-
The family of these late 80s 100W System PSUs share the same base, the 6632A is rated at 20V 5A, the 6633A at 50V 2A and the 6634A at 100V 1A - all have different transformer and a couple of different parts.
They are capable of two quadrant operation, so they can source or sink current and have programmable overcurrent and overvoltage protection on top of CV and CC mode.
Operator Manual and Service Manual with Schematics are awailable from Keysight website.
This HP 6632A is dated from mid 90s and is in reasonable condition.
RIFA and electrolytic caps look good, nevertheless, all these will be recapped to keep this unit healthy for next 25 years.
Fan and Heatsink had some dust build up, but nothing to worry about - after cleaning everything looks like new.
As this unit is in spec over full operating range everything is fine and we can enjoy the beauty of this top class system PSU - if we wear ear protection - the fan screams for some noise improvement
- see PART 2
This unit will be mainly used as a bench PSU, so the rear connectors are not that convenient - good that there was an option for Front Binding Posts, so this should be a simple addition
- see PART 3
LCD has not the best contrast - maybe we could improve that.
Edit 2020-05-23:
Recapped and added sense switch see Part 4
Added picture of PCB bottom to attachments.
-
PART 2: Temperature controlled Fan for <1$
Ingredients:
-TL431LP
-10k resistor
-2k2 resistor
-10k NTC (fully heatshrunk/insulated leads + insulation to heatsink)
-100 ... 330µF 25V (105°C recommended)
-two wires
-23x43mm protoboard
-angle bracket for mounting on board (4mm hole)
-thermal paste
Mix all ingredients together, put thermal paste on the contact point of heatsink and NTC (the heatsink has high potential - insulation between ntc & heatsink required), mount on board, add some thermal paste on top of NTC, solder the wires to VR101 and enjoy the silence
The Fan is now temperature controlled and gets ~4V@25°C to ~12V@80°C.
With shorted outputs and set to 5.1A the units dumps ~220W and the Heatsink temperature rises to ~70°C whilst the fan runs at ~10V and the noise is still acceptable
-
PART 3: Add binding posts to front panel
This addition is not to complicated, the Front Panel is hold in place by three screws, two on each side and one with a nut from the back in the middle.
It has already the cutouts in the plastic frame and with the help of a drillpress the suitable holes are easy to make into aluminium part of front panel.
Put a punch mark at both centers of the cutouts on back, drill a small hole to suit the subland twist drill (in german it is called a step drill) and from the front then drill to the required size to fit your favorite binding post - here SEB 2610F were used.
Solder two short flexible wires with reasonable gauge to the binding posts after adding cable shoes on the other end - soldering to the board connectors is possible too.
Add two thinner wires, put dupont jacks on the other end and plug the sense wires into the connector XT1P8 near output terminal (S- at Pin 8 and S+ at Pin 13, see p. 105 Service Manual).
Et voilá we have convienent access to the outputs including sensing on the front
If there is the need to use the rear terminal with sensing, then it is recommended to add a double pole switch into these sense lines and put it on the back panel (there are already optional cutouts for other options) to switch between front and back panel sensing.
-
Part 2 & 3 now finished.
Any ideas how to improve the LCD contrast? -
Hi MiDi,
Thank you very much for sharing the wealth of information on the 6632A as well as the gorgeous pictures. The 6632A was the first system power supply I ever bought, and I understand why and appreciate each of the modification you made. Noise has been a problem from day one on this unit, I have seen a number of fan control project on the web, and yours is by far the best. The rest put the sensor far from the heat sink, or used a temperature control fan with the sensor on the fan. None of these other methods can accurately and quickly respond to a change in temperature. Your approach is much closer to the one used in the later 664XA series of the Agilent supply (except the Agilent design monitors both the heat sink and ambient temperature), nice work.
I changed my fan to a modern Papst fan with slightly larger cfm, but 5dB quieter. I also added sound absorbing mounting gasket and vibration insulation on the screw. I used a sound meter to measure the noise level, it dropped my 6632A to the same level as my Agilent 6643A. But I am sure your approach would drop the noise even more at low speed.
I do have a couple suggestions:
* I just did a test on my original Papst 612 fan, I just happened to finish servicing it after I had it pulled from the unit. The startup voltage is right around 3.9V, it looks like your lowest voltage is at 4V. As the fan gets dirty and dusty, you may need to be careful with a stalled fan at 4V. The fan is rated from 6-15V, I would suggest setting the minimum voltage to 4.5 V to give you an additional margin of safety with minimal noise increase.
* With regard to the caps, I did a wholesale cap replacement a month ago. See my post on https://www.eevblog.com/forum/testgear/check-your-rifa-caps-in-your-agilent-bench-power-supplies-and-lessons-learned/msg2885786/#msg2885786
Bottom line, you should replace all the RIFA caps, mine had cracks on them. But I found all the old electrolytic caps to be right on spec, and there is really no reason to replace those. And if you do, make sure you use the top shelf caps because otherwise you are probably making it worse.
Nice work, and thanks for sharing. I may try your fan control suggestion someday. -
Hi MiDi,
Thank you very much for sharing the wealth of information on the 6632A as well as the gorgeous pictures. The 6632A was the first system power supply I ever bought, and I understand why and appreciate each of the modification you made. Noise has been a problem from day one on this unit, I have seen a number of fan control project on the web, and yours is by far the best. The rest put the sensor far from the heat sink, or used a temperature control fan with the sensor on the fan. None of these other methods can accurately and quickly respond to a change in temperature. Your approach is much closer to the one used in the later 664XA series of the Agilent supply (except the Agilent design monitors both the heat sink and ambient temperature), nice work.
Very pleasant to give valuable input back to this great communityI changed my fan to a modern Papst fan with slightly larger cfm, but 5dB quieter. I also added sound absorbing mounting gasket and vibration insulation on the screw. I used a sound meter to measure the noise level, it dropped my 6632A to the same level as my Agilent 6643A. But I am sure your approach would drop the noise even more at low speed.
I do have a couple suggestions:
* I just did a test on my original Papst 612 fan, I just happened to finish servicing it after I had it pulled from the unit. The startup voltage is right around 3.9V, it looks like your lowest voltage is at 4V. As the fan gets dirty and dusty, you may need to be careful with a stalled fan at 4V. The fan is rated from 6-15V, I would suggest setting the minimum voltage to 4.5 V to give you an additional margin of safety with minimal noise increase.
* With regard to the caps, I did a wholesale cap replacement a month ago. See my post on https://www.eevblog.com/forum/testgear/check-your-rifa-caps-in-your-agilent-bench-power-supplies-and-lessons-learned/msg2885786/#msg2885786
Bottom line, you should replace all the RIFA caps, mine had cracks on them. But I found all the old electrolytic caps to be right on spec, and there is really no reason to replace those. And if you do, make sure you use the top shelf caps because otherwise you are probably making it worse.
Nice work, and thanks for sharing. I may try your fan control suggestion someday.
The voltage at the fan is ~3.3V@21°C, with lower temperatures even lower - as you mentioned this is not recommended.
The main purpose of the 330µF cap is to blow the fan free at full speed on power up and then settle to regulation voltage - as a side effect you get audible feedback that the fan is working
Fans can have a lower working voltage than startup voltage due to break free moment.
The regulation voltage with weared/dusted fan may still be on the low side and the fan may stall, but the temperature of the HS will rise quickly and it should turn on at some point - if not, it is time to clean/exchange it and as a last resort there is the overtemperature protection kicking in and signalling at the LCD - so for these units this solution seems to be suitable.
Even the best electrolytic caps wear out in decades and without any indication may fail in the near future, so replacement is suggested after some decades.
Sure, the caps will be replaced with suitable first class low ESR, 105°C and longest lifetime.
It cannot be repeated enough:
-
Hei MiDi,
thank you very much for sharing this nice project.
Regarding the LCD backlight: There are some mods existing for the HP 3457A which can be used here as well, perhaps:
https://www.eevblog.com/forum/projects/hp-3457a-lcd-backlight-mod-(also-for-3478a)/
HTH
And keep us updated.
-
Nice catch, MiDi. Thanks for sharing your mods and all the great pics!
-
Part 4: Recapping & Sense Switch
Recapping is now finished, 7 electrolytic and 3 mains filter caps were replaced.
The RIFA X2 has an odd pin spacing of 20mm, so the pins of the WIMA X2 with 22.5mm had to be bent to fit.
I like the seviceability, it is not to complicated to get the pcb out - if you know how
Remove 3 screws from pcb, 2 from HPIB-Connector, 4 from metal bar, 1 from earth wire, slide pcb towards the front and carefully lift front side over front panel to get it out.
Overview of replaced parts:
Voltages in () are the values present on caps, without are taken from service manual BOM, Dimensions are measured by hand
Easy C&P to your favorite spreadsheet appDesignator Value Voltage Dimensions dxh Pin spacing Replaced by Remarks C26 6.800µF 16V 22x26mm 10mm LGY1C682MELZ25 C110 470µF 50V (35V) 10x20mm 5mm UHE1H471MPD C105 1.000µF 35V (50V) 15x26mm 7.5mm UHW1H102MHD6 C134 2.200µF 35V (50V) 22x26mm 10mm LGY1H222MELZ25 (alternative: 381LX222M050H012) C101, C103 7.400µF 63V 35x52mm 10mm LGY1J822MELC45 8.200µF C102 33.000µF 16V 35x52mm 10mm LGY1E333MELC50 C131 X2 220nF 250V 20mm MKX2AW32205B00KSSD 22.5mm C132,C133 Y2 2.200pF 250V 10mm MKY22W12203D00KSSD
As proposed in Part 3 a Sense switch Front/Back was added.
The hidden hole above back output connector was drilled to fit a DPDT switch.
Red & black Dupont jumper wire were cut to size and used to connect to XT1P8 for sensing and soldered to the switch - the sense wires to front post were reused.
Now front and back terminals are usable with proper sensing and it allows for remote sensing from back sensing terminals for front too.
-
The RIFA’s unique pitch is definitely a challenge for cap upgrade, I did not want to bend pins and I ended up with new RIFA caps in a lot of my refresh. I figure I can manage a 10 - 15 year refresh cycle.
Did you use a LCR meter to check the old caps you pulled? I found all the electrolytic caps to be exactly in spec after a quarter century of service, remarkable! With the exception that the RIFA caps were all cracked. I suspect the fan running at constant top speed may contribute to a low running temp throughout the case and subsequently minimizing cap deterioration.
As a result, I found no improvement in output ripple. Did you see change change in output characteristics?
-
Yeah, on an afterthought the sharp bending on X2 may reduce lifetime.
If I had to replace another one I would probably add a spacer and bend the pins slight at an angle - each pin needs only 1mm offset.
All caps measured fine, except X2 had ~4Ohms ESR, new one <10mOhm.
I would not expect significant ripple change with minor change in ESR.
Edit:
For mains capacitors that are mostly unfused & powered when unit is switched off and should be designed to be failsafe, I would only put first class caps in.
Unsure how these RIFA caps can be sold anymore - obviously unchanged
Even if the electrolytics measure fine, I would not trust them after decades.
-
Hi Guys.
I didnt want to start another topic relating 6633A as there are already many topics started.
I have 6633A with a faulty eeprom chip. I have purchased another 256Bit eeprom and was wondering if any one out there has the eeprom content for model 6633A. I saw a procedure provided by HP when replacing eeprom to run initial/program. But without a cable and software I am not sure if I will be able to do so.
I would appreciate any help
-
Hi!
Does anyone have the original hardcopy of HP 663xA service manual?
P.S. Or scanned copy in color. -
in the first post of this thread ?
http://literature.cdn.keysight.com/litweb/pdf/5957-6365.pdf
Quality is so so ... better than nothing ?
Check "Fleabay" for originals one ?? you may get lucky -
Hi Guys.
I didnt want to start another topic relating 6633A as there are already many topics started.
I have 6633A with a faulty eeprom chip. I have purchased another 256Bit eeprom and was wondering if any one out there has the eeprom content for model 6633A. I saw a procedure provided by HP when replacing eeprom to run initial/program. But without a cable and software I am not sure if I will be able to do so.
I would appreciate any help
Did you get it working?
If not, throw me a PM if you still need the content of eeprom... -
I'm specifically interested in the 4-output 6624A model.
Question: After each output has been configured for VSET and ISET, is there a way (front panel or via HPIB) to turn on/off more than one output at the same time?
The 662x series of power-supplies lack the hardware to do that, the outputs are only commanded via an internal serial connection from the hpib/display/keyboard processor.
You can send all four OUT commands separated by semicolon, and that is likely going to be as fast as it can go.
I can't remember if SAV/RCL saves the output on/off states, if it does, then you could also use that.
(But see also the DCPON command)
However, I want to point out that "at the same time" is under-specified, even for a supply which has HW-trigger facilities:
Do you want the four rails to turn on at the same instant, from where they ramp to their VSET voltage at whatever speed they do, depending on voltage, load, impedance etc ?
Or do you want them to reach VSET +/- some margin at the same time ?
If the latter case one would normally run the outputs through a four-pole relay.
-
in the first post of this thread ?
http://literature.cdn.keysight.com/litweb/pdf/5957-6365.pdf
Quality is so so ... better than nothing ?
Check "Fleabay" for originals one ?? you may get lucky
Thank you, I know about the official scans on the Keysight website. But the quality is not very good, especially the component placement image.
So I'm looking for a good scan of this page, at least in grayscale rather than black and white. -
Hi MiDi! Thank you for sharing!
I've managed to repair my HP 6633A (bought broken).
And I have implemented most of your modifications.
Today I wanted to finish this project, but something went wrong.
I made an identical fan speed control circuit, using 220uF/25V cap., 2k2 and 10k resistors, 10k NTC thermistor (Vishay NTCLE100E3) and TL431CP v.ref.
I checked the voltages without mounting the thermistor on the radiator, heating it with a hair dryer. And everything worked fine.
Finally, I secured the board and leaned the thermistor against the heatsink (I added a small amount of ARCTIC MX-4 thermal paste to the area of contact).
But as soon as I power on the power supply, I heard a sharp pop and saw a small amount of magic smoke escaped from the thermistor.
P.S. Fortunately, only the DIY board was damaged.
-
It looks like the thermistor needs additional insulation between its case and the heatsink. I am thinking of using a thermal pad.
In my case (for 6633A), there is a potential of ~85 volts DC between the the zener diode and the radiator, apparently this is enough to breakdown the insulation of the thermistor. -
Glad that the damage is minor
Seems your ntc was damaged before, but good point, the ntc should be fully heatshrunk / insulated against heatsink.
Will change it in my unit and add a notice to my post.
-
Seems your ntc was damaged before
You may be right, but before soldering, I examined the thermistor under a 10x magnifying glass and found no visual defects. Even now, with the exception of the black dot of burnout, it looks normal. And it even works, although the resistance has changed downward.
Perhaps this is a defect in my particular thermistor, but I think it's better to isolate it. -
Maybe it is only weird light reflection that looks like a scratch...
-
This is not a scratch, but a burnout trace in the thermistor, going further under the coating. As I said, initially I did not find any visual defects, but this does not mean that the thermistor could not be defective.
P.S. I scraped the coating off the damaged thermistor. As you can see, on both sides there are places where the contacts come out on the surface and are covered only with a very thin layer of colored varnish. Because both sides look the same, I think this is not a defect.