-
#3660 by xyzzy on 07 Jul, 2016 22:46
-
Basically, the pendulum, like any dynamic system has a natural frequency (period). This period is trivial to measure: all the DIY experimenter needs is a stopwatch to measure the period of oscillation.,
If the forcing function has the same frequency (period) as the natural frequency (period) of the pendulum, one should get a resonant response whose amplitude is only limited by the damping in the dynamic system.
The experimenter should pulse the EM Drive at the same frequency of the pendulum, just like pushing on a swing at the right frequency swings a child to ever higher amplitudes.
This is a very intelligent suggestion.
If there is no resonant response, there is no EM Drive thrust. Simple.
Really good suggestion! With some care, that may even allow the whole system to be operated under thermal steady state conditions for a prolonged period of time.
A lot of the measured side effects seem to be coming from the power cord supplying a magnetron - the wires have to carry a large current (for filament heating), so they themselves heat up, expand, untwist and generally tend to do whatever they like while being particularly difficult to control experimentally.
Compared to air currents and other forces, these wires may even turn out to be the "elephant in the room".
Therefore, one first step would be to keep these wires in a thermally and electromagnetically steady state. This could be done by keeping the magnetron filament heated steadily and only switching on the high voltage when needed. Thus the experimenter could allow the wires (and the magnetron filament) to reach a steady temperature, let the system settle, and then start pulsing the high voltage (e.g. with an appropriate high voltage relay).
The high voltage could then be pulsed synchronized to the pendulum's natural resonance while keeping the pulses themselves short (low duty cycle) and the average thermal power low enough for the magnetron to survive for an extended experiment duration. With some luck that might even allow for a steady-state operation of the whole experiment. -
#3661 by rq3 on 07 Jul, 2016 23:33
-
Just received the precision 250 Ω resistor, 0.1% tolerance, with ferrite choke from Dataq. This should greatly reduce the interference with the laser displacement sensor.
No. It won't. You have a ratiometric sensor, and the resistor will make no no difference at all, except to shift the scale of the output. The ferrite choke serves no function whatsoever. -
#3662 by FattyLumpkin on 07 Jul, 2016 23:39
-
Dielectric insert for Dave's build: Dave's build is 80.015% smaller than the NASA frustum by volume. 80.015% volume of the NASA dielectric insert = 833.283 ccs. Matching the reduced diameter/height ratio to the reduced volume was time consuming or (@#$%&*). At any rate came very close on the volume and a little off on the ratio: .00474% and .0345%. Glad both builds are shaped differently! LOL Insert dimensions = 14.4 cm D x 5.1408 cm L
PS: thanks to Monomorphic for exact dimensions!!! -
#3663 by dustinthewind on 07 Jul, 2016 23:50
-
Good idea, add X,Y level bubbles and "remove and wine" and I have a vertical laser line. Note, it's best to remove any fluids from the glass before you turn the stem horizontally to shine the laser into.
The only problem I see with using a line rather than a laser dot is that post processing the laser dot to extract information is easier than the straight line. It's more a complication than a problem. The signal to noise ratio will be higher and the process more tedious.
If you limit the perception of the sensor so that it only sees a small length along the ruler, such that the line appears as a dot, would this solve the problem for the sensor? It would be like putting blinders on a horse. -
#3664 by rfmwguy on 08 Jul, 2016 00:28
-
Just received the precision 250 Ω resistor, 0.1% tolerance, with ferrite choke from Dataq. This should greatly reduce the interference with the laser displacement sensor.
The tighter tolerance will avoid thermal drift/improve accuracy (just like the ADC manufacturer says) and the ferrite will help prevent from RF getting into the ADC input lines. Nice... -
#3665 by SeeShells on 08 Jul, 2016 02:07
-
I'm looking at another solution but keeping the vertical laser line although limiting what shines on the ruler/ It would be a smaller dot that will be stable and not show vertical movement.Good idea, add X,Y level bubbles and "remove and wine" and I have a vertical laser line. Note, it's best to remove any fluids from the glass before you turn the stem horizontally to shine the laser into.
The only problem I see with using a line rather than a laser dot is that post processing the laser dot to extract information is easier than the straight line. It's more a complication than a problem. The signal to noise ratio will be higher and the process more tedious.
If you limit the perception of the sensor so that it only sees a small length along the ruler, such that the line appears as a dot, would this solve the problem for the sensor? It would be like putting blinders on a horse.
Shell -
#3666 by Rodal on 08 Jul, 2016 03:25
-
Dielectric insert for Dave's build: Dave's build is 80.015% smaller than the NASA frustum by volume. 80.015% volume of the NASA dielectric insert = 833.283 ccs. Matching the reduced diameter/height ratio to the reduced volume was time consuming or (@#$%&*). At any rate came very close on the volume and a little off on the ratio: .00474% and .0345%. Glad both builds are shaped differently! LOL Insert dimensions = 14.4 cm D x 5.1408 cm L
PS: thanks to Monomorphic for exact dimensions!!!
FattyLumpkin, you sent me a message asking for the latest Cannae patent featuring a dielectric. Here it is:
http://www.google.com/patents/WO2016004044A1?cl=en -
#3667 by JonathanD on 08 Jul, 2016 03:38
-
FattyLumpkin, you sent me a message asking for the latest Cannae patent featuring a dielectric. Here it is:
http://www.google.com/patents/WO2016004044A1?cl=en
From that document, I pick out the phrase:
"The preferred embodiment is capable of generating 40-50 μΝ of force when 30 watts of power is transmitted into the waveguide as the EM energy."
Layman's question: Is that significant in the same magnitude that Shawyer was indicating? Just trying to understand the scale of claims here.
Thanks,
J -
#3668 by Tellmeagain on 08 Jul, 2016 03:45
-
From that document, I pick out the phrase:
"The preferred embodiment is capable of generating 40-50 μΝ of force when 30 watts of power is transmitted into the waveguide as the EM energy."
Layman's question: Is that significant in the same magnitude that Shawyer was indicating? Just trying to understand the scale of claims here.
Thanks,
J
It is nothing special, just a rephrasing of the EagleWorks' test results (28W, "between 31.7 and 45.3 (40.0 mean) micronewtons"). -
#3669 by keithpickering on 08 Jul, 2016 03:48
-
Dielectric insert for Dave's build: Dave's build is 80.015% smaller than the NASA frustum by volume. 80.015% volume of the NASA dielectric insert = 833.283 ccs. Matching the reduced diameter/height ratio to the reduced volume was time consuming or (@#$%&*). At any rate came very close on the volume and a little off on the ratio: .00474% and .0345%. Glad both builds are shaped differently! LOL Insert dimensions = 14.4 cm D x 5.1408 cm L
PS: thanks to Monomorphic for exact dimensions!!!
FattyLumpkin, you sent me a message asking for the latest Cannae patent featuring a dielectric. Here it is:
http://www.google.com/patents/WO2016004044A1?cl=en
This patent should not have been granted and is likely unenforceable. It's based on prior published work by White et. al., announced on July 28, 2014. Cannae filed the patent nearly a year later (June 30, 2015) while claiming a priority date of June 30, 2014, conveniently a few weeks before White's publication. -
#3670 by mikegem on 08 Jul, 2016 04:54
-
Dielectric insert for Dave's build: Dave's build is 80.015% smaller than the NASA frustum by volume. 80.015% volume of the NASA dielectric insert = 833.283 ccs. Matching the reduced diameter/height ratio to the reduced volume was time consuming or (@#$%&*). At any rate came very close on the volume and a little off on the ratio: .00474% and .0345%. Glad both builds are shaped differently! LOL Insert dimensions = 14.4 cm D x 5.1408 cm L
PS: thanks to Monomorphic for exact dimensions!!!
FattyLumpkin, you sent me a message asking for the latest Cannae patent featuring a dielectric. Here it is:
http://www.google.com/patents/WO2016004044A1?cl=en
This patent should not have been granted and is likely unenforceable. It's based on prior published work by White et. al., announced on July 28, 2014. Cannae filed the patent nearly a year later (June 30, 2015) while claiming a priority date of June 30, 2014, conveniently a few weeks before White's publication.
Just a comment on procedure, not the merits. It's a published patent application, not a granted patent. See "Publication Type" at the linked page. -
#3671 by xyzzy on 08 Jul, 2016 12:11
-
Therefore, one first step would be to keep these wires in a thermally and electromagnetically steady state. This could be done by keeping the magnetron filament heated steadily and only switching on the high voltage when needed. Thus the experimenter could allow the wires (and the magnetron filament) to reach a steady temperature, let the system settle, and then start pulsing the high voltage (e.g. with an appropriate high voltage relay).
...
Since high voltage relays rated for the power levels involved tend to be very specialized (transmitter applications and such) and therefore really expensive, another way may be better:
Transformers from old microwave ovens are cheap and easy to come by, so one can just use two of them. Let one transformer supply only the filament (with the high voltage winding left not connected), while a second transformer supplies only the high voltage (with the filament winding going nowhere). That way the problem of switching the high voltage can be moved to the primary (mains) side, where it's easy to solve with commonly available affordable relays.
As a positive side effect, using a separate transformer would make the filament current independent form the magnetron power. Microwave transformers have magnetic shunts in order to provide a crude way of current limiting. The output voltages on these transformers are not very "stiff", but rather (intentionally) the opposite: as the current being drawn is increased, the voltage will drop a lot. This also has a side effect on the filament - when the anode current is turned on, the voltages of all transformer secondary windings will drop - thereby dropping the filament current in the process.
For a microwave oven application this is beneficial. The filament current automatically drops as soon as the magnetron warms up and starts generating microwaves. But for an EM drive test, the benefit is rather doubtful. One would rather prefer a really constant filament current in order to keep the various uncontrolled forces (thermo-mechanical and electromagnetic) acting on the lead wires steady, so that the high voltage can be pulsed without unintended side effects. In this case, using a second, independent, transformer would be a clear advantage for thermal stability.
P.S.
IIRC, Shell was building a power supply where the filament current was provided by some independently controlled / operated circuitry. That feature might come in useful one day
-
#3672 by frobnicat on 08 Jul, 2016 13:27
-
Hello @rfmwguy,
thanks for the parameters in your previous post, and thanks to TheTraveller for reposting a pic and contextualizing your setup so that we clearly see that the center of mass (of moving assembly) is indeed much too low (compared to attachment point of suspension wire on it) to allow any swinging even remotely in the minute period.
( bigger )
Wikipedia is not the ultimate reference and I'm not immune to blunders, but I think the following is correct for a simplified model of a thin beam of length B swinging in pitch (horizontal axis orthogonal to beam, above it), neglecting any other mass (mast and guy wire for illustration only) please see attached picture and equations.
(1) The moment of inertia of axis orthogonal to beam through center of mass CM. Third entry here.
(2) Moment of inertia around pivot, through parallel axis theorem.
(3) Period of a compound pendulum and replacement by expression in (2)
(4) I wan't to consider a given length of beam B and ask about the needed distance L of CM under pivot to have a given T, so I use B as a natural unit of length and express the problem in the dimensionless ratio L/B noted as lambda. The smaller the lambda the closer is CM from pivot ( the more "flattish" the overall proportions will look ). As it stands, on my illustrations lambda is 0.5, but equations hold for whatever, I'm solving for lambda actually. We already see that for very big lambda the left term (in the root) rules and we'd get the usual T=2 pi sqrt(L/g) as a swinging period, as we approach the conditions of a punctual mass (B small relative to L), this means having the bigger lambda (longer L) possible to get a longer period. For very small lambda the 1/(12 lambda) term will rule and then we will want a lambda as small as possible to get a longer period. That's why I wrote in previous post that "... period can be made arbitrarily long without having km sized suspension wire, actually by lowering the distance of center of mass below the pivot ...". Hopefully this is now more rigorously substantiated.
(5) Solving for lambda with auxiliary dimensionless gamma introduced for commodity.
(6) Long story short, a high value of gamma (meaning : aiming at T much higher that the minimum one) allows as first order approximation (checked to be within at least 2 decimals with typical values used below) those 2 real (physically admissible) solutions :
-> lambda2 giving a very long distance of CM below the pivot (compared to a given fixed B), again we check that the usual T=2 pi sqrt(L/g) would then prevail.
-> lambda1 giving a very short distance of CM below the pivot (compared to a given fixed B).
So what does it mean ? Well, after all that I realize after reading again your post that we still lack the vertical positions of masses, and that most of the mass is not homogeneously spread in a thin beam geometry. Sigh. Oh well, I'll just find a roughly equivalent thin beam : let's say that the most mass is concentrated in the 2 lumps of the test article and the dampener and let's pretend they are at same altitude, that makes for a two points body with a moment of inertia ICM=mrČ where r is distance from axis r≈30''=0.762m. For all practical matter this is equivalent with a beam (of the simplified model above) with B such that mBČ/12=mrČ => B=2sqrt(3)r≈2.64m (total span).
So we can roughly solve your system for L (aiming at some high valued T) with modeled compound pendulum formulas using B=2.64m :
T=1minute=60s => gamma≈339 (this checks that this is high enough to make (6) good approximations)
-> lambda2≈339 => L≈895m that needs some hell of a high ceiling.
-> lambda1≈2.46Ś10-4 => L≈0.65mm (millimeter) that means some rather finicky tuning of height of masses really close to instability (like in some high precision balances). Clearly you are not in this case. Intuitively I expected something more in the cm, it always pays to check.
This confirms that in your experiment the only axis that can oscillate meaningfully on minute long periods is the yaw, driven by the weak restoring torque of the torsion wire. Of course this is ignoring damping, a severely over-damped system (damper immersed in thick honey for instance) could easily have time constants above 5 minutes in any axis, but those would no longer oscillate ! If it oscillates at all then we know that the period of such damped oscillations can't be far from un-damped ones, as calculated above. I also remained ignorant of the stiffness (in bending) of the suspension wire (meaning lower T), the pivot may be found higher (meaning even higher L => lower T for lambda1 solution), my approximations tend to over-estimate the reachable T. Upper bound is way below 1 minute. A rough indication of vertical placements of the masses of your setup would allow some estimation of expected T (for the swing in pitch).
Onward...
-
#3673 by rfmwguy on 08 Jul, 2016 13:51
-
Hello @rfmwguy,
Thanks Professor, your mechanical analytical skills are superb, far above mine. I know its difficult to judge without seeing it face to face. My description was very basic and I also thank Phil for the pic and artwork added
thanks for the parameters in your previous post, and thanks to TheTraveller for reposting a pic and contextualizing your setup so that we clearly see that the center of mass (of moving assembly) is indeed much too low (compared to attachment point of suspension wire on it) to allow any swinging even remotely in the minute period.
( bigger )
Wikipedia is not the ultimate reference and I'm not immune to blunders, but I think the following is correct for a simplified model of a thin beam of length B swinging in pitch (horizontal axis orthogonal to beam, above it), neglecting any other mass (mast and guy wire for illustration only) please see attached picture and equations.
(1) The moment of inertia of axis orthogonal to beam through center of mass CM. Third entry here.
(2) Moment of inertia around pivot, through parallel axis theorem.
(3) Period of a compound pendulum and replacement by expression in (2)
(4) I wan't to consider a given length of beam B and ask about the needed distance L of CM under pivot to have a given T, so I use B as a natural unit of length and express the problem in the dimensionless ratio L/B noted as lambda. The smaller the lambda the closer is CM from pivot ( the more "flattish" the overall proportions will look ). As it stands, on my illustrations lambda is 0.5, but equations hold for whatever, I'm solving for lambda actually. We already see that for very big lambda the left term (in the root) rules and we'd get the usual T=2 pi sqrt(L/g) as a swinging period, as we approach the conditions of a punctual mass (B small relative to L), this means having the bigger lambda (longer L) possible to get a longer period. For very small lambda the 1/(12 lambda) term will rule and then we will want a lambda as small as possible to get a longer period. That's why I wrote in previous post that "... period can be made arbitrarily long without having km sized suspension wire, actually by lowering the distance of center of mass below the pivot ...". Hopefully this is now more rigorously substantiated.
(5) Solving for lambda with auxiliary dimensionless gamma introduced for commodity.
(6) Long story short, a high value of lambda (meaning : aiming at T much higher that the minimum one) allows as first order approximation (checked to be within at least 2 decimals with typical values used below) those 2 real (physically admissible) solutions :
-> lambda2 giving a very long distance of CM below the pivot (compared to a given fixed B), again we check that the usual T=2 pi sqrt(L/g) would then prevail.
-> lambda1 giving a very short distance of CM below the pivot (compared to a given fixed B).
So what does it mean ? Well, after all that I realize after reading again your post that we still lack the vertical positions of masses, and that most of the mass is not homogeneously spread in a thin beam geometry. Sigh. Oh well, I'll just find a roughly equivalent thin beam : let's say that the most mass is concentrated in the 2 lumps of the test article and the dampener and let's pretend they are at same altitude, that makes for a two points body with a moment of inertia ICM=mrČ where r is distance from axis r≈30''=0.762m. For all practical matter this is equivalent with a beam (of the simplified model above) with B such that mBČ/12=mrČ => B=2sqrt(3)r≈2.64m (total span).
So we can roughly solve your system for L (aiming at some gigh valued T) with modeled compound pendulum formulas using B=2.64m :
T=1minute=60s => gamma≈339 (this checks that this is high enough to make (6) good approximations)
-> lambda2≈339 => L≈895m that needs some hell of a high ceiling.
-> lambda1≈2.46Ś10-4 => L≈0.65mm (millimeter) that means some rather finicky tuning of height of masses really close to instability (like in some high precision balances). Clearly you are not in this case. Intuitively I expected something more in the cm, it always pays to check.
This confirms that in your experiment the only axis that can oscillate meaningfully on minute long periods is the yaw, driven by the weak restoring torque of the torsion wire. Of course this is ignoring damping, a severely over-damped system (damper immersed in thick honey for instance) could easily have time periods above 5 minutes in any axis, but those would no longer oscillate ! If it oscillates at all then we know that the period of such damped oscillations can't be far from un-damped ones, as calculated above. I also remained ignorant of the stiffness (in bending) of the suspension wire, the pivot may be found higher (meaning even higher L => lower T for lambda1 solution).
Onward...
I am pretty confident this system will allow me to pin down the exact cause and perhaps locale of the horizontal displacement force I detected with the defective magnetron. With Mr. Li's advice, I am leaning towards harness twisting/untwisting in the drop loop that is aligned along the beam's axis. Good thing was, it was very repeatable and if I can't remove it, I can simply remove that component and search for any other forces present...as long as there are not too many other x-y-z variables.
Thanks for your continued interest and valuable data! - Dave
-
#3674 by SeeShells on 08 Jul, 2016 14:04
-
That's quite close to what I've done.Therefore, one first step would be to keep these wires in a thermally and electromagnetically steady state. This could be done by keeping the magnetron filament heated steadily and only switching on the high voltage when needed. Thus the experimenter could allow the wires (and the magnetron filament) to reach a steady temperature, let the system settle, and then start pulsing the high voltage (e.g. with an appropriate high voltage relay).
...
Since high voltage relays rated for the power levels involved tend to be very specialized (transmitter applications and such) and therefore really expensive, another way may be better:
Transformers from old microwave ovens are cheap and easy to come by, so one can just use two of them. Let one transformer supply only the filament (with the high voltage winding left not connected), while a second transformer supplies only the high voltage (with the filament winding going nowhere). That way the problem of switching the high voltage can be moved to the primary (mains) side, where it's easy to solve with commonly available affordable relays.
As a positive side effect, using a separate transformer would make the filament current independent form the magnetron power. Microwave transformers have magnetic shunts in order to provide a crude way of current limiting. The output voltages on these transformers are not very "stiff", but rather (intentionally) the opposite: as the current being drawn is increased, the voltage will drop a lot. This also has a side effect on the filament - when the anode current is turned on, the voltages of all transformer secondary windings will drop - thereby dropping the filament current in the process.
For a microwave oven application this is beneficial. The filament current automatically drops as soon as the magnetron warms up and starts generating microwaves. But for an EM drive test, the benefit is rather doubtful. One would rather prefer a really constant filament current in order to keep the various uncontrolled forces (thermo-mechanical and electromagnetic) acting on the lead wires steady, so that the high voltage can be pulsed without unintended side effects. In this case, using a second, independent, transformer would be a clear advantage for thermal stability.
P.S.
IIRC, Shell was building a power supply where the filament current was provided by some independently controlled / operated circuitry. That feature might come in useful one day
Using a "matched" set of microwave transformers (very important, if not smoke may ensue) in series, driven by a variable rheostat, which gives me a clean full wave rectified and filtered variable >4KV output.
The heater is driven by another transformer and then full wave rectifying the output. Filtering that for a clean DC voltage, to be voltage and current control no matter the loading on the dual transformers. The filament voltage and current can be switched off after a set time to further stabilize the water cooled magnetron. All we are interested in is heating the heater
to suck off electrons whether it's DC or AC, it doesn't matter it will just be more stable with no AC added to the magnetron's output.
Shell
added: cleaned up...
SAFETY FIRST! THESE ARE KILLER VOLTAGES, CURRENTS AND MICROWAVE ENERGIES, IF YOU DON"T KNOW WHAT YOUR DOING THEN DON"T DO IT!!! -
#3675 by zen-in on 08 Jul, 2016 17:20
-
...
That's quite close to what I've done.
Using a "matched" set of microwave transformers (very important, if not smoke may ensue) in series, driven by a variable rheostat, which gives me a clean full wave rectified and filtered variable >4KV output.
The heater is driven by another transformer and then full wave rectifying the output. Filtering that for a clean DC voltage, to be voltage and current control no matter the loading on the dual transformers. The filament voltage and current can be switched off after a set time to further stabilize the water cooled magnetron. All we are interested in is heating the heater to suck off electrons whether it's DC or AC, it doesn't matter it will just be more stable with no AC added to the magnetron's output.
...
Have you compared the two power supplies (AC Vs filtered DC) by recording the spectral output of the magnetron? Where is the noise and what kind of noise is it? Do you get a narrow band low phase noise output with the filtered DC supply? -
#3676 by keithpickering on 08 Jul, 2016 17:56
-
Since high voltage relays rated for the power levels involved tend to be very specialized (transmitter applications and such) and therefore really expensive, another way may be better:
Transformers from old microwave ovens are cheap and easy to come by, so one can just use two of them. Let one transformer supply only the filament (with the high voltage winding left not connected), while a second transformer supplies only the high voltage (with the filament winding going nowhere). That way the problem of switching the high voltage can be moved to the primary (mains) side, where it's easy to solve with commonly available affordable relays.
If you've ever taken apart a microwave oven (and I'm on #2 now), you will see that the power supply transformer already has two secondaries, one a low-voltage AC to keep the filament warm. The second is half-rectified into half-wave pulsed HVDC.
-
#3677 by SeeShells on 08 Jul, 2016 18:09
-
Yes I have and (it looks like Daves with just the single diode voltage doubler ckt ) it runs stable under 1MHz BW with little splattering or drift (takes about 15-20 seconds to start to stabilize). The key was the water jet cooling to keeping the signal from drifting. Dave got to see the thermal drift and was lucky that his cavity resonance was a lower TE013 and it could drift into it and lock....
That's quite close to what I've done.
Using a "matched" set of microwave transformers (very important, if not smoke may ensue) in series, driven by a variable rheostat, which gives me a clean full wave rectified and filtered variable >4KV output.
The heater is driven by another transformer and then full wave rectifying the output. Filtering that for a clean DC voltage, to be voltage and current control no matter the loading on the dual transformers. The filament voltage and current can be switched off after a set time to further stabilize the water cooled magnetron. All we are interested in is heating the heater to suck off electrons whether it's DC or AC, it doesn't matter it will just be more stable with no AC added to the magnetron's output.
...
Have you compared the two power supplies (AC Vs filtered DC) by recording the spectral output of the magnetron? Where is the noise and what kind of noise is it? Do you get a narrow band low phase noise output with the filtered DC supply?
The spectral harmonics have decreased as have the amplitude modulation splattering, but I still have some "noise" in the spectrum but it is down starting under 12-14db.
Shell
-
#3678 by SeeShells on 08 Jul, 2016 18:11
-
A simple yes but thermally cool the magnetron as well....
...
Do you get a narrow band low phase noise output with the filtered DC supply? -
#3679 by SeeShells on 08 Jul, 2016 18:18
-
LOL, I have 5 microwave skeletons. Yes, I know the microwave secondaries provide a heater tap but I needed a little more voltage (for regulation) and to keep the issues caused by the dual series HV transformers (http://sound.westhost.com/xfmr2.htm) out of the heater output and regulator.Since high voltage relays rated for the power levels involved tend to be very specialized (transmitter applications and such) and therefore really expensive, another way may be better:
Transformers from old microwave ovens are cheap and easy to come by, so one can just use two of them. Let one transformer supply only the filament (with the high voltage winding left not connected), while a second transformer supplies only the high voltage (with the filament winding going nowhere). That way the problem of switching the high voltage can be moved to the primary (mains) side, where it's easy to solve with commonly available affordable relays.
If you've ever taken apart a microwave oven (and I'm on #2 now), you will see that the power supply transformer already has two secondaries, one a low-voltage AC to keep the filament warm. The second is half-rectified into half-wave pulsed HVDC.
Shell
to suck off electrons