Is there a practicle way to reduce or eliminate the miller capacitance of a plate choke? I ask as I am wanting to use a plate choke to filter the output of a high voltage supply for an electrostatic loudspeaker. It has been suggested that the stator insulation of an ESL can store enough charge to modulate the resistively coated diaphragm. This then can couple back to the high voltage supply. So the choke has been suggested as a fix to this concern and found to be effective. The choke recommended is a ten H choke with six chambers to reduce miller capacitance. These chokes are expensive. I am wondering if there is a way to use an inexpensive choke and somehow neutralize the internal miller capacitance? Any suggestions help or advice would be welcome. Best regards Moray James.
"Miller capacitance" is usually understood to be feedback around an amplifier, which isn't the issue here. But nevermind terminology.
> ten H choke with six chambers to reduce.. capacitance .... a way to use an inexpensive choke and somehow neutralize the internal.. capacitance?
No such luck. Can't fool mother nature.
The maybe-cheap trick would be ten 1H chokes spread out along a very insulated board. Depends what bulk-price you can get on small chokes.
I'm not at all sure I understand the point, but then advanced condenser-speakers always baffled me. Still, if I understand where you plan to stick this choke: why can't you use a 10Meg resistor?
> ten H choke with six chambers to reduce.. capacitance .... a way to use an inexpensive choke and somehow neutralize the internal.. capacitance?
No such luck. Can't fool mother nature.
The maybe-cheap trick would be ten 1H chokes spread out along a very insulated board. Depends what bulk-price you can get on small chokes.
I'm not at all sure I understand the point, but then advanced condenser-speakers always baffled me. Still, if I understand where you plan to stick this choke: why can't you use a 10Meg resistor?
Now I'am really choked!
PRR; the bad guy here is the load resistor itself. The resistor generates a small amount of capacitance as a result of the resistive track windings (I am told). That capacitance permits the diaphragm to talk to the supply. Would the internal capacitance of a six chamber choke be substantualy less than the capacitance of a 500 M ohm resistor? As I think about this it sounds like a very dumb question. A 500 M resistor has a hand full or two of windings on a small ceramic core only a couple of inches long. How on earth could the ten H choke have orders of magnitude less internal capacitance? It does not make any sence. I wish I knew more about this.
What kind of layout would be required to do what you have suggested in terms of coil orientation and space? What about insulstion between chokes and air core vs cores? Thanks for your assistance. I kind of figured that this would be the case otherwise companies would not build six chamber chokes in the first place. Best regards Moray James.
PRR; the bad guy here is the load resistor itself. The resistor generates a small amount of capacitance as a result of the resistive track windings (I am told). That capacitance permits the diaphragm to talk to the supply. Would the internal capacitance of a six chamber choke be substantualy less than the capacitance of a 500 M ohm resistor? As I think about this it sounds like a very dumb question. A 500 M resistor has a hand full or two of windings on a small ceramic core only a couple of inches long. How on earth could the ten H choke have orders of magnitude less internal capacitance? It does not make any sence. I wish I knew more about this.
What kind of layout would be required to do what you have suggested in terms of coil orientation and space? What about insulstion between chokes and air core vs cores? Thanks for your assistance. I kind of figured that this would be the case otherwise companies would not build six chamber chokes in the first place. Best regards Moray James.
Moray James (or is it James Moray?),
As PRR started to say, Miller capacitance is the equivalent capacitance due to the multiplicative effect of gain and the plate-to-grid capacitance. Its effect is primarily felt in the grid circuit. The capacitance load of the ESL is effectively from plate-to-plate (which can be viewed as plate to ground with double the plate-to-plate capacitance, a sort of two-for-one Miller-like effect, if you will). The capacitance across the choke will be from plate-to-supply, which is effectively from plate-to-ground via the bigger supply capacitors. After all, for example, 50pF in series with 100uF is still essentially 50pF to ground. Your ESL’s capacitance will swamp the smaller choke capacitance – probably –especially if the choke is sectioned. The concern then is will PS noise get past the choke via the small capacitance? Possibly in theory, but probably not very much, since the cap is small and has a high reactance at lower harmonics of the line frequency. The fact that both PP tubes are fed by the same supply confers PS rejection too. There are always things that you can do to improve PS ripple fairly easily. Another choke concern is having enough choke voltage capability to handle the thousands of volts of plate swing needed for ESL drive without breakdown.
Advantages and disadvantages of resistors versus chokes other than mentioned above? A choke has the potential for giving a higher effective impedance load than a resistor which has to be small enough to pass standing DC plate current. A resistor wastes power, lots of it in this application. A resistor load will require a high B+, probably about twice the value with a choke. With a choke there will some parallel resonance frequency due to Cesl and Lchoke interacting, making a bump, possibly in the pass-band. This can be mitigated by a low (triode) plate impedance.
With either chokes or resistor loads, you can simply split the required value into several smaller pieces and achieve lower capacitance; plus it might be easier to find several decent 5 or 10 resistors rather than one big monster.
As PRR started to say, Miller capacitance is the equivalent capacitance due to the multiplicative effect of gain and the plate-to-grid capacitance. Its effect is primarily felt in the grid circuit. The capacitance load of the ESL is effectively from plate-to-plate (which can be viewed as plate to ground with double the plate-to-plate capacitance, a sort of two-for-one Miller-like effect, if you will). The capacitance across the choke will be from plate-to-supply, which is effectively from plate-to-ground via the bigger supply capacitors. After all, for example, 50pF in series with 100uF is still essentially 50pF to ground. Your ESL’s capacitance will swamp the smaller choke capacitance – probably –especially if the choke is sectioned. The concern then is will PS noise get past the choke via the small capacitance? Possibly in theory, but probably not very much, since the cap is small and has a high reactance at lower harmonics of the line frequency. The fact that both PP tubes are fed by the same supply confers PS rejection too. There are always things that you can do to improve PS ripple fairly easily. Another choke concern is having enough choke voltage capability to handle the thousands of volts of plate swing needed for ESL drive without breakdown.
Advantages and disadvantages of resistors versus chokes other than mentioned above? A choke has the potential for giving a higher effective impedance load than a resistor which has to be small enough to pass standing DC plate current. A resistor wastes power, lots of it in this application. A resistor load will require a high B+, probably about twice the value with a choke. With a choke there will some parallel resonance frequency due to Cesl and Lchoke interacting, making a bump, possibly in the pass-band. This can be mitigated by a low (triode) plate impedance.
With either chokes or resistor loads, you can simply split the required value into several smaller pieces and achieve lower capacitance; plus it might be easier to find several decent 5 or 10 resistors rather than one big monster.
> Would the internal capacitance of a six chamber choke be substantualy less than the capacitance of a 500 M ohm resistor?
I doubt it. I'd ballpark a single-winding choke as 300pFd end to end. so any reasonably sane 6-winding choke comes to about 300pFd/5 = 60pFd, maybe 10pFd if the winding is mostly air-space. I'd figure a single resistor as maybe 3pFd, so we are already ahead. For several reasons, I'd think of a 500Meg resistor in a "large" application (not, say, a condenser mike) as maybe twenty-four 22Meg resistors in series, and for high voltage and low capacitance laid end to end. The shunt capacitance then looks like 3pFd/24 or a small fraction of a pFd.
Using a baggie of common 22Meg resistors may be no more expensive than a special 500Meg resistor. Common resistors are only good for 300V-500V, but a series-string is good for 7KV-12KV. When we only had carbon-composition, such a string would be noisy (might not matter here), but carbon-film is solid stuff without the loose-contact flaws of composition. Note also that carbon-film was first developed for microwave use, where very low parasitic capacitance (and inductance) is essential, and in some uses excess-noise or intermodulation can't be tolerated.
I confess again: when I was a boy, we grounded or fixed-volted the diaphram and electrode(s), and they were good conductors. Since that era Quad and others showed advantages in high-resistance bias and diaphrams with such low conductivity that I would call them insulators. I've looked at the theory a few times and it is greek to me. So there may be some point in choke-feed rather than fixed or resistor feed, but I'm missing it.
Note that an ideal 10H choke is 3,000 ohms at 50Hz, 30K at 500Hz, far-far lower impedance than a Meg resistor. I'd expect a very low-capacitance 10H winding on an audio transformer (which is a choke with an output) to rise from there to a broad peak of less-than 300K at 5KHz and then fall due to the capacitance. The advantage of a choke is that it can pass DC at maybe 300 ohms loss. But if I understand the application, there isn't any DC, or maybe nanoAmps of leakage, so that is a non-issue.
> air core vs cores?
A 10H air-core would be a monster. Even 1H is never done in air-core: air-core tends toward milliHenries and fairly high resistance in the audio band (though here the resistance may be moot).
> What kind of layout
We just have to put in some breaks where the area/separation ratio is much greater than that inside the winding. Since choke windings use thin paper, like 0.010", and the winding area is similar to the external choke dimensions, we want much more than 0.010" between chokes. Ten chokes with an inch between bodies, first and last over a foot apart, should get you down toward 30pFd.
But overall: either the theory is flawed, you have repeated it incorrectly, or I'm misundestanding. If there is zero DC, a many-Meg resistor array will give extremely high impedance from DC up far above the audio band. Using any practical choke gives lower impedance for DC and bass, which might be good, but also lower impedance at the top of the audio band. Much lower than a single 500Meg resistor, even with a pie-wound choke.
I doubt it. I'd ballpark a single-winding choke as 300pFd end to end. so any reasonably sane 6-winding choke comes to about 300pFd/5 = 60pFd, maybe 10pFd if the winding is mostly air-space. I'd figure a single resistor as maybe 3pFd, so we are already ahead. For several reasons, I'd think of a 500Meg resistor in a "large" application (not, say, a condenser mike) as maybe twenty-four 22Meg resistors in series, and for high voltage and low capacitance laid end to end. The shunt capacitance then looks like 3pFd/24 or a small fraction of a pFd.
Using a baggie of common 22Meg resistors may be no more expensive than a special 500Meg resistor. Common resistors are only good for 300V-500V, but a series-string is good for 7KV-12KV. When we only had carbon-composition, such a string would be noisy (might not matter here), but carbon-film is solid stuff without the loose-contact flaws of composition. Note also that carbon-film was first developed for microwave use, where very low parasitic capacitance (and inductance) is essential, and in some uses excess-noise or intermodulation can't be tolerated.
I confess again: when I was a boy, we grounded or fixed-volted the diaphram and electrode(s), and they were good conductors. Since that era Quad and others showed advantages in high-resistance bias and diaphrams with such low conductivity that I would call them insulators. I've looked at the theory a few times and it is greek to me. So there may be some point in choke-feed rather than fixed or resistor feed, but I'm missing it.
Note that an ideal 10H choke is 3,000 ohms at 50Hz, 30K at 500Hz, far-far lower impedance than a Meg resistor. I'd expect a very low-capacitance 10H winding on an audio transformer (which is a choke with an output) to rise from there to a broad peak of less-than 300K at 5KHz and then fall due to the capacitance. The advantage of a choke is that it can pass DC at maybe 300 ohms loss. But if I understand the application, there isn't any DC, or maybe nanoAmps of leakage, so that is a non-issue.
> air core vs cores?
A 10H air-core would be a monster. Even 1H is never done in air-core: air-core tends toward milliHenries and fairly high resistance in the audio band (though here the resistance may be moot).
> What kind of layout
We just have to put in some breaks where the area/separation ratio is much greater than that inside the winding. Since choke windings use thin paper, like 0.010", and the winding area is similar to the external choke dimensions, we want much more than 0.010" between chokes. Ten chokes with an inch between bodies, first and last over a foot apart, should get you down toward 30pFd.
But overall: either the theory is flawed, you have repeated it incorrectly, or I'm misundestanding. If there is zero DC, a many-Meg resistor array will give extremely high impedance from DC up far above the audio band. Using any practical choke gives lower impedance for DC and bass, which might be good, but also lower impedance at the top of the audio band. Much lower than a single 500Meg resistor, even with a pie-wound choke.
> Maybe a CCS
Aside from the practical details (KV-rated parts that maintain high output impedance at those heroic voltages): the bias must arrive at a specific voltage. Too low, output is weak, too high, sparks fly. By "arrive", I mean to imply that it is OK if it takes a while to "charge up". I gather that the Quads took seconds or minutes to stabilize. So a very high resistance is OK. An infinite resistance current source would, in theory, charge to infinite voltage, bad. Or if it changes current to keep voltage from arcing, it defeats the idea of having a high resistance source.
There is a class of circuit called "gyrator" that, with a cap, can fake a "choke". Aside from the pesky KV details, such schemes can't build huge impedances easily. The current-gain of the active device sets a limit on the ratio of series resistance to shunt resistance. To get like the 300 ohms DC resistance of a 10H choke with 500Megs impedance at higher frequencies we'd need a current gain over a million. This might be possible with a MOSFET (which is also the go-to sand-state device for KV uses) but I'm dubious. Vacuum tubes may not be better: the series resistance will be higher, which may be a non-issue, but the grid leakage sets a limit on grid-circuit resistance.
If the problem really is diaphragm signal "couple{d} back to the high voltage supply", then to me the first-aid is to increase the size of the HV caps. 1,000pFd of panel capacitance might cause 1% change on a 1uFd HV supply cap; get 10uFd or 100uFd supply caps. True, your living room is full of cans of oil, and it could be a truly lethal stored charge, but I'd weigh that against the other options and a good review of theory and fair experiments.
Aside from the practical details (KV-rated parts that maintain high output impedance at those heroic voltages): the bias must arrive at a specific voltage. Too low, output is weak, too high, sparks fly. By "arrive", I mean to imply that it is OK if it takes a while to "charge up". I gather that the Quads took seconds or minutes to stabilize. So a very high resistance is OK. An infinite resistance current source would, in theory, charge to infinite voltage, bad. Or if it changes current to keep voltage from arcing, it defeats the idea of having a high resistance source.
There is a class of circuit called "gyrator" that, with a cap, can fake a "choke". Aside from the pesky KV details, such schemes can't build huge impedances easily. The current-gain of the active device sets a limit on the ratio of series resistance to shunt resistance. To get like the 300 ohms DC resistance of a 10H choke with 500Megs impedance at higher frequencies we'd need a current gain over a million. This might be possible with a MOSFET (which is also the go-to sand-state device for KV uses) but I'm dubious. Vacuum tubes may not be better: the series resistance will be higher, which may be a non-issue, but the grid leakage sets a limit on grid-circuit resistance.
If the problem really is diaphragm signal "couple{d} back to the high voltage supply", then to me the first-aid is to increase the size of the HV caps. 1,000pFd of panel capacitance might cause 1% change on a 1uFd HV supply cap; get 10uFd or 100uFd supply caps. True, your living room is full of cans of oil, and it could be a truly lethal stored charge, but I'd weigh that against the other options and a good review of theory and fair experiments.
Flawed Theory
PRR: Thak you very much indeed for the time you have taken and the cogent response. You are probably correct that the theory is flawed but that's all it has ever been a theory. I think that the pigtail of carbon film resistors is a solid practicle solution to the concern at hand. Given that I believe that my friend has heard substantial improvement useing the choke. So I think that the choke is probably filtering diode switching noise. Carlos reminded me that the choke will have a significant impact on supply ripple.
I have just obtained a set of old Hammond 7 H chokes. I have enough parts to do a bunch of experiments to see what all these thing sound like. The diodes in the multiplier supply each have a 0.33nf cap filtering them. What benefit would there be to increasing the cap value two or three times? Is this a waste of time and money because the caps necessary to make a really big improvement need to be huge in value as you mentioned?
Putting aside issues with breakdown at the diaphragm is there a benefit to placeing a cap to ground on the diaphragm side of the load resistor? should I keep all caps and snubbers on the supply side of the load resistor? Would it help to have a snubber on the diaphragm side of the load resistor?
Answers to these questions will I am sure lead to a safer series of experiments. Thanks again for the great consultation. Oh, one last question do you see any special reason or purpose for Acoustat to have used such a high value load resistor (500 Meg ohms) in thier supply? The supply and cap mod can be seen at the bottom of this page http://www.izzy-wizzy.com/audio/spkrif.html Best regards Moray James.
PRR: Thak you very much indeed for the time you have taken and the cogent response. You are probably correct that the theory is flawed but that's all it has ever been a theory. I think that the pigtail of carbon film resistors is a solid practicle solution to the concern at hand. Given that I believe that my friend has heard substantial improvement useing the choke. So I think that the choke is probably filtering diode switching noise. Carlos reminded me that the choke will have a significant impact on supply ripple.
I have just obtained a set of old Hammond 7 H chokes. I have enough parts to do a bunch of experiments to see what all these thing sound like. The diodes in the multiplier supply each have a 0.33nf cap filtering them. What benefit would there be to increasing the cap value two or three times? Is this a waste of time and money because the caps necessary to make a really big improvement need to be huge in value as you mentioned?
Putting aside issues with breakdown at the diaphragm is there a benefit to placeing a cap to ground on the diaphragm side of the load resistor? should I keep all caps and snubbers on the supply side of the load resistor? Would it help to have a snubber on the diaphragm side of the load resistor?
Answers to these questions will I am sure lead to a safer series of experiments. Thanks again for the great consultation. Oh, one last question do you see any special reason or purpose for Acoustat to have used such a high value load resistor (500 Meg ohms) in thier supply? The supply and cap mod can be seen at the bottom of this page http://www.izzy-wizzy.com/audio/spkrif.html Best regards Moray James.
> placeing a cap to ground on the diaphragm side of the load resistor?
Distortion will rise. 3rd harmonic is lowest when the diaphragm is "free" to find its own voltage between the signal plates (constant charge instead of constant voltage).
Some distotion may be euphonious; that can only be judged by the listener.
A practical problem is: if the panel arcs, constant voltage (or a capacitor) will let MUCH more energy pass through the diaphragm, doing real damage.
> the choke will have a significant impact on supply ripple.
Not significant in addition to 500 Megs. 10H at 120Hz is 10H*120Hz*2*3.14= 4K ohms reactance.
And what ripple? 1,000pFd at 60Hz seems small in ordinary work, but the R-C product determines ripple. R in "normal audio gear" runs 50Ω to 50KΩ. This bias supply has low drop in 500Megs, so the total load must be more like 5,000megs! There is additional filtering in the 500Megs and the panel self-capacitance. Ripple will be very-very low.
One oddity: ripple will not appear as "buzz" as in most work. Instead it modulates the sensitivity, which is equivalent to IM distortion with one tone being the ripple. However I don't believe the ripple is remotely large enough to cause audible IM. Far-far-higher IM is instrinsic in loudspeakers, especially ones smaller than EL panels.
Distortion will rise. 3rd harmonic is lowest when the diaphragm is "free" to find its own voltage between the signal plates (constant charge instead of constant voltage).
Some distotion may be euphonious; that can only be judged by the listener.
A practical problem is: if the panel arcs, constant voltage (or a capacitor) will let MUCH more energy pass through the diaphragm, doing real damage.
> the choke will have a significant impact on supply ripple.
Not significant in addition to 500 Megs. 10H at 120Hz is 10H*120Hz*2*3.14= 4K ohms reactance.
And what ripple? 1,000pFd at 60Hz seems small in ordinary work, but the R-C product determines ripple. R in "normal audio gear" runs 50Ω to 50KΩ. This bias supply has low drop in 500Megs, so the total load must be more like 5,000megs! There is additional filtering in the 500Megs and the panel self-capacitance. Ripple will be very-very low.
One oddity: ripple will not appear as "buzz" as in most work. Instead it modulates the sensitivity, which is equivalent to IM distortion with one tone being the ripple. However I don't believe the ripple is remotely large enough to cause audible IM. Far-far-higher IM is instrinsic in loudspeakers, especially ones smaller than EL panels.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.