(FIXED) Keithley 225 (current source) repair

[motocoder]

I picked up a Keithley 225 off of eBay. I probably paid too much for it given it's in need of repair, but I thought it might turn out to be one of those units where the seller just didn't know how to operate it.

There are at least two things I've found wrong with it so far. The output was pegged on -130V no matter what I set the controls to. In addition, the 100 ohm precision resistor on the main current range switch is toast.

Both output driver transistors have failed with collector shorted to emitter. I guess this is why the output was stuck on -130V. So first question I have is where to source a suitable replacement. The part seems to be a TO-66 package, which really limits the selection. I've attached a picture of it. I believe the part number is RCA 60761. This does not seem to match the number in the Keithley manual, so either they swapped parts in later generations of the product, or someone has already had a go at replacing this part.  The manual says Mfr. Code = 80164, Mfr. Desig. = 40318, Keithley Part No. = 24560A. It also has an asterisk next to it which leads to a comment about it being a specially selected transistor (selected for good hFE?).

So my first question is how to go about selecting a suitable replacement transistor. I did a search on "RCA 60761", and a page came up listing NTE124 as a replacement. I have a distinct mistrust of NTE, as they seem to sell whatever crap fell off the boat, with minimal data sheet info and no guarantee of consistency or quality.

The manual for this beast is here:

http://exodus.poly.edu/~kurt/manuals/manuals/Keithley/KEI%20225%20Instruction.pdf

[johansen]

they are likely matched hfe npn/pnp pairs

replacement should not require an exact match, but ymmv.

you've got the schematic.. what is there to complain about? 

[motocoder]

they are likely matched hfe npn/pnp pairs

replacement should not require an exact match, but ymmv.

you've got the schematic.. what is there to complain about? 

They are both NPN, and the manual says "specially selected, order individually". There is another part where it says to order in pairs. So my guess is they are just selected for good hFE.

No complaining here. I am actually quite happy with the condition of this thing. If I can get it working, it will be a beautiful addition to the lab.

My local electronics store stocks the NTE part, so will just give those a try.

[Grapsus]

The part you are looking for is "RCA 40318". I can't find the RCA datasheet, but there is one by "New Jersey Semiconductors":

http://download.njsemi.com/40318.pdf

All the characteristics seem to match the RCA part which is described here:

http://alltransistors.com/transistor.php?transistor=20627

There is even a french site selling the exact RCA part:

http://www.donberg.fr/descript/4/40318.htm

[motocoder]

The part you are looking for is "RCA 40318". I can't find the RCA datasheet, but there is one by "New Jersey Semiconductors":

http://download.njsemi.com/40318.pdf

All the characteristics seem to match the RCA part which is described here:

http://alltransistors.com/transistor.php?transistor=20627

There is even a french site selling the exact RCA part:

http://www.donberg.fr/descript/4/40318.htm

Thanks, Graspus. I just noticed a bunch of toasted resistors on the range switch. I also see signs of someone that tried to make some repairs. However, in at least one case they used the wrong component value, so the whole thing is suspect and will need to be checked...

[motocoder]

All the non-precision resistors on the range switch (R102, R103, R107, and R108) are toast, as well as precision resistor R185. Those are all easy to replace. However, the pot on the end of the switch assembly (R138) is also damaged. It has a mechanical coupling to the shaft connected to the center dial. No idea where to get a replacement that will fit properly.

The pot is from CTS corp, and browsing their data sheets, I think this might be work:

http://www.ctscorp.com/components/Datasheets/450.pdf
450TA20S501A1A1

However, that's now a custom part apparently. Their stock parts seem to use plastic shafts, but may work, so I've ordered one of those.

[motocoder]

Looks like I'm typing to myself here, but I'll continue to update the thread in case it helps someone with a similar repair in the future.

So, the monkey that owned this thing previously replaced the 1/4A fuse with a 2A fuse, which explains why the range switch resistors were destroyed. The resistors that were destroyed were:

R185 - This is a 100 ohm 0.1% resistor that is used for the second highest current range.

R107, R108, R102, R103 - These appear to be used, along with Q103 and Q106, for a range-specific gross current limit.

For the gross current limit, I believe it operates as follows: the voltage across R102 or R103 (depending on which range is selected) is Vbe of Q103. So as the current through R102 increases, eventually Q103 starts to conduct. This has the effect of pulling down the base of driver transistor Q102, which makes Q101 conduct less, dropping the output voltage. This limits the current to ~150mA or ~33mA, depending on which range is selected. There is a similar mechanism for R107/108 and Q106 (although I'm not sure why R108 is a different value there).  Because Q101 and Q105 failed in a short condition, this current limit was not effective, and large amounts of current flowed. Because previous owner-monkey replaced fuse with 2A version, those resistors cooked until they eventually failed. I guess he switched it around to different ranges to make sure he cooked all 4 resistors.

So I replaced those resistors, but also decided to check the transistors (Q102 and Q104) that drive the output transistors, and the current limit transistors (Q103 and Q106). The driver transistors look OK, but the current limit transistors were both bad one shorted, one open circuit).  Q103 is a 2N3565, and Q106 is a 2N3638. These both seem to be unobtanium, so I replaced them with 2N3904 and 2N3906 which I selected to have similar beta.

Checking the output voltage now, it's not pegged anymore, but it's still not working correctly. For example, it does not respond to changes in the output current selector switches. Also, switching between positive and negative output polarity does nothing; it is always outputting a positive voltage. It does seem to respond a little to the "compliance voltage" (voltage limit) pot, but even there it's not operating as  you would expect.

I suspect the op-amp that or something else in the feedback loop is fried. Will check that tomorrow.

[motocoder]

I checked QA201 output (voltage regulator). It need a little trimming, but it's outputting 24V. However, QA202 is not working correctly. This IC is an Op-Amp that derives a virtual GND half-way between +12V and -12V from the regulated output of QA201.

[cncjerry]

If you pull the plug on top of page 24203e that connects those transistors you should be able to check them in situ with a dmm diode tester. 

[motocoder]

If you pull the plug on top of page 24203e that connects those transistors you should be able to check them in situ with a dmm diode tester.

Hi cncjerry -

I am not sure what you mean about "page 24203e". However, I do have some good progress to report on the repair.

First off, I found a bunch of bad parts. It looks like some very high voltage or over-current situation happened on one or both of the output channels. I found a bunch more bad parts. And I'm not speculating when I say "bad". I mean they measured bad in-circuit, I removed them and they still measured bad.
So far this is the tally:

Resistors burned in half: R107, R108, R102, R103, R185,
Resistors, no visible damage but resistance way off: R138 (pot, just damaged on one end of the range), R101 (marked 12K, measured 20K), R120 (marked 10K, measured 17k).
Transistors: Q103, Q104, Q106, Q115, Q116
Also I was wrong when I said 24V regulator (QA201) was OK - that and the op-amp (QA-202) being used to generate the "virtual ground" for the +/- 12V supply were both bad.

This last issue was what really confounded me on the repair. I would measure various voltages in the circuit, but everything was way off. There would be no logical change in voltage at the appropriate test point as I adjusted the KV range switches. This turns out to be because the buffer op-amp I used to replace QA-202 did not have enough current drive. I had convinced myself there was another bad component dumping too much current into this net and causing the problem, but after carefully reviewing the schematic, and doing some testing out of circuit, I came to the conclusion that the op-amp just didn't have enough drive.

I've ordered something from DigiKey that should work better, but being impatient, I decided to just whip up a little circuit of my own. I used a 741 op-amp (it's what I had laying around) and a couple of transistors in a push-pull configuration to create a beefier virtual ground. After wiring this in, I can see a very steady ground node for the +/-12V supply - it changes less than a uV as I change the range selector and output selector switches. And wonder of wonders, I can actually see the KV divider working now.

So, I am waiting on a replacement for Q104, and then I will put that in, reconnect the +/- 130V rectifier and give the whole thing another try. My only real concern at this point is that all the transistor replacements will not turn out to really be a good match for the parts they replaced, and there will be some problems created because of that.

[TiN]

Great job.

Some photos of internals would juice up this thread even more 

[motocoder]

Great job.

Some photos of internals would juice up this thread even more 

Ok, I'll add some photos to this post.

I am starting to get a bit frustrated with this thing. I keep finding more bad parts. The latest is variable resistor R121. The manual says this pot is the adjustment for the bias current through transistors Q101 and Q102. However, after replacing the pot, I measured the voltage across the three diodes that are part of that circuit (D108, D109, and D110). The voltage across these is so low, that it's hard to imagine that this circuit is doing anything at all currently, and this voltage, nor that across the pull-up resistor (R120) does not change at all when I adjust R121. I don't really understand this part of the circuit, so hard to say what is going on there.

I spent a good deal of time looking at the very first differential stage, which uses a matched JFET pair (Q118A and Q118B). I took some measurements of D/G/S voltages for these JFETs as I adjusted the current value and range values. It appears that this circuit is completely biased over to one side, with the JFETs so saturated that they don't do much. This would explain why the output of the unit is always a positive voltage, regardless of which polarity I set the output selector to. One of the pics below shows the voltage measurements I took on that JFET - all voltages were made with respect to the 12V ground (this ground effectively "floats" on top of the voltage output by the series driver transistors, before the output current sensing resistors that connect to the output). I tried adjusting the "Amplifier Zero", which is a pot connected to the source of Q118A/B, but it has no effect at all. This isn't too surprising, because I think something else in the high gain stage is badly skewed to one side. I didn't capture it in the initial measurements, but I also took some measurements with the output polarity reversed. I could see the polarity of the gate voltage of Q118B reverse, but because of the aforementioned problem, this has no effect on the output voltage.

Finally, I have some questions for anyone who owns one of these. The manual has a diagram (Figure 16) with some points to use to measure various voltages during calibration. However, it's really hard to tell were the attachment points are for the "Ref or Zero" measurements. In Figure 16, it just points at the potentiometer, and at some unlabeled point on the range selector switch. Can someone clarify exactly where these attachment points are?

Range selector resistors burned in half (bottom left side of photo, just to the right of the black wires)


Top view of main circuit board


Measurements I took on Q118A/Q118B - first differential amplifier stage

[motocoder]

Ok, looks like something is wrong in the "transient absorber" section (Q401, Q402, R402 - R405). I measured the voltages on the transistors and the voltage on the emitter of Q402 is not what it should be (-3.30V instead of the expected -1.04V). This circuit is connected to the gate of Q118A, and so might be the source of the imbalance problem.

[motocoder]

Bad idea to remove the "Transient Absorber" transistors and power on. Now I'm going to need a new JFET pair...

[motocoder]

Big breakthrough this evening. I replaced the part I burned out with my dumb idea to remove the "transient absorber" transistors. I also replaced those two transistors since it wasn't possible to salvage them (leads too short).

Then I started walking through each transistor in the high gain stage. I discovered that the base of Q115 was pegged out at -12V, clearly not correct. The problem turned out to be Q116, which was shorted between collector and emitter. I replaced this and Q117 with a pair of ZTX550's, which I selected to have fairly close beta. Not sure those are a perfect match, but they can sustain more current than the original part, which seemed to be the primary reason it was selected.

Now I can see the output voltage change as I turn the selection dial. I've got more to check to confirm whether or not it's all working, and eventually need to get my questions about the calibration test points answered so I can do a calibration, but this is good progress.

[motocoder]

Well, was all working!

But then while I was trying calibrate it, R105 went up in smoke. I figured those current-limiting transistors had gone out. Checked and sure enough, Q103 was bad. Replaced Q103 and Q106 with beefier ZTX450/550 parts, put it all back together, and it burned through another R105. Quick check of output transistors and I think they might have gone too Output transistors are still good.

Not sure what is causing this.

[cncjerry]

I cant see how the base resistor r105 would burn out unless q105 is shorted.   also, wrt  my earlier note, there is a plug with the reference 'see dwg 24203e...' on it.  if you pull that plug you can diode check all those transistors possibly from the plug itself. 

one other thing on r105, did you scope it after changing the transistors to another type?  Maybe it is oscillating.  Current sources are known to have problems with certain reactive loads.

[motocoder]

I cant see how the base resistor r105 would burn out unless q105 is shorted.   also, wrt  my earlier note, there is a plug with the reference 'see dwg 24203e...' on it.  if you pull that plug you can diode check all those transistors possibly from the plug itself. 

one other thing on r105, did you scope it after changing the transistors to another type?  Maybe it is oscillating.  Current sources are known to have problems with certain reactive loads.

I pulled Q105 out of the socket, and it is fine (not shorted). I checked Q101 as well while I was there. What's interesting is that the negative polarity is not working at all anymore. It did work before the R105 problem. I don't have a thermal camera, but using one of those little infrared thermometers, it appears to me that Q104 is heating up quite a bit. Q104 has been replaced with a 2N5416S, whereas Q102 (the symmetrical/complimentary transistor on the positive side) is still the original part (MM3003). Maybe the overall gain between the two polarities is too mismatched now?

Regarding your comment about plug dwg 24203e - there is no plug anywhere on the board. What you're referring to is just how they've split the schematic into two parts. There is no physical connector between the two parts.

[motocoder]

Something is definitely fishy here. I measures the voltage across R105 and it was ~55V. THen just a few minutes ago, before connecting my scope to it, I measured again, and it's just a few 10's of millivolts. And there is a fair amount of noise on the Vbe of Q101 (~60mV), and to a lesser extent Q105, so there is some oscillation looks like.

Also, I can definitely see that Q104 is heating up, even though when I measure Vbe across it, it is just a few tenths of a volt, and Vce is the full 130V. Maybe enough leakage current there that with the high Vce value it generates some heat.

[cncjerry]

did r105 pop again?  Sometimes the burn out and don't flame.  I don't remember the value, but do the power math at 55v to be sure it isn't too much which would point you somewhere.  when I had a resistor popping in my keithley 228 VI source, one cool product by the way, I put 1/4w in to make it easier to blow.

I'm going to go look at the schematic again.

[cncjerry]

So r105 is 180ohms.  At 55V my ohm's law says it is handling 16.8W  (55^2)/180 = 16.8W.  So that means the bias is wrong on Q105 which is determined by Q104, those diodes, R104, the cap I can't see and what looks like a ladder network of some type.  I assume you checked Q104 and those diodes as well as the cap.  It could be just a gain issue with the transistor you put in.  But Q105 probably won't last long with all that base current.  Without digging in to far, I would go back and do some ohm checks.  If it get tricky to figure out, you can throw the circuit into LTSpice and try to figure out what would cause the 55V across r105.

I've been thinking about a 225 to complement my 228.

[motocoder]

did r105 pop again?  Sometimes the burn out and don't flame.  I don't remember the value, but do the power math at 55v to be sure it isn't too much which would point you somewhere.  when I had a resistor popping in my keithley 228 VI source, one cool product by the way, I put 1/4w in to make it easier to blow.

I'm going to go look at the schematic again.

It's actually a little toasted, but I've measured it a few times with the ohmmeter and it's still measuring 179 ohms.

[motocoder]

So r105 is 180ohms.  At 55V my ohm's law says it is handling 16.8W  (55^2)/180 = 16.8W.  So that means the bias is wrong on Q105 which is determined by Q104, those diodes, R104, the cap I can't see and what looks like a ladder network of some type.  I assume you checked Q104 and those diodes as well as the cap.  It could be just a gain issue with the transistor you put in.  But Q105 probably won't last long with all that base current.  Without digging in to far, I would go back and do some ohm checks.  If it get tricky to figure out, you can throw the circuit into LTSpice and try to figure out what would cause the 55V across r105.

I've been thinking about a 225 to complement my 228.

I can't reproduce the 55V across R105. Maybe I just misread a "mV" on the meter instead of "V". Probably not, though, given it burned out already once.

I've been staring at the schematic for a couple of weeks now, and I think I have most parts of it figured out. I think the diodes you're referring to are D108, D109, and D110. These, along with R121 bias the voltage across Q104 and Q102 such that they are separated by a couple of volts. This last stage is basically a push-pull amplifier, so they're biased apart a little so there isn't a dead zone near zero. A lot of times people use a "Vbe multiplier" to do this, but here they just used 3 diodes and a potentiometer (an inferior choice, IMHO). The bias point difference, however, means both complementary stages will be conducting some current at 0 output. The current can go from +130V, through Q101, out through R102 or R103 (depending on range switch setting), and then back through R107 or R108, and then through Q105 to -130V. All of that can happen without changing anything in the feedback loop, so without some limiting mechanism, it could easily run away. That's what Q103 and Q106 are for. Using Q103 as an example, and assume range switch has seletected R102. Q103's base is on one side of R102, and its emitter is on the other. R102 is 3.9 ohms. Q103 will be conducting pretty good when it's Vbe is around 0.6V. That happens when the current through R102 is 0.6 / 3.9  = 153mA.

Now I think the beta of Q101, the replacement I put in, was around 48. So if ~153mA is flowing through Q101 emitter, that means Ib is around 153/(1+48) = 3.2mA. Let's assume most of this current is coming from Q102. At 0 current output, it's dropping about 130V minus a couple of Vbe drops from collector to emitter on this transmitter. So that's about 0.4W being dissipated. So I did the calculation for the wrong side , but let's assume they're similar for the negative side. I replaced Q104 over there with a 2N5416S, which looks to have a slightly lower power dissipation rating. Maybe it's just overheating?

But either Q104 is now burned out, or there is some other part that went out because the negative polarity is not working at all - even when I start with everything cold. It does seem to work on positive polarity just fine, although as I turn up the current, Q104 seems to be heating up faster.

I did check all the transistors and diodes in this part of the circuit with the diode check function on my meter (in circuit), and they all showed up fine. However, I've had that mislead me on a couple of parts already. I hate to desolder anything, though, as the traces on this board are in rough shape after all the rework...

[motocoder]

Yes, Q104 was bad. My transistor tester said it was a "common cathode diode network", so it still looked like valid PN junctions, but just wasn't operating like a transistor. This is why I didn't detect the issue in circuit.

I've replaced it with a NTE397, which NTE lists as an equivalent part for the MM3003. I am going to go very slowly and make sure nothing is overheating this time.

[motocoder]

Ok, it's *mostly* working, but have a few problems:

- I am not able to set the bias for transistors Q101 and Q102 as described in the manual, table 6.2d on page 19. I was able to set this before, so something is definitely still not right. The bias voltage is way too high, which means there is too much current flowing at equilibrium. I think the problem might be that the gain for the two polarities is too imbalanced.
- There is definitely some oscillation that was not there before when it was briefly working perfectly.
- I can't set the positive output voltage past 30V. I can set the negative polarity, but just not positive. I think this is related to the gain problem as well.
- I noticed that the package for the power transistor I am using (NTE124) says maximum Ic is 0.1A. Now the NTE data sheet says 1A, but I'm not sure I trust that. I did find a source for the Central Semiconductor parts, mentioned earlier on the thread, that are a very close match. I may try those.

[cncjerry]

Good job with q104.  Maybe if you had the correct parts it would work now.  The transistor shouldn't get too hot unless it has a heatsink then I think you should be able to hold your finger on it. I went back a looked at your pictures.  Are those paper caps still good? Did you hang a scope or meter on the PS?  I've replaced those in most of my old radio gear as they might show the right capacitance but leak like hell.

I can make a coupler, new shaft, ect for that switch coupled pot if you get in a jam.  Cost of shipping only.

[motocoder]

Good job with q104.  Maybe if you had the correct parts it would work now.  The transistor shouldn't get too hot unless it has a heatsink then I think you should be able to hold your finger on it. I went back a looked at your pictures.  Are those paper caps still good? Did you hang a scope or meter on the PS?  I've replaced those in most of my old radio gear as they might show the right capacitance but leak like hell.

I can make a coupler, new shaft, ect for that switch coupled pot if you get in a jam.  Cost of shipping only.

I think the caps are good, but I may replace them anyway once I have the other issues resolved.

I spent about an hour troubleshooting it early this morning before I had to head out. I think I understand in part why it's not 100% working, but I don't understand the root cause  yet. The negative polarity side of the amp is not shutting off when the input selector is set to a positive voltage. The positive side polarity working fine. This is why Q104 is heating, why the bias current is out of range and you can't adjust it, and why it isn't able to hit the full positive output voltage (the negative side is dragging it down). I want to retake some measurements and I'll post those later tonight.

Regarding the coupler, thank you very much. I might ask your advice on some related mechanical issues (I am just awful with  mechanical stuff), but the existing shaft and coupler seemed to work with the potentiometer I lucked into at the local electronics shop. The only loss over the OEM equipment seems to be that this pot doesn't have a built-in switch like the original.

[motocoder]

To determine why Q105 is not shutting off, I measured the current going into Q105's base by measuring the voltage across R105 and dividing by it's value (180 ohms). With the output current set to a large negative value, it's 1mA. With the output set to a large positive value, it's as high as 0.47mA!  The transistor I selected for Q105 had a beta of around 48, so that's 48mA flowing when this thing is off - unacceptable!

Where is this current coming from? Q102, which I've replaced with an NTE154, has a typical value for emitter cutoff current of 1nA. Q104, which has been replaced with an NTE397, if the NTE "data shit" is to be believed, has a value for collector cutoff current of around 50uA. We're clearly well above that. Where is the current coming from? Is Q104 not performing up to snuff? It  definitely should be cutoff hard; Veb on this PNP transistor when output is set to the positive polarity is -0.67.

I've removed C101, so it's not leakage from that. The schematic shows a cap across the base-collector of Q104. I can't read the designator for the cap on the schematic (C110?), and I can't find any caps in that area of the PCB that might be this.

It's a mystery.

[motocoder]

Ok, I think I finally have it sorted out. The amplifier stage is really sensitive to the characteristics of Q104.

Now when I had it working great for a few minutes the other day - before the big flame-out of R105 - I was using a 2N5416S transistor that I thought was going to be a really good match. Those transistors are quite pricey, so after the flame-out, I replaced it with one of those fell off the boat" (NTE) parts, an NTE397. The NTE transistor apparently has really bad Icbo leakage at these high voltages. So once I was pretty confident this was the source of my over-current problems in the negative channel, I replaced it with the second 2N5416S that I had on hand. I've put some heatsink compound between it and its heatsink, and I'm carefully watching its temperature while I run at max output. I'm hoping the first burnout of this part was just a fluke, or a side effect of the resistor problem (see below).

The second thing I did was to replace the output driver transistors with ones with lower Hfe. I'm using the same NTE part, an NTE124, just two I selected that have Hfe of 17 and 23. Based on the specs for the OEM transistor, I think this will be a bit closer, and I am hoping to get the amp gain back down to closer to the original amount, so that these oscillations will go away. If not, we'll have to look at some additional caps to frequency compensate it.

Oh, also forgot to mention earlier that R104 was bad as well. It's tucked away behind the big yellow filter caps, so I didn't initially check it. When I did, I noticed it was reading about 30% high, so I went to replace it. As soon as I touched it with the hemostats, it crumbled.

I've had the whole thing assembled, semi-calibrated, and running now for about an hour, with no issues. I think next up will be the electro-mechanical clean-up - clean the PCB, clean the switch contacts with De-Oxit, etc.

[motocoder]

Looks like you are correct about those caps (Edit: caps replaced). Tons of rectified 60 Hz ripple bleeding through. (Edit: This was mostly common mode noise being picked up by my o-scope differential probe)

[cncjerry]

Are you not able to get the actual parts?  For my keithley parts I found a place called something like westfloridasupply.  They shipped the old transistors I needed pretty quickly.

[motocoder]

Are you not able to get the actual parts?  For my keithley parts I found a place called something like westfloridasupply.  They shipped the old transistors I needed pretty quickly.

I emailed Keithley before I started the repair - still no reply from them. If you know of a place that stocks the parts, can you give me a link or phone number? I searched for "westfloridasupply", but all I found was some company selling trash bags.

BTW - every thing is working now, except for the power supply ripple and some oscillation. Swapping the big yellow caps out for some Nichicon 220uF helped some, but there's still what I would describe as a ridiculous amount of 60/120Hz noise, and a lower level oscillation.

[motocoder]

There is more noise on the output then I would like, but it's not the dire situation I posted earlier. A lot of that was noise I was picking up on my differential probe.

[motocoder]

After the clean and De-Oxit on the switch contacts, it's pretty well. Output noise is still around 60mV. I'll poke around with it and see if I can figure out the source there, but in general I am pretty happy with it.

I'll prepare some notes and get a scan of my annotated schematic, so that hopefully someone else that buys one of these turkeys doesn't have to go through as much effort on the repair.

(Late breaking update on noise)
I soldered a 1uF film cap on the underside of the board on the +/- 130V transformer secondary. That dropped the mains noise way down. Measuring on my scope, into a 50 ohm terminator (not the one inside the scope - too scary) the peak-to-peak noise is around 14mV. RMS is 240uV.  Spec in the manual says to measure it using a meter into a 10 ohm load. I can't even measure anything using that method.

[cncjerry]

https://www.westfloridacomponents.

They had the mosfets I needed for my 228.  I only dealt with them that one time.

I'm thinking I'll just buy another 228 though Keithley made a lot of current and voltage sources.

Glad tee someone other than me stick with a difficult restoration.

[Vgkid]

Good job 

[motocoder]

https://www.westfloridacomponents.

They had the mosfets I needed for my 228.  I only dealt with them that one time.

I'm thinking I'll just buy another 228 though Keithley made a lot of current and voltage sources.

Glad tee someone other than me stick with a difficult restoration.

They do have some interesting, older, parts, but I don't see any match for the challenging parts in that 225. I did find that Central Semiconductor has a part, 2N3585, which is pretty much an exact match for the output transistors. Mouser carries those if you need them, although I think the NTE parts I'm using are fine. It's really just Q104 that is a challenge. Even now that I have it working, it is the one part that heats up quite a bit under operation - up to 80C when I was watching it, and that was with the case open.

I don't think I'll be buying anything else from Keithley. This thing is a somewhat fragile design. Compared to the HP precision and Power Designs power supplies, it doesn't stack up very well.

[motocoder]

I've scanned in the main schematic page with my annotations. I also added some notes and a list of replacement parts that I found to work. Hopefully this will save someone some time.

I have a higher res version in Photoshop format, with the annotations on separate layers. However, it is too big to upload here. PM me if you want a copy.

[TiN]

I would be happy to rehost it on my site in Keithley section, if you may upload here.
 

[motocoder]

I would be happy to rehost it on my site in Keithley section, if you may upload here.
 

Thanks. They are in a folder called Keithley_225.

[motocoder]

Ok! I figured out why Q104 was heating up so much. It is because of R104. I believe this resistor is just there to pull the base of Q105 down so that the transistor is not on when Q104 is off. However, 100 ohms is a very low value when you consider the voltage across it can be ~130V. The current this was generating through the collector of Q104 was around 15mA. When you combine that with the 5mA needed to drive Q105 at full output current, that's 20mA flowing through Q104. At 130V drop across this transistor, that's about 2.5W, way too much.

This is why the first transistor I used for Q104 failed. I see no reason why R104 needs to be 100 ohms. I have replaced it with a 12k, everything still appears to be working, and Q104 is nice and cool. 

[TiN]

Added to Keithley section, many thanks! 

Keep up good work.

[motocoder]

Added to Keithley section, many thanks! 

Keep up good work.

Thank you. You should add a link to your web site to your EEVBlog profile.

[motocoder]

Here's a final pic of the whole thing buttoned up and working.

[cncjerry]

Nice job, I'm sure you are very satisfied with the results.

[motocoder]

Nice job, I'm sure you are very satisfied with the results.

I still have a few quibbles, mostly with the "rail splitter" circuit that I built to replace the broken one in the source, but I'm forcing myself to leave it alone. Thanks for your advice and help on the restoration.

[motocoder]

I'm selling one of these 225s (not the one discussed in this thread, but a second unit I purchased to use as a reference). Details here:

https://www.eevblog.com/forum/buysellwanted/keithley-225-current-source/

[dom0]

they are likely matched hfe npn/pnp pairs

replacement should not require an exact match, but ymmv.

The electronics of the 225 are quite nice, actually.

Output stage: on the positive side you have your usual Darlington (alas - directly driven from the voltage gain stage of the error amp), on the negative it's a Sziklai pair. Both have additional instantaneous current limiting (Q103 / Q106 ... and switched with range!). So the available drive current is rather low, so I concur with the others - they are probably just selected for high current gain.

Pass element leakage doesn't play a role here, the error amp compensates it. The matching between the two is not very relevant, since they are biased to some cross-over current anyway.

I personally find the clarity and "well-thoughtness" of the schematic very nice. No surprise that the 225 is a lab classic. What surprises me a bit is that this is actually a voltage source controlled to output current. I would have expected a current-controlled current source.

[motocoder]

they are likely matched hfe npn/pnp pairs

replacement should not require an exact match, but ymmv.

The electronics of the 225 are quite nice, actually.

Output stage: on the positive side you have your usual Darlington (alas - directly driven from the voltage gain stage of the error amp), on the negative it's a Sziklai pair. Both have additional instantaneous current limiting (Q103 / Q106 ... and switched with range!). So the available drive current is rather low, so I concur with the others - they are probably just selected for high current gain.

Pass element leakage doesn't play a role here, the error amp compensates it. The matching between the two is not very relevant, since they are biased to some cross-over current anyway.

I personally find the clarity and "well-thoughtness" of the schematic very nice. No surprise that the 225 is a lab classic. What surprises me a bit is that this is actually a voltage source controlled to output current. I would have expected a current-controlled current source.

I wasn't able to measure the gain on this supply's output transistors because they were both blown. I have another 225, the one that I am selling, but I didn't want to remove it's output transistors as I was planning on selling it and wanted to leave it as untouched as possible. However, I found with repair of another, similar, current source (HP 6177C), that the output transistors were relatively high gain (hFE > 120). I wasn't able to source transistors with gain that high, so perhaps the replacements I am using here are not as high gain. Nonetheless, the performance of my modified 225 is quite good. In the noise/ripple department, it exceeds that of the unmodified version now. This is mostly attributable to the improved circuit I built for the virtual ground on the +/-12V supply.

I agree that the schematic on the Keithley is good. In fact, I would say that the whole manual is great. I love that they include a section on theory of operation.

If you're interested, take a look at the HP 6177C/6181C design. The manual for that also has a very good theory of operation and a decent schematic (you need to pay the money for a good scan, the free scans are awful). Many aspects of the design are similar. For example, there is a -1V reference voltage (HP calls it the "guard voltage") that is generated via one set of circuitry, and that is compared against a high-side voltage drop across sense resistors. All the control circuitry "floats" on top of the high side voltage right before the current sense resistors. The HP supply is unipolar, has a different circuit for the voltage limit, has an active circuit to shunt unused current, and has a meter for adjustment instead of the KV decade switches. Also, the HP surrounds the output conductor with the "guard voltage" as a way of minimizing leakage current. I am not sure how much not doing that impacts the Keithley, as I really don't have a good way to measure things at the very low currents necessary to see the impacts there.

[Ckid]

I just want to add to this Keithley 225 repair thread. I am repairing a similar failure-mode Keithley 225 with similarly failed parts, but a different original cause to the failure.

In mine, like in this thread, R108, R107, R103, R102 are burned. Someone attempted repair by replacing a couple of these resistors. Further checking finds Q101 and Q105 are bad.

The cause of these failures is the plastic on the shaft insulator of the rotary switch has deteriorated and cracked. The insulator shorted the switch shaft to the case resulting in a similar set of failed parts. The insulator is a Millen 39016 which Keithley modified by drilling out the plastic center for the fine-adjust inner shaft.

Millen is out-of-business but I found the part on eBay.  I'm not finished with repairs yet, but this thread will be an excellent help completing the repair.

[bap8guy]

I have tried to access the link provided for the manual, however it is no longer accessible.  The manual I downloaded from Tektronix is missing pages 13 and 14, crucial to the calibration steps.  Does anyone have them that can share?