High power high voltage variable DC supply

[rhb]

I've discovered a bunch of 872A mercury rectifier tubes among Dad's stuff.  These are 10K PIV 5A half wave  rectifiers.  I also found a 2.5 KVA autotransformer.

I'm contemplating constructing a variable voltage HV PS using a couple of matched microwave oven transformers to form a center tap full wave bridge fed by the autotransformer.  Other than feed a high power laser, no real idea what I would do with it.  These were meant to provide the plate supply for 10-25 KW radio transmitters.  On the radio side I'm more interested in QRP.  This would really just be a lab instrument.

Obviously a very dangerous toy in the wrong hands.  And finding suitable capacitors and wires will be challenging as it should produce up to 4-5 KV DC at 1-2 A.

I'm contemplating switching the primaries in parallel or series to give me a high and low voltage range.

[Wolfgang]

First, and most important:

- Your first error is your last error. Play safe

- MOTs as HV tranformers have a lot of issues
  Made for resonant operation
  One secondary lead is attached to the core/ground, so no isolation there
  To run on full mains, they need their iron shunt removed and/or a stray inductor in series
  They need fan cooling.

[rhb]

Obviously both ends of the HV secondary will need to be isolated from the core. The one I have on hand has two secondaries, a filament winding and an HV winding with one side of the HV grounded via a screw terminal.

I don't see that a microwave oven PS is anything but a very low quality half wave bridge HV supply.  I'm old enough to have repaired tube equipment.  The last time was when the CRT section HV supply capacitor in my Heathkit IO-18 scope took out the rectifier tube and the replacement tube. 

I may well have a WW II era HV transformer.  But the MO transformers are easy to get and transformers from a 1200-1500 W oven will be more than stout enough for anything I'll ever do.  There is no way I'm ever going to strain the ratings of an 872A.

Not quite sure what you are referring to by "resonant operation".  They are designed for 60 Hz which is what it will be fed.  I'm not aware of a power transformer that is not designed for resonant operation.

Also puzzled by "iron shunt".  They run on full mains by design.

The 872As are much more fussy about cooling than the transformer.

As far as the design goes, it's  a standard plate supply project, just a bit over the top relative to the ham power limits.   It will need a couple of chokes in addition to the capacitors to filter out the AC ripple.  It *is* just a plate supply for a tube based linear amplifier.  A very common ham project.  This one just happens to be absurdly large, 3 KW continuous plate power with transformers from 1500 W ovens. 

Hmmm, maybe I should try to find 4 and go for 6 KW ;-)

[Wolfgang]

Some extra thoughts on this:

in a MOT, one end of the HV winding is attached to core (and ground). You can make a half-bridge with two MOTs where the primaries are driven with opposite polarities.

in a real microwave oven, the MOT arrangement is not a low-quality half-bridge, but a pulsed voltage doubler. Try simulate this. Dont forget the capacitor in your simulation.

It is an absolute misunderstanding that all normal line transformers are running in a resonant mode. They simply dont. Also, normal line transformers have no extra saturable magnetic elements (i.e. shunts), either. Google "ferroresonant transformer" to see how it is done.

You probably wondered why a MOT has such a small core size compared to a classic line transformer of the same power rating. It is specifically designed for a pulsed voltage doubler circuit as in a microwave oven and *not* for a general rectifier application.

The ham beast I know of dont use MOTs, but Peter Dahl anode transformers with a CCS rating.

[rhb]

I'm familiar with the voltage multiplier circuit already.  I've replaced a few diodes. But I'd not thought about the way that worked in a long time.   

I used a 10-20 KVA ferroresonant transformer to feed a MicroVAX II. That was a very different beast.  Boy did it generate a large field.  I had to move it away from the terminals because the text oscillated on the CRTs.

On close inspection I see that there are shunts between the windings.  They are so thin I had not noticed them.  Five laminations.

I'm far more likely to want to power a laser tube than a linear amplifier.  I'll have to read up on that application and the PS requirements.

Thanks for the heads up.

[N2IXK]

I wouldn't recommend using 872As nowadays. They were phased out in broadcast transmitters as soon as it became possible to do so, at first using long strings of silicon diodes with equalizing resistors and capacitors across each one, then , as silicon diodes improved, direct plug-in solid state replacements for mercury rectifier tubes were developed, and almost universally adopted nearly overnight.

Mercury vapor rectifier tubes were always problematic devices, requiring careful handling and upright operation to prevent splashing mercury onto the anode or other internal parts, a perhaps hour-long "conditioning" period prior to initial operation to condense all the liquid mercury into the bottom portion of the tube,  a several minute warmup period before application of plate voltage to prevent internal arcback, a limited range of ambient temperature to assure proper mercury vapor pressure, and careful shielding and filtering to remove the RF hash that they generated.

Then there is the whole hazmat disposal issue at end of life.

High voltage rectifier diodes are widely available, including from the microwave ovens you plan to scavenge the transformers from.  Much more reasonable design choice than using an 872A nowadays...

[Gyro]

...
I used a 10-20 KVA ferroresonant transformer to feed a MicroVAX II. That was a very different beast.  Boy did it generate a large field.  I had to move it away from the terminals because the text oscillated on the CRTs.
...

You sure you mean a MicroVAX II?    When I worked for DEC in the '80s, I had one of those sat under my desk plugged into an ordinary socket strip, as did all the folks in adjoining cubicles.

[Wolfgang]

I agree. Mercury rectifiers dont have advantages over silicon HVRs nowadays.
Silicon HVRs are cheap (the ones from microwave ovens are fine) and much easier to handle,
need no preheat period, can take more abuse (within limits) and last forever.

A microwave oven type HV fuse is a good idea in both rectifier legs, too.

[rhb]

Well, 10 KV PIV 5A diodes are over $1500.  But I generally agree.  I think the 872As are probably more trouble than they are worth.  Especially considering I have no need for that sort of power level unless I decide to start shooting down satellites in space from my secret lair in a volcano ;-)

A MicroVAX II in a BA123 worldbox.  The  terminals connected to it did not like the field it put out.  This thing was a monster.  Well over 100 lbs.  It was lying around after the Seismograph Services Phoenix system was replaced by an 11/780.  It had powered the Phoenix.

I had two terminals besides the console running on the uVAX despite not having a multiuser license.  We just logged in on the console and spawned a terminal process.  I was rather proud of that.  I also ran the system at 100% CPU utilization for weeks at a time.  But if an interactive user logged in it appeared to be completely idle.  I set the working set extents and quotas so that if paging was required, the performance hit was on the batch queues, not the interactive user processes.  Something you could do with VMS that you can't do with Unix.  The advantage of a transient process space vs fork-exec.

[Wolfgang]

... Well, you could still use microwave oven rectifiers in parallel (with equalization resistors, of course).
10/12kV units with half an amp cost less than a $.

[N2IXK]

An 872A is only good for 1.25A average. The 5A rating is a non-repetitive peak current rating.

[Gyro]

A MicroVAX II in a BA123 worldbox.  The  terminals connected to it did not like the field it put out.  This thing was a monster.  Well over 100 lbs.  It was lying around after the Seismograph Services Phoenix system was replaced by an 11/780.  It had powered the Phoenix.

I had two terminals besides the console running on the uVAX despite not having a multiuser license.  We just logged in on the console and spawned a terminal process.  I was rather proud of that.  I also ran the system at 100% CPU utilization for weeks at a time.  But if an interactive user logged in it appeared to be completely idle.  I set the working set extents and quotas so that if paging was required, the performance hit was on the batch queues, not the interactive user processes.  Something you could do with VMS that you can't do with Unix.  The advantage of a transient process space vs fork-exec.

Sorry, wasn't doubting you, it just seemed a huge lump of Iron for a MicroVAX.  Yes, that was the little beastie - quite futuristic looking for DEC even if the grill at the top did have a bit of a storage heater vibe. Sad that all that, VMS included, has almost faded from memory. Yes reliability was the reason MicroVAXen (PDP11s for that matter) lived in industrial setting for so long - I'd lay odds that there are quite a few lurking in dark recesses. Personally, I'd have ditched the iron, but I guess you had environmental reasons for having it there.

Anyway, I'm diverting you OT.

[rhb]

The motivation was the uVAX locking up every couple of months.  I'd come in and turn on the console and it would show the login prompt, but nothing worked.  I'd reboot.  No crash dump, nothing.  This went on for about 18 months.  Support would come out each time and could find nothing.  We put a power line monitor on the circuit, but found nothing.  But the UT Comp center wouldn't let me keep the power line monitor for more than a couple of weeks so I dragged the resonant transformer from the basement and plugged it it.

I finally called and asked for the support manager.  He sort of was blowing me off until I started reading the service call dates almost 60 days apart. So he sent the wildest character I have ever seen in such a setting.  He wore a grey suit, purple shirt and had hair down to his waist.  He started with DEC working on PDP-8s.  This all took place around 1988.  He spent a day poking and prodding but could find nothing.  So he left and one of the regular level staff came.  We had the skins off and were trying to figure out what was wrong.  We were about to replace the backplane as that was the only thing that had not been replaced when it went down.  One side of the PS had shut down.  As it turned out, the fault was a combination of a bad thermal sensor in the top of the BA123 and poor design so it shut down  one rail rather than generate an interrupt, record a message and shutdown properly.

I think doing something with the 872As is probably lunacy.

[PointyOintment]

Well, 10 KV PIV 5A diodes are over $1500.

Why not use ten 1 kV diodes (any of the ones in the last row of the table on this page) in series? It lists parts rated up to 10 A. (Or you could put lower-current ones in parallel (with current-sharing resistors).)

[rhb]

This was a "What do I do with this?" project.  The fact that each tube includes a GE form for making a claim on defective tubes suggests I really don't want to mess with them.  If I can't sell them I'll salvage the mercury.

A constant current laser power supply is so much cheaper than a tube, building my own PS is not all that attractive.  For other uses I have no need of such high voltage or such large currents.

[Wolfgang]

Strings of 1kV (or 2kV) diodes work, but only with HV grade voltage equalization resistors (not small). It all gets a bit bulky, especially if you have 10 diodes and minimum 10 resistors per leg.

[Hexley]

I'm contemplating constructing a variable voltage HV PS using a couple of matched microwave oven transformers...

You might be interested in the article, "Build a High-Voltage Power Supply at Low Cost", by Randy Henderson, WI5W, published in QEX for Jan/Feb 1998. It is available with a little digging through archive.org at http://web.archive.org/web/*/http://www.vaxxine.com/phil/QEX_HV_Power_Supply.pdf

He used four microwave oven transformers to make a 2700 volt power supply that  "loafs along with loads drawing 1000 to 1400W..."

[Wolfgang]

Hi,

I remembered this article and now you found it before I did, thanks !!

Even when "unshunted", the MOTs could draw more current on idle than the could take, thermally. That depends on the type of MOT and needs to be tried out experimentally, because data on MOTs is incredibly sparse and only describes the data with the shunt in place.

1000 to 1400 W out of 4 MOTs sound reasonable, however.

My alternative would be 4 x 500VA isolation transformers with their secondaries in series and a voltage doubler. This should be good for ca. 2kV, with the same power.

[mzzj]

Strings of 1kV (or 2kV) diodes work, but only with HV grade voltage equalization resistors (not small). It all gets a bit bulky, especially if you have 10 diodes and minimum 10 resistors per leg.

Avalanche rated diodes would be alternative to equalization resistors. Most? of the high voltage diodes today are more or less controlled avalanche capable but types with guaranteed avalance  like MUR8100, BYV26, SF4007 are safer bet.
Orginal 4007 from seventies are probably failing with destructive breakdown without balancing resistors.

[Wolfgang]

The idea behind avalanche-rated components is that components in a series connection will break thru in a controlled way as long a maximum current is not exceeded.
While this should work in principle, I have never seen a HV PSU that did not rely on balancing resistors. You need a string of resistors anyway to bleed away the charge of your filter caps (It should be down to less than 60Volts after not even a minute). So, I see no big advantage in leaving the resistors out.