3456A Repair: Sticky keys, digital board and ... Current Source?

[don.r]

I have an HP 3456A that suffers, from among other things, the dreaded sticky HP push button syndrome. This is due to bent/creased spring steel in the buttons. I have new springs but need to remove the caps to get the old ones out and the new ones in. Some of the caps came off easy but most refuse to budge and I'm loathe to put force lest I rip the button clear off the board (heat staked in). I tried freezing the board over night but had no luck.

[wn1fju]

Am I missing something here?  I've replaced the springs in dozens of HP models and haven't had to remove any buttons if I recall.  One slides the spring
in from the low point (the usually red housing) and curls it up until it meets the ridge on the button shaft (usually white).   I usually use a combination of
tweezers and dental picks because one has to bend the flat spring so it climbs up the shaft.  Then you have to press on the ends of the spring to seat it
fully.

Takes a bit of practice and often takes me 5 or 6 tries to get it right.  But I've certainly had my share of trying to pull the button off only to have the
entire switch mechanism break free of the heat staking and totally lift off the board.

I know I've seen some (probably YouTube) videos out there demonstrating this procedure.

[don.r]

In the videos, the key caps are off the buttons so I assumed that was standard procedure. I'll have go at doing it with the caps in place and see how it goes.

[MarkL]

You could try gently heating the key caps with a heat gun.

I haven't tried it on HP key caps, but it work on Tektronix knobs which are plastic knobs on plastic shafts.

[don.r]

Damn that is painful. Takes me about 5 minutes to replace a spring, mostly putting the new bugger in. I have noticed a couple of the heat stakes are snapped off leaving only one leg on the switch. Would an epoxy from the underside be enough to hold it in?

[don.r]

I got all the sticky springs sorted out and have now run the self tests.

All voltage ranges appear to work for both AC and DC with reasonable results (haven't checked accuracy yet). Resistance measurements (4 or 2 wire) always show around 10 ohms when any resistor is connected.

All tests except 2, 10, 11, 12 pass. 2 is the CPU and likely some bit rot in the ROM. Both CPU and RAM have been replaced in the past (flux residue). I'll address that later probably with a ROM swap (7 digit firmware?). 10, 11 and 12 are all out by at least a factor of 10. This corresponds with the ohms readings being bad. According to page 8-59 of the service manual, the last 3 tests failing all point to the ohms function being bad so I'm guessing the current sources. When inputs are open 10G and 1G ohm ranges both show 0, other ranges show OL. Service Group D is the next step.

[don.r]

I'm trying to understand how the current source works but I can't figure out what the purpose of Q207 is (upper left JFET).

[don.r]

Current source selector signal H4 is inactive caused by  U701 CD4555 demux having a shorted output on pin 7. Will need to wait for replacement to arrive. I suppose a 74LS139 will work as a substitute. Getting low resistance values across all the current source output JFETs. One or more of them may be leaky. I might try going through them one by one. In the meantime, I'm converting it to fanless by moving diode CR1 and regulator U1 to rear panel on a bracket I'll make. Old fan was 120V @ 2A!!!!

[Kleinstein]

The JFET Q207 is used as a low leakage diode. It should prevent the current regulating  OP to go to far into positive saturation, e.g. if the DUT resistance is hight.

[don.r]

The JFET Q207 is used as a low leakage diode. It should prevent the current regulating  OP to go to far into positive saturation, e.g. if the DUT resistance is hight.

Thanks. So the circuit is relying on the FETs leakage characteristics to operate. It appears to be working because the voltages on the op amp pin 2 are valid and stable.

By the way, if I need to replace any of the current switching JFETs do they need to be the original HP part numbers (I can get them)? They are listed as matched JFETs. Not sure if there would be a suitable 3rd party substitute.

[Kleinstein]

I don't think the fets need to be matched. It is more that leakage needs to be low. For most of them the 2N4118 should be OK. Q209 may need to be a little higher current (e.g. J202 or similar).  The expected signal at TP203 could give a hint on the threshold level the fets are supposed to have.

There is no need for special leakage characteristic - just the lower the better.  I don't see any compensation of leakage currents for the FETs.

[don.r]

I don't think the fets need to be matched. It is more that leakage needs to be low. For most of them the 2N4118 should be OK. Q209 may need to be a little higher current (e.g. J202 or similar).  The expected signal at TP203 could give a hint on the threshold level the fets are supposed to have.

There is no need for special leakage characteristic - just the lower the better.  I don't see any compensation of leakage currents for the FETs.

Thanks again. Most of the FETs in the 3456a seem to be of two types given by two different part numbers and look to be more or less general purpose.

[don.r]

It looks like U205 is the culprit in the current source as it outputs about 7V even when the inputs are equal. So looks like it needs replacing. Schematic calls for an LM208. I have an LM108 on hand or I could use an LT1012 but not sure if it would see any benefit.

[don.r]

The saga continues with a new twist. I powered the unit up after removing the drain leg of one of the current source FETs and got the all dash display "------". Turns out the 5V rail on the Inguard PS now measures 0V and the 33V appears like a 50V AC line i.e. no rectification or filtering.  Either I did something horribly wrong or Murphy struck with impeccable timing. I'll check the PSU next out of circuit and then disconnected from the Inguard boards.