Territory folks should stick together*
*from Oklahoma! (About how ranchers and cowboys should be friends)
The tragic aspect of this sort of thing is the degree to which speaker designers insist on acting in isolation, and amplifier designers doing the same, and in the rare cases where a holistic approach is attempted, the hifi press and most customers reject them, and the integrated product is a big flop.
I had an idea for an updated motional-feedback system using, primarily, capacitative feedback along with a sort-of built-in linearization scheme to determine instantaneous position, and I presented it to a transducer designer. His instinctive response at this potential incursion into his territory was as if to the proverbial red flag, and he scrambled to tell me he could do all of this by further fine-tuning of transducer parameters. Of course the obvious question was Why haven't you done this already? My answer: because you can't, and in particular the aging of the materials assures that you never will have a long-term solution.
*from Oklahoma! (About how ranchers and cowboys should be friends)
The tragic aspect of this sort of thing is the degree to which speaker designers insist on acting in isolation, and amplifier designers doing the same, and in the rare cases where a holistic approach is attempted, the hifi press and most customers reject them, and the integrated product is a big flop.
I had an idea for an updated motional-feedback system using, primarily, capacitative feedback along with a sort-of built-in linearization scheme to determine instantaneous position, and I presented it to a transducer designer. His instinctive response at this potential incursion into his territory was as if to the proverbial red flag, and he scrambled to tell me he could do all of this by further fine-tuning of transducer parameters. Of course the obvious question was Why haven't you done this already? My answer: because you can't, and in particular the aging of the materials assures that you never will have a long-term solution.
Give or take...
Thanks for the impedance graph.
I note that the settling time for a .3 ohm load at the end of a reasonable length zip cable of any gauge will be tremendous, way up in the 20usec plus range with respect to content driving into a 10 or 20 ohm load.
This is something that is trivially simulated should anybody bother..
Remember, there is a significant distinction between settling time, and LCR phase shift of steady state sines.
John
I'm surprised, we have initiatives at some big computer speaker companies ("L") for this stuff.
Stirring the pot again?
Yup.
As I recall, one of the reasons given against it was "there is no such thing as a speaker that drops to one ohm".
So clearly, .33 ohms must be a measurement error..
What struck me odd was the order of magnitude change one octave to either side..whoa.
John
A series LC with no damping, what's not to understand? The driver alone is not doing this.
I didn't say "not understand", I said "odd".
It's just odd to see a speaker load that drops that much in a critical midrange band.
OTOH, a crossover design that tries for ruler flat impedance... well, the amp may be happy, but what about the stuff that comes outta the cones?
John
I am sceptical in the willingness of speaker designers to iron out their crossover networks, not because I doubt they can do it, but because of the time, trouble abd resources required to achieve that. At the time, when my friend and myself were developing my speakers, we had an actual living prototype inside of two weeks - but then spent a better part of six months ironing out the crossover network. It's painstaking work, get one thing just right, and something else goes askew. Then you get a "near miss" situation, something close to what you want but not quite IT, and it's up to you and your instincts to recognize the point which is indeed IT. It's a hell of a lot harder than it sounds, but in return, the sound you end up with is in my view well worth the time and trouble.
And no matter what you do, it will never quite equal a competent active speaker, with its electronic XO preceding the power amp, and the fact that vetween the power amp and the driver there's like 2 inches of connection wiring.
All this is far too much trouble and reserach budget for most modern companies.
And no matter what you do, it will never quite equal a competent active speaker, with its electronic XO preceding the power amp, and the fact that vetween the power amp and the driver there's like 2 inches of connection wiring.
All this is far too much trouble and reserach budget for most modern companies.
I'll drink to that, John. When making an amp, the designer has no idea what it might be asked to drive and at what levels.
The odd thing is I was trying out Nelson's simulated load and almost fainted when the result came in. I was registering considerably lower distortion levels with his load than with my own totrure bend load, which is a simple resistor in parallel with 1uF and 2uF. With 4 Ohms in parallel with 2 uF, I got 0.05% THD 20Hz-20kHz with a passband of 100 kHz. With Nelson's load, I got just 0.018% THD under the same conditions. With my load, the amp was terying to deliver like 6A of current at 100 kHz, in fact to be expected.
Other interesting things happened. Using ON Semi's 200W NJW3281/1302 as default ouput devices. Using 3 pairs returned relativel high distortion figures, but when increasing that to 4 pairs the whole thing settled down (NFB was around 26 dB, nominal output 28.3Vrms continuous, passband 100 kHz, input filter with -3 dB at 300 kHz, open loop bandwidth at 75 kHz. When the output devices were changed to MJ21195/21196, which offer more power per device (250W in TO-3 package), the power level remained the same, but THD went up rather sharply - these are older devices rated at Ft 4 MHz, versus NJW's 30 MHz.
Work in progress.
The odd thing is I was trying out Nelson's simulated load and almost fainted when the result came in. I was registering considerably lower distortion levels with his load than with my own totrure bend load, which is a simple resistor in parallel with 1uF and 2uF. With 4 Ohms in parallel with 2 uF, I got 0.05% THD 20Hz-20kHz with a passband of 100 kHz. With Nelson's load, I got just 0.018% THD under the same conditions. With my load, the amp was terying to deliver like 6A of current at 100 kHz, in fact to be expected.
Other interesting things happened. Using ON Semi's 200W NJW3281/1302 as default ouput devices. Using 3 pairs returned relativel high distortion figures, but when increasing that to 4 pairs the whole thing settled down (NFB was around 26 dB, nominal output 28.3Vrms continuous, passband 100 kHz, input filter with -3 dB at 300 kHz, open loop bandwidth at 75 kHz. When the output devices were changed to MJ21195/21196, which offer more power per device (250W in TO-3 package), the power level remained the same, but THD went up rather sharply - these are older devices rated at Ft 4 MHz, versus NJW's 30 MHz.
Work in progress.
Agreed John, but I still think the older devices, despite all of their shortcomings, have that something in their sound the newer devices are much harder to get from. That said, I also use NJW 200W devices (Ft 30 MHz) as my default value, they are simpler to implement, and are cheaper by a wide margin. Thankfully, still robust as I have come to expect from ON Semi/Motorola.
Anyway, my playing around with simulated loads shows that to be truly safe for a 28.3 Vrms output into whatever, one needs to use at least three pairs, and preferably even four pairs, of 200W devices, as worst case current requirements will go up to 14-16A, so each device has to be able to contribute its share. Let me make that definite 4 pairs, admittedly into my sadistic load of 3/4/6/8 ohms in parallel with 2 uF. By my reckoning, I should be using a close tolerance toridal transformer of 300VA at the very least per channel, in reality 500VA would be better if followed by at least 20,000 uF per line, or 4*20,000 uF for a stereo amp, and assuming separate electronic regulation of the voltage amp PSU lines.
Because of the split power supplies, the voltage amp is held at +/- 62V, while the current amp sits at +/- 56V. This gives me a better SOAR protection point since my overall losses are reduced while allowing me to keep the current stage lower. Using my custom toroids, this will drop by 3% only at full blast, meaning that my transient output power will be around 180/350W into 8/4 Ohms.
Contrary to popular expectation, a non-zero phase angle can give you
lower distortion at the same impedance - maximum current does not
occur at the minimum voltage point (Vce or Vds), so the gain holds up better.
Agreed, Nelson. That's what makes the dimensioning of the output stage so interesting, the number of possibilities for a load is practically unlimited.
Must be an LC shunt over the tweeter that causes this.
Elliptic filter sections should be made illegal in crossovers, also for their phase behaviour.
Call it what you want: mistake, oversight, whatever. Dave did not realize what he was doing to the speaker impedance curve when he added this added series resonance. Apparently it sounded and measured acoustically better (he used Heyser's TEF in those days) and it did not affect his amps, so he put it in. He was very embarrassed when I pointed out to him that he made a mistake, but he took it well. It took years for him to 'work around' this aberration, and in some ways for the worst. However high current supply amps with limited output protection will drive this load OK, and even tube amps do OK with it. They apparently 'ignore' the dropout.
- Status
- Not open for further replies.