Spellman DXM70N600X3547 power supplies on ebay -70kV 8.56MA

[Peter Schmelcher]

Andrew I powered up the supply using the 24V fan idea and my supply works as expected. Low voltage test points are fine plus ethernet is good. The hobbs meter reports 225.9 hours HV On.

I think everyone should inspecting C149 33uF 450V for any signs of venting electrolyte before applying 120VAC!

The capacitor can top is a mechanical fuse that is designed to tear open under failure conditions. The top of mine bulged slightly and leaked. C149 might be the root cause of my damaged unit.

-Peter

[Dan Knapp]

Pin 15 is not available as a spare on the custom version of the supply. See Andrew's Nov 22 post. Pin 15 is an additional interlock input. I need to post a revision of my analog controller circuit to reflect this. It also appears that the other (other than voltage control) inputs cannot be left floating as in my circuit diagram. My DXM is not yet producing HV, but I seem to be making some progress.

[Jim Kovalchick]

Sarvesh,
You need to jumper the interlock pins to proceed. See Andrew's posts and the manual. Nice job getting the applet to run.

Rich Feldman,
You got to the same point I did with getting the applet to start and stall. I did the network settings the same way. I think the only way the applet will run is on an old version of JAVA.

Yesterday, I tried to use an app called Packet Sender to send commands as TCP packets to the unit. As I suspected, I really struggled with the syntax of the strings. You need to know a good portal number, and it seems that Andrew is correct that 50001 will work because I can get return strings back from that portal. My trouble is getting coherent instructions to and from the unit. The app is nice, but it want''s to translate ASCII strings into two digit hex and I'm sure that I am failing miserably at this.

After the holiday, I may try to teach myself enough Visual Basic to try that approach. ...and Kuba, no fair saying that it is the programming language of your grandfather. I don't know how old you are, but probably back before you were born I was entering machine code into an Apple one hex digit at a time because I was too poor to buy an assembler. Visual Basic is seems like a dream compared to that.

[Rich Feldman]

A terminated ribbon cable and PC board header are in hand for my analog control "fixture". Thanks for the reminder about pin 15.

>> Pin 15 is an additional interlock input. I need to post a revision of my analog controller circuit to reflect this.

Thanks for drawing that up, Dan. While you're editing, please consider:

1. Change -75 kV to -70 kV in two places, to match the nominal voltage of DXM70N power supplies instead of their output connectors.

2. Change DB-15 to DA-15 in three places. The letter B was mistakenly carried over from DB-25 in the real RS-232 standard. (Both are D-sub connectors.)

dabcde.JPG (25.31 KiB) Viewed 14473 times

3. Add the Zo values (5k,5k,10k) of the three analog monitor pins. That's important for controllers that have analog or digital panel meters, easily calibrated to indicate kV and mA.

[Richard Hull]

Thanks Andrew for the full testing and the full, realized output of a working one of these units. It will be great to see someone use one of these to do fusion. I am a little concerned about the ultra low current capability of the device and whether there is an "auto kill" at exactly 8ma. This was frustrating to the VCU nuclear engineering folks who spent $$$$ on a brand new rack mount 30kv 10ma spellman supply. The moment they started to do fusion, the supply tripped offline. This forced a frustrating reset and re-approach to the fusion threshold. They have now spent more $$$$ for yet another beefier supply.

Super electronic supplies by the best manufacturers seem to auto protect as they hit the honestly rated limit on current and can catch and trip on 15ma bumps and burps in the gas conduction regime. This is one of their most nasty faults and crippling limitations for the amateur fusioneer. I guess we will wait and see.

The one good thing about a home brew HV supply that is metered very well is that there is not current trip point, be the supply a home made switcher or a classic hard-nosed linear type. The human being is in total control and not some engineered safety trip point working in the micro-second or millisecond time frame to shut the entire thing down. The homebrew will take the burps and bumps and the end user will never even see the metering jiggle.

My statement in many FAQs about one absolutely having to have 10ma of current in any HV supply before lower level, detectable fusion begins was a full running, stable, 10ma. Excursions and burps and bumps of 100ma or even more are possible as the user learns to apply the voltage needed to do fusion. The gas conduction regime rules the roost and a sudden strike of the glow at 28kv in a good system can easily draw a multi-millisecond 250ma bump! Your supply can't see this and trip out or you will never get there.

Sound defeatist? It's not me, it is the physics of what we are doing.

Richard Hull

[Andrew Seltzman]

I've compiled the matlab m files into a stand alone program that ramps the supply up to 10kV, it should start ramping up the voltage when executed.

DXM_test.zip

(3.16 MiB) Downloaded 569 times

You will need the matlab runtime(free) on your computer to run the program
Get the Windows 64-bit version of the MATLAB Runtime for R2016a
https://www.mathworks.com/products/compiler/mcr/

[Rich Feldman]

Attempted AC power-up was not so successful, but was a good learning experience.

Fuse board silkscreen says F1 and F2, both 10 A 250 V. My PS arrived with F2 blown. The fuse was ceramic, so not easy to judge the category of overcurrent that had made it operate. I used glass replacements.

Special tool 1: a nine-inch extension cord with male and female tab terminals. Allows power-up with full visibility of main board and fuse board. A continuity check, all the way from wallplug end of the IEC cord, showed line and neutral reversed. Spellman would never do that, would they? Problem went away when I substituted a different IEC cord. First cord had been miswired at the factory, before final molding of the connectors.

With new 5 amp fuse at F2, and main board connected, DC resistance measured between wallplug blades was more than 10 k ohms each way. So no DC short circuit. Plug it in. F2 blows instantly, with small noise and light, leaving a couple of fused metal balls visible. I think that indicates a fault current at least a few times greater than nominal fuse value, but probably not 100 A, and certainly nowhere near a metallic short circuit current.

With new 10 amp fuse at F2, plugging in made F1 (as received) blow instantly. It's a BUSS GLH 10, with zigzag-shaped link. Two adjacent narrow points had opened, freeing one whole zig to rattle around in the tube. F2 appears to be undamaged. Did this just demonstrate the difference between fast-blow and very-fast-blow models?

Special tool 2, made decades ago: my "ballasting outlet". A regular duplex grounded receptacle in a regular metal box, with a 3-wire cord and plug for use as an extension cord. Except the two outlet sections are no longer wired in parallel, they are in series. Front plate is marked to show which wide slot is Neutral, which narrow slot is Hot, and which left and right slots are connected to each other. I tested this by plugging in two 300-W halogen worklights & getting dim light from both.

When I powered up the DXM70 with two new 10A fuses, and one 300-W lamp in series, the lamp came on & stayed on at almost full brightness. I forgot to measure the voltage drop across the DXM. Am suspecting a semiconductor problem, though as mentioned before, nothing looks or smells burned.

If the circuit board assembly isn't repairable, I would keep the unit just for the bottom half. Probably no active components down there, just transformers & capacitors and diodes. Connections from the upper half are a bundle of 4 similar black wires, a bundle of 2 similar black wires, and a cable of 5 colored wires to J86 "FIL FDBK". Would building a driver from scratch be much harder than building a turbopump motor driver from scratch, as reported here many times?

[Andrew Seltzman]

Rich,

As far as possible places for shorts in the supply, check:

CR35 (the rectifier)
Q6 (STW30NM60D mosfet in the PFC system)
Q10 (STW30NM60D mosfet in the PFC system)
Q7 (IRFBG30 mosfet for a flyback converter that supplies power to the digital logic)
C161 (big blue electrolytic)
C160 (big blue electrolytic)
C149 (smaller electrolytic near the 2 big blue caps)

Notes:
Q6 and Q10 are in parallel
C161 and C160 are in parallel(the big blue caps)

[Dan Knapp]

Thanks to Rich for picking up the errors on my previously posted circuit diagram for an analog control for the DXM. I've corrected the errors on the attached revised diagram as well as added the input and output impedances as suggested. In this revision I've added jumpers so as not have any of the control inputs floating. I have not actually built this yet pending a decision on whether my DXM is working OK and thus whether to keep or return it. At this point I'm still playing with a 9 volt battery/pot divider for the 0-10V inputs.

[Dan Knapp]

There is a cable with a CA11 connector listed on eBay:
http://www.ebay.com/itm/Claymount-13025 ... SwXeJYL0sI
The seller is asking $400. It has a CA1 on the other end if anyone needs both. The price was lowered from $500, but it is still more than would interest me.

[Andrew Seltzman]

I finished the HV cable for the supply

and have successfully tested with plasma and a neutron output

At 45kV, 8.56ma about 1.5e6 n/s

At 50kV, 8.56ma about 1.8e6 n/s

Some notes about operating the supply on a fusor:
The supply does have an arc fault protection system that is fairly sensitive, when triggered it will momentarily reduce output voltage and then bring it back up to the set point. After a certain number of arcs in a given time window(about 5 arcs in a few sec) it will shut off the power supply. This does not present a problem with an occasional grid arc on a well conditioned grid, but is an issue with charge buildup on the surface of an unshielded HV cable(like the silicon one I have built). As the charge builds up on the outer surface of the insulating cable it will be attracted to a conductor and discharge, causing the arc fault to reduce voltage, this will cause the cable to pull away from the conductor and then attract once HV is re-applies causing another discharge. This repeated mechanical oscillation of the cable will trip the arc fault circuit.

Arc fault trip due to repeated static discharges from cable

The HV cable will need a shield to prevent this effect, and possibly a semiconductive anti-corona tape layer between the silicon cable and the shield before higher voltages are attempted. Other then this effect, the supply operates very stably with a fusor and should help most people significantly increase their neutron output.

[Dan Knapp]

Has anyone been able to figure out the login and password for the Spellman interface at 192.168.1.4?

[Dan Knapp]

To answer my own question, after overlooking it multiple times, I found the login and password on p. 15 of the DXM Digital Interface Manual: username: admin, password: SHV_Applet.
These worked on my DXM using Internet Explorer on an old laptop running XP. I was able to access the Web Monitor and Control Applet and read the hours meter. My unit reads 173.1 hours.
With a high voltage probe inserted into the CA11 receptacle and the enable pins 11 and 12, and 12 and 15 on J2 connected, I can step up the voltage using the web interface. Starting at 1000V, I get the right voltage on the output and can step up to 1400 volts in 100 volt increments with the correct output voltage. The current setting was 0 ma; with the current set to 1 ma, I get an "under current" fault both on the web interface and the front panel LED. When I try anything over 1400 volts, I get an "under voltage" fault both on the web interface and on the front panel LED.
Trying Andrew's standalone Matlab program, it starts scanning and fails the same way. I've not yet been able to get Andrew's DXM Control2 program to run under Matlab, possibly because I'm running a 2011 versionof Matlab whereas Andrew is using a 2016 version of Matlab.
I am unable to get any output using the analog control inputs via J2. This is perhaps a clue to where the fault lies.
At this point, my DXM is close enough to running that I think I'll keep it and continue to play with it.

[Sarvesh Sadana]

After connecting pins 11 and 12 on the J2 connector, the interlock closed and I was able to enable the high voltage.

Despite that, I still got a message stating that my DXM is incorrectly configured, and I was not able to set the voltage or current. What am I doing wrong here?

Also, has anyone made any further progress regarding the group purchase of the HV connectors and cables? I'd much rather buy one than try my hand at making it.

[Dan Knapp]

Sarvesh
Have you tried connecting pins 12 and 15? Pin 15 is a second interlock on this version of the DXM (see Andrew Seltzmann's post).
Have you logged in to the web interface? Login: admin, Password: SHV_Applet
Are you getting the "x-ray on" LED on the DXM?
Your post gave me a clue to my problem. You are showing a reading on the "LVPS-15V" where mine is showing zero. I'll try to check the 15V power supply.
Keep sharing; between the various owners of the two dozen supplies maybe we can turn up enough clues to make them all work.

[Andrew Seltzman]

As preventative maintenance, I replaced the C149 electrolytic capacitor on both supplies. One had started to bulge and the other has leaked a little out the top. The replacement was(thanks to Peter Schmelcher for finding a better version of the cap):
http://www.digikey.com/product-detail/e ... ND/1246008

During the upgrade I also noticed that come of the 450V capacitor pins on the bottom of the board were beginning to rub through the plastic insulating sheet. If these do puncture the sheet they may short with the grounded insulating frame.

These areas were reinforced with kapton insulating tape, and the pins were trimmed a little closer to the board, and the solder balls rounded a little more. Everyone with a short in their supply should check these. In addition I added some heatsink compound to the rectifier to make better thermal contact with the frame.

Additional pictures of the bottom of the board:

After the modifications both supplies were tested with the dummy plug to 70kV without any problem.

[Richard Hull]

Andrew nailed it and proved the supply can work in a good fusor provided the arc fault cutoff is properly handled. He climbed higher on the cross sectional curve via voltage. This curve is rather steep up to a little over 100kv, (get more bang - fusion- for increased voltage). Few are the folks who have drifted over the 40kv applied mark. I can get to 1.5e6 n/s sec, but I am at the limit of both my X-ray supply and the fusor insulator at about 43 kv where I hit that high neutron mark. There is negligble boost in fusion in going to 200kv applied as the curve flattens quickly

Andrew, can you report on the x-radiation at those higher voltages? The shell starts to go transparent at about 35kv applied and at 41kv, I am getting about 100mrem/hr at the glass view port, but at the shell, up close, the fusor only puts out about 20mrem/hr At 6-7 feet from the shell it is down to 1or 2mrem/hr and during a normal run will not effctively move my 100mr pen dosimeter off zero at that range.

I would imagine at 50kv applied the shell, acting as a filter shoots out only the most penetrating x-rays which are only a small portion of the x-radiation produced by the fusor, (and trapped in it). The x-ray energy at 50 kev applied only has a tiny fraction at that energy level.

To others here with these wonderments, Andrew's setup can handle the higher end voltages..... Can yours? Most fusioneers fall a bit short in their insulators. Of course, you need not run this supply full out, voltage wise. Most will discover their "rope limit." I know mine and it is about 43kv applied.

Richard Hull

[Andrew Seltzman]

I modified my cable with some resistive anti-corona tape salvaged from another HV cable and some braided shield, this improved the tolerance to arc faults, but it still is a problem over 50kV.

X-ray rates were measured ~1ft from the fusor, my system has a 1/8" thick stainless shell, the view port has a 1/4" thick lead glass shield.
40kv <1mrem/h
45kv 5mrem/h 1e6n/s
50kv 15mrem/h 1.2e6n/s
55kv 50mrem/h 1.8e6n/s
Brief operation at 60kV possible before the arc fault system tripped, but not enough to to get a measurement. These were brief runs with minimal prior conditioning and no optimization of pressure / ion source tuning to get maximum neutron output. The runs were at 8.56mA current.

[Richard Hull]

Thanks Andrew, That is good data and shows how the shell gets real transparent real quick with increasing voltages. As I have noted many times over the years. You do not have to shield for neutrons. Shield for X-rays!....And then only if you are wandering over the 35kv mark in an SS chamber of normal .0625 inch thickness. Andrew had double that thickness which is rather unusually thick for a fusor and his x-ray really shined through over 50 kv applied.

Note: normal, inexpensive, spun SS hemispheres are all .0625 inch thick. .125 inch is rather rare

Richard Hull

[Andrew Seltzman]

Does anyone know a distributor for an impregnated polyester anti-corona tape like this:
http://www.dexlu.com/woven-tape/anti-corona-tape.html
or something like it?

3m makes semi-conducting tape for field grading on the ends on HV conductors:
http://www.ebay.com/itm/NEW-Scotch-3M-E ... SwhcJWO5Br

But I can't seem to locate a distributor for resistive tape.

[Rich Feldman]

I found C149 and CR34 both shorted. With those parts out, the DXM70 board DC rails are no longer shorted. Haven't yet installed replacements. For use exclusively on nominal 120 VAC mains, does anyone think the boards need or benefit from 450 volt rated capacitors?

The offending capacitor from my board doesn't appear to have bulged or leaked. Smoking gun was toasted FR-4 around the diode attachment holes. Sometimes that happens on working designs that have poor cooling or insufficient component derating.

That's a little snip from a full-board pdf file off a flatbed scanner, here in case anyone wants to carry on with reverse engineering.

dxm70_bot.pdf

(112.54 KiB) Downloaded 509 times

I experimented with Visio as a tool for reverse engineering. Took the flatbed scan and overlaid circles and rectangles, keyed by color, on various layers. Looks promising, and I already learned some practical details of half-bridge drivers etc.

[Andrew Seltzman]

The supply uses a PFC/boost converter system which increases the voltage to a few hundred volts, so the 450V cap is required.

[Andrew Seltzman]

Here are 2 good references:

Spellman power supply reference
http://www.spellmanhv.com/-/media/Files ... manual.pdf

How to terminate cables for low corona/leakage
http://www.spellmanhv.com/Technical-Res ... age-Cables

[Rich Feldman]

Those are great references, Andrew. Thanks for the pointers.

Speaking of group buys, how about some repair kits with the nickel-and-dime components that many of us found dead? If you have to order an individual item by mail, you can buy ten for about the same total money (after shipping charge).

As Peter said a while back, "The supply is conservatively laid out and the RTV is a good indication of vibration testing for harsh industrial environments. Honestly it would be a pleasure to work on if I had a schematic. So it’s just finding the hours to do a methodical investigation/repair with a budget of $100 for parts."

I don't think Cliff would mind if we continue to share notes about the bill of materials, and sourcing of replacement parts. Here's a piece of a schematic learned by inspection, as others here have done. So far un-reviewed.

Who wants to draw the power path from AC input through PFC circuit (Q10 and Q6), so we can see if it matches my scribbles? Can't see the label on my RV2 without removing the part, which measures about 24 ohms at room temperature. Can't see the label (or even the ref des) of rectangular pink capacitor next to the main bridge rectifier.

Almost forgot to mention, I made a DC resistance model of the whole interface coming up from the DXM70N's basement level. If the fancy inverter and control board were totally dead, or missing, why not use the potted "basement unit" to make a dumb HV supply, substantially derated of course?

[Liam David]

I replaced the blown IRFBG30 mosfet on my power supply board, but it still does not function. I checked all of the capacitors, and C149 has a slight bulge but has no continuity. Otherwise none of the components on the board appear damaged. Removing the fan revealed that there is only about 2V on the bus, which I assume should be 24V. Any ideas what could be causing this problem? Some kind of bad voltage regulator? All six red warning lights on the front panel light up and blink repeatedly, so some power is going somewhere.
The other supply functions properly and I was able to turn the voltage up to 28kV. Still working on making a CA11 connector.