Medium voltage (~800v) SMPS

I think you made a typo in there. Which one 'ya want?

Tim

-- Deep Fryer: A very philosophical monk. Website @

Look up Cockcroft-Walton circuits, there's a wealth of literature. These are usually driven from step-up transformers running at moderately-high frequencies to keep the capacitor and transformer sizes down.

Be careful not to kill yourself.

Remember, one hand behind your back.

Don't post to SED with HV wires dangling over the keyboard or you miiiiiiiiiiiiiiiiiiiiiiiiiiiiiiiii

D from BC

I will be driving a cascade multiplier with the SMPS. Every other generator I've seen does it this way (10x cascade multiplication) so I figure I should too.

Lessee here... 800V * 10

10x 800V is 8kV, not 80kV

Best regards, Spehro Pefhany

--
"it\'s the network..."                          "The Journey is the reward"
speff@interlog.com             Info for manufacturers: http://www.trexon.com
Embedded software/hardware/analog  Info for designers:  http://www.speff.com

I feel like an idiot now... I had a big reply typed up but Google ate it. I'll spend some time tonight to write my thoughts down...

Grant

Err...*if* the multiplier had zero loss, a 10X would give 8KV from an

800V source, not 80KV. For 80KV from an 800V source, perhaps a 120X would do in practice.

deep one's at that.

--
"I\'m never wrong, once i thought i was, but was mistaken"
Real Programmers Do things like this.
http://webpages.charter.net/jamie_5

Aha, bitten by a Google web-based usenet message- posting feature.

If you spend too much time composing your message, Google times you out. When I fear this happening, I do a ctrl-A to select everything, and a ctrl-C to copy it to the clipboard. Then I can DISCARD the message, refresh the topic so Google's timer is reset, select REPLY, ctrl-A, and then paste my carefully edited text in the fresh message window.

Anyway, I've found that if you really loose your message, and if you have the energy, recomposing it immediately after losing it is usually a fast task, because you effectively have a mental copy.

I actually had copied it and was about to paste it when I accidentally pressed ctrl+c!!!

I am done with Verizon math! : )

-12v primary

-1250v secondary

-6 cascade diode doubler sets for a nominal 80,000v (I'd be happy with

70kV)

-25kHz switching frequency

-final output maximum of 80kV@600uA

-adjustable output from 30kV to 70kV nominal

-uA feedback

-kV feedback, used as switching feedback source

-10 turns of 14ga, ~1317 turns of 26ga, ETD29 core

-Full bridge Push-Pull converter

I'm still a little concerned about how the switching supply would work without an output inductor. Does anyone have an idea on how this type of switcher would react to a cascade Cockcroft-Walton type doubler being connected directly to the secondary?

The voltage after the cascade is where commercial machines get the feedback to the switching controller. I imagine that the load of the diodes and capacitors should keep the converter stable even when open circuit. When testing commercial machines we use blanking plugs to ramp them up to full potential without a load to check for converter faults and power supply operation. Since there is no beam current it takes a long time for the HV to decay.

I'll draft a schematic in the next few days and upload a PDF to my website. The HV part is the first obstacle. The next part would be designing the filament isolation transformer. If I can get the HV working I could at least test the tube with a battery and rheostat.

Grant

As in 2^6 multiplication? Sorry, doesn't work that way. Cascades are linear, not exponential. You put AC in and get DC out, so you would need an inverter to feed the next doubler and so on. You need about 60 stages to do what you want.

You'd be much better off with an off-the-shelf flyback transformer and a more reasonable cascade, like 4x or so. Note the FBT needs to be an AC output type; many have the diode(s) and focus resistors internally potted.

Tim

-- Deep Fryer: A very philosophical monk. Website @

Thanks for that warning... I don't know how that happened. Those technicians giving me "help" shouldn't be trusted. : ) I guess I'll have to only look to google for information. They never did explain how the mA feedback worked.

The commercial machine with 160kV 20mA output has 10 stages. The

225kV 10mA machine has 14. This I can trust because it is in the official maintenance manual schematics.

I was hoping for something smaller. I should be able to get at least

5,000-10,000v out of a small hand wound transformer. I've seen it done on an x-ray machine by hand. Maybe I need special enameled wire?

Using the same ETD29 core I would need 10 primary turns and 5266 secondary turns for a 5000v secondary. I'll have to figure out exactly how many cascade stages I want and then figure out the transformer.

I'm going to have to make a machine to wind that thing.

You're running your primary at 10 volts?? I don't recall reading in earlier posts in this thread any mention of that. In any event, for a particular core and frequency, you're going to be stuck with some minimum number of turns per volt, so, as they say, "deal with it." I don't mean that in a mean or sarcastic way; it's just the way things are.

Note that there's a big difference between "flyback" and "full bridge push pull": the output peak to peak voltage will be the primary peak to peak voltage times the turns ratio, and with a full bridge push- pull, the primary peak-to-peak voltage will be twice the supply voltage, nominally. With a flyback system, the primary peak to peak voltage can easily be several times the supply voltage; during the flyback portion of a cycle, you generate a relatively narrow high voltage spike. OK?? Now, note that a voltage multiplier circuit multiplies the peak-to-peak voltage. This means a couple of important things: you won't have to wind as many turns with the flyback, AND perhaps more importantly, you won't have to control the primary-side input voltage to control the output voltage. Note that with the push- pull excitation, the output peak-to-peak depends almost entirely on the primary side power voltage and the turns ratio. In such a SMPS, the output voltage is controlled by varying the pulse width, which controls the _average_ output voltage, and a choke-input filter turns the pulsating average into a smoothed DC average; but your multipliers won't be able to use that.

Another way to do the job: use a resonant tank circuit to step up the AC voltage generated by the SMPS; vary the frequency to vary the output voltage. It's quite possible to get a large voltage stepup with a resonant tank. For practical frequencies, you'll need reasonably fast diodes for the multiplier... And the output of the resonant tank will be fairly strongly dependent on load.

Cheers, Tom

(Sounds like you ought to be using SPICE to look at expected waveforms. You can learn a LOT about things that are obviously unclear to your right now, if you do that. LTSpice from the Linear Technology website is free and easy to use.)

12 volts. I used a calculator at to come up with the core and turns. It suggested that as a "good" but not "very good" match.

I'll try to find more information on the minimum number of turns per volt. 10-15 turns sure isn't a lot. Is there a general rule?

I chose the SMPS mode based on the one used in a commercial x-ray generator. I figured if it was good enough for them, it would be good enough for me. However, that system is 4kW and the fly-back method isn't as suitable for that amount of power. The flyback method does seem to reduce component count/size which is one of my goals. It appears at least by the calculator that the core size increased. That must be because of the peak current difference.

The automatic calculator did cut off about a thousand turns for the transformer calculation, but still left me with 11-13 primary turns for all example core configurations. I'm thinking about running the system off of 24v nominal instead of 12v.

Thanks for the jump start on heading in the right direction! Even if I never finish the project I will still have learned something in the process. : ) Winding the transformer sounds fun. I'm thinking of ways to make a machine to count the number of turns as I spool in the wire.

I'll try out the SPICE program you mentioned tonight.

Grant

Oh, color of a different horse. 600uA @ 8KV is a lot easier than umpteen mils @ 80KV. Not i purposely mis-read your "80KV" in that list... That load is no problem.

Well, i have about nine pounds of #32 double build Kapton MW16C which might be useable (if not, it will last at least 20,000 years at present rate).

Where did you find that calculator? All of those i have seen limit one to 250V or less at any output. Nobody wants to talk about high voltage unless they define high voltage as (maybe as much as) 36V input on a buck converter with maybe

2V output at 50 amps (exaggerating by not much here).

I did a Google search for SMPS calculator and it was the first result