Hi Earl, no argument aboiut correcting a bad crossover the new one is far superior to my first effort (though the original one I was going to put in line level baffle step compensation).
I disagree though that it is not baffle step that I have compensated for.
The attached is the simulation I did of the baffle step of the drivers on the baffle before I made the speakers. The main component in the crossover for dealing with the baffle step was a notch filter at about 1.3Khz... I wanted a narrow baffle (this is 200mm). Some might say this is extreme, but I'm very pleased with the results. The actual crossover of course was made based on real measurements not the simulation.
Tony.
Attachments
Shouldn't it then be directly measurable as distortion? And if the difference between drivers is 1000 to 1 shouldn't it be obvious?
I'm seeing a lot of "interested" people commenting here . . . perhaps if you suggested some way to actually measure this effect someone here would find the time to actually test for it (even more if it were a simple test that rendered the effect audible).
I'd advise "caution" here.
What Earl thinks of "dynamics" and what others tend to think of it aren't necessarily the same thing.
I think he is describing dynamics in a more mechanical context.
Because compression drivers are so efficient, coupled with their extreme INsenstivity to thermal change - they have no problem altering intensity without various effects like thermal compression.
In home use compression drivers are pretty much "loafing-it" at any reasonable domestic sound pressure.
Traditional dome tweeters on the other hand may (..largely depending on the crossover, power input as a result, and VC size + venting), be heating up the vc and motor so much that it actually "blunts" (or reduces) the intensity during higher intensity passages with certain program material. (ie. "squashing" a dynamic peak of +15 db to only +9 db.)
That may or may not "translate" into "dynamics" as a subjective description for any particular person. (..or with the ex. above - a lack of "dynamics".) On the other hand it might.. (..don't know).
Last edited:
Scott,
That is a very eloquent way to describe the difference between musical dynamics, and electrical power compression in a dome tweeter. One being a mechanical phenomena due to the rising resistance of the voicecoil wire and musical being what I would refer to as the rise time of the device in question reproducing a sound.
That is a very eloquent way to describe the difference between musical dynamics, and electrical power compression in a dome tweeter. One being a mechanical phenomena due to the rising resistance of the voicecoil wire and musical being what I would refer to as the rise time of the device in question reproducing a sound.
Yes, "rise-time" on a transient basis seems to be pretty much the component people are describing when saying "dynamics". I think mms is also a factor - heavier mass altering intensity "quickly" tends to produce a more force-full sound, than many seem to associate with being more "dynamic".
Of course a compression driver has both physical qualities, and usually has similar subjective descriptions.
I have to disagree with this strongly. VC heating doesn't happen that fast. I spent a lot of time modeling this the last time it came up and posted the results . Even in the limit of no cooling the thermal response time is just too long to have temperature follow the transient behavior. At the bottom of the page are some predictions based on actual samples I took from different musical passages.
Lots of emphasis on the "may" there, I would hope. It's easy enough to look up the temperature coefficient of Copper wire and get a sense of the magnitude of possible short-term resistance change, and that is easy to test with a simple DC pulse and a bridge (to get both magnitude of change and time constant for a given driver).
It would also be easy to avoid, by using current drive. In fact I've heard it claimed as a benefit of current drive amps . . . but I've never heard it demonstrated.
One more item for the "place in history, opportunity for" file . . .
John K,
I would think that the impedance rise of the voicecoil would be very dependent on the input level you were using the device and also the gap dimensions and also the differences of short and long coils vs gap length. There are many variables that would affect the temperature rise and usually power compression is only at high levels of input power. Under normal listening levels it is probably only a minimal factor in the overall picture.
I would think that the impedance rise of the voicecoil would be very dependent on the input level you were using the device and also the gap dimensions and also the differences of short and long coils vs gap length. There are many variables that would affect the temperature rise and usually power compression is only at high levels of input power. Under normal listening levels it is probably only a minimal factor in the overall picture.
He does mention it in his book, although only briefly: paragraph 19.4 is called "power compression".;
I'm sure a compression driver on a waveguide may well go 30 dB's louder than the average dome. I'm however still not convinced dynamic range is of much influence at sound levels that are normal in a living room. I rarely listen at more than 90 dB average, where the loudspeaker maybe puts out peaks of a bit over 100 dB, of which the tweeter usually does only a fraction.
If dynamic power-compression really is a big issue, I think it shouldn't be too difficult to measure it. I'd say just feed it some very loud sinewaves and put the output through a spectrum analyzer. I imagine dynamic thermal compression should lead to subharmonics, right? I'm curious what you actually measured in your preliminary tests. Would you share your findings?
Power compression is something we talk about in pro-audio where you are running very high voltage and current through a device on a long term basis. The magnetic system can only sink so much heat before it stops cooling the voicecoil and you have to remember that it is a heat sink and not a heat radiator like a finned aluminum plate used for transistors and such. The temperature rise is to fast to be able to remove the heat over the limited surface area of the magnetic steel and the magnet itself. This is where the term power compression comes from. It is not so much from a quick transient, it is from long term power output at elevated levels.
It's also worth noting that thermal issues obey diffusion equations and modling by R's and C's has limited utility. We were surprised at the self heating of transistors having some very fast (usec) local heating issues even though the long term die/package time constants are on the order of seconds.
Scott,
I understand what you are saying on the micro sized die that a transistor must face. I have had to have this discussion in tooling design many times where someone would fall for the hype that you could use epoxy with aluminum filler to transfer heat and dissipate heat in tooling used to mold thermoset materials that had an extremely fast exponential heat rise. You could add all the aluminum you wanted to the epoxy but each and every grain of aluminum was surrounded by the epoxy which was an insulator so that the heat transfer would not and could not keep up with the heat output from reaction. People just couldn't follow this simple phenomena and it lead to many unsuccessful tooling applications by misinformed people.
I understand what you are saying on the micro sized die that a transistor must face. I have had to have this discussion in tooling design many times where someone would fall for the hype that you could use epoxy with aluminum filler to transfer heat and dissipate heat in tooling used to mold thermoset materials that had an extremely fast exponential heat rise. You could add all the aluminum you wanted to the epoxy but each and every grain of aluminum was surrounded by the epoxy which was an insulator so that the heat transfer would not and could not keep up with the heat output from reaction. People just couldn't follow this simple phenomena and it lead to many unsuccessful tooling applications by misinformed people.
If the VC is essentially "cooking" for a while at a higher spl - yet still reasonably linear because of continually low-level transients (from the average), but then accepts a short burst of power from a strong transient..
-would that cause an "instantaneous" problem?
Also, (..in the realm of "cooking" - and I say this just having grilled a few hamburgers
), might there be other sources heating-up and "draining" into the VC of a tiny tweeter? (..or is there no effect like that happening?)
Last edited:
Oh yeah.. caveats galore!
..I've also heard that about current amps as well, that with a reduction in either 2nd or 3rd order non-linear distortion (..don't remember which order though.)
Some SOI transistors face an initial transient heating of 1000-3000 C/W because the oxide pocket is to the first order a thermal as well as electrical insulator.
Yes that is some heat rise Scott. I can understand how an instantaneous rise like that would happen when moving electrons around through such a small junction. I don't know what the actual temperature was that was reached in a fast urethane reaction but it is plenty and for a chemical reaction is fairly fast. Not as fast as pouring water into an acid, but fast enough. I always used aluminum tooling unless it had to be steel. All these tools were water temperature controlled, but that was of little consequence if the tool was epoxy. All the final material properties would be out the window if you got a decent looking part at all. I can completely understand what you have to go through on a much lesser scale by many factors different, but bad results in both situation.
Yes, all that matters. The heat generation is the easy part and knowing the heat generation rate it is pretty easy to figure the maximum rate of temperature increase possible for any given power level. It's sort of a thermal slew rate. So even if you ignore cooling you can still look at how fast the temperature can rise and the result is that the temperature simply can not follow the transients in music. The temperature as a function of time is an integral of the heat generation rate minus the heat lost to the surrounding. The integration process, even in the absence of cooling, is a smoothing process and short duration, high generation rates don't contribute much to the temperature increase. Cooling effects limit the max temperature than can be reached and slow the rate of increase. Think of heating like interest rates. If I loan you $100 at an annual rate of 36.5% and you pay back the loan in one day you would owe only $0.10 interest. But if I charge you 3.65% and you take a year to pay it back $3.65. It's the longer term average power that yields the temperature increase and the resulting thermal compression. The sharp spikes in power have their contribution, but you don't see jumps up and down in temperature following those spikes.
John,
Yes in a worse case situation you can look at this as only the temperature rise of a voicecoil in air. The transfer to air is much less than the transfer when you have the voicecoil and the steel structure there. Yes you can say that it will be an average temperature rise and that it smooths the average peaks, but there are still extremely fast rise times of heat only they are delayed from the signal. they are let's say as I don't know how to say it, out of phase with each other. The heating of the coil would lag the voltage rise I think. It is just that the, I will call it black body, of the steel absorbs the heat spikes by radiation. So no matter what the heat rise is whether it is less than or higher than either one of us thinks, the resistance value of the voicecoil is changing with the signal but, thermal rise is lagging behind, and this is a modulation of the signal that would happen in time, not at a given instance.
Yes in a worse case situation you can look at this as only the temperature rise of a voicecoil in air. The transfer to air is much less than the transfer when you have the voicecoil and the steel structure there. Yes you can say that it will be an average temperature rise and that it smooths the average peaks, but there are still extremely fast rise times of heat only they are delayed from the signal. they are let's say as I don't know how to say it, out of phase with each other. The heating of the coil would lag the voltage rise I think. It is just that the, I will call it black body, of the steel absorbs the heat spikes by radiation. So no matter what the heat rise is whether it is less than or higher than either one of us thinks, the resistance value of the voicecoil is changing with the signal but, thermal rise is lagging behind, and this is a modulation of the signal that would happen in time, not at a given instance.
Objectice data easily become subjective opinion if we insist on looking only to the part of data that supports our preconceptions. In this regard Earls look at objective data is a bit positivistic.
So please look at the green part of the diagram too and just keep in mind, that the lower frequencies are still part of the broadband signal:
Doesn't it appear to you how the 0.5 and 1 kHz signals follows the precedence "rule" for the first 0.5 ms quite exactly and then get all over the place? In contrast the 2 kHz signal gets nearer to the broadband signal with rising delta t.
What I am getting from this: For larger time intervals between first and second signals (like in room reflections - and that is what we are talking about here) lower frequencies contribute less and less to the forming of an (annoying) image shift or source broadening. It is the frequencies from 1 kHz up which "fix" the perceived direction of the precedence effect at larger ITDs.
Rudolf
Last edited: