A precipitator PS: some observations & meaurements

[Rich Feldman]

I've joined the club of precipitator power supply experimenters.
Hope to contribute a few more bench measurements with resistive loads, and reverse engineering notes,
following the footsteps of Finn Hammer and others. viewtopic.php?p=77912#p77912

First note: these "inexpensive Chinese power supplies" need to be identified by more than number of flybacks in series.
On ebay right now there are obvious differences between product series called "300W/400W/600W 30/40/60KV".
What I received looks practically identical to the latest postings from William E,
except for the angle at which the secondary coils are glued the cores.

I snipped pictures of a competing product that's different about number of trimpots, presence of small plug-in connectors, etc.
It doesn't help that there's no product manual or even model number.
Has anyone received a PPS that's different from its own online listing, as if the sellers consider them interchangeable?

[Rich Feldman]

>> Has anyone received a PPS that's different from its own online listing, as if the sellers consider them interchangeable?
Turns out that happened to me! The listing images above, showing a board layout obviously different from mine, could have been snipped from listing from which I bought my PPS. I had unpacked the board without noticing.

It's obvious why so many recipients report cracked cores. ( I was lucky: just a couple of ferrite flakes loose in the box, and a broken nylon tie-wrap around the secondary windings. ) It seems like the distributor takes an individually boxed unit out of a wholesale carton, covers it with plastic bag and shipping label, and hands it to Fed Ex. For the low prices and free shipping, does anyone expect seller to put the unit box into a real shipping box, much less place some padding on all six sides? Remember, a parcel service dropping box onto pavement doesn't count as abuse!

In my case, all four bolt ends had penetrated the box _and_ the white plastic wrapper. Might have damaged other parcels. Two of the bolt ends are significantly bent. Maybe with a service like ebay (as opposed to Amazon or Walmart), buyer could have a private dialog with seller asking for at least an extra layer of corrugated cardboard.

Moving along the road to posting feedback on ebay:
I had some temporary setbacks with a 1 Meg resistor bank previously used for NST characterization.
Decided to start with a secondary circuit resistance measurement. Recorded the voltage drop with currents ranging from 10 uA up to 10 mA. Here shown with linear and logarithmic Y axes.

The exponential current up to a few dozen volts is expected for HV rectifiers. On top of that, we see voltage drop from resistance of about 1,780 ohms. With 80 volts applied in reverse direction, current was less than 1 uA. Be interesting to see similar measurements from other users, esp. with more secondary coils.

None of my digitally metered bench supplies could deliver 80 volts DC, so I dusted off an old vacuum tube based supply. (regulated 0-150 and 0-400).

[Rich Feldman]

Got my PPS (CPPS? CCPPS?) running, at 5 kV unloaded, in the third lab session. That followed a couple of frustrating failures, including test of an identical PPS borrowed from Sid and powered for the first time.
Step one was unboxing, and assessment of shipping damage. Then, to avoid risk of starting at 30 kV and getting destructive arc between ferrite and the root of HV wire, I turned the primary voltage control all the way widdershins. When 240 V AC was switched on, the green LED came on and there was no smoke, but output voltage was 0.00 kV. Even after disconnecting my 1 Meg 1000 W load. Same result with Sid's PPS.

Easily explained, starting with a post from 2017:

The voltage is regulated with 2 pots, where the one of them sweeps the output through the range, the other is a kind of bias pot, to the effect that in one extreme, the output is adjustable from 0-10kV, in the other extreme from 4-18kV. These numbers are with the 1.6Mohm load. With a proper 4Mohm load, I am sure it would regulate up to 30kV as advertised.

I made some resistance measurements before experimentally turning the voltage control knob back to its as-received position.

First studied Rex's diagrams from 2019 of a PPS apparently identical to Finn's and mine (and Sid's. Maybe Joe's and Mark's et al).
viewtopic.php?t=12919
Let's identify the 3/4 turn voltage trimpot as VR1, in series with 10-turn pot VR2. Current limit pot is VR3. All three are 10 k ohms.
Found VR2 at 0.47k in my specimen, 1.11k in Sid's. With VR1 all the way down, totals (VR1 + VR2) were 1.64k and 1.13k.
After putting VR1 back to pictured position, total R was 6.49k. Switched on, we got green LED, then after a second or two, red LED and 5.19 kV.
The DC power at U1 (switcher control) went from 8.something volts to 11.29 V. A DMM in frequency mode, probing DC bus Common (it would be wrong to call it Ground) and gate drive pins, indicated 99.5 kHz. Consistent with chart in the datasheet of a chip similar to U1.

In other news, my previous post had charts of V/I curve measured between HV terminals of inactive PPS. Rex's notes show similar voltages measured at 1 mA and 10mA. His sketch of a coil and diode inside each secondary unit matches mine. But then Rex cut one open and found a 4-diode bridge. So our measured voltages are not for 2 diode drops + 2 coils, they are for 4 diode drops + no coils. The resistive term in my data is higher than I'd expected, even for 4 diodes.

[Rich Feldman]

When characterizing PPS with resistive loads, charting details like trimpot resistance and switching frequency and RMS input current (unless I find it's already been done and reported here), I'd prefer to start with everything dry. How far can we safely go, without messy oil immersion of PS or load? (The latter would also create need to cool a container of oil, or limit the run time.)

In a fusor we need to ground the positive output.
On the bench, I expect the voltage stresses are halved if we keep the HV output centered around ground.
That's trivial to achieve with my balanced kV meter, whose 1000:1 voltage divider has its "little R" in the middle of 90 megohm string.
viewtopic.php?p=80678
It's also easy if the load is made of many resistors in series. (Beware of connecting an unbalanced KV meter between + and - terminals.)

Can anyone tell us if the intended precipitators use these power supplies immersed in oil?
Are the plates grounded, or isolated so the HV potential can be balanced between negative corona wires and positive plates?

My PPS came with ferrite cores and their U-bolts electrically isolated from everything else. Doesn't that mean their electric potential is defined only by leakage currents? Anybody see problem with tying both U-bolts (through a high ohm resistor and perhaps a gas tube shunt) to ground, or DC bus common, or one side of the AC input? And how about the black anodized heatsink? I forgot to check if it's connected to the circuit, and PPS is not at hand here.

[Rich Gorski]

Rich,

My guess on safe HV output without oil immersion would be something like 15kV max. Unless someone has cranked one of these in air we just don't know how good the potting insulation is. Old TV flybacks for the 27" and larger picture tubes would run up to 30kV without oil immersion but are these secondaries insulated well enough to go that high? Since the secondaries are connected in series the transformer opposite ground will be the one to worry about. My guess on the u-bolt potential is that it is running near ground since they don't appear to be very well isolated from the primary. Also I think there is a HV rectifier on the secondary so the output will be half wave pulsed DC.

Rich G.

[Bob Reite]

The outputs are totally floating on these supplies and the load is also totally floating. Once one end is grounded, I can see insulation resistance becoming a problem. However I played with one of these "precipitator" power supplies a few years back and had it working with the positive side grounded. It only died when I tried to extend the voltage control off the board. It did not like the extra lead wire length, due to inductance and capacitance I guess and the magic smoke came out of a poorly marked IC.

[Rich Feldman]

Thanks for the suggestions. For first loaded tests I kept voltage very low and made no effort to ground any part of the HV circuit.

First note: PPS puts out rectified AC, and it'd be nice to measure the ripple amplitude. Load capacitance tends to smooth the voltage, and load inductance tends to smooth the current. At what point do they matter?
Suppose R were 3 Meg (30 kV / 10 mA) and frequency were 50 kHz (from Finn Hammer's OP).
Then any shunt C of more than 1.06 pF would conduct more AC current than the R.
A series L would have to be 9.55 henrys to match the impedance of the R.
Arithmetic review would be welcome!

A tale of two loads:
My NST power characterization (up to 450W) used a bank of ten 100 kΩ 100 W 6.5-inch-long wirewound resistors, found at the flea market long ago. Each mechanical attachment clip had been electrically tied to closest resistor terminal, I guess to prevent undefined potential difference across the small air gap. Three R's had failed open after a few years of storage outdoors, in spite of vitreous enamel coating and the natural corrosion resistance of, uh, resistance wire. One had healed itself (as measured with ohmeter) after being unsoldered; the other two were replaced with bigger R's of same nominal resistance. This was tested a week ago with a 15kV 30mA NST. (When driving a single R the NST delivered about 28 mA, 2800 V, 78 watts.) Then it was tested with PPS, getting only 0 volts, later explained by voltage setting knob being too low.

For more resistance, flexibility, and portability, I started a Low Cost HV Load project at the end of 2015.
Bought a full reel of 1000 2-watt 350-volt axial-lead metal oxide resistors, brand new from Digikey, for about $0.038 each. (In today's catalog they are $0.0536.) Stackpole RSF2JT47K0.
I chose 47K resistance, which was in stock at the low price. About 21 in series per megohm. (This week I realized that 64 in series is almost exactly 3 megohms.) Would hit nominal power at 6.5 mA with voltage margin to spare. 7.5 mA would hit the voltage limit, but not overheat if cooled with a fan. Might try 10 mA (4.7W !) with aggressive air cooling, unless someone says to forget about it.

Finn Hammer showed a resistive load with string of R's in a helix around a vertical cylinder.
In 2016 I sketched a zigzag configuration. Tried it today with 32 R's (measured total 752 kΩ + 752 kΩ, cost total $1.22), soldered on a jig that aligns 16 at a time. Full length axial leads are copper, not plated steel, and will help with cooling. It's too soon to tell about corona behavior at high voltage.

The units will be mechanically supported with strips of dielectric material that have small holes every 1/2 inch.
For initial test with PPS, i used a found object made of stiff plastic foam.
PPS got a test winding of 6 turns on one core, to get induced voltage waveform.
Voltage knob setting same as before (about 6500 ohms VR1+VR2).

DC output 5.11 kV with meter only (56 uA) and 2.98 kV with 1.504 megohm load (2.01 mA total). Core magnetization frequency 99.5 kHz with both load states. Sense winding voltage 320 mV RMS both ways, which is much lower than expected. I bet that's due to high frequency rolloff in the AC Volts mode of digital multimeter. (How many pF shunting 10 megohms? How fast is the True RMS comverter? ) Time to bring over an oscilloscope.

Rich G mentioned rectifiers built into the HV winding modules. Rex A showed that each module has a 4-diode bridge rectifier, so I bet the output voltage ripple frequency will be 2x the transformer AC frequency. ANd relative polarity of primary windings on the 2 cores won't matter; it might even be different on different specimens.

[Rex Allers]

For those who haven't seen it in my documentation... Here's the pic of the HV rectifier inside the insulating casting of one of the HV transformers from a precip supply. Supply was bad so I sacrifaced this one. Took quite a while for a slow archaeological dig into this hard material.

I was quite surprised to find a 4-diode full wave bridge. In the labels I added, the two AC are the outputs of the HV secondary coil inside this module.

[Rich Feldman]

Thanks for sharing the picture, Rex.

Here is some news from the "low cost HV load" front.
I found that a string of 47 kΩ 2-watt resistors can be operated at 10 mA (4.7 watts per resistor) with forced-air cooling.
My 3 megohm, 300 continuous watt goal could be reached with the 64 resistors already soldered into strings, plus a challenging and noisy air-blowing project. Or reached with silent natural convection, using two 6 megohm strings in parallel at 5 mA each, after soldering and supporting 192 additional resistors (12 more jig loads). Or reached using a smaller number of resistors with higher power rating. Or, probably, oil immersion with cooling of the oil.

That was learned by experiment, after futile searches of the Internet.
From Stackpole RSF2 datasheet, we know these devices could operate continuously in an oven at 235 °C if their power dissipation were close to zero.

(That's hot enough to melt even lead-free SAC305 solder.)

Resistor surface temperatures were measured using a thermal imaging camera.
A single resistor in free air, operating at 300 volts (1.9 watts) reached apparent temperature of 176 °C, for theta_JA of about 81 K/W.
Blowing air with a quiet shaded-pole-motor fan reduced temp to 84 °C. About 33 K/W.
Increasing power to 4.7 watts (470 V) brought indicated temperature to 186 °C, about 35 K/W, with fan still on. Seemed promising.

(Other resistors in picture, partly shielded from fan air, have relative voltages of 0, 1/3, and 2/3. Power dissipation of 0, 1/9, and 4/9 of the air cooled R. There's some misregistration between thermal IR image and background digicam image, due to parallax of targets that are close to the lenses.)

Next step was 0.75-megohm string over same fan. Power from NST on Variac; measured with AC/DC kV meter and a DVM on one resistor near center.
At 404 V (3.5 W) per R, the temperatures ranged from 175 to 213, apparently due to uneven airflow.
Time to try big noisy blower seen in 110:220V transformer thread, with resistor string centered on the exhaust port, later shifted to the inlet side of exhaust port. Got voltage up to 470, so whole string has 7500 V and 75 watts. In 16 resistors, none hotter than 146 °C. 27 K/W, with blower on its lowest speed setting.

We can still see significant temperature differences from resistor to resistor, especially after switching off electric power and watching the string cool down, so I think the cause is nonuniform air velocity. The threaded standoffs extending toward camera add enough weight that the resistor unit isn't levitated by the air current.

Plan for now is to mechanically assemble 3M string or 4.5M string, and use that on PPS for currents up to 6.5 mA with casual cooling.