Another Fluke 5440B/AF Repair

[essele]

I'm a bit of a muppet ... the relay that I destroyed I put back together as best I can and actually installed it in the unit (sample string K9) because I wanted the "off state" to be properly reflected.

Firstly the off-state of K9 is basically unconnected to anything, so having it in there was pointless ... and more importantly my reassembly of it actually caused some of the contacts to make when they shouldn't have, therefore messing up the 2M resistor string and therefore impacting the feedback look when on 275V and 1100V ranges. With that relay removed all output looks good.

The analog test now fails at M33 presumably (hopefully) because of the missing K9 relay.

I'm assuming it got past M33 last time because the broken relay was incorrectly making contact.

[e61_phil]

Hi,

very interesting thread!

My 5440B has a quite high TC ~0.1ppm/K. That is the specified value and I'm sure it was better some years ago, but in that region it is hard to find the problem. Maybe also a relay problem?

Another question I always had: Does anyone know if there is something harmful in the foam used in the 5440B for thermal insulation? Something like hazardous flame retardents or asbestos?

Btw: All the relays in my 5450A got quite high in resistance (some up to 100R), but I was able to cure them by cleaning up the contacts with a business card and some IPA. They are still fine since at least 4 years.

Best regards
Philipp

[essele]

So first attempt with a 3 x DPDT relay PCB seems to be successful.

I used some round headers that are thinner on one side and they go into sockets reasonably well. The socket pitch isn't actually 0.1" but it's close enough. I've also cut off the top of the pins so that they don't stick through too far and then covered them in solder mask.

This 6-Pole PCB is jumperable to cope with either side being positive (thanks for the heads up wn1fju) although they all seem to be the same on the sample string PCA.)

This PCB can also be used for the 4PDT versions which fit in the 6-Pole socket just by omitting the middle relay, and I have two other PCB's ready for the 4PDT and DPDT versions.

I am slightly worried about isolation although I've used a 4-Layer board to make sure I have as much space as possible ... and even with this version the smallest clearance is quite a bit bigger than the actual gap between the old relay contacts. There are a few changes I'm going to make though ... the holes for the pins could be slightly smaller and have a smaller pad and then I'll go from 50mil tracks to 40mil to increase clearance at the cost of a few milliohms of extra resistance.

I was also hoping to use the existing relays springs to keep these in place along with a 3d printed "cover" (hence the cuts in the ends) however that doesn't really work so I think I'll just resort to cable-ties around the socket.

Interested in any feedback or suggestions before I cut the next version ... also very happy to share the designs and/or get more boards made.

Back to the repair .... well all was going well, after some warm up it passed all the analog test.

The HV tests were going well, but then there was an arcing sound (I think at the 1100V test) and now I'm back to no sensible output on any range again! This is the gift that keeps giving (but I'm enjoying it!)

Code: [Select]

JOHN FLUKE MFG. CO., INC.                          5440  ANALOG DIAGNOSTICS

---------------------------------------------------------------------------

     ADM1: +.000000V
     ADM2: -.001698V
     ADM3: +2.52722V
     M41: -273.483V
     M42: +273.483V
     M43: +549.544V
     M44: +870.112V
     M45: +115.711V

END OF ANALOG DIAGNOSTICS

[MegaVolt]

Beautiful board. Do you plan to share the files for making it?

[essele]

Beautiful board. Do you plan to share the files for making it?

Thanks MegaVolt -- Absolutely ... I'm just going to make the clearance adjustments I mentioned, then very happy to share all three sizes.

I've just confirmed that the arcing isn't actually caused by my board ... I've found that U7 on the preamp board has blown again, and replacing it (the ua714 does seem to work in this one) I can reproduce some arcing at 500V and turn it off again without re-destroying it.

It sounds like its coming from a relay on the output board, although I'm not quite sure how a fault there would lead to HV at U7 without several other things being destroyed along the way .. but this has happened twice now. Not sure whether to revisit all the relays again or just focus on getting the boards right and everything replaced.

[MegaVolt]

Interesting. Is it possible to use contacts from the original relays for the output contacts. They can be taken out of faulty relays and soldered onto the board if you change the holes on the board a bit.

[essele]

I thought about taking the contacts, but each one is a single piece that runs all the way to the movable contact, so you could do it, but it would take some cutting!

It also means destroying the relays completely, so I gave up on that idea.

[essele]

I've gone through the PCB's again and have achieved a 40mil separation at all points, and generally I've managed with 50mil tracks with just a couple of small areas where I've needed to shrink slightly to get through a gap.

40mil (1mm) separation seems to match exactly the gap on the existing relay socket where it's soldered onto the PCB so anything more is unlikely to be beneficial.

However ... the pins I'm using are 0.5mm diameter for the side that goes into the socket, on the other side they are 0.65mm diameter but have a thicker support which is 1.8mm in diameter, on a 2.54mm pitch that gives only a 0.74mm clearance which is only 30mils. I'm going to find an insulation tester to do some tests, however it doesn't feel right.



Normal square pin headers don't have the extra support and so have plenty of clearance, however the pins are 0.65mm thick and just won't go into the socket.

So ... I need to find some 0.5mm thick (or diameter) headers that don't have extra meta supporting bits!

[dietert1]

For our oximeter modules we use 0.5 mm pins that come in a black plastic strip similar to the one you show. But the pins are flat /like wire, so we pull the plastic carrier after soldering through-hole. Looks like a hybrid.

http://cadt.de/spo/Views_3D.jpg

Regards, DIeter

[essele]

Hi Dieter ... do you have any details? that sounds/looks like it might be perfect.

I've also found these https://www.farnell.com/datasheets/2340109.pdf which might work if cut off, but they're quite expensive.

[dietert1]

The 40x strip had the product code 952-08-040-1-00 and we got them from W+P PRODUCTS GmbH in 32257 Buende/Germany. Certainly made in China..

Regards, Dieter

[essele]

Thanks Dieter ... I've found a few length variations on Conrad, so I've ordered a selection to see how they work out.

[essele]

Here are the KiCad files...

There are three variations....

6 Pole ... this can be used for full 6-Pole (602953) as well as the 4-Pole-in-6P-Socket variant (519405)
4 Pole ... this only seems to be used on my A7 Preamp PCA despite not being called out on in the service manual??
2 Pole ... this is a normal 2P version (514240) -- (EDIT: see post below, may not be suitable)

The 2 Pole variant is just double sided, and I've moved the power selection jumpers to the other side so they are clear of the socket.

CAVEATS: I have not size tested the 4 Pole or 2 Pole variants yet, and I haven't insulation tested anything (but do intend to)

In my 5440B I need 20 of the 6-Pole board (15x6P and 5x4P), 1 of the 4-Pole board, and 4 of the 2-Pole. If my numbers are right, that's 446 pins and 61 relays!

I'm just about to place an order at JLCPCB and follow it up with some HFD27/005-S from LCSC since they seem to be good value and specs look comparable to the data I can find about the American Zettler or Potter and Brumfield relays.

EDIT: updated with the correct relays after I ordered the wrong ones. Also worth noting that these relays are not polarised, so no need to have the polarity switching jumpers, but will leave them in to allow more flexibility.

[essele]

My insulation tester just arrived ... I've been waiting for a good excuse to get one for a while 

A very rudimentary test of the original header pins I was using show's that they are *probably* ok ... on the first pass through the voltage ranges I didn't see any arcing and resistance off the scale until 5kV. At 5kV there was considerable arcing which started at the narrow point but then travelled along the length of the pins (and it didn't start at my shoddy solder peak!)

Interestingly on a second run through there was also some arcing at 2.5kV, much less, and short lived. Not sure if this is to do with surface damage or something in the air post the previous 5k test.

The same test with some standard 0.1" 0.65mm square headers showed considerable arcing at 5kV but I was unable to get anything at 2.5kV with resistance off the scale consistently.

The 0.5mm pins will have a greater clearance still, so they are unlikely to be the weak point .... once the new boards arrive I will do an insulation test to make sure that they don't negatively impact the relay dielectric strength.

BTW ... I have dishwashered the first couple of boards, and apart from having to spend a fair amount of time getting the water out of the relay sockets, they look 100 times better. It's just a shame I can't fit the whole chassis in there!

[essele]

I've just been staring at the output board after the cleaning, re-capping, and sorting out a couple of rusted screws in the shields ... and I've noticed that the sockets for K9 and K10 are very close together and because of the orientation of the relay on the 2-Pole board I don't think it's going to fit properly!

Also K5 (6-Pole) is right up against one of the shields and the current design has an overhang on that side which I also don't think will fit ... and even if it did I suspect the relay pins would be too close to the shield.

I've done another 2P board design with the relay oriented differently and the power selection jumpers removed ... it's much simpler, and only has a very minor overhang at the end rather than the side, so this should be no problem on this PCA.

K5 is going to be a pain ... anyway, I'll work on that now, just wanted to highlight.

[essele]

New PCB designs attached ... there was another problem in addition to the above, the 4-Pole in a 6-Pole socket is actually not as simple as just removing a relay as the pin configuration is different to the normal 4-Pole design. So quite a few errors overall, thankfully JLCPCB let me ditch the boards that were in progress and refund the shipping (which is the bulk of it!), which means not an overly expensive mistake.

So the new designs are as follows:

1. 6-Pole -- this is as before, just with power choices removed and some minor tidying.
2. 6-Pole-K5 -- this is a new design specifically for K5 on the output board.
3. 4-Pole-in-6 -- this is a new design for the 4-in-6 configuration. This is just double sided.
4. 2-Pole -- this is a new 2-Pole design which avoids the "too close together" problem.

I've also left the old 2-Pole and the 4-Pole designs in the zip file but I don't intend to use them. For the single actual 4-Pole relay I'm just going to use the 6P board without the last relay fitted, there is plenty of space.

I've checked all the designs to ensure I maintain 40mil clearance on all non-power traces, and they should all have at least 20mil clearance to the edge of the board.

Fingers crossed these ones are right ... waiting game now ... 8 days or so.

BTW -- I will have loads of spares of the 6-Pole board, I need 15 but there was very little difference in price between 20 and 50, so if they work ok I'm happy to send some out at postage cost if anyone wants any.

[essele]

So the various size pin headers have arrived from Conrad (thanks for the pointer Dieter) ... as always not a clear cut decision.

In all cases the plastic part of the header is 1.7mm thick with the various size pins protruding different amounts from either side. Unfortunately none of them are "perfect" in an out-of-the-box way.

The -08 variant protrudes 2.3mm and 2.5mm and they really don't feel like they get much bite at all into the socket when the plastic bit is still in place, they are fine with the plastic removed, however this then brings the solder joints for the relays very close to the metal receptacles in the socket which I wasn't comfortable with, so I really need to keep the plastic bit in place.

The -09 variant protrudes 3.3mm and 3.4mm. The 3.4mm side does a reasonable job of engaging with the socket, but it just doesn't feel quite far enough.

The -12 variant protrudes 3.3mm and 6.3mm. These give a really secure mating with the socket, however I don't know what the mechanical arrangement is internally and I'm worried that the excessive length may bring pins too close to other bits of the socket internally.

The -13 and -15 are even longer so not viable.

So I've opted for a modified approach to the -09 variant. I've printed a small jig that allows me to setup and hold all the pins before starting soldering, and it allows the pins to sit 1.3mm further down than they normally would so there is a 1.3mm gap between the plastic and the PCB. Once they are soldered the plastic bit can be pushed up to effectively extend the pins by 1.3mm bringing them to pretty much the same length as the original relay pins. This then gives a really nice mating with the socket and has the added advantage that there is less pin exposed on the other side of the PCB to interfere with the relays.

The relays also arrived today and look good. Just waiting for the PCB's now ... they're due to be finished tomorrow, so hopefully I'll have them mid next week.

[essele]

Some progress on the front panel keyboard ... I managed to get my hands on a Fluke Y1700 keyboard which uses the same (almost) switches as the 5440B front panel.

So the keyboard PCA is now cleaned up (as best I can) and I've replaced the two broken switches, I've cleaned the key caps that I have and I have a masterplan to re-create the missing ones (just waiting for a few things to arrive.)

The keyboard switches are Datanetics DC-60 series, you can see some of the history here: https://deskthority.net/wiki/Datanetics_DC-60_series

There is a very slight variation between the switches from the Y1700 and the original ones in the 5440B, both seem to the right hand variant from the "evolution" pictures in the above URL, but the 5440B ones have a much shorter tab, the long tab gets in the way slightly, but it's very easily removed by 10 seconds with a file.

I have quite a few spares now, along with all of the Y1700 key caps, chassis, and PCB if anyone has a need. I'd like to recoup a little of the cost from the switches if possible, but the other Y1700 bits I'm happy to see go to a good home for postage costs. I'll post on the selling forum when I get a chance.

The PCB's have also arrived .. super impressed with JLCPBC turnaround and delivery times, the order included 4-layer PCBs, was placed on the evening of 18th and arrived this morning.

[essele]

Ok, all relays replaced and it was looking really good. The arcing stopped once I'd done the pre-amp, so it was presumably a problem with a relay on there.

The output all looks good, and I've tested up to 900V so far.

The analog tests fail on a DAC problem which only started since I had the arcing issue before, so presumably that's damaged something.

However, I've just powered it up again and I'm getting a front panel comms error! Everything was completely fine just a few seconds ago, so I'm hoping that this is a dodgy connection somewhere.

[dietert1]

Your relay replacement work looks great.
If the contacts inside the relays were unreliable, the contacts of the backplane are at least as unreliable. Also it looks like those gold plated edge connectors of the boards may need attention. Yesterday i read about using vinegar with salt to remove oxide layers from copper parts - as a preparation for gold plating. But i don't know whether the 5440 backplane has copper or brass springs. Probably they are gold plated anyway, so it's just removing dirt.

Regards, Dieter

[essele]

Thanks Dieter ... I'm very pleased with the relay replacements, just a bit annoyed I can't continue testing it because of this comms problem.

It actually looks like it's the UART on the front panel PCA ... I couldn't see any output signals at all, the pin is just sat at about 2V with a small amount of ripple that looks like the CLK. The TBRL (transmit buffer register load) line was going low which should then load the data and cause it to go out of the output .. but no output.

I tried cleaning and reseating it, to no avail.

So I've temporarily replaced it with the one from the A16 Controller and I can see some data on the output ... so looking fairly conclusive.

I can't find a sensible place to get a replacement other than eBay ... hopefully should arrive by the end of next week! More waiting!

[essele]

OK ... UART replaced and we're back in business ... not sure what caused that to fail.

The unit now seems to be functioning all normally, however I'm getting an error during the analog self tests that's to do with the zero amp. The zero amp is only used during internal calibration and helps to figure out the offset of the output.

The test that's failing is a "shorted input" test (M23) that should see a result between -0.25mV and +0.25mV and I'm seeing results that range from 1.5mV to 1.7mV.

The zero amp is a fairly simple OP07 configuration with a gain of either 100 or 1000, with the feedback resistor being switched by a JFET when needed, x1000 is the default, but x100 is used for the particular test.

I've tried swapping the OP07D with an OP07CPZ (which should be lower offset) and actually saw the numbers get worse ... I'm hoping this is because the OP07D is performing better than spec.

I've tried swapping the analog mux just in case there was some leakage there, but it made no difference.

So I'm left with two theories:

1. The JFET Q23 is leaking Gate/Source (do they do that?) which I think would then pull the inverting input slightly lower, thus increasing the output.

Or hopefully much more likely....

2. The transition from x1000 to x100 isn't working and the result is 10 times bigger than it should be. This could be an issue with the switching signal (ZCA) or Q21, Q22, or Q23?

Does that sound feasible? Or is there anything else that could be doing this?

I've attached the schematic ... the zero amp is towards the bottom-centre surrounded by the dotted shield line.

This did only start happening after I had the issues with arcing, so I was really expecting something to be fried by excessing voltages caused by faulty relays, but everything seems ok, at least at very initial testing.

[Kleinstein]

Gate leakage is possible with a broken JFET.

A not working JFET and thus always the x 1000 setting makes sense.  The JFET should not be something special - likely some 2N4393 or J113.

[essele]

Thanks Kleinstein ... I explored this idea a bit.

I managed to power up the board on the bench with +/- 17V and the +5V relay power, and it behaved pretty much exactly as expected, I saw 4mV at x100, and 40mV at x1000 which all seemed nicely within tolerance and ruled out any issues with the jfet.

However, once back in the 5440B, at standby (which should be x100 shorted inputs) I was getting 250mV ... took me ages to figure it out [note to self: use a scope as well as a dmm], but it was ultimately a couple of electrolytics I hadn't yet changed on the PreAmp board, whenever that board was plugged in I saw over 500mV of noise/ripple on the 17V supplies.

There were three caps I changed, one for the oven circuit which was definitely faulty, but didn't improve things, but then when I did the two on the 17V supplies it all cleaned up nicely.

The analog tests still fail, but further on now at M33, I'm getting -7.7mV instead of something +/- 1.6mV, this is to do with the HIGHV INTCAL and seems to be pointing to a problem with the sample string PCA.... job for tomorrow!

EDIT:  Problem solved ... M33 got me thinking back to the similar problem I had when I had one of the sample string relays making contact incorrectly ... a quick inspection of that replacement relay PCB for that relay and I found three pins on one of the relays that I hadn't soldered (rushing too much obviously!). With this fixed, the analog checks all pass ok, other than a sample string oven issue which I need to look into, and I'll wait until tomorrow to run the HV ones, end the day on a high.

[baof]

Is the output voltage stable? I replaced the relay with Panasonic. It was found that the output voltage was affected by the heating of the relay coil. The thermoelectric potential was very obvious. After 10V voltage was output, it would drop by more than 1.5ppm after the coil was heated