Precipitator HV Supply Hacking

[Mark Rowley]

Rex,
I just sent you a PM.

Mark Rowley

[Mark Rowley]

Rex and I were chatting earlier about how to disable the automatic current limiter on the 60kV supply. The method ended up being quite simple. The red wire going through the current sensing transformer just needs to be rerouted around it instead of through it. This stops signal voltage that’s proportional to the current from being sensed in the unmarked 16 pin processor chip.

Other than shutting down the limiter function, I have yet to see any adverse effects. Just make sure the supply doesn’t go over 10ma.

Just a couple reminders....
These supplies do not respond well to flashover arcing. I’m a strong proponent of potting the whole thing in oil. Take every precaution and engineering effort to make the fusor arc/flashover proof.

Extending the voltage control pot with twin lead may have been the culprit regarding the earlier IR2153 failure. Speculation at this point but there’s been no issue since I removed the extension. I’d advise against attempting it until a better arrangement can be determined.

The following picture shows the location of the current sensing transformer (circled in yellow). There’s another one near the 220v input that’s not relevant to this modification.

Thanks Rex for all the work you’ve put into this!

Mark Rowley

[Rex Allers]

Mark,

I never heard back from you. Did you try just turning that current level pot max clockwise?

Moving that wire out of the sense transformer definitely does disable current sensing in the primary switching of the supply, but I didn't actually suggest this.

Well, what works, works -- at least for now. Just wondering if there was a change that would minimize tripping but still let some extreme current kill the output.

Since I have yet to put one of these into any real test environment, only looking at the circuits, I can't know a better way.

Good to know that my work has helped, even if I wouldn't highly recommend this way.

[Mark Rowley]

Exactly, we had no definitive conclusion as to a method and you made no recommendation to reroute the line. I really didn't want to move the discussion to personal emails but I respected your wish to not go further with the fusor website messenger emails. All is good and very appreciative of the communication. I'm still thinking that upping the value of the resistor from 1k to 2k may allow for higher current capabilities but still offer a shutoff in the case of a dead short. I may try that at a later date.

Full clockwise on the current limiter puts it into shutdown mode. 50% clockwise allows for very little current flow...still almost instant shutdown. In fact, just barely tweaking it clockwise wont allow for much more than a few seconds of operation without tripping. I confirmed this is normal with these supplies as my spare 60 does the same and a friend who has one reported the same as well.

I'm not observing much evidence of the supply being taxed from the modification. Probably the only risk on my end would be if the tungsten grid melted and shorted to ground.

Mark Rowley

[Rex Allers]

Interesting, Mark.

That current pot behavior is exactly the opposite of what I expected.

You mentioned possibility of changing a resistor value for current sense. In this pic is where I think the current from the sense transformer (bypassed in your change) is converted to a voltage that is passed to the micro.
Bigger current = higher voltage, because E = I * R. So E which is passed to the micro goes up with current.
So if we want the micro to be less sensitive to the current we want E to be less for a given current. So we actually want R to be lower, not higher.

Changing the the current sensing R to a lower value might work. Here's where it lives from my circuit hacking. The 1K R just above the C circled in this image.

After thought: You could tack-solder another resistor across this sensing one. Another 1K tacked in parallel would lower its sensitivity by 50%.

[Mark Rowley]

I think I have a few 510 ohm 1/4 watt resistors in the shop. Or, I most certainly have a few 1k pots which may work better at determining the effectiveness of a value change.

I'll give it a shot and report back.

Mark Rowley

[Mark Rowley]

Rex,
Last week I picked up a second 60kV supply for an upcoming ion gun project. Low and behold, the numbers on 16 pin mystery chip weren’t fully scraped off.

It’s a cheap and relatively abundant 15W408AS 16 pin microcontroller.

Here’s an eBay auction for one;

https://rover.ebay.com/rover/0/0/0?mpre ... 2256556105

Hopefully this helps solve the final questions about how the 60kV unit operates.

Mark Rowley

[Bob Reite]

Now to figure out the programming of the microcontroller.

[Rex Allers]

Mark,
Yes, correct on the chip, but I already figured that out. See my post in this thread of Aug 29. It is yet another version of the 8051 chip design.

Edit: Going back, I now see there were lots of posts on Aug 29
Let's see if this link goes more directly to the post I meant...
viewtopic.php?f=11&t=12919#p83946

Bob. I never felt inclined to try to read the chip's code. One would have to unsolder and put it in a reader. If the code is not protected then maybe one could begin to disassemble and start to figure it out. A lot of work if possible at all.

Here is another approach for these boards. I wanted to try this but never got it to the top of my list.

A while back I made a prototyping test board for playing with switcher supply circuits. It is a 3525 switcher controller chip that is very popular in many switch mode supplies. The one I used is SG3525A but there are many clones from different vendors out there.

Here's what I made:

Proto board

Proto board Front Panel

The two pot adjustments on my device set:
The frequency -- between about 28 and 90 kHz
The duty cycle -- between zero and 47%

The 47% is one half the cycle for each of the two phases, so essentially full maximum drive.

The output of this board is two signals with pulses for the A and B or High and Low switch drive into the two power switch devices of the supply.

On the design of both the 30 KV and 60 KV versions of the precip supplies, the main switching control chip is an IR2153. This has H and L outputs that drive the gates of the two switch MOSFETs. The 2153 also has an oscillator like a 555 chip that makes the frequency of these gate drive pulses. These supplies seem to control the output by adjusting the frequency of this oscillator within the 2153.

This seems a strange way to control the output vs the many typical switching supply configurations. PWM is the most common method to vary output.

My idea for trying a new hack of these supplies is to replace the IR2153 chip with an IRS2106. The 2106 is a simple driver chip with no oscillator. The power pins and the output drive pins are the same between the two chips. The inputs (pins 2 and 3) on the 2153 control the oscillator frequency and on the 2106 are just input drive pulses.

My though is to unsolder the IR2153 (Mark has done this for repair) and to put in an 8-pin socket. Then I would put a 2106 into that socket. The traces on the board would need to be cut around pins 2 and 3 to isolate them so they can be driven just by the two H and L signals from my 3525 board.

Next would be experimenting to find the best operating frequency and learning about setting duty cycle vs output V and I. After a good frequency for operation is found, this could be set with fixed components on the 3525 circuit.

The control abilities of the 3525 could then be used to take feedback of the HV output to regulate the supply.

Something like that. Should be fun to try and hopefully could make these precip supplies have a lot more finesse.

[Mark Rowley]

Rex,
That post was somehow integrated into the missing posts/site migration discussion that was specific to this thread. For whatever reasons it never showed up on my feed, hence the subject matter of my followup reply on the link you provided.

I have a nearby friend who's hobby is microprocessor programming. I'll see what he says about decoding....way beyond any of my interests and water under the bridge concerning any fusor application. The 60 has been operating great as-is and can provide very decent numbers for the cost.

Mark Rowley

[Mark Rowley]

Rex,
I like the 2106 idea. Sounds like a cool project and may give it a try after the first of the year. By then I'll have two 60's operating on different machines.

Mark Rowley

[Alan Sailer]

First I'd like to thank Rex for his terrific write-up on the presipitator supply. It allowed me to repair a blown up 30kV unit.

I would like to say that the "mystery" half bridge driver chip can be replaced with an IRS2153 with no obvious problems. I have not
yet re-attached the unit to the fusor but the voltage does go through it's paces.

Cheers.

[Rex Allers]

Alan,

Thanks for the good words on my shared results from hacking these supplies.

The two mystery chips (markings sanded off) were only mysteries for a while after the hacking began.

In one of my zips for the 60 KV unit there is a file with this:

------
Two chips used on the 60 kV supply

1) 8-pin driver for the switching transistors is an
IR 2153
it is a combined half-bridge driver and is switched by its own
internal oscillator, similar to a 555 timer.

2) 16-pin chip is a microprocessor, seems to be an
STC15W408S (STC = chinese chip mfgr)
a member of the 8051 family
----

So, yep, IR 2153 is a good replacement.

I found there were newer versions of this chip. I'd have to dig into my notes to get the numbers. My thought was that at least one other PN should work but I never tried to prove that.

[Rex Allers]

Oops, I see you used IRS... vs original IR...

Good to know that worked. As I said, I found a few chips that seemed they might work. I'd have to go look at the datasheets again to remember what the differences were.

I'm no power supply design engineer but the function of these supplies seems unusual to me. If I got the circuit right, the voltage adjust pot is controlling the frequency of the 50/50 switching pulses. There is no PWM. If I remember right, lower frequency gives higher voltage output.

That's a key reason why I wanted to try replacing the sw driver chip with one that has direct inputs rather than the internal oscillator. That replaced chip could be driven by an external (off board), more typical, switcher chip with adjustable PWM and separately frequency. -- Another thing on my list of -- not done it yet.

[Alan Sailer]

Rex,

I don't have a complete explanation but I did work a bit with flybacks for an earlier project.
All the flybacks I messed with* had a resonance somewhere in the 10's of kilohertz. The
resonance was due to the parasitic capacitance resonating with the secondary inductance.
At that point the impedance of the transformer is at a minimum and large currents and
voltages are created.

The voltage at this resonace is at a maximum but it is not a good place to operate the transformer.
I killed two by doing this.

The area you want to operate is below the resonance in the area where the transformer is
acting as an inductor ( above resonance it acts capacitive). You can vary the output voltage by
changing frequency which moves you along the slope, changing the output voltage as you do.

I did one circuit that used PWM at a fixed frequency to change the voltage but the range was
not as high as the frequency shifting technique. Maybe 9-12kV.

As a final note none of this operation is true flyback operation (which I don't understand). It apparently
uses a sawtooth drive to ramp up the magnetic energy in the transformer core and then
abruptly release it, causing an output current spike. Which results in a large voltage spike V=L(di/dt).

When I was a kid my Dad showed me how to make a self-oscillating circuit using a relay. When the relay
contacts opened, causing a collapse of the current in the relay coil, you would get a painful shock if you touched
the coil.

Cheers.

* Vintage off Ebay and new from China.