TTi TSX1820 repair

[lukier]

I recently scored this power supply from a dumpster. Usually T&M stuff that I get from the dumpster from time to time doesn't need much repair, just replace a fuse, clean everything, maybe replace caps. Unfortunately, this power supply is in a very bad shape. It looks like it was standing in water, at least inch deep, for a long time.









I'm determined to repair it, as I don't have this kind of PSU in my lab, most of my lab power supplies go to 5A and that's it. I have HP server PSUs that provide 12V and a lot of current and they have overcurrent protection but it is fixed at 100A 

Here's the insides. It was even worse, this photo was taken after some air dusting (rust particles) and brushing. Analog control in the top left corner has a lot of oxidation (copper or tin). TO-220 tabs often have green copper patina as well.



My ultrasonic cleaner is nowhere big enough for the whole board, but the side that was in the worst condition (analog control & feedback) managed to fit there and got some cleaning with demineralized soapy water and isopropanol. The rest of the board was cleaned as well, but manually with a brush. I could probably replace some parts (ICs) that are in a really bad shape, but I'm not desperate enough to replace everything. Also, I have very little THT parts (even resistors), everything I have is mostly SMD.



This power supply seems like a solid design, no cutting costs on filtering and protection. I also like that it is oversized enough that it doesn't have a fan (of course it has DC-DC switching backend, but even switching PSUs often need fans). It is fairly old school design, no bells and whistles, but all the parts are pretty common and easy to get.

Here a quick power-up test without heatsinks:



Fortunately nothing blew up. Electronics seems to sort-of work. Yesterday I cleaned the heatsinks and put them back with new silpads, isolators and screws. I only have Agilent 6632B that I can use as a load for the moment and I managed to get 18V @ 5A without trouble. The only issue is that the digits on the display jump up and down, but not only during readback, even with the output off (programming) so I suspect it is just a fault on the control board - I didn't clean it yet, maybe dirty pots, intermittent switches etc. The voltage as measured with a multimeter is OK. I need to test output noise with poor man differential probe as well.

Couple of days ago I've send an email to TTi support asking for the service manual and I got it day later - very impressed! The manual itself is very well written, extremely detailed description of the circuit, theory of operation, detailed schematics. It's a pity other companies don't do service manuals like that anymore.

Manual was very helpful as I don't have isolation transformer or a differential probe (only RCD) and the manual describes clearly dangerous areas, how to measure them, which heatsink is connected to HVDC  etc. I should start saving some money for Pintek DP-25 anyway 

So now the worst bit - restoring the case. The only idea that I have for the moment is to get rotating brushes, steel brushes, a lot of sanding blocks and so on, spend hours scrubbing rust and then paint it with Hammerite, put new nutserts etc. Retrobright treatment for the front panel.

Do you have any ideas how to deal with such amounts of rust? Sometimes I think there might be more rust than metal left underneath 

[philpem]

Looking at that... I wonder if its current state is a bit, how should I put this, "British Leyland"?

As in, the rust is pretty much what's holding it all together?   

Props to you for repairing it - looks like it's in good nick electrically, but I wish you the best of luck with the mechanical refurb!

[david77]

Looking at that... I wonder if its current state is a bit, how should I put this, "British Leyland"?

  I love that.

I don't think the rust is so bad. I'd start to sand it down to get the loose bits of rust off, then apply rust converter. This stuff http://www.fertan.com/products/derusting-products/fertan-rust-converter/ works very well.
After it's done its thing paint the case in any colour you like, personally I'd probably use Hammerite.

edit:
Thinking about it I reckon I wouldn't bother with rust converter as it's unlikely to get wet again.

[macboy]

The display board is not part of the control of the regulator. Even fault conditions are only inputs to it. The display of voltages and currents are done via a single ADC with a pair of CMOS analog switches as an input mux for its differential input. The PIC micro controls the MUX and reads the ADC, displaying results on the displays. My first step in debugging this would be to measure the voltage at the input to the ADC. Measure also the voltage reference input to the ADC. The Vref will be around 190 mV. The full scale ADC input voltage is then 190 mV = 4096 counts, so at 18 V (or 18 A), the input will be only around 84 mV. Both the Vref and input are is very stable but the display is not, then the ADC is faulty. If the input is not stable, then measure again at the inputs to the MUXes. If all the displays modes are unstable then that is about as far back as the fault will be since nothing beyond the MUXes is common (except ground... check for a good noise-free ground relative to the V+ output).

[lukier]

Thinking about it I reckon I wouldn't bother with rust converter as it's unlikely to get wet again.

I didn't know about these rust converters, thanks! If it'll save my a lot of scrubbing and sanding that might be still a sensible option.

The display board is not part of the control of the regulator. Even fault conditions are only inputs to it. The display of voltages and currents are done via a single ADC with a pair of CMOS analog switches as an input mux for its differential input. The PIC micro controls the MUX and reads the ADC, displaying results on the displays. My first step in debugging this would be to measure the voltage at the input to the ADC. Measure also the voltage reference input to the ADC. The Vref will be around 190 mV. The full scale ADC input voltage is then 190 mV = 4096 counts, so at 18 V (or 18 A), the input will be only around 84 mV. Both the Vref and input are is very stable but the display is not, then the ADC is faulty. If the input is not stable, then measure again at the inputs to the MUXes. If all the displays modes are unstable then that is about as far back as the fault will be since nothing beyond the MUXes is common (except ground... check for a good noise-free ground relative to the V+ output).

Yup, the manual says it as well (especially about this strange Vref value). Displays are fluctuating even with the output off, when just programming voltage/current so it might be just some problem around the MCU, maybe as simple as dirty pots and switches.

What's important that the main PCB with the switching and linear converters and regulation circuitry seems ok and that was the part I was most worried about.

[SparkyBruce]

If you ask TTi nicely they might be able to quote you for a new case for it.

They have always been great help when it comes to service manuals and general support for even their older units.

I am lucky in that I drive past their place twice each day.

Bruce

[lukier]

If you ask TTi nicely they might be able to quote you for a new case for it.

They have always been great help when it comes to service manuals and general support for even their older units.

That's a good idea, I might try that route but I'm worried what the quote will be. TTi stuff is quite pricey. I don't think I would buy this PSU new even if I wasn't cash-strapped. Good supply, no rocket science inside, but price/performance ratio not that great. I'd probably buy Digimess SM5020 or something like that.