Time to give this its own thread. Got some progress to report on Version 3.
Immediate goal is a load that can be configured for 3 megohms, 30 kV, 10 mA, 300 W, continuous operation. Be nice if it's got some margin and/or is extendable to fusor supplies bigger than the PPS we've been talking about.
Version 1 is the bank of 100 KΩ 100W tubular resistors, from flea market long ago, pictured in my CCPPS thread:
viewtopic.php?p=104418#p104418
To use this style in 10 mA application, the high power-per-R is mostly wasted because the maximum ohm value is 100K. To get 3 megohms we'd need 30 resistors at total cost of about $500. String could handle 31 mA and dissipate 10 times more power than we need.
Version 2 is a series string of 47 kΩ 2W resistors, pictured later in same post. We can get 3 megohms with 64 R's, but it hits rated power at 6.5 mA. 10 mA is possible with big noisy air blower, airflow management fabrication, and accepting operation far above the rated working voltage per resistor. Could handle 13 mA with casual cooling by using 256 R's in two 6-megohm strings; then the project would be about mechanical support and tedious soldering. Or get a smaller R value, say 20 kΩ 2W. Then 10 mA uses full power rating of each R and has plenty of voltage margin. Qty 150 in series would match today's target for load bank, and cost less than $20.
Version 3 is a test of 5 watt cement resistors, to reduce the number of parts. At a local electronics store I got qty 24 of 47 KΩ value. 10 mA puts them very close to both power rating and voltage rating.
Tried an inexpensive way to mount them with good air circulation. 16 had been pre-soldered into strings of 4 on a fixture, for a plan with fewer support points. The other 8 were pushed into saw slots and soldered in place.
Assembled string measured 1.13 megohms. In initial power test, with a quiet fan blowing air gently across the array, variac was turned to get 473 V measured across one resistor near the center. Current 10 mA, total voltage 11.3 kV, total power 114 watts. Thermal imaging camera indicated max temp of 172 °C, which won't bother the resistors. (Cement resistor datasheet shows power derated from 100% to 0% as ambient temperature increases from 70 to 275 °C.) Plastic support strips (PVC?) showed no damage, even from the soldering.
Caveat: I haven't studied the measurement accuracy of FLIR camera and how it deals with target emissivity differences.
Unless a better idea comes up, I want to expand this to 64 or 96 R's, for 3.0 or 4.5 megohms. Mounting strips will probably be thin FR-4 material since I don't have enough of the plastic angle stock. Will improving the soldering fixture be worth the trouble?
Will learn about corona behavior with all those skinny wire ends. Load resistor arrays could operate under oil, but then we'd need to cool the oil. For the floating output of PPS under test, we can hold the load center at ground. That trick won't help when testing HV supplies that inherently ground one side of output.
The X and Y pitch are driven by resistor size & no cutting of leads. It's possible to get twice the density & waste less airflow area, while keeping all R's in the same plane. But that makes physical support and string-to-string connections more complicated, not to mention air gap issues. Anyone know rules for minimum spacing of parallel wires at, say, 4000 volts AC?
HV Resistive Loads at low cost
- Rich Feldman
- Posts: 1559
- Joined: Mon Dec 21, 2009 6:59 pm
- Real name: Rich Feldman
- Location: Santa Clara County, CA, USA
HV Resistive Loads at low cost
All models are wrong; some models are useful. -- George Box
- Bob Reite
- Posts: 605
- Joined: Sun Aug 25, 2013 9:03 pm
- Real name: Bob Reite
- Location: Wilkes Barre/Scranton area
Re: HV Resistive Loads at low cost
Figure anything less than one inch per 30 KV will arc over in air.
The more reactive the materials, the more spectacular the failures.
The testing isn't over until the prototype is destroyed.
The testing isn't over until the prototype is destroyed.
- Rich Feldman
- Posts: 1559
- Joined: Mon Dec 21, 2009 6:59 pm
- Real name: Rich Feldman
- Location: Santa Clara County, CA, USA
Re: HV Resistive Loads at low cost
Thanks, Bob. We know it's very sensitive to sharpness of the conductors. I think for smooth 20 cm spheres a 1 cm gap can hold off about 30 kV.
[edit] Here's one reference: https://www.eeeguide.com/sphere-gaps-ar ... asurement/ . It says "in air at 20°C and 760 ton pressure".
Anyway, I got a practical answer by experiment!
DId a high voltage proof test after populating the first two strings in a second bank of 32 cement resistors, using an easier mechanical support scheme. Bent resistor leads are slipped into holes of perf-board and soldered in place. If the drilled holes had a little more clearance for resistor lead pairs, it would be easier to change the configuration without major re-soldering. Harder to keep the wires close and parallel before soldering, and we would depend on gravity to keep the strings in place.
The design voltage is 500 V per resistor, so serpentine path for current can have as much as 4 kV between parallel resistor leads.
Is 0.6 inch spacing enough?
The two strings were connected to a NST on variac, putting same voltage on all five dielectric gaps.
Indicated current was 0.0 mA until we reached about 8,400 volts AC, when arcs would start between various resistor lead tips. I figure peak voltage divided by gap width exceeded 19 kV/inch, 7 kV/cm.
For slightly higher breakdown voltage, we might cut the resistor leads shorter, bend them to make little U-turns, or insulate them with heat-shrink tubing. Or use much thicker support strips. Or, uh, increase the pitch!
There's no need to fix a thing that's not broken, eh?
It looks like there's a reasonable factor of safety, especially since the load is intended for DC operation. We will keep the 0.6" pitch and stuff the remaining resistors. It wouldn't be hard to do a "HIPOT" test with full voltage stress on all 35 gaps, before making the end connections for serpentine path. Who knows what we will learn after both banks (64 resistors) are all wired in series and connected to 30 kV?
Perf board is handy, but solid FR-4 sheets without unnecessary holes would be better. They would be easier to keep clean, to minimize surface leakage current. Making the support strips narrower would increase the airflow area, but if there's a fan involved then larger area might mean less pressure drop & air velocity at resistor surfaces.
We could get more voltage margin with same area density, using a slight diagonal zig-zag, but that would crowd some pairs of resistors. Could that increase their temperature, from restricting airflow on one side? I guess eventually, depending on boundary layer thickness.
[edit] Here's one reference: https://www.eeeguide.com/sphere-gaps-ar ... asurement/ . It says "in air at 20°C and 760 ton pressure".

Anyway, I got a practical answer by experiment!
DId a high voltage proof test after populating the first two strings in a second bank of 32 cement resistors, using an easier mechanical support scheme. Bent resistor leads are slipped into holes of perf-board and soldered in place. If the drilled holes had a little more clearance for resistor lead pairs, it would be easier to change the configuration without major re-soldering. Harder to keep the wires close and parallel before soldering, and we would depend on gravity to keep the strings in place.
The design voltage is 500 V per resistor, so serpentine path for current can have as much as 4 kV between parallel resistor leads.
Is 0.6 inch spacing enough?
The two strings were connected to a NST on variac, putting same voltage on all five dielectric gaps.
Indicated current was 0.0 mA until we reached about 8,400 volts AC, when arcs would start between various resistor lead tips. I figure peak voltage divided by gap width exceeded 19 kV/inch, 7 kV/cm.
For slightly higher breakdown voltage, we might cut the resistor leads shorter, bend them to make little U-turns, or insulate them with heat-shrink tubing. Or use much thicker support strips. Or, uh, increase the pitch!
There's no need to fix a thing that's not broken, eh?
It looks like there's a reasonable factor of safety, especially since the load is intended for DC operation. We will keep the 0.6" pitch and stuff the remaining resistors. It wouldn't be hard to do a "HIPOT" test with full voltage stress on all 35 gaps, before making the end connections for serpentine path. Who knows what we will learn after both banks (64 resistors) are all wired in series and connected to 30 kV?
Perf board is handy, but solid FR-4 sheets without unnecessary holes would be better. They would be easier to keep clean, to minimize surface leakage current. Making the support strips narrower would increase the airflow area, but if there's a fan involved then larger area might mean less pressure drop & air velocity at resistor surfaces.
We could get more voltage margin with same area density, using a slight diagonal zig-zag, but that would crowd some pairs of resistors. Could that increase their temperature, from restricting airflow on one side? I guess eventually, depending on boundary layer thickness.
All models are wrong; some models are useful. -- George Box
- Bob Reite
- Posts: 605
- Joined: Sun Aug 25, 2013 9:03 pm
- Real name: Bob Reite
- Location: Wilkes Barre/Scranton area
Re: HV Resistive Loads at low cost
Make big round solder blobs where the resistors join. From what I can see of the photos, that may be a problem. It will get worse the further from ground that the component is. Might have to consider small brass balls at the junction points.
The more reactive the materials, the more spectacular the failures.
The testing isn't over until the prototype is destroyed.
The testing isn't over until the prototype is destroyed.
- Rich Feldman
- Posts: 1559
- Joined: Mon Dec 21, 2009 6:59 pm
- Real name: Rich Feldman
- Location: Santa Clara County, CA, USA
Re: HV Resistive Loads at low cost
Bank 1 and Bank 2 are now fully stuffed and tested at almost 10 mA. Together they amount to about 3 megohms and employ $16 worth of resistors.
Before HV breakdown tests, I tried blunting a few paired wire ends with balls of solder, as suggested by Bob. It was hard to make them pretty and the benefit is limited in this configuration (see below). For other small projects that want balls to mitigate corona and HV breakdown, how about split shot from the fishing tackle shelf, or hollow metal balls from bead chain ?
All new string-to-string gaps were proof tested by NST with digital metering of AC voltage and current. Bank 2's breakdown voltages ranged from 8 to 9 kV before onset of arcs, generally between wire ends. Sleeves of plastic hose suppressed those, but then at 9.4 kV arcs would start next to perfboard surface, and move along the 0.6-inch-separated parallel wires until they reached a pair of resistor bodies. The final string-to-string gaps in Bank 1 both withstood 8.9 kV without arcing, while current meter indicated 0.0 mA. Now if we apply 16 kV to a normally wired bank, 7 gaps get 4 kV each and the other 28 gaps get 3, 2, 1, or 0 kV. Time to aim for 30 kV test, using PPS, where we have to worry about voltage between the corner resistors and the rest of the world. Maybe balls will help then. If something fails it will be a learning experience. As practiced by SpaceX with rocket development, or Tesla with FSD feature in manned automobiles.
After HV breakdown tests, each bank got end connections to put all 8 strings in series. Measured resistance of Bank 2 was 1.507 Meg at room temperature, 1.477 Meg when hot, and 1.432 Meg when very hot; Bank 1 was 1.47 Meg when hot. Apparent tempco is very roughly -200 ppm/K.
Heating was by a 15-30 NST slightly overclocked (125 V input), for indicated 13.5 kV and 9.9 mA on one bank. Thermal imaging camera indicated that hottest resistors reached 200 °C (which is tolerable indefinitely) with no forced air; a quietly running fan reduced that to 125 °C.
The thermal IR view showed warm spots on each perfboard strip where the resistor leads touch, but no indication that the wire leads were warm. I bet the wires are hotter than the perfboard, but not seen because of their thin-ness and low emissivity. Perfboard heat isn't from leakage current or dielectric loss because it's about the same at all resistor contact points. I'm not in a hurry to check the kV and mA meter calibration. Both are AC/DC, with bridge rectifiers (See old NST Power thread), and long ago calibrated for DC. When measuring AC I've multiplied the readings by 1.11 to get RMS of a sinusoid.
Before HV breakdown tests, I tried blunting a few paired wire ends with balls of solder, as suggested by Bob. It was hard to make them pretty and the benefit is limited in this configuration (see below). For other small projects that want balls to mitigate corona and HV breakdown, how about split shot from the fishing tackle shelf, or hollow metal balls from bead chain ?
All new string-to-string gaps were proof tested by NST with digital metering of AC voltage and current. Bank 2's breakdown voltages ranged from 8 to 9 kV before onset of arcs, generally between wire ends. Sleeves of plastic hose suppressed those, but then at 9.4 kV arcs would start next to perfboard surface, and move along the 0.6-inch-separated parallel wires until they reached a pair of resistor bodies. The final string-to-string gaps in Bank 1 both withstood 8.9 kV without arcing, while current meter indicated 0.0 mA. Now if we apply 16 kV to a normally wired bank, 7 gaps get 4 kV each and the other 28 gaps get 3, 2, 1, or 0 kV. Time to aim for 30 kV test, using PPS, where we have to worry about voltage between the corner resistors and the rest of the world. Maybe balls will help then. If something fails it will be a learning experience. As practiced by SpaceX with rocket development, or Tesla with FSD feature in manned automobiles.
After HV breakdown tests, each bank got end connections to put all 8 strings in series. Measured resistance of Bank 2 was 1.507 Meg at room temperature, 1.477 Meg when hot, and 1.432 Meg when very hot; Bank 1 was 1.47 Meg when hot. Apparent tempco is very roughly -200 ppm/K.
Heating was by a 15-30 NST slightly overclocked (125 V input), for indicated 13.5 kV and 9.9 mA on one bank. Thermal imaging camera indicated that hottest resistors reached 200 °C (which is tolerable indefinitely) with no forced air; a quietly running fan reduced that to 125 °C.
The thermal IR view showed warm spots on each perfboard strip where the resistor leads touch, but no indication that the wire leads were warm. I bet the wires are hotter than the perfboard, but not seen because of their thin-ness and low emissivity. Perfboard heat isn't from leakage current or dielectric loss because it's about the same at all resistor contact points. I'm not in a hurry to check the kV and mA meter calibration. Both are AC/DC, with bridge rectifiers (See old NST Power thread), and long ago calibrated for DC. When measuring AC I've multiplied the readings by 1.11 to get RMS of a sinusoid.
All models are wrong; some models are useful. -- George Box