(FIXED!) Keithley 705 - PSU, ROM and GPIB

[don.r]

I have a Keithley 705 Scanner with a burnt resistor in the PSU. The resistor in question is parallel to the 7805 regulator. Its R132 33 ohm at the top of the schematic coming out of the bridge rectifier. I replaced the burnt carbon resistor with a 2W metal film and its dropping about 7.5V constantly so burning about 1.6W. It gets quite hot and thus the death of carbon. I am curious as to the purpose of this resistor other than acting as a heater. Does it allow the battery circuit to kick in to back up when the 5V rail drops out? I assume it would be OK to replace this with a 100 ohm resistor to lower the chance of another burnout. The battery is old and hold very little charge, btw.

[Gyro]

To reduce the power dissipation of the 7805 regulator? The resistor would probably cook and fail if the 5V rail was shorted.

Just a thought.

[Ash]

Hi Don.R,

Yes, I had to replace the ones on my two 705s. I went with a higher power wire wound, probably 10W, but not sure (I'd have to open it up and look).

It is an interesting design choice for sure. Like Gyro, I thought it was probably something to do with additional current capability "around" the 7805. From memory it was similar on my 740 scanning thermometer, but I don't have schematics to check that without opening it up again..

The Memo Power 3-101 FT-P NiCd cells were common to the two 705 and the 740 I have, all of which failed long ago. They work fine without them, but obviously you settings get lost. If you find a suitable replacement let me know! I pulled the batteries out to avoid leakage damaging the boards.

Also the big filter cap near the 33R resistor was very dead on all my units - must have been cooked slowly. I went through and replaced all the electros anyway.

Ash.

[don.r]

Thank you, gentlemen. It certainly operated as a fuse in mine! The filter cap looks OK. The caps are all either electrolytic or ceramic. The filter cap tests good for ESR and capacitance. I added another 33 ohm in series and its running a lot cooler. I'll pull the NiCd out for safety. I wonder if a 3.7V LiPo with a 5V charging circuit would work? Now I just have to get the rest of the system working. Front panel is dark on power up with occasional gibberish. Likely the CPU or ROM.

[Ash]

Hi Don.R

Good luck! I have B6 and B7 revision firmware EPROM images if you need them.

I've attached a "before" photo of one of my units for posterity..

[edit]: By the way, I was wrong about the K 740 scanning thermometer being similar, it has a different more "normal" supply circuit.

Ash.

[don.r]

Thanks so much, Ash. I may take you up on the firmware. I think my images are the same. THey have the original metal tape cover that is not wide enough to cover the window . My board looked in very similar condition to yours. The resistor was also heavily discoloured and when I touched it it broke in half. I also replaced the jumper header next to the resistor as the clasp had become yellow and brittle from the heat and snapped off. Not one of Keithley's best designs, IMO.

[David Hess]

The resistor just bypasses some of the load current around the regulator to limit power dissipation in the regulator itself.

[don.r]

The resistor just bypasses some of the load current around the regulator to limit power dissipation in the regulator itself.

Given the downsides I'm surprised Keithley didn't just use a pass transistor instead. The placement of the resistor next to the main filter cap is also poor since its dissipating 1.6W continuously.

[Ash]

The resistor just bypasses some of the load current around the regulator to limit power dissipation in the regulator itself.

Given the downsides I'm surprised Keithley didn't just use a pass transistor instead. The placement of the resistor next to the main filter cap is also poor since its dissipating 1.6W continuously.

On the positive side, I guess it keeps the humidity from condensing on your very high quality relays.. I ended up with a 7058 low current card in one of my units - triaxial connectors and > 10G isolation.. Wash with Freon the manual says.. hum, maybe not 

Ash.

[David Hess]

The resistor just bypasses some of the load current around the regulator to limit power dissipation in the regulator itself.

Given the downsides I'm surprised Keithley didn't just use a pass transistor instead. The placement of the resistor next to the main filter cap is also poor since its dissipating 1.6W continuously.

Using a pass transistor does not improve the situation as far as power; the extra power has to be dissipated somewhere no matter how.  And a pass transistor added to a three terminal regulator requires say 3 resistors and a diode to extend the current limiting and thermal protection of the integrated regulator to the transistor which is a lot more expensive than just a power resistor.

The only disadvantage of using a resistor is that the minimum load current is raised but if you have a well defined load, that is not a problem.

[don.r]

The resistor just bypasses some of the load current around the regulator to limit power dissipation in the regulator itself.

Given the downsides I'm surprised Keithley didn't just use a pass transistor instead. The placement of the resistor next to the main filter cap is also poor since its dissipating 1.6W continuously.

Using a pass transistor does not improve the situation as far as power; the extra power has to be dissipated somewhere no matter how.  And a pass transistor added to a three terminal regulator requires say 3 resistors and a diode to extend the current limiting and thermal protection of the integrated regulator to the transistor which is a lot more expensive than just a power resistor.

The only disadvantage of using a resistor is that the minimum load current is raised but if you have a well defined load, that is not a problem.

The other disadvantage being the resistor burns out (seems to be common to these) placing full load on the regulator so endangering it. Looking at the build quality and original price of the 705 (~$3000), I doubt cost was a big consideration. The 5 and 6v regulators sit on their own offboard heatsink already. Considering that scanners are meant to be on for long periods of time, putting a carbon resistor continuously running near its power limit and then placing it where they did seems rather odd to me.

Anyway, thanks to Ash for supplying me with a copy of the B6 firmware, I found that one of the ROMs was corrupted. I flashed a new one but no luck yet. The CPU appears to be working fine as clock, data and address lines are all functioning. Its something further up the chain like the interface chip, glue logic or display shift register. Hopefully not the latter as that means I have to pull all the button caps off to take the board out.

[David Hess]

The other disadvantage being the resistor burns out (seems to be common to these) placing full load on the regulator so endangering it.

Presumably the resistor should be power derated for this not to be a problem which oddly enough they did not do.  Larger resistors are not that much more expensive.

Looking at the build quality and original price of the 705 (~$3000), I doubt cost was a big consideration. The 5 and 6v regulators sit on their own offboard heatsink already. Considering that scanners are meant to be on for long periods of time, putting a carbon resistor continuously running near its power limit and then placing it where they did seems rather odd to me.

That *is* odd.  My guess is that the design was finished before they realized there was a problem with power dissipation or current rating for the +5 volt regulator and they did not want to make significant changes.  The regulator should have had a more effective heat sink and if current was a problem, a higher current regulator used.  An 3 amp LM350 or 5 amp LM338 is the usual solution since TO-220 fixed voltage regulators were not common above 1.5 amps although TO-3 ones were.  Alternatively, a partially discrete solution which includes foldback current limiting would have been my first choice.  No reference would be needed since it could have tracked one of the other regulated supply voltages which has its own advantages.

The situation reminds me of the story where the experts work on the digital and analog parts of the design while the new guy is assigned to the "simple" power supply.  I remember when I was the new guy.

[don.r]

That *is* odd.  My guess is that the design was finished before they realized there was a problem with power dissipation or current rating for the +5 volt regulator and they did not want to make significant changes.  The regulator should have had a more effective heat sink and if current was a problem, a higher current regulator used.  An 3 amp LM350 or 5 amp LM338 is the usual solution since TO-220 fixed voltage regulators were not common above 1.5 amps although TO-3 ones were.  Alternatively, a partially discrete solution which includes foldback current limiting would have been my first choice.  No reference would be needed since it could have tracked one of the other regulated supply voltages which has its own advantages.

The situation reminds me of the story where the experts work on the digital and analog parts of the design while the new guy is assigned to the "simple" power supply.  I remember when I was the new guy.

I would have opted for a larger regulator or a pass trnasistor. The supply only has two regulated outputs, one 6V for the analog and one 5V for the digital. Both of these have to be made available to up to 2 I/O cards.

If anyone wants to check the signal on pin 39 (CA2) of the 6522 (U119) that would be appreciated. This should be the 1khz clock for the display shift register. Mine is always on logic high (5V). Either the 6522 is not being programmed correctly or its defective. Since the EPROMs now check out and the CPU appears to be functioning, I'm going with the latter.

[Ash]

The other disadvantage being the resistor burns out (seems to be common to these) placing full load on the regulator so endangering it. Looking at the build quality and original price of the 705 (~$3000), I doubt cost was a big consideration. The 5 and 6v regulators sit on their own offboard heatsink already. Considering that scanners are meant to be on for long periods of time, putting a carbon resistor continuously running near its power limit and then placing it where they did seems rather odd to me.

Note sure they were aware of the "common" failure mode of these carbon composition resistors back then.. 

It appears obvious to us in hindsight, after fixing similar issues, but may not have been obvious then. Same goes for the electrolytic right next to the heat source I guess - not sure if it was "common knowledge" back when this unit was designed. I could very well be wrong about that though.

Anyway, thanks to Ash for supplying me with a copy of the B6 firmware, I found that one of the ROMs was corrupted. I flashed a new one but no luck yet. The CPU appears to be working fine as clock, data and address lines are all functioning. Its something further up the chain like the interface chip, glue logic or display shift register. Hopefully not the latter as that means I have to pull all the button caps off to take the board out.

Your welcome. I've uploaded the firmware images I have to the XDevs documentation site (Thanks @TiN), I've got some other equipment eproms and manuals I'll add there as I get a chance over the next week..

I remember having to pull down the front panels of my units when I was repairing them. The plastic can be a little brittle on the buttons I discovered, I broke part of the stem from one button. Take it easy when you pull them off. I glued mine back together and its fine now Also watch out for the mounting lugs that screw into the bottom of the box.. one of mine came broken.. Its a bit fiddly to get it disassembled the first time.

Sorry I didn't record my repairs in any detail on these units.. I've been trying to be better at that since..

Ash.

[don.r]

Yes, my mounts on the bottom panel were broken when my unit arrived. Age and plastics are not a good mix. Luckily, I was able to re-glue enough pieces to restore the bottom panel. My front panel button caps all came off but only after demanding a sacrifice to the repair gods: a blister for my thumb. 

[David Hess]

That *is* odd.  My guess is that the design was finished before they realized there was a problem with power dissipation or current rating for the +5 volt regulator and they did not want to make significant changes.  The regulator should have had a more effective heat sink and if current was a problem, a higher current regulator used.  An 3 amp LM350 or 5 amp LM338 is the usual solution since TO-220 fixed voltage regulators were not common above 1.5 amps although TO-3 ones were.  Alternatively, a partially discrete solution which includes foldback current limiting would have been my first choice.  No reference would be needed since it could have tracked one of the other regulated supply voltages which has its own advantages.

I would have opted for a larger regulator or a pass transistor. The supply only has two regulated outputs, one 6V for the analog and one 5V for the digital. Both of these have to be made available to up to 2 I/O cards.

The Keithley 705 design is old enough that the LM309 5 volt 3 amp TO-3 packaged regulator was available but I think they were pretty expensive even then.  The only common higher current regulators I remember from then were the still available TO-220 packaged 3 amp LM350 and 5 amp LM338 variable regulators which would have been my choice and I would have used an LM317 (or 7805 with the common pin raised by 1 volt, ha!) instead of that unusual 7806 Keithley chose.

A more discrete design would have been price competitive but more complex and take more space.  Adding the bypass resistor to the TO-220 7805 was certainly the cheapest option.  Another interesting option is to parallel a pair of 7805s with an operational amplifier used to enforce current sharing; this has the advantage of decreasing the junction to heat sink thermal resistance over a more powerful TO-220 part like the LM350 or LM338 while keeping the protection features of an integrated regulator.  Adding an external pass transistor to the 7805 is a good option also and what I have done in the past; if you do it right, then the integrated regulator's current limit and thermal protection protect the transistor as well.

[don.r]

Although the parts list specifies an LM7805, my unit came with an LM340A and it looks factory. Pretty certain these were available in TO-3.

[SeanB]

LM340 was a second source for the LM7805, though there were later versions with higher current capability, and which came in both TO3 and TO220 packages.

With the resistor I would replace it with a 5W ceramic version, and use some stand off ceramic spacers to have the resistor as high off the board as possible, lead length permitting, so it ran cooler and did not cook the connector and the capacitor. the placement says this was a near last amendment to fix a power dissipation issue in the voltage regulator circuit, which was either failing regulation during thermal cycling, or with voltage input range testing at the upper limits for the input mains voltage range along with the thermal cycles as well.

The regulator probably was running at 130C on the die at the top end of ambient temperature in an enclosed test rack, and with the input voltage being at the upper mains limit for hours, the dies on certain would go into thermal shutdown and cycle the power supply, especially if the unit was delivering a high internal dissipation at the same time.

[don.r]

Yes, its quite possible that they added this as an attempt to keep the regulator cooler but at less than 250mA (7.5V/33ohms) its not a lot of extra headroom. Schematic looks like it was intended from the beginning but that could have been redrawn.

[don.r]

So I can get the unit to display the initial startup screen (full illumination) but it stops there. I can do this fairly repeatedly by pressing down on the 2732 ROM IC. I swapped the 6808P and 6522 from my Keithley 230 with no difference so they are both functional. I'm suspecting the 2732 and will replace it when a new one arrives.

[don.r]

Finally resurrecting this project after a long pause. The 2732 checks out and I replaced the 2764 (It was a mess) using Ash's B6 firmware. I still get the random gibberish on the display (which appears to otherwise work fine) followed by a lockup. I suspect it could be the SRAM (yikes!) as the CPU and peripheral controller are both fine. Any further help would be appreciated.

[Ash]

Finally resurrecting this project after a long pause. The 2732 checks out and I replaced the 2764 (It was a mess) using Ash's B6 firmware. I still get the random gibberish on the display (which appears to otherwise work fine) followed by a lockup. I suspect it could be the SRAM (yikes!) as the CPU and peripheral controller are both fine. Any further help would be appreciated.

Hi Don,

Do you have a copy of the manual? Ram and Rom checks are done at start up. If you get something that looks like "-1AAAAAA", that's a RAM error. "1oooooo" (all on top half of segments) is a ROM error. This is in section 5.6 of the manual (about page 62 in my copy of the pdf). There is also self test that can be run using "program 8" on the front panel. It will do a RAM check and a ROM CRC check and show "rA FAIL" or "rO FAIL" for the rAm and rOm failures.

If it is the RAM, have a check that the CS/WE is being driven and that they are getting good power - they are powered from the battery circuit, not the normal 5V rail. There seems to be some safety circuits around the RAM to prevent bad writes when powered down and to isolate the power domains a little. Q101 and Q102 in the charge circuit should switch 5V onto the +BAT lines when there is power, Q102 would isolate it when there is no power so the battery isn't drained.

After checking the voltage rails, I'd be tracing out signals starting from U112 (the address decoder), then U115 (a quad switch that drives the RAM chip selects) and the area around U113 (quad NAND) is correctly driving the write enable pins on the RAM. There is also a RAM Enable signal that comes from U104 (4094 8 bit shift register) - it seems the CPU must set that up pretty early during boot I guess.

Hope that helps.
Ash

[don.r]

Hi Ash,

Right now the screen goes to the startup (all LEDs lit) and then random garbage. Power rails (+5v, +BAT) look good. I do notice that the all data lines have long tails on the down slope, like something is trying to pull them all up. Both ROMs read correct on my EPROM programmer.

[don.r]

PROGRESS! I pulled the GPIB chip TMS9914A and low and behold... she works! 0 1 F 0 and everything. Must have been the 9914 messing up the data bus. I ran program 8 and it passes so RAM/ROM is good. When I have time I'll swap the TMS9914A from my 730 to confirm that it is the culprit. YEEHAW! To think I was so close all this time...

[don.r]

Confirmed: the TMS9914A is bad and everything works fine with a good one. Thanks to everyone for the help, especially Ash. On to the next project!