flyback cores question

[rick]

I have quite a few somewhat interesting cores from x-ray flyback transformers
that I would like to use in some experiments. It would be nice to be able to
categorize them somewhat, so I could compare predicted and actual results. Is
this a realistic (sane) thing to do, or are cores like this so cheap and readily
available that I would be better off tossing these in the garbage and getting
something with specs?

I have a picture of one of the cores here:

http://www.skyko.com/xfrmer/flyback.JPG

The plastic discs that make the gaps are 0.043 inches thick. The pieces are
0.783 inches in diameter and the overall length is 5.21 inches and the width is
2.52 inches. It seems to be a pretty decent sized core, IMO. The primary was a
0.325 inch wide, 0.050 inch copper strip wrapped 5 times around the core, while
there seemed to be multiple secondaries of many turns of extremely fine wire.
Anyway, it was easy and non-destructive just to slip the windings off the core.

I don't have any books yet on flybacks, smps design, etc. I was thinking about
ordering A. Pressman's "Switching Power Supply Design". I am interested in high
voltage (3kV to 10kV) medium low current (10mA to 60mA) switching power
supplies. I am not sure how much his book goes into this type of supply.

Thanks,

Rick

 

[John Popelish]

rick wrote:

I have quite a few somewhat interesting cores from x-ray flyback transformers
that I would like to use in some experiments. It would be nice to be able to
categorize them somewhat, so I could compare predicted and actual results. Is
this a realistic (sane) thing to do, or are cores like this so cheap and readily
available that I would be better off tossing these in the garbage and getting
something with specs?

I have a picture of one of the cores here:

http://www.skyko.com/xfrmer/flyback.JPG

The plastic discs that make the gaps are 0.043 inches thick. The pieces are
0.783 inches in diameter and the overall length is 5.21 inches and the width is
2.52 inches. It seems to be a pretty decent sized core, IMO. The primary was a
0.325 inch wide, 0.050 inch copper strip wrapped 5 times around the core, while
there seemed to be multiple secondaries of many turns of extremely fine wire.
Anyway, it was easy and non-destructive just to slip the windings off the core.

I don't have any books yet on flybacks, smps design, etc. I was thinking about
ordering A. Pressman's "Switching Power Supply Design". I am interested in high
voltage (3kV to 10kV) medium low current (10mA to 60mA) switching power
supplies. I am not sure how much his book goes into this type of supply.

Thanks,

Rick

With that much gap, the core properties do not matter much, except for
saturation flux, and that doesn't differ, much from one power ferrite
to another, so you should get very representative results for that
core size, regardless of the material you might replace it with,
later.

Keep in mind that 10 ma at 10 kV is 100 watts of continuous power,
with the instantaneous power during flyback operation being something
like 4 times that. So you are talking about pretty beefy stuff.

--
John Popelish

 

[Terry Given]

"John Popelish" <jpopelish@rica.net> wrote in message
news:4070DD66.A3A4CE21@rica.net...

rick wrote:

I have quite a few somewhat interesting cores from x-ray flyback
transformers
that I would like to use in some experiments. It would be nice to be
able to
categorize them somewhat, so I could compare predicted and actual
results. Is
this a realistic (sane) thing to do, or are cores like this so cheap and
readily
available that I would be better off tossing these in the garbage and
getting
something with specs?

I have a picture of one of the cores here:

http://www.skyko.com/xfrmer/flyback.JPG

The plastic discs that make the gaps are 0.043 inches thick. The pieces
are
0.783 inches in diameter and the overall length is 5.21 inches and the
width is
2.52 inches. It seems to be a pretty decent sized core, IMO. The
primary was a
0.325 inch wide, 0.050 inch copper strip wrapped 5 times around the
core, while
there seemed to be multiple secondaries of many turns of extremely fine
wire.
Anyway, it was easy and non-destructive just to slip the windings off
the core.

I don't have any books yet on flybacks, smps design, etc. I was
thinking about
ordering A. Pressman's "Switching Power Supply Design". I am interested
in high
voltage (3kV to 10kV) medium low current (10mA to 60mA) switching power
supplies. I am not sure how much his book goes into this type of
supply.

Thanks,

Rick

With that much gap, the core properties do not matter much, except for
saturation flux, and that doesn't differ, much from one power ferrite
to another, so you should get very representative results for that
core size, regardless of the material you might replace it with,
later.

Keep in mind that 10 ma at 10 kV is 100 watts of continuous power,
with the instantaneous power during flyback operation being something
like 4 times that. So you are talking about pretty beefy stuff.

--
John Popelish

good comments so far. Obviously you understand the 100W, but why 4 times?!?

If using DCM (Discontinuous Conduction Mode - the inductor current returns
to zero before the end of a switching period) with a 50% duty cycle [aim for
this - it gives the best "balance" between primary and secondary RMS
currents - low D gives high primary RMS current and low secondary RMS
current (reflect it back to the primary for this comparison); high D is the
opposite] then you can caluclate:

Iavg = Pout/Vout
The secondary current is a sawtooth, height Ipeak, width Toff. In one
period, its average value must be the same as Iavg, i.e.
0.5*Ipeak*(Toff/Tperiod) = Iavg

then calculate Iavg. If Toff = Tperiod/2 (boundary of DCM, D=0.5) then
Iavg=0.25*Ipeak i.e. Ipeak = 4*Iavg. For any other D use Ipeak =
2*Iavg/(1-D) (for secondary currents; for primary use D not 1-D).

Also the RMS current is Ipeak*sqrt(D/3) - easy enough to prove.

This is actually the big downside to flyback converters - the output peak &
RMS currents are quite high. Mind you I have built 400W 24V DCM flybacks
running at 50% duty cycle - Iavg = 16A, Ipeak = 64A, Irms = 26A......but you
need to pay close attention to diode & capacitor specs/temp.

perhaps you should look at genomes website.....


Cheers
Terry

 

[R.Legg]

I have a picture of one of the cores here:

http://www.skyko.com/xfrmer/flyback.JPG

Looks like UR64/40/20

http://www.ferroxcube.com/prod/assets/urcores.pdf

The plastic discs that make the gaps are 0.043 inches thick. The pieces are
0.783 inches in diameter and the overall length is 5.21 inches and the width is
2.52 inches. It seems to be a pretty decent sized core, IMO. The primary was a
0.325 inch wide, 0.050 inch copper strip wrapped 5 times around the core, while
there seemed to be multiple secondaries of many turns of extremely fine wire.
Anyway, it was easy and non-destructive just to slip the windings off the core.

You should not disassemble it unless you are convinced that its
original configuration will not serve the intended purpose. If it was
not impregnated, the wire and stiffer insulation layers might also be
reusable, if of suitable guage for this or other tasks.

I don't have any books yet on flybacks, smps design, etc. I was thinking about
ordering A. Pressman's "Switching Power Supply Design". I am interested in high
voltage (3kV to 10kV) medium low current (10mA to 60mA) switching power
supplies. I am not sure how much his book goes into this type of supply.

It's a fairly large part for this power level, but this should only
make HV isolation a little easier to achieve.

Reading up is always a good idea. If the existence of the part is the
only reason for your interest, then reading is mandatory - as you have
yet to set an application to aim for. Without this goal, you have no
starting point for a design procedure.

RL

 

[Bill Sloman]

"rick" <rick@skyko.com> wrote in message news:<_Y3cc.10431$Zw.8665@nwrddc02.gnilink.net>...

I have quite a few somewhat interesting cores from x-ray flyback transformers
that I would like to use in some experiments. It would be nice to be able to
categorize them somewhat, so I could compare predicted and actual results. > Is this a realistic (sane) thing to do, or are cores like this so cheap and > readily available that I would be better off tossing these in the garbage
and getting something with specs?

Farnell certainly stocks a useful range of transformer cores, and they
aren't particularly expensive. I'd go for something where you can get
a data sheet.

I have a picture of one of the cores here:

http://www.skyko.com/xfrmer/flyback.JPG

The plastic discs that make the gaps are 0.043 inches thick. The pieces are
0.783 inches in diameter and the overall length is 5.21 inches and the width > is 2.52 inches. It seems to be a pretty decent sized core, IMO. The
primary was a 0.325 inch wide, 0.050 inch copper strip wrapped 5 times
around the core, while there seemed to be multiple secondaries of many turns > of extremely fine wire.
Anyway, it was easy and non-destructive just to slip the windings off the
core.

I don't have any books yet on flybacks, smps design, etc. I was thinking
about ordering A. Pressman's "Switching Power Supply Design". I am
interested in high voltage (3kV to 10kV) medium low current (10mA to 60mA)
switching power supplies. I am not sure how much his book goes into this
type of supply.

I can't find much in the way of explicit references in my copy. One of
the numerous problems with this sort of supply is that the secondary
windings have relatively high inductance and self-capacitance, and
tend to be resonant at relatively low frequencies.

Peter Baxandall invented his class-D oscillator precisely to deal with
just this problem, and while Pressman does discuss resonant
converters, he doesn't seem to discuss that particular configuration,
which does not appear to be well known in the U.S. Jim Williams of
Linear Technology uses pretty much exactly the Baxandall circuit in
his famous application note 65 on driving fluorescent back-lighting in
lap-top computers

http://www.linear.com/pdf/an65f.pdf

but describes it as a "Royer" inverter, despite the fact that Royer's
inverter was not resonant.

The sort of high voltage inverter you are talking about is quite
difficult to design and build - protecting the secondary side
connections from arcing over is an art in itself. Cambridge
Instruments built their own high voltage (30kV)power supplies for
their electron microscopes for some years, but by the time I started
working for them in 1982, they'd gone over to buying them in from
people who had specialised in making these sort of supplies.
Brandenburg and Hunting HiVolt come to mind.

----------
Bill Sloman, Nijmegen

 

[John Popelish]

Terry Given wrote:

"John Popelish" <jpopelish@rica.net> wrote in message
news:4070DD66.A3A4CE21@rica.net...
(snip)
Keep in mind that 10 ma at 10 kV is 100 watts of continuous power,
with the instantaneous power during flyback operation being something
like 4 times that. So you are talking about pretty beefy stuff.

good comments so far. Obviously you understand the 100W, but why 4 times?!?

If using DCM (Discontinuous Conduction Mode - the inductor current returns
to zero before the end of a switching period) with a 50% duty cycle [aim for
this - it gives the best "balance" between primary and secondary RMS
currents - low D gives high primary RMS current and low secondary RMS
current (reflect it back to the primary for this comparison); high D is the
opposite] then you can caluclate:

Iavg = Pout/Vout
The secondary current is a sawtooth, height Ipeak, width Toff. In one
period, its average value must be the same as Iavg, i.e.
0.5*Ipeak*(Toff/Tperiod) = Iavg

then calculate Iavg. If Toff = Tperiod/2 (boundary of DCM, D=0.5) then
Iavg=0.25*Ipeak i.e. Ipeak = 4*Iavg.

That is the 4 I was thinking of. Both the switch and the rectifier
has to deal with a current peak 4 times the average current, at least,
and at least full voltage, so an instantaneous power or 4 times the
average power, at least.

For any other D use Ipeak =
2*Iavg/(1-D) (for secondary currents; for primary use D not 1-D).

Also the RMS current is Ipeak*sqrt(D/3) - easy enough to prove.

This is actually the big downside to flyback converters - the output peak &
RMS currents are quite high. Mind you I have built 400W 24V DCM flybacks
running at 50% duty cycle - Iavg = 16A, Ipeak = 64A, Irms = 26A......but you
need to pay close attention to diode & capacitor specs/temp.

I contributed a fair part of the design effort on a 4 phase, 1600 watt
85 to 125 volt in, 42 volt out, power factor correcting commercial
flyback design. (so about 4 of yours in parallel).

A tiny bit of leakage inductance between primary and secondary do
nasty things to the voltage requirements of the switch. Passing
conducted and radiated noise specs is a bitch, too.

perhaps you should look at genomes website.....

I await its completion with an......

ticipation.

--
John Popelish

 

[Genome]

"rick" <rick@skyko.com> wrote in message
news:_Y3cc.10431$Zw.8665@nwrddc02.gnilink.net...
| I have quite a few somewhat interesting cores from x-ray flyback
transformers
| that I would like to use in some experiments. It would be nice to be
able to
| categorize them somewhat, so I could compare predicted and actual
results. Is
| this a realistic (sane) thing to do, or are cores like this so cheap
and readily
| available that I would be better off tossing these in the garbage and
getting
| something with specs?
|
| I have a picture of one of the cores here:
|
| http://www.skyko.com/xfrmer/flyback.JPG
|
| The plastic discs that make the gaps are 0.043 inches thick. The
pieces are
| 0.783 inches in diameter and the overall length is 5.21 inches and the
width is
| 2.52 inches. It seems to be a pretty decent sized core, IMO. The
primary was a
| 0.325 inch wide, 0.050 inch copper strip wrapped 5 times around the
core, while
| there seemed to be multiple secondaries of many turns of extremely
fine wire.
| Anyway, it was easy and non-destructive just to slip the windings off
the core.
|
| I don't have any books yet on flybacks, smps design, etc. I was
thinking about
| ordering A. Pressman's "Switching Power Supply Design". I am
interested in high
| voltage (3kV to 10kV) medium low current (10mA to 60mA) switching
power
| supplies. I am not sure how much his book goes into this type of
supply.
|
| Thanks,
|
| Rick
|
|

Very meaty....

Converting to SI units gives the following approximate figures.

Effective area, Ae = 310mm^2
Effective length, Le = 392mm
Length of gap, G = 1mm
Primary copper area = 10mm^2

If you assume it's something like 3C80 material then intitial
permeability is ui = 2000. That gives an effective permeability of

ue = ui/(1+G.ui/Le) = 327.

5 turns of your copper gives

Lpri = uo.ue.N^2.Ae/Le = 8.12uH

If you've got a way of measuring inductance then you can reverse
engineer for a closer figure on ui and natch it up. I thing it's a
reasonable guess though.

Assuming Bsat of 300mT then the peak current is

Ipk = Bpk.Ae.N/Lpri = 57A (ouch)

If the thing goes discontinuous 50% duty at 40KHz, not unreasonable,
then

VIN = Ipk.L/Ton = 37V

Not too far away from a standard 48V bus allowing for regulation.

This is in the realms of 500 Watts worth. I don't know about effciency
of X-ray tubes but I'd expect them to be fairly poor

10mm^2 of copper at 4A/mm^2 is worth 40A RMS. 57A worth of 50% duty
triangle is worth 23A average. Not unreasonable given the likely AC
losses. (skin + proximity effect)

As to using one in anger........ not for the faint hearted.

DNA

 

[rick]

"R.Legg" <legg@magma.ca> wrote in message
news:e715b5cc.0404042336.29866db2@posting.google.com...

I have a picture of one of the cores here:

http://www.skyko.com/xfrmer/flyback.JPG

Looks like UR64/40/20

http://www.ferroxcube.com/prod/assets/urcores.pdf

Dang! You are good. That looks spot on.

The plastic discs that make the gaps are 0.043 inches thick. The pieces are
0.783 inches in diameter and the overall length is 5.21 inches and the width
is
2.52 inches. It seems to be a pretty decent sized core, IMO. The primary
was a
0.325 inch wide, 0.050 inch copper strip wrapped 5 times around the core,
while
there seemed to be multiple secondaries of many turns of extremely fine
wire.
Anyway, it was easy and non-destructive just to slip the windings off the
core.

You should not disassemble it unless you are convinced that its
original configuration will not serve the intended purpose. If it was
not impregnated, the wire and stiffer insulation layers might also be
reusable, if of suitable guage for this or other tasks.


I don't have any books yet on flybacks, smps design, etc. I was thinking
about
ordering A. Pressman's "Switching Power Supply Design". I am interested in
high
voltage (3kV to 10kV) medium low current (10mA to 60mA) switching power
supplies. I am not sure how much his book goes into this type of supply.

It's a fairly large part for this power level, but this should only
make HV isolation a little easier to achieve.

Reading up is always a good idea. If the existence of the part is the
only reason for your interest, then reading is mandatory - as you have
yet to set an application to aim for. Without this goal, you have no
starting point for a design procedure.

RL

Thanks for the info and tips. I sorta have an application in mind, but I just
wanted to learn about boost converters too. Perhaps cutting my teeth on
something less than 3KV at 50mA would be wise (but not as much fun?).

I have a dual cathode, center anode CO2 laser tube that I would like to power.
The tube is about 30 inches in length between the cathodes, with the anode in
the exact center. I am not exactly sure what the design voltage is supposed to
be, but I believe it is in the neighborhood of 5kv to 6kv at 50 to 60mA. I do
know the tube should output about 70 watts, so that would be an input of 350
watts at theoretical max. eff. of 20%. Depending on the switching frequency, I
believe it is possible to get a good output from the laser just driving it from
the rectified high voltage, possibly even without ballast resistors. Also, if I
want to modulate the laser electrically, it would seem that a high frequency
switching design would allow for easier/faster modulation than a 60hz rectified
and filtered chunky tranformer design.

Rick

 

[rick]

<snip>

I can't find much in the way of explicit references in my copy. One of
the numerous problems with this sort of supply is that the secondary
windings have relatively high inductance and self-capacitance, and
tend to be resonant at relatively low frequencies.

Peter Baxandall invented his class-D oscillator precisely to deal with
just this problem, and while Pressman does discuss resonant
converters, he doesn't seem to discuss that particular configuration,
which does not appear to be well known in the U.S. Jim Williams of
Linear Technology uses pretty much exactly the Baxandall circuit in
his famous application note 65 on driving fluorescent back-lighting in
lap-top computers

http://www.linear.com/pdf/an65f.pdf

but describes it as a "Royer" inverter, despite the fact that Royer's
inverter was not resonant.

Thanks. That is a very interesting ap note.

The sort of high voltage inverter you are talking about is quite
difficult to design and build - protecting the secondary side
connections from arcing over is an art in itself. Cambridge
Instruments built their own high voltage (30kV)power supplies for
their electron microscopes for some years, but by the time I started
working for them in 1982, they'd gone over to buying them in from
people who had specialised in making these sort of supplies.
Brandenburg and Hunting HiVolt come to mind.

----------
Bill Sloman, Nijmegen

Perhaps that is why the whole assembly (flyback, diodes, caps) was immersed in
oil in the x-ray power supply? I assumed it was because they were running in
the neighborhood of 100kV or so. I had hoped something around 6kV might be
quite a bit simpler.

Thanks,

Rick

 

[rick]

"Genome" <Genome@nothere.com> wrote in message
news:eqdcc.117$Xi.99@newsfe1-win...

snip
Very meaty....

Converting to SI units gives the following approximate figures.

Effective area, Ae = 310mm^2
Effective length, Le = 392mm
Length of gap, G = 1mm
Primary copper area = 10mm^2

If you assume it's something like 3C80 material then intitial
permeability is ui = 2000. That gives an effective permeability of

ue = ui/(1+G.ui/Le) = 327.

5 turns of your copper gives

Lpri = uo.ue.N^2.Ae/Le = 8.12uH

If you've got a way of measuring inductance then you can reverse
engineer for a closer figure on ui and natch it up. I thing it's a
reasonable guess though.

Assuming Bsat of 300mT then the peak current is

Ipk = Bpk.Ae.N/Lpri = 57A (ouch)

If the thing goes discontinuous 50% duty at 40KHz, not unreasonable,
then

VIN = Ipk.L/Ton = 37V

Not too far away from a standard 48V bus allowing for regulation.

This is in the realms of 500 Watts worth. I don't know about effciency
of X-ray tubes but I'd expect them to be fairly poor

10mm^2 of copper at 4A/mm^2 is worth 40A RMS. 57A worth of 50% duty
triangle is worth 23A average. Not unreasonable given the likely AC
losses. (skin + proximity effect)

As to using one in anger........ not for the faint hearted.

DNA

Thanks for the calculations. I was not expecting someone to do that much work,
but I guess Sunday night *is* pretty crappy for TV.

It sounds like a core this size *could* support a 350 to 400 watt design then.
Also, this gives me an idea what the untouched oil filled transformer/multiplier
units might expect as a drive voltage. Not that I really have a use for 100kV
at 5mA....


Thanks,

Rick

 

[legg]

On Mon, 5 Apr 2004 14:46:46 +0100, "Genome" <Genome@nothere.com>
wrote:

"rick" <rick@skyko.com> wrote in message
news:_Y3cc.10431$Zw.8665@nwrddc02.gnilink.net...
| I have quite a few somewhat interesting cores from x-ray flyback
transformers
| that I would like to use in some experiments. It would be nice to be
able to
| categorize them somewhat, so I could compare predicted and actual
results. Is
| this a realistic (sane) thing to do, or are cores like this so cheap
and readily
| available that I would be better off tossing these in the garbage and
getting
| something with specs?
|
| I have a picture of one of the cores here:
|
| http://www.skyko.com/xfrmer/flyback.JPG
|
| The plastic discs that make the gaps are 0.043 inches thick. The
pieces are
| 0.783 inches in diameter and the overall length is 5.21 inches and the
width is
| 2.52 inches. It seems to be a pretty decent sized core, IMO. The
primary was a
| 0.325 inch wide, 0.050 inch copper strip wrapped 5 times around the
core, while
| there seemed to be multiple secondaries of many turns of extremely
fine wire.
| Anyway, it was easy and non-destructive just to slip the windings off
the core.
|
| I don't have any books yet on flybacks, smps design, etc. I was
thinking about
| ordering A. Pressman's "Switching Power Supply Design". I am
interested in high
| voltage (3kV to 10kV) medium low current (10mA to 60mA) switching
power
| supplies. I am not sure how much his book goes into this type of
supply.
|
| Thanks,
|
| Rick
|
|

Very meaty....

Converting to SI units gives the following approximate figures.

Effective area, Ae = 310mm^2
Effective length, Le = 392mm
Length of gap, G = 1mm
Primary copper area = 10mm^2

This thing has been assembled with added sections, to provide four
gap locations and to increase the Le of the original assembly to
approximately 500mm. (Assumed the tape measure is inches, as no
reputable metric measure would use divisions of 8ths or 16ths)

It might be assumed that the added sections are the same core
material, from the physical appearance, but there's some pretty
'ferrite-looking' iron/mpp dust ferrules out there.

Depending on the number of spacers used (not mentioned), there may
have been an air gap as small as 2mm (the spacer gap appears in both
arms) or larger than 6mm if a spacer were located in each available
position. If the spacers were only used as end washers, it could alsso
have been solely dependant on the physical dimensions of a lower
permeability insert.

Equations that assume UR64 will be valid without the additional
magnetic inserts, but should recognize the double gap dimension.

Doubling the gap will halve the Lcalc value, reducing the supply
voltage required in the end use for any specific Ipk.

RL

 

[rick]

<snip>

This thing has been assembled with added sections, to provide four
gap locations and to increase the Le of the original assembly to
approximately 500mm. (Assumed the tape measure is inches, as no
reputable metric measure would use divisions of 8ths or 16ths)

It might be assumed that the added sections are the same core
material, from the physical appearance, but there's some pretty
'ferrite-looking' iron/mpp dust ferrules out there.

Depending on the number of spacers used (not mentioned), there may
have been an air gap as small as 2mm (the spacer gap appears in both
arms) or larger than 6mm if a spacer were located in each available
position. If the spacers were only used as end washers, it could alsso
have been solely dependant on the physical dimensions of a lower
permeability insert.

Equations that assume UR64 will be valid without the additional
magnetic inserts, but should recognize the double gap dimension.

Doubling the gap will halve the Lcalc value, reducing the supply
voltage required in the end use for any specific Ipk.

RL

It assembles just as pictured. There are only two 1mm thick spacers total.
They each go in the place shown in the photo and everything else is bolted
together with the brass rods. The length of the bolted together unit is about
132mm.

 

[Bill Sloman]

"rick" <rick@skyko.com> wrote in message news:<Bpfcc.11339$I66.3516@nwrddc03.gnilink.net>...

<snip>

The sort of high voltage inverter you are talking about is quite
difficult to design and build - protecting the secondary side
connections from arcing over is an art in itself. Cambridge
Instruments built their own high voltage (30kV)power supplies for
their electron microscopes for some years, but by the time I started
working for them in 1982, they'd gone over to buying them in from
people who had specialised in making these sort of supplies.
Brandenburg and Hunting HiVolt come to mind.

----------
Bill Sloman, Nijmegen

Perhaps that is why the whole assembly (flyback, diodes, caps) was immersed > in oil in the x-ray power supply? I assumed it was because they were
running in the neighborhood of 100kV or so. I had hoped something around
6kV might be quite a bit simpler.

I've not had much to do with voltages around 6kV. From what I have
seen I'd guess that all the high-voltage side of the power supply
would be potted in de-aired potting compound with just a 6kV-rated
coaxial connector sticking out.

People like Radial and Greenpar make specialised high-voltage
connectors, but my experience was that you had to go to the
manufactureres to get them - the market was too small for even
specialised distrubutors to find it worth their while to stock them.
Pasternak Enterprises in California do seem to stock a bunch of SHV
connectors, but they are only good to 5kV

http://www.caton.com/21.htm

One great advantage of using coaxial connectors and cables is that if
a point in the circuit does arc over, the consequent huge burst of
discharge current (as the cable capacitance discharges) does not
induce large external magnetic fields. Cambridge Instruments electron
microscopes for years had a nasty habit of blowing up board to board
links when the 30kV across the electron gun flashed over, which went
away when we went over to proper coaxial connectors on either end of
the cable linking the high voltage supply to the electron gun.

-----
Bill Sloman, Nijmegen