Teardown : Keithley 2304A High-speed PSU (service and pimping up)

[TiN]

Well, time to get first PSU in the lab. Yes, that's right, I manage without PSU so far all this time..

As being KI's cheapscate fanboy, which PSU shall it be? 2304A for 99$ will do!



Sold as is, for parts or not working.. Not scared,  let's see what mice may hide inside?



Hmm, no visible damage...



Digital board looks dirty but ok. Datecode on chips are 2002-2003. ADC is well-known ASIC from K2000.



Front panel board and LCD... I'm feeling urge to upgrade it to either OLED or VFD. What shall I do?



And...power applied, it works! Measured with 34970A, output accurate and matching LCD reading down to 1mV.

So, feel free to tell me what you want to see happening to it.

I'll have to buy resistors for it to calibrate.

As always, article on the my site will be permanently hosted.

[picburner]

This time you didn't have to work hard to fix it!
I used EZGPIB to calibrate my 2303 with a 4 ohm and 3Kohm resistors using a K2001.
Check that there are no errors in the program because I used it a long time ago and do not remember
if I made then last minute changes.
For the rear connector you can use the Farnell 1792860 cutting 2 clips, it costs much less than the original.

[TiN]

Got OLED 16x2 today...

Well, utter fail. Did not work well, top line is scrolling randomly and ain't working right. 

[Short Circuit]

Looks like it has been repaired before. The long arrays of ceramics are originally SMDs.
pictures here: https://www.eevblog.com/forum/reviews/keithley-2304a-high-speed-power-supply/msg91147/#msg91147

[nidlaX]

God, why did Keithley have to go with these puny ass 2 line displays? I think the Agilent 66900 series and the R&S NGMO2 are much better designed. The whole front panel on these supplies is just terrible for usability.

[Kleinstein]

This supply is supposed to be used with computer control, not by manual setting. So the primitive LCD and the keys - just as a minimal backup. The LCD is not that bad - not problem running this 24/7 without getting dim like VFD. Also lower noise (but does not matter in this case).

The baord looks a little different from the one of Short Circuit - the area near the output connector is different, especially what looks like a bodged in Rectifier/diode (protection for a shunt ?).

[nidlaX]

This supply is supposed to be used with computer control, not by manual setting. So the primitive LCD and the keys - just as a minimal backup.

That is basically true for all the high speed DC measurement supplies, but R&S and HP/Agilent still chose to make their products more usable. I think it's a shitty cost-cutting measure on the part of Keithley. They had better front panels they could utilize for the job.

[TiN]

Installed crappy negative LCD for time being...



More teardown pics before recapping and assembly PSU back.











And first use case:

I had phone laying on the shelf for few weeks without charger connected, and it died, as battery got depleted and charger could not start starge.
Plugging to mains charger did not do anything, phone still had no reaction to power button. Solution - take pesky thing apart and give a kick to a battery.



After 5 min of charge phone was assembled back and charged well using native charger. It's working OK now.

[TheSteve]

Hmm, I think the original LCD looks better then the crappy negative one.

[Kleinstein]

One was to careful with charging LiPo batteries that got to empty: they may develop an internal short or damage the separator if discharged to far - don't know the exact limit, but something like less than 2 V is critical. They may seem to work after charging, but there is a certain risk the catch fire after that - not only during charging, but also later on.

So there is a good reason good chargers refuse Li packs that got to low.

[TiN]

Hmm, I think the original LCD looks better then the crappy negative one.

It's actually better in real life. Excuse my photo skills.
Old one had horrid viewing angles.

Kleinstein

I'm keeping eye on it.

[TheSteve]

The OLED looked beautiful of course, wonder why it didn't work. Generally you can swap the 16/2 displays around all day long and they work fine.

[TiN]

Yah, that what I expected as well. Seems timing issue, but I don't plan to waste more time on this for now.

[TiN]

Pre-calibration performance test results.



Not bad, I'd say, considering fact that PSU is ~13 year old. 

Test setup:

PSU source and sense terminals connected at HP 3458A HI/LO posts.
NPLC100, ramp from 0V to 20V in 0.1V steps, each point is averaged from 3 readings.

[ManateeMafia]

Looks good.

Is there enough space to mount binding posts through the front cover?

[TiN]

Plenty. But I decided not to do so in my unit, terminal block is more suitable, as I can change loads with one quick snap instead of fiddling with posts. And since it will likely be used with remote control only, I'll have it with back face on the shelf

Only need to deal with fan speed/noise but for than need to do some thermal testing first.

[ManateeMafia]

Good to know. Maybe it would be a good project for another 2304?

Reading the user manual, there is a mention of a relay control port on the back of the 2304A. I have found photos both with and without them. Any idea if this was an option or if it was added later?

I cannot tell if there was space on your boards for an unfitted option.

[TiN]

Only need to find all that time somehow...

I saw commands yesterday in gpib during python bodge. There is empty IDC connector near CPU on bottom board, maybe its for that.

Btw, psu is reporting full 6.5 digit resolution over GPIB, which is no surprise given use of K2000 ADC. Could be handy as limited DMM (with single non-isolated 20V range).

[amc184]

Great teardown, thanks for posting it.

I recently bought one of these myself.  It was just a chance purchase, happened to search Keithley on eBay, and couldn't resist it for USD49.  Mine was advertised as for parts or repair.  I took it apart on receiving it, found no line fuse and a disconnected display cable.  With them sorted it works fine, both as a supply and as a load.

I was looking at the differences between our two units, and I noticed that yours seems to be a hybrid of two units.  The base board in mine is a revision E, the top board is a revision F, both built in '99.  Yours has a revision H base board from '03 and revision E from '97.  I only really see differences in the top board, slight layout changes as well as a board pattern for the diode bridge and diode that in yours is soldered to the resistor near the output block.

Mine also has the relay control option fitted, it connects to J1015, near the processor on the base board.  I can't see any hardware differences between mine and yours here, so unless it's software disabled you should be able to add it.  The cable for is made with this pinout:

J1015	1/4" Jack
1	Tip
2	Ring
NC	Sleeve

The sleeve is connected to chassis earth directly by the uninsulated socket.

[TiN]

Great, good info. Seems my guess was correct then about that port. Perhaps you can make few photos of differences? Can be handy to someone. Maybe I can even include this into my article?

[amc184]

Sure, next time I've got my camera out I'll take a few pictures.

[TiN]

One of calibration steps / performance test setup:

.

[TiN]

This PSU is a power hog
Judge youself:

Power mains : AC source Chroma 61504
DC load : Chroma 6314A+63106A

33W when output off, (0.98PF, 110VAC, 60Hz, 0.3A). No wonder it need that noisy fan.
60W = 20V 1A
81W = 20V 2A
108W = 20V 3A
136W = 20V 4A
166W = 20V 5A

[bdivi]

Similar with the HP 663x series - in order to achieve the fast reaction the output is two quadrant and can both source and sink current. Idle consumption is in the order of 50W.

[amc184]

Here's some photos of my 2304A.  Upper board:









Lower board: