Fusor Forums

Paul, let me suggest a different tack.

A. Your OP says "starting at 2.5 megohms", and later you say things like "per leg". Are you planning to make a 2.5-MΩ 1000-watt string for use alone, to get 20 mA at 50 kV? To be followed by identical strings in series for lower current, or in parallel for full current at lower voltage? The resistors will be wastefully under-loaded, power-wise, in anything but the just-one-by-itself case.

Better to build the max configuration as a series-parallel combination of identical sub-strings, which can be reconfigured when you want overall R values higher or lower than 2.5 M. The simplest example is four strings, each 2.5-MΩ 250-W (max 10 mA, max 25 kV).

B. Do you have student labor and want to teach soldering skills? Think about using lower power resistors in large numbers -- they cost less per watt. At Digikey, 2 watt 500-volt Yageo FMP-200's are in stock for $0.02272 each if you buy a whole reel of 2,500 resistors ($56.80 plus shipping). Staying within the ratings, you can meet the OP requirement with 525 resistors ($11.93 worth).

The 120kΩ value has a good balance between max power and max voltage rating. Let's design for 4 mA, 480 V, 1.92 W per resistor. A series string of 105 resistors would draw 4 mA at 50.4 kV. Exactly divisible by 3, 5, or 7 if you want to build it in sections. Five of those strings would draw 20 mA at 50.4 kV.

You can find creative ways to spread out the resistors for good cooling and high-voltage withstanding. Blowing air over them with a fan can give you lots of power margin. I am not sure oil immersion gives any benefit that outweighs the hassle.

@Hank Ball: Good points. I did find a data sheet for it (http://www.newark.com/vishay-dale/rh050 ... dp/41K9147) that says it's rated at 2000 Vrms, and I don't think in normal use they'd ever exceed that (across each), so it would just be a matter of keeping them physically isolated. I hadn't thought about heat sinking them, though. The data sheet says without heat sink the dissipation is devalued to 40%, or 20W. I wonder if forced air (which I was planning to use) would get some of that back.

But maybe I should get these cheaper 20W ceramic cement ones instead: http://www.ebay.com/itm/5pcs-Ceramic-Ce ... 5adf0b3270. I could push 14mA through without violating the power rating, though I'd still like to test the supply out up to 20mA if I can get there. That would require them to dissipate 40W. OK for short periods of time?

Rich Feldman wrote:taying within the ratings, you can meet the OP requirement with 525 resistors

I like your design idea, but the thought of soldering and mounting 525 resistors in this network gives me hives, esp. since the dummy load is really just an accessory and not the main event. I wouldn't want to ask the students to do it for a 10-week class either - maybe if they were working on it as a summer research project, or if I had a couple dozen of them working on it, but they can't afford to invest that much time for the advanced lab when I'd rather they get to work trying out configurations of the power supply.

However, you have convinced me of the wisdom of (4) strings, each 2.5Mohm, 250W. I was considering maybe 3 strings, with at least one at 1000W, so it could be used by itself, but also get other resistances by combinations in series and parallel, but you're right - for the sake of an extra string I can drop the power requirements and the cost enormously. So now I'm thinking (4) strings of (25) 100 kohm, 10W ceramics like these:
http://www.ebay.com/itm/151744883697?_t ... EBIDX%3AIT

100 resistors to solder and mount is worse than 25 but a lot better than 525, it gives a lot of flexibility, and ends up being quite reasonably priced. And I don't have to deal with metal cases or heat sinks. Thanks for the suggestions!

-- pwf

Hank, you are right!

I just happened to connect some dots in a way that makes an old ebay prank more sensible.

Paul pointed to some nameless ceramic cement power resistors on ebay. U.S. component distributors like Mouser and Digikey offer familiar brands like Ohmite and Vishay. Turns out there's a Japanese maker called Fukushima Futaba Electric Co. Ebay always has resistors for sale under Fukushima title, bearing the Fukushima logo as seen on company website http://www.fu-futaba.co.jp/english/prod ... up_01.html and this image example: That helps to explain this ebay offering that I captured in 2012, not long after a famous tsunami event in Japan. Can anyone think of a reason to seek resistors of type "Film, Uranium" from FUkushima or FUtaba for a FUsor project?

As per my experience Fusor power supplies are generally lethal. Demo fusor's using NST transformers can be relatively safe when used by someone familiar with high voltage.
As for a test load for a high voltage, high current power supply, a torr range gas is an excellent ballast if one is in the right range.
Many lasers operate at 1 - 20 torr with 30 kV supplies and argon is an excellent gas.

I think resistor would be out, at V2/R and Pr*r its about 1kv per resistor, at the first foumla you will be looking at 1-5 watt for 50% tollernce, making the price $300-500, a switch and inductor might be better

@Andrew You're right, 1 kV per resistor max, so I would need 50 per string to go all the way up to 50 kV. (I think the tolerance would probably be in my favor, though - I could probably push the heat dissipation higher for short times.)

I just bought these instead:
http://www.ebay.com/itm/25pcs-Ceramic-C ... V#shpCntId

110 kOhm at 20W, so voltage drop of up to almost 1500V across each. That's the highest P*R I could find, and they're less than $1/ea. With a single string of 30 I can go up to 44kV without pushing the rating. That will at least let me test a load of up to 13 mA; for a bigger load I can add a second string, but that should be plenty to start with. Two of these resistors in parallel will also make a nice ballast resistor for the fusor and the other glow discharge tubes.

Now the challenge will be mechanical - I need to build a mount that will let them dissipate all that heat without starting a fire, and connect them in a way that doesn't produce corona.

-- pwf

OK I'll post a few pics of how I make the ballast and divider resistor arrays @ work
The blue resistors are VR68 and are arranged in a spiral picking up each accel plate to set the plate voltage. 120kV total approx 15kV between each plate these resistors are rated @ 10kV each and are cheap by the boxful off digikey about $180 for 500
The metal clad (gold) are RS 25watt 10ks as Ballast for 20kV working note also the array of 26 100 V transil diodes that creates the anode cathode differential
The two big green are WW 30K each ballast for the accel tube these are used @ 120kV hence the corona rings
The group of four green ones are WW 7.5K ohm each all in series these are run up to 40kV note the remoted Glassman supply that floats @ 40kV

Thanks for the story & pictures, John.
By coincidence, this week I chose & ordered some resistors for a HV load spec'd about like Paul's: 2-3 megohms, 200-300 watts. Am looking forward to comparing notes.

Andrew and Paul should know that resistors have model numbers, data sheets, and voltage ratings independent of the R value. In DC service, tolerable voltage is sqrt(P*R) or the voltage rating, whichever is less. For high resistance values, the voltage rating gets you first. Even sooner in AC or low duty cycle cases.

For example,
a 1 watt 10 kΩ resistor is power limited to 100V DC or RMS AC. The power rating allows 316 volts at 10% duty cycle, which is OK for most 1 watt series.
A 1 watt 1 MΩ resistor is power limited to 1000V DC or RMS AC, but most 1-W series have voltage ratings of 350 or 500 volts.
http://www.ohmite.com/cat/res_od_of_oa.pdf John's stuff has some axial lead 1 watt R's with 10 kV rating, from a special series designed for high voltage. Don't know the voltage rating of R's like Paul's, but would guess 1500 V is OK. Try browsing similar-looking components at non hobby-oriented stores.

@John Thanks for the photos! Where do you get those nice corona rings? I was thinking I would make one out of copper tubing, but if I wanted to get a commercial one I have no idea where I would look.

Speaking of, with all the HV experience here someone will be able to tell me - am I likely to need more than one corona ring if I'm going up to 40 to 50 kV? My plan is to connect my resistors in a helix like in the Glassman example Preston posted early in this thread. Probably six resistors per turn, approx. 15 cm diam., with 2.5 cm turn-to-turn separation. That makes about 5 turns for a stack of 30 or so resistors about 13 cm tall. Then a corona ring on top, and a terminator where I can connect my HV cable. Is there a rule of thumb about how often to include corona rings? Seems like I might need one mid-way, at the 25 kV point? Or would it be sufficient to try to keep my connections between resistors as smooth as possible? I can just try it and see and add one as needed, but thought there might be some engineering principle I should learn.

@Rich Thanks for the tip! I knew resistor voltage was limited by flash-over in addition to dissipated power, but wasn't aware that these power resistors could also break down internally even if the power/current rating was not exceeded. A quick study online seems to indicate that exceeding the voltage rating causes a gradual decline of resistance value over prolonged use rather than catastrophic failure. I couldn't find any ratings for resistors like these, so I'll just give it a go and keep an eye on it. Good to be aware of.

-- pwf

My eight stage CW multiplier has a corona ring at each stage, but that is probably overkill, unless I decide to push it to 100KV. So maybe have a ring for every 12.5 KV?

The corona rings are 12mm aluminium rod rolled into a circle and welded

Bob
a corona ring every 12 kV is conservative but you will not regret it
I seem to have more trouble after 20kV with the odd sharp edge

On a normal PCB, around 300 V per mil is a conservative design point with uncoated traces. More than that and you need to start looking into adding slots and so on.

In air, breakdown starts at electric field strengths of around a megavolt per meter, or about 10kV per cm. The best geometry for between a given voltage drop is two spheres. Planes, or worse points, will concentrate the field and lead to breakdown at lower absolute voltages.

Past a few 10's of kV, design rules are hard to pin down. Everything goes non-linear and becomes much more empirical. If you have the capability on your setup, adding one stage at a time while appropriately scaling the load voltage and recording the drain current will show you exactly where any corona losses are happening. This will show up as an anomalous current not accounted for by Ohm's law and the basic resistor circuit to ground.

A classic designer's trick is to imagine stretching a rubber sheet over the HV section towards ground. Where the sheet would break first is where you should expect the most arcs.

Always avoid sharp points and keep the voltage gradient headed in one direction without folding it back on itself.