There is much talk about a certain loudspeaker design having better "dynamics" than another, and other concepts along these lines.
I have been recently doing some tests using dynamic signals on loudspeakers rather than the mostly static tests that are performed. Some very interesting results have occured.
The point of this thread is to ask if anyone knows of or has done dynamic tests of a loudspeakers performance. How was the test performed? What were the results? Conclusions?
Please, analytical tests only. I'm not really interested in subjective interpretations of a listening test.
To give you an idea, here is the test that I performed.
I sent out a white nose signal at a very low level. I use this to represent the steady state linear response, lets call it the reference.
Then I blasted the loudspeaker with a very high level noise signal, just about at its thermal limit (the 3 dB compression point) for about 20 seconds. Then the signal is dropped back down to the initial low level. I calculated the frequency responses of the system at the very high level burst and then down into the low level as it cooled. I plot the differences from the reference as a range of curves for different times - 4 sec, 8 sec, 12 sec, etc.
In some designs the frequency response differences from the reference would remain below about 2 dB everywhere and typically at about 1 dB through the large signal burst and would recover almost immediately.
In other designs the frequency response difference would be as high as 4-6 dB with an almost 2 dB variation across the bandwidth. The recovery time was slow.
Does this seem like a viable test? Any comments? Its fairly easy to perform, although you can destroy the speaker if you are not careful and it also gets real loud!!
I have been recently doing some tests using dynamic signals on loudspeakers rather than the mostly static tests that are performed. Some very interesting results have occured.
The point of this thread is to ask if anyone knows of or has done dynamic tests of a loudspeakers performance. How was the test performed? What were the results? Conclusions?
Please, analytical tests only. I'm not really interested in subjective interpretations of a listening test.
To give you an idea, here is the test that I performed.
I sent out a white nose signal at a very low level. I use this to represent the steady state linear response, lets call it the reference.
Then I blasted the loudspeaker with a very high level noise signal, just about at its thermal limit (the 3 dB compression point) for about 20 seconds. Then the signal is dropped back down to the initial low level. I calculated the frequency responses of the system at the very high level burst and then down into the low level as it cooled. I plot the differences from the reference as a range of curves for different times - 4 sec, 8 sec, 12 sec, etc.
In some designs the frequency response differences from the reference would remain below about 2 dB everywhere and typically at about 1 dB through the large signal burst and would recover almost immediately.
In other designs the frequency response difference would be as high as 4-6 dB with an almost 2 dB variation across the bandwidth. The recovery time was slow.
Does this seem like a viable test? Any comments? Its fairly easy to perform, although you can destroy the speaker if you are not careful and it also gets real loud!!
That is a curious test and an interesting result.
Two things that come to my mind are
1) measure something like "Slew Rate" ie how fast can the output go from zero to something "big" If there is something akin to compression (which I think of as the opposite of "dynamic"), then there would be a depressed slew rate. The problem, however, is that you may end up simply with a measure that is contaminated by the device's bandwidth (less bandwidth then a seemingly lower slew rate).
2) The other possibility is to measure some akin to a frequency response using white noise or another "steady state signal" and then perform the following two measures. First measure the frequency response using an MLS sequence or a Golay code (that is a series of impulses). Do this at a "reasonable level" and then do it again but this time increase the level by 20 or 40 dB. Is the response comparable? The obvious disadvantage would be that at 40 dB higher you may simply be running the device into clipping or distortion and any difference may simply represent that. IOW, a change in response could be either caused by the introduction of distortion or by the inability to reproduce "dynamics". Difficult to distinguish. Another difference is that with the extra 40 dB, you simply have a cleaner measure since you are now out of the noise floor (environmental etc).
A very different possibility is look at some of the statistics associated with band passed "white noise". One of those is the "crest factor" (there are others that correspond to the "peakedness" of the waveform). Here you would look at whether the peakedness (perhaps crest factor) on the output measures the same when it is manipulated (again at an level of an extra 40 dB or so).
The caveat on all this is that when you start putting big pulses through your device you could easily ruin the components if you are not careful.
There is a technique used in psychoacoustics, that employs signals that have their peakedness manipulated using "Schroeder Phase". Perhaps that is worth pursuing (varous references in JASA) as a signal of choice
Interesting to speculate,
-Tom
Two things that come to my mind are
1) measure something like "Slew Rate" ie how fast can the output go from zero to something "big" If there is something akin to compression (which I think of as the opposite of "dynamic"), then there would be a depressed slew rate. The problem, however, is that you may end up simply with a measure that is contaminated by the device's bandwidth (less bandwidth then a seemingly lower slew rate).
2) The other possibility is to measure some akin to a frequency response using white noise or another "steady state signal" and then perform the following two measures. First measure the frequency response using an MLS sequence or a Golay code (that is a series of impulses). Do this at a "reasonable level" and then do it again but this time increase the level by 20 or 40 dB. Is the response comparable? The obvious disadvantage would be that at 40 dB higher you may simply be running the device into clipping or distortion and any difference may simply represent that. IOW, a change in response could be either caused by the introduction of distortion or by the inability to reproduce "dynamics". Difficult to distinguish. Another difference is that with the extra 40 dB, you simply have a cleaner measure since you are now out of the noise floor (environmental etc).
A very different possibility is look at some of the statistics associated with band passed "white noise". One of those is the "crest factor" (there are others that correspond to the "peakedness" of the waveform). Here you would look at whether the peakedness (perhaps crest factor) on the output measures the same when it is manipulated (again at an level of an extra 40 dB or so).
The caveat on all this is that when you start putting big pulses through your device you could easily ruin the components if you are not careful.
There is a technique used in psychoacoustics, that employs signals that have their peakedness manipulated using "Schroeder Phase". Perhaps that is worth pursuing (varous references in JASA) as a signal of choice
Interesting to speculate,
-Tom
I assume Earl is talking about power compression. This should not be affected by an amps slew rate since both are worlds apart in terms of time.
I did indeed take care of that (though I didn't measure anything) by using a HiFi driver with P.A. driver sized voice coil (3/4" for 180 Watts). But I think this is a bigger issue in the PA busines, isn't it ?
Regards
Charles
Hello Earl
You might want to take a look at the Vented Gap Cooling Tech sheet in the JBL Tech library on the Pro site. They test a group of 15" tranducers with a 300watt input signal and ploted the power compression with time. Not what you are doing but you might find it interesting if you have not seen it already.
They have second one called Distortion and Power Compression which looks like it was tied to the release of SFG Motor. Here's the AES paper cited in it.
2. M. R. Gander, “Dynamic Linearity and Power
Compression in Moving-coil Loudspeakers:’ presented
at the 76th Convention of the Audio Engineering Society,
8-11 October 1984, New York; preprint number 2128 (E-11).
Rob🙂
You might want to take a look at the Vented Gap Cooling Tech sheet in the JBL Tech library on the Pro site. They test a group of 15" tranducers with a 300watt input signal and ploted the power compression with time. Not what you are doing but you might find it interesting if you have not seen it already.
They have second one called Distortion and Power Compression which looks like it was tied to the release of SFG Motor. Here's the AES paper cited in it.
2. M. R. Gander, “Dynamic Linearity and Power
Compression in Moving-coil Loudspeakers:’ presented
at the 76th Convention of the Audio Engineering Society,
8-11 October 1984, New York; preprint number 2128 (E-11).
Rob🙂
I am talking about power compression in loudspeakers and not slew rate (in a linear system slew rate is bandwidth).
I will have to look at the JBL stuff, although I do recal it.
Clearly the higher the power level the bigger this issue is. For low level signals like background music, its not much of an issue. But I suspect that these factors take place at any realistic listening level where one has the playback level in the 90's.
I was testing pretty big stuff at pretty high levels, but I suspect that a 1" tweeter would be affected by what I measured at normal listening levels. These little devices can't take much heat without seriuos changes in their performance.
I think that my point here is that this seems to be an ignored area as only the high output pro guys do anything with it. But I suspect that its an issue even in a home setting.
I will have to look at the JBL stuff, although I do recal it.
Clearly the higher the power level the bigger this issue is. For low level signals like background music, its not much of an issue. But I suspect that these factors take place at any realistic listening level where one has the playback level in the 90's.
I was testing pretty big stuff at pretty high levels, but I suspect that a 1" tweeter would be affected by what I measured at normal listening levels. These little devices can't take much heat without seriuos changes in their performance.
I think that my point here is that this seems to be an ignored area as only the high output pro guys do anything with it. But I suspect that its an issue even in a home setting.
"I was testing pretty big stuff at pretty high levels, but I suspect that a 1" tweeter would be affected by what I measured at normal listening levels. These little devices can't take much heat without seriuos changes in their performance."
There was an article in Stereophille where they used the temperature rise in the voice coil as a method to prove that power compression was not an issue with music played at "average" levels. You might want to look at what they did. The idea of measuring the temperature rise and tracking the temperature as it decreases tied into the measured response changes might be interesting. Just food for thought.
Rob🙂
There was an article in Stereophille where they used the temperature rise in the voice coil as a method to prove that power compression was not an issue with music played at "average" levels. You might want to look at what they did. The idea of measuring the temperature rise and tracking the temperature as it decreases tied into the measured response changes might be interesting. Just food for thought.
Rob🙂
" .....
I assume Earl is talking about power compression. This should not be affected by an amps slew rate since both are worlds apart in terms of time.
...."
You have completely misunderstood my comment. I probably was not being clear. I was certainly not referring to measuring an amplifier's slew rate. Consider the analogy of what a mechanical system is doing: going from zero to some degree of excursion. It would be difficult to measure this at a mechanical evel (perhaps laser interferometry?). But one could measure the response to a step input of various step sizes. This of course would simply be measuring the faithfulness of scaling in terms of a linear system.
The other suggestions were simply descriptions of measuring frequency response with test signals of various degrees of "peakedness" re: to a baseline (steady-state).
-Tom
I assume Earl is talking about power compression. This should not be affected by an amps slew rate since both are worlds apart in terms of time.
...."
You have completely misunderstood my comment. I probably was not being clear. I was certainly not referring to measuring an amplifier's slew rate. Consider the analogy of what a mechanical system is doing: going from zero to some degree of excursion. It would be difficult to measure this at a mechanical evel (perhaps laser interferometry?). But one could measure the response to a step input of various step sizes. This of course would simply be measuring the faithfulness of scaling in terms of a linear system.
The other suggestions were simply descriptions of measuring frequency response with test signals of various degrees of "peakedness" re: to a baseline (steady-state).
-Tom
My second thought is that you are correct, I am not referring to power compression, I was talking about compression of instantaneous amplitude (IOW clipping).
I still think the output to step response would "change" if it were now riding on a carrier frequency that was either at a "big level" or a "small level". If the carrier (steady state) were narrow band, its effects could be separated from the response of the step function (or the alternative, "peaked" signals).
-Tom
I still think the output to step response would "change" if it were now riding on a carrier frequency that was either at a "big level" or a "small level". If the carrier (steady state) were narrow band, its effects could be separated from the response of the step function (or the alternative, "peaked" signals).
-Tom
WithTarragon said:
This is correct, but this is a system nonlinearity and not a thermal compression issue which has a much longer time constant. Nonlinearities act instaneously, but thermal issues have very large lag (compared to the signal changes).
Interesting thoughts - a good subject.
What struck me right away is the time period. 20 seconds at high level? That's not what I would consider dynamics. Tho it may be some measure of ultimate dynamic power handling, it's not what most people refer to as "dynamics."
Isn't the subjective impression of dynamics a much shorter loud burst? Say 1 second or less? The exception might be going into a fortissimo section for a few bars- and being able to maintain it. But my impression of a dynamic system is one that can jump up to a much higher level with ease. Maintaining that level just adds to the sensation, but does not initiate the sensation.
How quickly do the ears start to compress?
I know that Dr. G. was asking for no subjective listening tests, but it seems to me that 20 seconds on-off is not what most people experience as "dynamic." Shorter peaks are.
FWIW, I can use the same driver, same amp, but a different box or baffle and get a different feeling of dynamics. Why? Is it simply different efficiency, or something else?
What struck me right away is the time period. 20 seconds at high level? That's not what I would consider dynamics. Tho it may be some measure of ultimate dynamic power handling, it's not what most people refer to as "dynamics."
Isn't the subjective impression of dynamics a much shorter loud burst? Say 1 second or less? The exception might be going into a fortissimo section for a few bars- and being able to maintain it. But my impression of a dynamic system is one that can jump up to a much higher level with ease. Maintaining that level just adds to the sensation, but does not initiate the sensation.
How quickly do the ears start to compress?
I know that Dr. G. was asking for no subjective listening tests, but it seems to me that 20 seconds on-off is not what most people experience as "dynamic." Shorter peaks are.
FWIW, I can use the same driver, same amp, but a different box or baffle and get a different feeling of dynamics. Why? Is it simply different efficiency, or something else?
".....
What struck me right away is the time period. 20 seconds at high level? That's not what I would consider dynamics. Tho it may be some measure of ultimate dynamic power handling, it's not what most people refer to as "dynamics."
Isn't the subjective impression of dynamics a much shorter loud burst? Say 1 second or less? The exception might be going into a fortissimo section for a few bars- and being able to maintain it. But my impression of a dynamic system is one that can jump up to a much higher level with ease. Maintaining that level just adds to the sensation, but does not initiate the sensation.
How quickly do the ears start to compress?
....."
Interesting points. I agree that a transducer that has "good dynamics" (like a horn loaded cabinet, for instance) is a "short term phenomenon. That is the confusion (at least mine) about the term dynamics. Slow changes over 10s of seconds seems to be a different phenomenon, but maybe I am getting caught up in semantics. The original question was about this slower process, which seems to be a slower build wher the system then goes into a non-linearity.
Re: compression in the ear. There are two issues. First is how long it takes to initiate the process (it is an active process and not all its mechanisms of initiation are understood). This happens on the order of 10s to 100s of milliseconds. Secondly, how fast the compression can act on a dynamic signal, once initiated, is certainly much faster. I am being deliberately vague since this system is not fully understood.
-Tom
What struck me right away is the time period. 20 seconds at high level? That's not what I would consider dynamics. Tho it may be some measure of ultimate dynamic power handling, it's not what most people refer to as "dynamics."
Isn't the subjective impression of dynamics a much shorter loud burst? Say 1 second or less? The exception might be going into a fortissimo section for a few bars- and being able to maintain it. But my impression of a dynamic system is one that can jump up to a much higher level with ease. Maintaining that level just adds to the sensation, but does not initiate the sensation.
How quickly do the ears start to compress?
....."
Interesting points. I agree that a transducer that has "good dynamics" (like a horn loaded cabinet, for instance) is a "short term phenomenon. That is the confusion (at least mine) about the term dynamics. Slow changes over 10s of seconds seems to be a different phenomenon, but maybe I am getting caught up in semantics. The original question was about this slower process, which seems to be a slower build wher the system then goes into a non-linearity.
Re: compression in the ear. There are two issues. First is how long it takes to initiate the process (it is an active process and not all its mechanisms of initiation are understood). This happens on the order of 10s to 100s of milliseconds. Secondly, how fast the compression can act on a dynamic signal, once initiated, is certainly much faster. I am being deliberately vague since this system is not fully understood.
-Tom
For those interested here's the link to Stereophille and the experiments that were published there. They didn't measure temp they monitored the rise in VC resistance and used that to determine the VC temperature.
http://www.stereophile.com/reference/1106hot/index.html
Rob🙂
http://www.stereophile.com/reference/1106hot/index.html
Rob🙂
Awhile back I did some calculations of thermal rise in a VC. It happens almost instantaneously. This means that the VC resistance is being modulated at a very fast rate. The magnet heating and other effects take much much longer. But I became interested in the effect of temperture rise of the voice coil since this happens very fast.
These are good points though and I will look to see if I can track the time aspects faster.
These are good points though and I will look to see if I can track the time aspects faster.
Robh3606 said:
All I can say about Mr. Howards results is that they contrast with mine. He tested for long enough that he should have seen a slow but steady degradation in VC resistance - but he didn't. My test was loud, but not much louder than what he claims, and in the two way system that I tested I saw 4 dB changes. He only indirectly measured VC resistance while I directly measured system response - so my test included all factors.
Looks like I cannot confirm his results.
Hello Earl
He used a music signal. Is that the same power distribution as your test signal? If you used a noise signal wouldn't your test get more power to the tweeter than a music signal which tends to roll off with frequency?
Rob🙂
He used a music signal. Is that the same power distribution as your test signal? If you used a noise signal wouldn't your test get more power to the tweeter than a music signal which tends to roll off with frequency?
Rob🙂
gedlee said:
Well that should certainly affect dynamics, fast or slow. No argument there.
Robh3606 said:
Indeed. Why white noise? Just to get a better idea of the effect? Some other reason?
You're right Tom, it may be a question of semantics or simply the definition of the term "Dynamics." I've generally heard it used to describe short bursts, but maybe that's not what Dr. G was looking for.
Robh3606 said:
Actually I used pink noise, I may have mistated. Pink noise is much closer to a music signal than white noise.
Well I always thought "dynamics" was abbreviated from dynamic range. In a driver, dynamic range is limited by power handling and hysteresis losses in the suspension.
A test we used to apply to amps can also be used on speakers.
A square wave at lower frequency is modulated with a higher freq (say 8x) at 1/8th, or so, the voltage. If the generators are synced, the resultant leading edge is vertical. Any variation in the waveform as the power is being increased is due to slew rate and dynamic range limitations, power supply sag etc. In speakers, non linear suspensions etc
I built a unit using digital dividers and adjustable master osc, but have misplaced it over the years. It's best to sample on a digital storage cro, too much time spent gazing at an analogue cro will cook a voice coil.
Geoff
A test we used to apply to amps can also be used on speakers.
A square wave at lower frequency is modulated with a higher freq (say 8x) at 1/8th, or so, the voltage. If the generators are synced, the resultant leading edge is vertical. Any variation in the waveform as the power is being increased is due to slew rate and dynamic range limitations, power supply sag etc. In speakers, non linear suspensions etc
I built a unit using digital dividers and adjustable master osc, but have misplaced it over the years. It's best to sample on a digital storage cro, too much time spent gazing at an analogue cro will cook a voice coil.
Geoff
Geoff H said:
To me this is a different phenomina than dynamic range. Its an effect that makes the sound seemed compressed even when its not. This is vastly different than the standard deffinition of dynamic range.
This is an interesting question as it relates to the typical 1" dome. I spend a lot of effort doing distortion testing of the typical 1" domes out there. While most of my test data is centered around the nonlinear distortion, I will occasionally observe an interesting phenomena. At high power levels and relatively high excursions, there will be a slow rise in the distortion signature of a driver which would fit the pattern of a thermally induced phenomena. I've really only looked at this in terms of nonlinear distortion, but I could easily look at FR changes. Praxis is pretty flexible in this way. We'd have to think out the details of a reproducible methodology though to make comparisons across drivers/systems valid.
www.markk.claub.net
www.markk.claub.net
- Status
- Not open for further replies.