I'm referring to dynamic range compression, not distortion. Two different things.
What's the difference?
You can have a loudspeaker producing low non-linear distortion, but that exhibits power compression.
Even though it'd be great to see a standard measurement for power compression added to stereophiles reviews, I would much rather see them add in measurements for non linear HD.
Sure, and equally offensive to say, that sounds like the defense of one who can't afford it/them. It's amazing how polarized the audio community can be.
Fortunately, I'm not offended and don't fall into either camp.
Exactly. Some of the mechanisms that produce dynamic range compression also produce distortion, for example over excursion, but some don't, for example thermal compression. (Re rise with instantaneous voice coil temperature)
But how audible is the harmonic distortion versus the dynamic range compression, when dynamic range compression is typically frequency dependant in speakers, resulting in changes in the frequency response with SPL ?
Dynamic range compression is distortion at higher SPLs. Toole shows the frequency response of a bunch of speakers at 80 db and at 100 db, IIRC. At 100 db, the frequency response of some samples changes drastically. To me, that's dynamic range compression.
I doubt power compression is a problem at home SPLs.
I doubt power compression is a problem at home SPLs.
Oh I wasn't trying to say how audible either of them are, more that HD measurements are easy enough to do and would help show any others flaws inherent to the design, beyond what stereophile normally test for.
Power compression I'd guess is a lot harder to meaningfully test for, although even if the test waa flawed, even a simple one should show the differences between one design and another.
Well anything that's contrary to the input signal is distortion, unless the effect is wanted. And yes, that is dynamic range compression but only half of it. You could have a loudspeaker who's frequency response changes by naught when driven by 1 watt, 100 watts or 1/10th of a watt, but when faced with a dynamic signal the instantaneous heating of the voice coil results in the dynamic being squashed by 1dB. This is of course just another face of the same coin mind you.
Indeed, I agree with you, most people listen at levels where this wouldn't be a huge concern, but it would help highlight weaknesses in an otherwise well put together design, just like testing for non linear distortion does.
I used to measure power compression curves on all the JBL systems I designed. They looked pretty in the brochures. Do we really think that is one of the key distortion mechanisms?
Note that there is a time constant involved in all the heating effects so a unit needs to be at a high level for a TC or 2 to see the effect. Curves need to be measured fairly slowly to see it at all.
The mechanism is that frequency response will drift slowly when sustained high power is applied. I can't imagine a more benign type of distortion.
The difference between class A and non-class A?
David S.
Note that there is a time constant involved in all the heating effects so a unit needs to be at a high level for a TC or 2 to see the effect. Curves need to be measured fairly slowly to see it at all.
The mechanism is that frequency response will drift slowly when sustained high power is applied. I can't imagine a more benign type of distortion.
The difference between class A and non-class A?
David S.
As 5th Element says anything that changes the signal is a distortion. The most fundemental thing about DDR is that it is a way of looking at things from a different direction.
dave
Indeed, this was the trouble I was referring to when doing these measurements meaningfully
It would be an issue if the maintained level results in frequency response shift that edges towards making a loudspeaker sound forward or fatiguing. This might not be an issue to lots of potential buyers, but for the ones who are going to listen predominantly at higher power levels it very well could be.
Besides that though, when testing systems for JBL did you find that large transients have an instantaneous affect on power compression? Or is the thermal mass too large to allow this to happen?
I am more talking of you're listening at an average power of say 10 watts with the coils sitting at say 80 degrees, then along comes a massive 300 watt transient, would the frequency response become compressed due to the instantaneous heating of the voice coil?
I can't comment about cost. That's irrelevant for DIYer
All listed parameters are important in my experience, but I'm aware that all parameters can't be 'the best globally exists' in the same loudspeaker. Some weighting must and could be done. All parameters are not equally significant in all rooms, speaker/listener locations, sound pressures and listener preferences of course. For example if room is acoustically very damped or listening distance is very short and listening spot is constant, smoothness of off-axis and power responses are not critical. Same story with low listening volume; there is no need to maximize SPL capacity. But if room sucks acoustically, listening distance is long, listening area is large and sound pressure high, it's better to optimize as much and many parameters as possible. This gives also flexibility if system is relocated in a different room, house etc.
The JBL tests were just done as slow frequency response sweeps as shown in most of the brochures in the 80's.
We typically did 1 and 100 watts (sometimes 1 and 10) and of course dropped the 100 watt curve by 20 dB so they should plot one on top of the other, less the compression effects.
You are measuring heating so it has to be real watts in the passband of a particular driver. As inductance rises the effect trails off. Bass input causes no HF compression, etc.
For a 300 Watt transient, you would need to calculate the amount in tweeters range and then consider the time constant. At KEF I measured time constants on a number of drivers for the KM1 project. Tweeters were about 5 second, mids maybe ten and woofers 20 to 30 seconds. That essentially means that the 300 watt transient gets smeared by the time constant. In fact, unless it lasts for 2 to 3 time constants it never gets close to its ultimate temperature but drops back to ambient as soon as the input stops. Temperature can only rise as fast as the time constant allows, which is how a small unit can (thermally) survive a 1000 watt input, if it is of short duration.
For the KM1 we eventually used an analog computer approach to model voice coil tmeperature: Input signal goes into a full wave recitifier and log converter for True RMS conversion and detection. The resultant signal goes into an RC averaging circuit with the measured time constant (actually two, one for the coil and a much longer one for the magnet circuit) to give a voltage proportional to real time VC temperature. The particular channel was calibrated to cut out at a certain voltage (temperature).
The earlier KEF approach was to inject some DC through each voice coil and monitor resistance rise to measure VC temp directly. We found that short transients did little to bump temperature above ambient.
It was when you said "turn it up some" and gave it another 10dB that things started to take off. Then a long sustained treble burst and the temperature would soar.
That is when voice coils are burned out.
David S.
We typically did 1 and 100 watts (sometimes 1 and 10) and of course dropped the 100 watt curve by 20 dB so they should plot one on top of the other, less the compression effects.
You are measuring heating so it has to be real watts in the passband of a particular driver. As inductance rises the effect trails off. Bass input causes no HF compression, etc.
For a 300 Watt transient, you would need to calculate the amount in tweeters range and then consider the time constant. At KEF I measured time constants on a number of drivers for the KM1 project. Tweeters were about 5 second, mids maybe ten and woofers 20 to 30 seconds. That essentially means that the 300 watt transient gets smeared by the time constant. In fact, unless it lasts for 2 to 3 time constants it never gets close to its ultimate temperature but drops back to ambient as soon as the input stops. Temperature can only rise as fast as the time constant allows, which is how a small unit can (thermally) survive a 1000 watt input, if it is of short duration.
For the KM1 we eventually used an analog computer approach to model voice coil tmeperature: Input signal goes into a full wave recitifier and log converter for True RMS conversion and detection. The resultant signal goes into an RC averaging circuit with the measured time constant (actually two, one for the coil and a much longer one for the magnet circuit) to give a voltage proportional to real time VC temperature. The particular channel was calibrated to cut out at a certain voltage (temperature).
The earlier KEF approach was to inject some DC through each voice coil and monitor resistance rise to measure VC temp directly. We found that short transients did little to bump temperature above ambient.
It was when you said "turn it up some" and gave it another 10dB that things started to take off. Then a long sustained treble burst and the temperature would soar.
That is when voice coils are burned out.
David S.
Last edited:
I can't comment about cost. That's irrelevant for DIYer
But if room sucks acoustically, listening distance is long, listening area is large and sound pressure high, it's better to optimize as much and many parameters as possible. This gives also flexibility if system is relocated in a different room, house etc.
Thats why these high end speakers cost $20,000 plus. The designers assume that every parameter must be perfected, rather than finding which are most crucial and where the thresholds of detection are.
Not my definition of good engineering.
David S.
If the studio is a recording studio this is very unlikely to ever happen because:
a) There probably are no million dollar speakers used in studios. One of the most expensive monitors I know of are Quested HM415 which cost £70k for the pair. This includes active crossovers made by XTA, a number of amps made by MC2 Audio and Roger Quested himself supervising the installation (even if the studio in question is in Mexico!).
b) They are likely to be soffit mounted.
Granted there are Westlakes which cost a fair bit more but I have never heard anything good about them from pros (I have not heard them myself though).
The speakers in the Stereophile's list don't look very well engineered, fully optimized and great sounding in my opinion - just very expensive. Optimizing of some basic parameters shouldn't be too expensive - at least for DIYer. Top quality finishing, special construction/materials and statement product image add $ to commercial product.
In what room size and room acoustics, and with what placement of speakers and listening position?
We will have to come up with a standard for all this before we can answer that question. And we know that typical bass response in a regular room can vary a lot. +/- 20 dB is not uncommon.
That is why speaker manufacturers design their speakers for either 4pi (acoustic free-field) or 2pi (half-space), or even somewhere between.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.