HP E3615A power supply slow ramp up

[gaminn]

Hi,
recently I noticed that when I turn on my E3615A set e.g. for 5V, it takes it some time to reach 5V. Immediatelly after power the voltage is 4.30 V and after approx 1 minute it reaches 4.98V. After several minutes it finally gets to 5V. No load connected to the power supply. What can be wrong with it?

[MathWizard]

What's it like at other voltages, are there are relays that click on for voltage ranges? It sounds like something warms up and drifts back to good. Have you tried it with a load, and see what happens when stuff starts to heat up quicker ?

I wonder what the PSU section is doing, maybe some rail is slow to regulate, like some transistor that's slow turning on or whatever.

I don't see the E3615 schematic, here's the E3610-3613 manuals, I hope they are similar, there's not much to them really.
https://xdevs.com/doc/HP_Agilent_Keysight/HP%20E3610A,%2011A,%2012A%20Operating%20&%20Service.pdf

[pqass]

Service manual is here.  Schematics on PDF page 34.
The CV and CC amplifier blocks are more complex than the E3610-12A supplies.

On first power up (after hours of being off) is the CV LED on or does it start with the CC LED on first (or both LEDs on)?    Is the OVP LED on at any time?
Try turning the current knob fully clockwise and try restarting (with no load attached).

Does the same slow ramp up occur even after the unit is warmed-up?
ie. After being on for an hour, turning the unit off for 10 seconds then on again.

[gaminn]

Thanks guys.

1) CC LED is not on during the slow ramp up.
2) After 90 minutes of being powered off, without load, the unit reaches 4.95 V after approx 30 seconds. With 5 ohm load, it is seems to be the same amount of time.
3) When the unit is warmed up, i.e. 1 hour on, 10 seconds off then on, it reaches set voltage immediatelly.

[pqass]

First,lets confirm that the built-in voltmeter isn't the culprit here. 
Connect an external trusted multimeter/bench meter (in VDC mode) to the PS output and try powering again from a cold PS (many hours off).   If the attached multimeter exhibits the same slow ramp-up then it confirms that something in the PS voltage feedback loop is the culprit.  Proceed to the next test below.

Secondly, attach a trusted multimeter/bench meter (capable of 1 millivolt resolution@10V, ie. 4-1/2 digits or 20,000 count; in VDC mode) between TP6 (black) and TP7(red).  Again, after a cold start, confirm that the reading is a steady +10V (or close to it).  Confirm that this reading DOES NOT ramp-up/down (should be stable just seconds after power-up).

If the TP7/TP6 reading experiences a similar ramp-up/down over 30 seconds (or whatever until stable at a similar rate to the built-in voltmeter), then that points to U15 (LM336BZ-5.0); the voltage reference.  Confirm again, this time TP6(black) and R109 (red on opposite of the -12V resistor lead).  It should be -5V (or close to it). If you see the same ramp-up/down (opposite polarity and half the magnitude of previous TP7 reading, over the 30 seconds 'till stable) then it's definitely U15.

[gaminn]

First,lets confirm that the built-in voltmeter isn't the culprit here. 
Connect an external trusted multimeter/bench meter (in VDC mode) to the PS output and try powering again from a cold PS (many hours off).   If the attached multimeter exhibits the same slow ramp-up then it confirms that something in the PS voltage feedback loop is the culprit.  Proceed to the next test below.

Secondly, attach a trusted multimeter/bench meter (capable of 1 millivolt resolution@10V, ie. 4-1/2 digits or 20,000 count; in VDC mode) between TP6 (black) and TP7(red).  Again, after a cold start, confirm that the reading is a steady +10V (or close to it).  Confirm that this reading DOES NOT ramp-up/down (should be stable just seconds after power-up).

If the TP7/TP6 reading experiences a similar ramp-up/down over 30 seconds (or whatever until stable at a similar rate to the built-in voltmeter), then that points to U15 (LM336BZ-5.0); the voltage reference.  Confirm again, this time TP6(black) and R109 (red on opposite of the -12V resistor lead).  It should be -5V (or close to it). If you see the same ramp-up/down (opposite polarity and half the magnitude of previous TP7 reading, over the 30 seconds 'till stable) then it's definitely U15.

Thank you. Built-in voltmeter is OK, external multimeter shows the same ramp up. Reading from TP6 TP7 is 10.669 V without any ramp up, once I push On/Off button, it goes to 10.669 V.

[pqass]

Assuming all the switches in the back of the unit are in the up (master, local) position and nothing is inserted in the terminals...

See attached (open in new window, zoom in) for a highlighting of part of the voltage feedback loop and this partial simulation (and attached screenshot).

Background before discussion below:
All the op amps are powered via +15V/GND/-12V isolated supply (second transformer winding) with GND attached at the +OUTPUT terminal! Note the purple circles.
Q:Why reference GND to +OUTPUT?
A:So a voltage can drive the MOSFET gate(s) WRT its source(s) without considering the currently set voltage between +OUTPUT and -OUTPUT.
You might think this is a variable negative power supply, ie. +OUTPUT=0V and -OUTPUT=0..-20V

Schematic Error Found?:
The generation of C (label) voltage via U6D (found on the SCR Control schematic) may be incorrectly documented since the same C is used in the CC amp section and screams its purpose is to limit any remote source to max +5V or +10V (not sure which). I've disconnected it from CR26 in the simulation, otherwise, the voltage feedback loop won't work.

How voltage regulation works:
Front-panel voltage setting pot R84 emits a 0..10V voltage at its wiper.  This voltage (at CR26 anode) is divided in half due to R68+R80 to GND.  That voltage (0..5V) is then buffered by U9C (Vout=Vin).  U9C output drives the top of R82/R62 (26.1K/105K or 20/80 ratio) divider.  The bottom of the divider (yellow) is connected to -OUTPUT (via inconsequential 1K R9).  Purple is connected to +OUTPUT/GND. Pink is the mid-point of the R82/R62 (20/80 ratio) divider. U9B compares purple with pink and drives its output such that it brings pink=purple.  Q:How?  A:Assuming front-panel voltage setting is at max (20V and therefore U9C output is +5V) and +OUTPUT to -OUTPUT difference is 0V (just ramping up). Then U9B +Vin > -Vin because the mid-point of the R82/R62 is above GND. U9B output is fully positive (acts like a comparator).  CR21 is reverse biased/not conducting thus allowing Q9 to be shut off via R67, which allows R65 to pull MOSFET gates to +15V (B label) turning them on. The +OUTPUT to -OUTPUT difference starts growing to the point where 20V over R62 and 5V over R82 makes the mid-point 0V. This matches the U9B -Vin (GND/0V).  If -OUTPUT overshoots to -20.1V WRT +OUTPUT then U9B -Vin > +Vin which drives the U9B output negative through CR21 turning on Q9 and dialing-back MOSFET gate drive, decreasing +OUTPUT to -OUTPUT voltage.  An equilibrium is struck such that the U9B output is at a value that will maintain 20V (front-panel setting knob) across R62.

So what could go wrong to gradually grow the set voltage 700mV over a minute?
If the +10V (TP7) stabilizes seconds after power-up (your last post), then test further along: at R84 wiper, at U9C +Vin, at U9C output.  If you find at one point it takes one minute to ramp up, then you've found the culprit. Although I don't see the R86/R68+R80 divider the likely culprit, but the R84 pot does have 10mA flowing through it; still, not likely.
There is virtually no current through the purple, pink, yellow, green traces and therefore, the resistors and switches (S1, S2, S5A, S3) in series should not change value (over a minute) due to heating.
The transistors in-line with the cyan trace will experience heating and thus their outputs can change (if given the same base/gate voltage) but this is compensated by U9B servo action; similarly, any other (minimum) load resistors that may heat up like R60 or leaky SCR CR20 (OVP schematic).

This leads me to think U9 (B or C) may change over that minute period.  Although, you can't tell with B as it may be responding to Q9 and MOSFET heating changes.

[gaminn]

All switches are in their default state (up), no connections at the rear of the unit used.

U9C output (pin 1) is stable immediatelly, no drift here.

Please see U9B output (black is +OUTPUT, red is U9B pin 7) waveform during power on. Using multimeter I can see it starts at 3.094V and it slowly drifts to approx 3.114V.

[pqass]

Yeah, U9B has some instability at the start while the op amp supplies ramp-up (probably).  Then U9B allows MOSFETs to turn on (output going fully positive) while B (label) is initially off for ~70ms (Q10 in the Over Shoot schematic). After C57 charges up then B is energized (MOSFETs now turn on) and U9B immediately goes negative (direction) then recovers to 3.094V when the set voltage is homed-in across R62.

As for the ramp-up from 3.094V to 3.114V (20mV)... that is actually lowering the Ic in Q9 which allows R65 to pull the MOSFET gate up and thus +OUTPUT rises.  It could be compensation* for heating in the MOSFET but if there is no load attached to the output then only R60 (1K) is across (pre-shunt) +OUTPUT and -OUTPUT (just 25mW,5mA@5V set voltage).  That doesn't seem like much.  I'd replace U9 (add a socket too).


* Sorry, I just checked (I'm not an EE): MOSFET Vgs(th) has a negative tempco so a warmer device should require less gate voltage for the same output.  If compensation for heating was happening then Vgs should be going down, which needs Q9 Ic to go up for that to happen, which needs Q9 Vbe to go down (less than 3.094V).  This isn't happening.

[gaminn]

FYI I have another good E3615A power supply. This is the startup waveform of U9B output. U9B output also experiences slow drift of approx. 15 mV within the first minute (3.466V to 3.480V).

[pqass]

FYI I have another good E3615A power supply. This is the startup waveform of U9B output. U9B output also experiences slow drift of approx. 15 mV within the first minute (3.466V to 3.480V).

By "good" you mean the +OUTPUT to -OUTPUT is stable during the first minute after a cold start yet U9B output still experiences a 15mV drift?
I'm at a loss to explain this.

The only other thing I can think of is R62 or R82 changing in the first minute.  But the mW=3.8 and 0.9, respectively (@20V across R62 and @5V across R82).  That's "half a bees dick." --D. Jones

[gaminn]

By "other good power supply" I mean the other unit I have which doesn't have slow ramp up, the output voltage is stable within a second after I turn it on. U9B output in this unit still has similar drift as in the faulty unit with slow ramp up.

[pqass]

By "other good power supply" I mean the other unit I have which doesn't have slow ramp up, the output voltage is stable within a second after I turn it on. U9B output in this unit still has similar drift as in the faulty unit with slow ramp up.

My guess would be that's due to the MOSFETs settling in then.

The additional 5mV drift could be from R62, R82 or C29 changes.

Try replacing R62 with a 100K+4K7 or C29.

[gaminn]

I'm now waiting to receive LF347BN from Mouser, hopefully it is this opamp that is faulty. Meanwhile I disconnected all components that are not "necessary" - CR22, R19, C36, C29, CR28, CR27, C37. I replaced R82 and R62. It didn't change anything. I can't even find some components (R125, C62, C67) on the PCB of my E3615A power supply , so I think the schematic is not 100% accurate.

[gaminn]

Unfortunately, replacing U9 (LF347BN) didn't help. Another idea how to fix this issue?

[pqass]

R125, C62 aren't suppose to be fitted (see Table A-11 (page A-19)).

I think the only other components you didn't touch are R87 and R88. Individually, try shorting their leads together.
They're also not necessary* in the case of an ideal comparator configuration.


* I'm sure someone will disagree here and will tell you why. But worth a try.

[gaminn]

Nothing changed. R87 and R88 replaced with new resistors. When R87 is shorted, then voltmeter is chaotically changing values, CV and CC LEDs are on, ammeter shows approx 0.3A.

To summarize current state: Nothing connected to rear connectors, all rear switches up. U9C pin 7 is immediatelly steady within 100 uV (measured between U9C pin 7 and OUTPUT+). These parts were removed from PCB: CR22, R19, C36, C29, CR28, CR27, C37. These parts were replaced: U9, R87, R88, R82, R62. Still no change, there is significantly long ramp up when the power supply is turned on (after it was off some time).

[gaminn]

I'm a little desperate about the slow ramp up, so I tried this:

1) Checked power supply rails for stability - they are stable immediatelly.
2) Tried remote +sense and -sense - nothing
3) I tried to figure out whether the cause of the slow ramp up is electrical or thermal. Normally I have to wait several tens of minutes to again have full length slow ramp up. I tried to forcibly cool down all components on PCB with big fan - this didn't help, the ramp up was not reset. I also tried to discharge all electrolytical capacitors and small yellow (ceramic?) capacitors. Also, this didn't bring full length ramp up without waiting tens of minutes.

[pqass]

A long shot...
Try reflowing the solder pads of U9, R88, R87, R82, R62, C29, S1, S2, S5A.
Just to be sure that there isn't a problem with the PCB traces (keeping back switches in up (master, local) position),
(a) tacking a temporary wire from R87 directly to +OUTPUT (+front post); purple highlight in my diagram, bypassing S1.
(b) tacking a temporary wire from R88 directly to mid-point of R82/R62 divider; pink highlight in my diagram, bypassing S2.
(c) tacking a temporary wire from R62 (bottom) to -OUTPUT (-front post); yellow highlight in my diagram, bypassing R9.

If the reference is stable (U9C), the comparator (U9B) is working properly, and the CC amp isn't interfering (CC set knob fully CW),
then it can only be the sense elements (R82, R62, C29) or their connection to U9B.

[gaminn]

A long shot...

Thanks again. You can watch this video I recorded: https://youtu.be/Vd9mJ_f1A_Q . The multimeter displays the reference (red U9C.1 and black +OUT). All unnecessary components in CV control circuit disconnected. CC and shutdown circuit also disconnected (R19 and CR22 disconnected). R82, R62, R87, R88 replaced by new resistors. U9 replaced by LF347BN purchased on Mouser. These connections are wired with new temporary wire:
U9C.1 - R82
R82 - R62
R82 - R88
R87 - +OUT
R62 - -OUT
R87 - U9B.6
R88 - U9B.5

Unfortunately still no change...

Voltage on main capacitors C8 and C7 is stable immediatelly (16 - 22 V), so is voltage between U9.4 and U9.11 (U9 power supply - approx 28 V).

[pqass]

Okay, so far we been assuming the problem was somewhere before the U9B inputs.  I think we've exhausted all possibilities (bad U9, +10V ref., change in sense elements R82/R62, interference from CC amp, protection, and bandwidth limiting elements, trace resistance).

However, maybe the transistors following U9B aren't reacting fast enough (cyan highlighting).  The symptom of a gradual ramp up smells like a capacitor charging (I believe are C1 and C3).

Add a parallel 2k2 resistor across R65 to reduce its resistance and see if there is a change in ramp speed.  This shouldn't stress Q9 by much.


FYI: I have an E3611A and is almost identical to yours (after U9B,  although it uses BJTs; not MOSFETs) and I'd managed to cure an issue with it (I don't remember the symptom) by increasing the base current.

[gaminn]

Add a parallel 2k2 resistor across R65 to reduce its resistance and see if there is a change in ramp speed.  This shouldn't stress Q9 by much.

Adding 2k2 accross R65 has no effect, the ramp speed is the same.

If there is an issue with Series pass transistor and current balancing circuit causing the set and real output voltage to differ during the ramp up then I would expect the output of U9B.7 would try to compensate for that (saturated output ?). But U9B.7 never saturates during the ramp up.

FYI If i connect external higher voltage to the power supply output, then U9B.7 saturates (approx -9V) trying to compensate the error it sees.

[pqass]

Adding 2k2 accross R65 has no effect, the ramp speed is the same.

Maybe then the issue is higher up; how B is generated. Or the Shut Down Circuit. Or even the bulk power source (TP1,TP2)!
(a) Lift one leg of R19 to eliminate the Shut Down Circuit from interfering.
(b) Short Q10 collector to emitter to remove the Turn-on Overshoot Circuit from delaying power to the MOSFET gate pull ups.
(c) Confirm how long the bulk source of power takes to ramp up. There could be an issue with the SCRs or bad bulk capacitors C8, C7.
Whichever the above causes a different ramp up, we'll work back from there to identify the offending component.

If there is an issue with Series pass transistor and current balancing circuit causing the set and real output voltage to differ during the ramp up then I would expect the output of U9B.7 would try to compensate for that (saturated output ?). But U9B.7 never saturates during the ramp up.

FYI If i connect external higher voltage to the power supply output, then U9B.7 saturates (approx -9V) trying to compensate the error it sees.

An important distinction needs to be made...
just after turn-on while the output is ramping up, U9B.7 doesn't directly drive Q9.  All U9B.7 actually does during this time is go fully positive (close to +15V rail) which doesn't go past CR21.  So the ramp up drive voltage comes from R67 (shutting down Q9) allowing R65 to pull up the MOSFET gates hard.  Only when the output overshoots the set voltage does U9B.7 go negative to turn on Q9 enough to pull back on the MOSFET gates.  If the B voltage takes a long time to ramp up, or maybe even the bulk source (TP1,TP2), then U9B.7 will just wait at positive until the output overshoots the set voltage.

[gaminn]

Maybe then the issue is higher up; how B is generated. Or the Shut Down Circuit. Or even the bulk power source (TP1,TP2)!
(a) Lift one leg of R19 to eliminate the Shut Down Circuit from interfering.
(b) Short Q10 collector to emitter to remove the Turn-on Overshoot Circuit from delaying power to the MOSFET gate pull ups.
(c) Confirm how long the bulk source of power takes to ramp up. There could be an issue with the SCRs or bad bulk capacitors C8, C7.
Whichever the above causes a different ramp up, we'll work back from there to identify the offending component.

(a) I have already disconnected R19. CR22 as well. No effect.
(b) No effect
(c) With approx 12V output voltage set, C8, C7, R11 voltage is stable immediatelly at 16.9V after power on. With max output voltage set, C8, C7, R11 is stable immediatelly at 22.2V after power on.

[pqass]

I apologize for not responding sooner....

We've gone from the end (+/-OUTPUT) to source (before the MOSFETs) without identifying the problem so I'm sure we've missed it.

Unfortunately, a few hundred mV on the output are probably micro-volts on the MOSFET gates, Q9 base, or U9B.7.  This would make it hard to detect just monitoring with a multimeter (hard to distinguish from noise).

One way to narrow-down the problem is to divide the path into two to determine which half has the issue.  For example, if you disconnect Q9's emitter (or remove Q9 entirely), and you still see a ramp up issue on the output posts, then you know the problem is downstream of Q9; if not, the problem is upstream.  Of course, since you've disconnected the voltage setting parts, you'd expect the output voltage to be close to the source voltage with the difference being does it still have the ramp up of few hundred mV over a minute.  Eg. use two multimeters with one red lead on drain (TP1), another red on source (right side of R1,R2), with blacks on -OUTPUT, to determine if there's narrowing voltage difference (TP1=23.0V and R1 right=22.2V increasing to 22.6V) or stable difference?