound quality of horns or waveguides.
His claims that non linear distortions have no undesirable effect on sound quality (until they suddenly do) simply don't match what I can still hear after 41 adult years in the sound reinforcement and recording industry.
It amazes me that Earl still is amazed "people will report THD numbers as if they had some meaning" when the Klippel distortion tests (as well as the many reports DDF listed in post #210) clearly show that they do have meaning to thousands of respondents.
Art
Agreed. THD vs frequency (with individual harmonics plotted), THD vs level /it is a lot of plots, at any frequency vs. level and at any level vs. frequency/, but let's not forget plots vs. complex load impedance - this is sometimes extremely interesting. Below noise - any narrow band spectral analysis sees below noise integrated over audio band.
the result of JAES articles investigating of "Interface Distortion" is pretty clear - for common audio power amp topologies the nonlinearity with complex impedance load looks no worse than can be extrapolated from loading the output to the same current with a resistive load
sustained continuous tone at large amplitude with complex load can stress thermal design of the output stage - again can be designed for - you just have to spec what signals, loads, amp behavior is acceptable and do the numbers
so yes, you have to have required peak current capability - various assumptions can give over 5x the nominal speaker name plate resistive load calculated peak currents
Self considers only 2x as practical for real music, "not insane" speakers unlike the early Watt Puppy which notoriously dropped to 1 Ohm Z at one point in the audio frequency range
but this really doesn't address the thread topic - we could easily just use a active XO, dedicated power amp for a horn driver with whatever overratings that floats your boat
sustained continuous tone at large amplitude with complex load can stress thermal design of the output stage - again can be designed for - you just have to spec what signals, loads, amp behavior is acceptable and do the numbers
so yes, you have to have required peak current capability - various assumptions can give over 5x the nominal speaker name plate resistive load calculated peak currents
Self considers only 2x as practical for real music, "not insane" speakers unlike the early Watt Puppy which notoriously dropped to 1 Ohm Z at one point in the audio frequency range
but this really doesn't address the thread topic - we could easily just use a active XO, dedicated power amp for a horn driver with whatever overratings that floats your boat
Last edited:
Cross-over distortion is more audiable in any speaker where the amplifier is outputting a level of power that is in the range of cross-over.
dave
Good point, it may be more audible with more sensitive speakers, as lower output voltage is used.
RuneStone said:
Well, from a strict physical angle I think it's fair to say it's not a resonance until energy is being reflected back to build the amplitude to a higher level, and that takes time just as when the wave is travelling in air to be diffracted and reflected on an edge. Only the actual medium is different.
Yes, something will happen when the force initially is hitting the medium, in this case the cone material, but the wave is propagating to the edge of the cone, more or less reflecting back when reaching surround and the fixation point of the surround in the chassis and after som time energy is building up to a higher amplitude and we have a resonance.
A TDR measurement of cable network is more or less the same thing. If a transsmission line has infinite length or proper termination nothing is reflected back to build a resonance. If the cable has a finite length and is poorly terminated reflections and a standing wave is a fact. But there is a time factor involved and when input of energy is stopped there will be some time before decaying to zero. This time can easily be longer in a speaker cone than the time involved for an edge reflection/diffraction. And that was my point, there's nothing instant in a cone resonance more than a reflection based on a change of load impedance of the air in the horn.
It takes time for a resonace to build up to full amplitude and it takes time for the resonance to decay to zero after turning off the input signal.. = not instant, but delayed.
Since the output is delayed vs input we have a phase shift and.. group delay.. (delay as in.. delay ;-)
Last edited:
- Status
- Not open for further replies.