Old HP 6294A 1A 60V PSU calibration/repair question (SOLVED)

[Paul Moir]

I went out to pick up a HP 3312A signal generator and this power supply followed me home.  It seemed to work so I cleaned up the knobs and since the meter was off a bit, I thought I'd do a quick calibration.  Unfortunately I got hung up at the zero constant current setting.

Quickly, the manual has you crank the voltage to max and short out the constant current knob so it's set to dead zero.  Then it has you check the output voltage from the supply and tinker with the bias resistors on the constant current amplifier (matched transistor discreet differential amplifier) until you get "0 volts".  The thing is it's drifting around way too much to say I have zero volts.  One way and the supply goes down to -.5v (where it hits a diode), bump it up a few ohms and it climbs to the full output voltage.

I know it doesn't really matter since it's really only putting out or taking in something under 50uA but it's driving me nuts.  Is this actually a problem or are the calibration instructions a little optimistic?  These have pretty much the same control circuitry as the 620x so I think anything like that would be the same.  If you have one, does the output go to zero volts when you set the current control all the way down?

Everything else appears to be working perfectly.  Even the transient recovery is way beyond spec.  Perhaps a little too good?

Side note:  I thought it was the current amplifier since the hFE of one side of Q2 (the current amplifier) was 200 while the other side was 280ish, but I replaced it with an NXP BCM847BS and the problem persists.  I'm pretty confident the current amplifier circuit is working as intended.  What I can't see is any real feedback when there's no current across the current shunt resistor.  Unless maybe it's the voltage clamp circuit that does it...

Service manual:
http://exodus.poly.edu/~kurt/manuals/manuals/HP%20Agilent/HP%206294A%20Operating%20&%20Service.pdf

[oldway]

Check first if the connections of the rear barrier strip are correct and if there is no bad contact.
Check also if the voltages of the reference regulator are not drifting.

[Paul Moir]

Thank you.  I have checked these but only superficially.  I will double check and report back.  The negative reference is slightly low so there may be a problem there.
My meter is 6000 count so I'll have to setup a resistor divider to see what's going on down at the bottom of the references (6.2v).

[Paul Moir]

Ah, it looks like the references are just fine.  No drift over an hour or two that I could measure (below 1mV).  And the jumpers are all fine and tight.


I've got the current dialed in as close as I could get it, but still when it drifts say +0.050mV it starts charging up the output cap.  When it decides to drop 0.050 it starts discharging it and eventually goes down to -0.5.
But I still don't properly understand the design.  I really don't see how the current limiter could "know" the output had hit 0v and not keep telling it to go down.  I'll keep studying it though.

[oldway]

I think you are looking for problems who does not exist.

One way and the supply goes down to -.5v (where it hits a diode), bump it up a few ohms and it climbs to the full output voltage.

This is a normal behaviour, nothing wrong with this.
As there is no load, there is no current feedback. It's an open loop.
For this reason, it goes to 0 or to Vmax, there is no possible regulation.

Before tempting to calibrate the power supply, you should have to do a functional test to be sure if a calibration is needed or not.
This kind of power supply is very simple and is not a high precision one.

[Paul Moir]

Thank you, I'm inclined to agree!  The only voltage feedback into the current amplifier I see is R15, which is 220K.  This is divided down by a 3.3 ohm resistor, which doesn't leave much signal left.  I can't imagine that operating down under 1V.

[Paul Moir]

Well, I couldn't need well enough alone.  To solve the problem, I hung a constant ~1mA load on the output.  This had a nice side effect in limiting how negative the supply would go in constant current mode since the constant current amplifier had a nice load to measure with.  Rather than going all the way to -.5v it goes only to about -.1v now.  (I had previously considered replacing or paralleling the anti-reverse diode with a schottky.)

The reason I did this was really just for fun.  If I have to present a practical reason, it would be that I would like to use the current limiter to turn the supply "off" while using it, since these supplies don't have an output switch.

Construction is, in honour of the recently deceased, on a Radio Shack perfboard of the kind where the pads lift off for no reason.  For the differential pair of the current sink I used two somewhat matched 2N5401 PNP transistors harvested from a CRT neck board.  These were superglued together for some thermal matching.  Incidentally, it didn't work at all without gluing them together and adding the emitter resistors, since the heating in one is drastically different than the other.  Resistors where mostly just what was handy.

This is sort of upside down because everything in these HP power supplies is referenced to the +S rail, including the 2.2V reference I hijacked. 






Here's the replacement for the original constant current amplifier with the poorly matching hfe.  Note the use of matched length traces; this is very important on these old power supplies and often overlooked.  Uneven lengths cause distortions if not present.