GW Instek PSS-3203 Programmable Power Supply

[wblock]

Just finished repairing this used unit, and because they seem somewhat rare, I took some pictures.

This is a 32V, 3 amp DC linear CV/CC single output power supply.  An encoder allows adjustment in increments of .01V and 1 mA.  The programmability consists of the ability to accept commands via RS232 or optional GPIB interfaces.  The one shown here appears to be from around 2007, although they are still sold today and still command high prices.

This unit was used, with a display problem.  The display was dim, but also had a contrast problem.  It is a standard 1602 LCD with an LED backlight.  In this case, the backlight is actually two SMD LEDs on a tiny board.  I removed the backlight from under the LCD panel and removed the tiny backlight board.  It is shown in the photo below, with a Molex connector above it for scale.  When probing those LEDs, one lit, and one disintegrated into layers.  I found a white SMD LED in the scrap parts pile, soldered it onto .1" headers, and glued it into the backlight with hot glue.  A standard 1602 would easily replace this, but the only normal one I had available was a low-profile version that was way too far back in the case.  After repairing the backlight, I added a resistor to get the default contrast value into the visible range.  The contrast can be adjusted from the front panel, but for some reason the value is not saved.  The voltage and current settings are saved.

GW Instek likes to use Jamicon capacitors, which are reasonably good.  While not the level of my favorite Panasonics, I can't recall ever finding a failed Jamicon.

Unfortunately, Instek also likes to use brown glue to secure large components.  This glue degrades with heat, becoming darker brown (no problem) and conductive (big problem).  The larger capacitors had been secured with this glue, and it had begun to darken.  This is visible on the power board photo.

In this case, the glue could have been left, since it was not touching any conductors.  But it was not possible to see that.  I ended up removing those capacitors, carefully removing the glue (just a fingernail was usually enough to scrape it loose), resoldering the caps, and then regluing them with neutral-cure silicone (not shown in these photos).  Funnily enough, the brown glue had not actually stuck to several of the caps.

Note the internal heatsink with the fan sucking air through the cooling fins and the supercap on the processor board.

[wraper]

The contrast can be adjusted from the front panel, but for some reason the value is not saved.  The voltage and current settings are saved.

Might be EEPROM with faulty memory cell.

This glue degrades with heat, becoming darker brown (no problem) and conductive (big problem).

I seriously doubt there is only one type/vendor of this glue and that all of them become conductive.

[wblock]

The contrast can be adjusted from the front panel, but for some reason the value is not saved.  The voltage and current settings are saved.

Might be EEPROM with faulty memory cell.

I figured it was some sort of volatile memory powered by the supercap.

This glue degrades with heat, becoming darker brown (no problem) and conductive (big problem).

I seriously doubt there is only one type/vendor of this glue and that all of them become conductive.

That could very well be, and then I just would not have seen those boards because they didn't fail.  There seemed to be less use of that glue over time, and it looked like it had gone away.  But there it is again, on the power supply of the new Instek GDS-1054B.  Unless it gets really hot, even the bad variety will probably outlast a 1- or 3-year warranty.

[wblock]

The fan is a temperature-controlled 24V sleeve bearing fan.  The original sounded a little sick.  Not grinding, just making a bit of ticking which added oil did not fix.  So I ordered a ball-bearing replacement, swapped it in, and it sounds really rattly.  Could be air/blade interference with the holes in the case.

Other than the fan and occasional clicking of relays as settings are changed, the supply is very quiet.

[wraper]

I figured it was some sort of volatile memory powered by the supercap.

Isn't there just some kind of clock built in? Storing cal data in the RAM powered from a supercap would not be a smart move. Leave it for a few months unpowered and cal data most likely will be gone.

[wblock]

I figured it was some sort of volatile memory powered by the supercap.

Isn't there just some kind of clock built in? Storing cal data in the RAM powered from a supercap would not be a smart move. Leave it for a few months unpowered and cal data most likely will be gone.

No clock available from the front panel, and I did not see anything inside that looked like one.  My guess is they use EEPROM for calibration data, but supercap-backed memory for user settings and persistent things that change all the time and would wear out something with limited write cycles.  If it forgets the last voltage and current settings, no big deal.  It comes up with the output turned off anyway.

...Just looked at the processor board again.  That QFP is the processor, a Winbond 78C032 ("ROMless operation") 8-bit micro.  Right next to it is an Atmel 93C66 4Kbit EEPROM.  Next to the supercap is a 62LV256 256Kbit static RAM.  This is in addition to the big ROM with the stick-on label.

While looking, I also saw an AD7541, a couple of TL071s, and a TL074.  Most of the rest is 74HC, with one or two 4000-series.  The supercap is 0.47F at 5.5V, says "Made in Japan", and has the Matsushita "M".  There is also an AS431 2.5V voltage reference diode in a TO-92 package, and six 817D optoisolators.

[wblock]

The noise from the fan is because it is only getting 6V at idle, not enough to start spinning.  Here is a better picture of the processor board, removed to trace the fan power:

[wraper]

Right next to it is an Atmel 93C66 4Kbit EEPROM

I would read it and save backup file. Then program a new IC and put it in the PSU and see if the contrast setting is saving.
There is nothing in the manual what would suggest that contrast setting is not permanent. BTW is buzzer setting saving?

5-3. The Display Contrast Setting
Set to Contrast Set window by pressing [MENU], modify its setting
value with the wheel knob and press [ENTER].
5-4. The Buzzer Setting
Set to Buzzer Set window by pressing [MENU], turn on or off the
buzzer with the wheel knob, then press [ENTER].

[wblock]

Right next to it is an Atmel 93C66 4Kbit EEPROM

[
I would read it and save backup file. Then program a new IC and put it in the PSU and see if the contrast setting is saving.

Well, a 1.1K resistor got it to the right place (see photo, taken earlier but forgot to include), and this particular display does not seem to have a wide contrast range anyway.  And it would require removing the SOIC chip and finding or making a programmer for it.  For right now, not a big problem, but longer-term, an interesting question to answer.

There is nothing in the manual what would suggest that contrast setting is not permanent.

Agreed.

BTW is buzzer setting saving?

The unit is not reassembled yet, and I only messed with that a bit, but yes, I think it was.

[wblock]

The noise from the fan is because it is only getting 6V at idle, not enough to start spinning.  Here is a better picture of the processor board, removed to trace the fan power:

Following up: actually, the fan does get enough power to spin.  The fan noise is due to PWM.  For future reference, the fan is connected through a 220 ohm power resistor to +18.9V.  Q202 is the transistor connecting the fan to ground.  Pin 3 of the optoisolator U206 is connected to pin 3 (base) of Q202 through a 4.7K resistor.  The base of Q202 is also tied to ground with a 3M resistor.

As per Suppressing Acoustic Noise in PWM Fan Speed Control Systems, adding a capacitor to ground between the 4.7K resistor and the base of Q202 can reduce that noise.  However, the amount of capacitance is tricky.  That app note suggests between .47uF and 1.0uF, but values that small do not make a noticeable difference.

10uF sounds better but does not eliminate the noise.  Too much capacitance and the fan just runs at full speed.