23rd June 2019 20th June 2019 By Bzzz In project 6640a6641a6642a6643a6644a6645aagilentapparent powerball bearingbotchcapacitorchemiconebm-papstesrfanhpkeysightmagic smokenichiconpanasonicpapstparts listpcbpower measurementreactive powerreal powerreplacementrifax class capacitory class capacitor

Post #100 HP 6644A refurbishment (#P14)

Not that anybody cares, but this is actually post number 100 already. I prepared something else that failed (to be reported!), so this one has to do instead

My “lab” power supply situation is ageing and often no longer up to the task, think of LEDs north of 35V or anything that takes more than one amp at voltages that cannot be provided with common computer supplies (3.3V, 5V, 12V, 19/20V) easily. So this is part one of the refresh.

I bought a HP 6644A power supply around six weeks ago off eBay, at 245€ including shipping and rack mounts via best offer. The unit was shown with real photos but not dolled up, it was dusty, had snipped off wires at the output terminals and had a shortcoming on the front presented as such. Minimal written description. The company selling this has lots of other random stuff, probably some liquidation trading or similar business models. By chance they had another 6644A in much better condition, newer unit (Agilent instead of HP branding, I don’t think there’s a Keysight version of it, as it’s discontinued for RoHS crap reasons?), for about double the money. Other offers are currently in the high 300s and up, strongly influenced by age.

Of course I went with the cheapo. I asked for more detail pre purchase and I was told the fan is a bit loud – I can live with that. Everything else should be fine. This is what arrived – heavily de-dusted and cleaned inside and out:

Crinkles in the detached front sticker were cause by me, and I guess this is also the reason nobody bought this earlier: In the seller’s photos, it looked like this was damage to metalwork, i.e. this thing has seen rapid deceleration on the floor. However, knowing these power supplies from videos like EEVblog #667 (Dave skipped #666! Bastard!), one could estimate that this was just torn plastic film from rough removal of a cal sticker or the like, there’s still superglue residue visible on that photo. Video for reference, Dave pokes holes into that at around 12 minutes, so clearly no metalwork involved here:

Fixing that up and also doing the front terminal conversion is a thing for a follow-up post, as I’m slowed down in developing the metal parts due to slow China shipping of important bits.

However, this thing has a couple more things to fix, and they are more pressing than front terminals. When cleaning this, I found …suspicious capacitors. X and Y class caps in yellow-ish clear plastic, RIFA branded. EEVblog #1182 or this one made replacing these ASAP an easy decision:

Also, the fan indeed is a bit loud, not totally necessary to replace right now, but when you have the unit open for extensive modifications anyway, why not do that now.

Let’s go!

Capacitor refresh:

When I first opened this for cleaning AFTER powering on for a bit, I discovered a bunch of these evil bastards right away. Three of them looked at me:

And when I removed the GPIB daughter board, even more of those meanies were present:

As I said, these just HAVE to go.

C452 – the big one. Part no. 0160-4413, 600nF (10%) 250V X2, 29mm (limited) x 15mm, H22mm (limited). Only replacement would be the Kemet PME271M660KR30 which is an exact fit (slightly higher voltage), 400V/µs. Others only make 680nF caps at 22.5mm or 27.5mm spacing, I haven’t seen a 25mm one. The Wima MP3-X2 series is higher at 26mm (250V/µs), the Würth WCAP-FTX2 and WCAP-FTXX series have much smaller dimensions including 15mm or 20mm spacing, and are rated 230V/µs. Reichelt order number “PME271M 600N 275”, 2,25€)

C645, C692, C693: Part no. 0160-4355, 10n (10%) Y2, 18mm (limited) x 5mm (limited), spacing 15mm. Wima MPY20W2100FC00MSSD (“MP3-Y2 10N” at Reichelt, 0,46€). Epcos B32022A3103M000 would be suitable as well (lower grade PP caps like Wima MKP-Y2)

C450, C451: Part no. 0160-4183, 1n0 (20%), 250V Y2, 13.5mm (limited) x 4mm (limited), spacing 10mm. Replacement: Wima MPY20W1100FA00MSSD (“MP3-Y2 1,0N” at Reichelt, 0,30€). Epcos B32021A3102M289 would be suitable as well (again, lower grade PP caps)

Here’s the old crap – the 600nF and also the 10n look undamaged, however, both 1n show cracks on both sides. Not immediately blowing up in your face, but when it comes to Y/X class mains capacitors: Better safe than sorry. I don’t want a Dave moment when I cannot even remove power quickly. I left a note on the replacement Kemet/RIFA so that the future owner doesn’t feel the need to replace these again.

Speaking of cap replacement: Here are some electrolytics that could use a refresh after 20 years of service as well…:

C695: Part no. 0180-4439 (different for 654x/664x models!), 68µF / 100V, Dia 19mm, was a Cornell Dubilier 105°C 300680M100GE2 type. I replaced it with Panasonic FCA series EEU-FC2A680L (Conrad 1481042, 0,85€), much smaller diameter/spacing – totally unsure if that is a good idea. Will keep the original part, won’t buy a 30€ Cornell cap as a replacement. Dude, that’s insane pricing for a rolled up strip of paper in a metal can.

C602, C603: Part no. 0180-3298, 2200µF / 50V, Dia 22mm, was a 85°C Chemicon type. Replace with Panasonic FCA series EEU-FC1H222 (Reichelt “RAD FC 2.200/50”, 1,80€) or Chemicon LXZ series ELXZ500ELL222MM35S, 105°C low ESR each

C405: Part no. 0180-4463 (again different for 654x/664x models!) 2700µF / 150V, Dia 35mm, H66mm (limited), was a 85°C Chemicon type. Best match would be a Cornell Dubilier 380LQ 380LQ272M200A052 (RS 871-4094 at a whopping 10,26€) – thanks, I’ll pass for now.

C416: Part no. 0180-3963, 6800µF (says 17000µF in the service manual)/ 16V, Dia 22mm, was a 85°C Nichicon LK(M) type. Best match probably a Chemicon KY series EKY-160ELL682ML40S (RS 811-8685 at 2,98€ for 2pcs)

C122 (connectivity daughterboard): Part no. 0180-3325 (surface mount version), 10000µF / 25V, Dia 30mm (limited), H32mm (limited), was a 85°C Chemicon type. Replace with Vishay 158 (snap-in) series MAL215836103E3 (RS 877-3703 at 4,99€). Pin spacing is 12.5mm, so larger than this Vishay one at 10mm.

I only changed the two 2200µF and the 68µF caps due to availability at the moment. The unit did work before and it works after the change, so I don’t see immediate need for an entire cap swap.

The 2200µF C602/603 fared pretty well, their ESR went up to something around 0.3Ω. Their replacement is now an order of magnitude better (again). Replacement is pretty easy in terms of pin spacing, see the various footprints for both of those caps. You can basically plug in everything that fits the case, you just need a decent iron to heat up those pads, they suck away a lot of heat due to internal layers.

The 68µF from Cornell is basically good as new, not saying this is expected behaviour, but for a cap that costs 30€ nowadays I demand a little more service life than for Wan Hung Lo “that’ll do” products. Have a look of the size difference of old and new and guess why these last such a long time:

For now the unit runs totally smooth, but if there were regulation instabilities or increased output noise, I’d go for the 2700µF capacitor that sits in the center of the transformer – linear regulator stage. The others feed the buttloads of logic, so given their oversized capacities, that’ll probably even work with leaky caps for quite some time.

Fan refresh:
Of course the fan has never been changed before, so for a mechanical part, it’s about time. It’s a quality brand one, a Papst (not ebm-papst, these folks united in 2003), model number 612. Too bad the 612 is now a series with dozens of models, so I asked customer support for advice.

The request got routed to HDS GmbH, a company that sells and services ebm-papst products. The general manager replied within 40 minutes…

Turns out the old 612 has been moved to the 622 series, specifically the 622N. “6” is the case size (60mm width), the last “2” is the voltage rating (2 for 12V, 4 for 24V, 8 for 48V), not sure about the center “2”.

I don’t have much data on the old 612, the little sticker only says 6-15V (12V nominal), DC 2.5W (= 0.21A). Built in 08/1998 and still stating “West-Germany” in the mould
The new 612N says 8-15V, at 1.9W or 0.16A, 6100 rpm, -20 to +70°C, 40m³/h at 35dB(A). There’s also a pressure-flow graph on the product page, giving this model a typical rating of 30Pa, up to 75Pa at minimum airflow. The HP heat sinks aren’t that restrictive, so 30Pa should be in the ballpark.

How did I get a replacement? Well, these fans are usually around 30€, but luckily I got some old stock (unused, week 29/2009) from eBay for below 10€ plus shipping. If you aren’t that lucky but can get some of these, I think the fan regulation should be capable of working with them: 612NN, 612NH, 632NU, 622H. All of them are about the same 30-35€ new except the NU (which is double that?!), but of course would drop significantly from private sellers, especially if old stock or slightly used. They all exceed the specs of the 622N but draw less power than the original 612, which is always a hard limit for me when replacing the fans. Note that the 6644A has fan speed regulation with a steady minimum airflow, these fans never turn off entirely when the power supply is on.

Anyway, the German post tried to sabotage but failed:

Shorten the leads, crimp on two new contacts, re-use the old (standard!) fan connector, and we’re good.

While the load regulation is fine due to temperature of the heat sinks being the main variable in controlling fan speed, I think having an RPM signal from the fan to get idle speeds (and therefore noise) down would have helped tremendously. Wasn’t standard in the 1990s, so I just have to live with that.

Another option to deal with noisy fans would be to replace the bearings. Contrary to the cheap computer stuff nowadays, these high power fans have replaceable ball bearings (plural). I cannot tell you how to get them out without damaging the top part that holds everything in place (and I don’t even know my bearing vocabulary to properly name that part), but someone who has successfully serviced a fan before probably would know. This very unit has two 8mm diameter bearings, 4mm height, 2.9 to 3.0mm inner diameter, other/newer models obviously could use something different.

Now that I think about it, the service dude didn’t mention this. Well, would have been cheaper to replace it compared to a brand new fan, but I doubt two quality bearings in retail would have been cheaper than old stock from a private seller. Still, if anyone out there has the same problem: Replacement is possible if you know what you’re doing and the electric side of the fan is still top notch.

Also note the botch (taken from one of the first images), there’s a bit of PCB material between fan and heat sink, acting as an air guide. That’s a bit cheap, HP! I replaced this with a cut piece of antistatic ESD matting which should be okay with the heat, but it’s also rubbery and therefore dampening to fan-> heat sink -> chassis noise transfer and amplification.

Other (miscellaneous) thoughts:

This post also applies to the other versions of the 6640A series, which consists of the 6641A, 6642A, 6643A, 6644A and 6645A models – same wattage for all of them, but different volt/amp settings, with different hardware in the key areas. Not sure if you could switch between models after replacing capacitors (and transistors?) to ones that can handle the entire range by modifying jumpers on the main PCB, though. I think this was discussed in several electronics forums but I don’t know if there’s a working solution.

Wiring of the front panel is nice, but not perfect. There’s the AC input cable that is trapped in the metalwork, as the main board connector doesn’t go through, and unplugging all four individual terminals from the front switch is certainly more troublesome than just unplugging one keyed connector.

Upgrading the display with a backlight probably isn’t an easy thing to do, and without a replacement on hand, I’m not going to try. The front PCB is fixed in place similar to the main PCB (see next), but with plastic instead of metal hooks that slide out and some plastic hooks that are bent (?). I’m too scared to break 25 year old plastic that has lost parts of its plasticizers just to get in and see that any illumination angle doesn’t work all that well. There are a few spots where one could get light into it from the top, but that doesn’t illuminate evenly at all. So sadly this display will stay on the dark side.

This great power supply is a terrible power supply itself. It is in fact SO bad that I finally bought an upgrade to my old crappy power meter, the ELV “Energy Master Basic 2” which displays real power, reactive power and apparent power in addition to having much better resolution on the lower end. To my surprise, it confirmed the measurements of the hard switched OFF 6644A: 12VA! There’s enough circuitry before the hardware switch to create a 12VA load at some ridiculous 1%ish power factor, so purely capacitive plus losses – remember the X2 and Y2 caps? Yup. ALWAYS ON.

Furthermore, idle power draw is in the order of 30W at some really shitty power factor of <0.4 as well. No wonder the fan control circuit never powers down, there's a baseline 30W to get rid of at any given time this little monster is turned on. I'm aware that there's losses in good (non-switched) power supplies, but this was a bit above my expectations. Upon RTFM'ing, I stumbled across the wee spec line at the very end: "Input power: 480 VA, 400 W at full load; 60 W at no load". Looks like it exceeds idle specs by a factor of two

And here’s the outlook to the follow-up posts: As I knew from Dave’s videos, there’s slots in the front housing that would basically take front terminals without problems. No idea why HP/Agilent never did offer that, but it’s certainly there.

Of course I’m going to gild the lilly here: There’s space for six terminals pre-moulded, but I’m going to drill in a 7th above that. I’m always annoyed by power supplies that only offer one terminal, so my current setup in a 19″ blank plate has 2+2 terminals plus one for grounding. You know, most standard cables are stackable, but many crocodile clips and more specialized cables aren’t. Having a similar output situation was indeed one of the requirements for my new gear, and with all that space on the right of the device front, I can certainly do that. So there’ll be two (+) jacks, two (-) ones and also (+) and (-) sense inputs, as having the option to do remote sensing is valuable to me. I’ll have standard 19mm spacing so standard short circuit adapters will fit in there, connecting front sense terminals to the front output per default, but with the advantage of being removable in an instant (sensing has to be set up in the back of the unit if I’m not mistaken, so I need to be in remote sensing mode all the time). These jumpers for proper wattages (™) are available in closed “]” configuration, as well as “Y” models where you wouldn’t lose the output. And of course, as there’s no other grounding link on the entire front – a grounding post, on top of all of them. 3D printed mockup as shown above – final plate will be made from aluminium. I’m glad the design tool allows for exports that can be converted to 3D printable files, so that I can be sure the final parts will fit just perfectly.

Well, that’s it for blog post #100 – this power supply (and the auxiliary stuff) isn’t done yet, so stay tuned for the final build. I’d say this could be the case in one month, as all standard parts are ordered and most design files are ready to order or close to completion, it just needs a suitable time slot between work, family, taxes and a commercial side project to get it done.

See ya!