Fluke 8840A bench multimeter teardown

[DRT]

Following Kibi's teardown and repair of a Fluke 8840A multimeter (https://www.eevblog.com/forum/product-reviews-photos-and-discussion/fluke-8840a-teardown-and-repair/msg121986/), here are a few pictures of mine. I had opened it up to investigate a seriously fading VFD display. Sadly, there is nothing electrically wrong (filament volts are fine at 4.5V AC, drive signals are fine at 0V/30V), so I conclude its just worn out. I might see if I can boost the filament volts a little.

This multimeter was made in 1986.

Picture descriptions:
1. The front panel. The VFD seems bright, but it isn't really. It has a tendency to fade up and down - a sharp tap seems to recover it (this is some mechanical effect, not a contact problem). [Edit: On further investigation, there was a dodgy joint on the filament output of the PCB-mounted mains transformer. Reflowing the pins fixed the fading up and down - it is still a bit dim though, but perfectly usable.]


2. The rear panel. This unit is fully optioned up with AC measurement board and the GPIB board.


3. Initial view - the GPIB add-on is on the left, the AC board on the right.


4. The main 5V rail electrolytic. My first reaction was 'uh-oh', as a distinct bulge is apparent. However, it seems that a domed plastic disc has been placed over the safety relief valve - you can push it down very easily into the void underneath it.


5. The three 40-pin DIP devices, from top to bottom, are
8279 Keyboard and display controller,
Z8 microcontoller (mask programmed with Fluke's code I think)
Custom Fluke Analogue to Digital converter



6 & 7. The GPIB option board. Again, the big electrolytic exhibits the scary 'false bulging top'.



8 & 9. The AC RMS option board.



10. Top view with both option boards removed. Of note, both front and rear inputs pass through an RF suppression toroid. To the left of the micro are ROM and SRAM ICs.  Upper middle are the VFD driver ICs (TTL to 30V level conversion), lower right is the front end circuitry.


11 & 12. More close-ups.



13. A close-up of one of the precision resistor networks. This one is three 33k resistors in series.


14. Another precision network. Some unusual values in this one - it is a series network of 5.00M, 5.007k, 45.52k, 505.3k and 19.54k.


15. Every rail and other significant reference voltage has it's own test point. Lovely!

[T4P]

Careful about the caps in the last picture, they look slightly bulgy

-Imagine the time taken to design this, BACK IN 19eighty-something!

[DRT]

Those caps are fine. It does have a few tants though - I'm more worried about those 

[SeanB]

The dipped caps are a lot more reliable than the SM ones in my experience. Only ones i have seen fail were due to overvoltage or reverse voltage, and that for long periods. I accidentally wired one in reverse one day, and it took around 2 minutes to blow up on a 5v rail.

[Fraser]

I have three of these multimeters and they are definitely 'keepers'. Excellent accuracy and performance despite their age.

Your unit is 'fully loaded' (RMS + GPIB) so worth some effort to restore to a decent display brightness. Mine have nice bright displays indicating quite low usage. As you probably know VFD's do fade/dim with hours run and you can actually start to get a metallised film forming on the interior of the front face if they have very high hours on them. It would be worth searching ebay for a half decent unit that is faulty but has a good display. I have seen several come up over the years. Fluke will probably still sell you a display, but sadly the price could be unpleasant. They were willing to sell me a new handle when I asked last year but it was GBP40 

Aurora

[FenderBender]

Careful about the caps in the last picture, they look slightly bulgy

-Imagine the time taken to design this, BACK IN 19eighty-something!

Nah those are good.

[lowimpedance]

I have three of these multimeters and the are definitely 'keeper'. Excellent accuracy and performance despite their age.

Your unit is 'fully loaded' so worth some effort to restore to a decent display brightness. Mine have nice bright displays indicating quite low usage. As you probably know VFD's do fade/dim with hours run and you can actually start to get a metallised film forming on the interior of the front face if they have very high hours on them. It would be worth searching ebay for a half decent unit that is faulty but has a good display. I have seen several come up over the years. Fluke will probably still sell you a display, but sadly the price could be unpleasant. They were willing to sell me a new handle when I asked last year but it was GBP40 

Aurora

The display and keypad PCB were a 'one piece' only as a spare. Incidently I contacted the local fluke agents out here in the "sticks" about this exact spare part and NO longer available, howerver they could sell me new DMM  .

[reagle]

These things are awesome- I am using 8842 from early  1990s and the thing is still reading exactly the same as a brand new Agilent meter that's still in factory calibration. The amount of attention to details that went into them is pretty crazy

[free_electron]

A couple of tidbits...

Underneath the 8279 keyb/ disp co troller there Are two more chips hiding....
The same for the external firmware rom and the eeprom. One chip each under the socket.

The eeprom is a xicor device with parallel operation. These things were so unreliable that the calibration data is stored twice with double checksums....
The main z8 cpu co tains 4k rom with a mini 'operating system'. That is : the low level functions to drive the 8279, uart( for gpib expansions) and other simple i/o tasks that are 'static' sit in that mask chip. The measurement algorithms sit in the external rom. If correction were necessary , or hardware revision would change in the analog world it could be altered in software there.
Never ever imstall a random firmware in an 8840. You may find a 2.1 and think . Oh mine is a 1.8... That will not work. The roms are tied to pcb revisions...

Looks like you have a second generation gpib option. The original one used a upd7210 plus a z8. In later revs they bitbanged the i/o.

The a/d convertor in that beast is a weird system. Actually it is not an a/d as we know it but a charge recycler...

For people that remeber my explanation on the multislope convertor in a 34401 this will sound similar , but it isn't. For people that haven't read that one. Don't worry you can read that later.

The asic has a 6 bit precision dac made using an external resistor network ( thin film hybrid next to the fluke asic ). An external sample and hold opamp takes a sample of the input signal and stores it on a capacitor. The asic now does a binary search with six analog switches switching in and out the resistors in the hybrid. A precision comparator tells the asic 'too high' or 'too low'. When all bits have been tried and the highest value, that does NOT produce a 'too high' output of the comparator, has been found, this analog voltage is substracted from the cap voltage, the result multiplied by 16 ( analog gain of 16) and stored in another capacitor.
And then the process restarts.
This is done 5 times. Giving 30 bits of overlapping information.
They can now mathematically compensate the errors in the dac ( errorin the resistor ratio's ) , errors in the x16 multiplier, and other factors.
The software calculates the result and spits it to the display.

Using this trickery they can get a good comvertor that can be self calibrated, pretty much like a multislope can self calibrate, and is only dependent on its reference voltage.

They have one advantage: the overrange is faster. If the first cycle fails to produce a ' too low' they immediately know: overrange !
I dont know if fluke aborts the opertion at that point. I believe they do. This makes autorange faster as we do not need to wait for a full conversion.

[reagle]

Silly question- how do you remove the fasteners holding the various option boards?
I know it's probably somewhere in the manual

[DRT]

Just pull the ends of the fastener out (while ensuring the lower bushing stays put)

[Fraser]

You pull the 'toggle' top.

Its attached to a center pin that holds the three retaining spurs open so when you pull the toggle the spurs can move inwards.

Warning, when I removed mine for an inspection I found the plastic to be very brittle and some of the spurs broke off.

I'll take a picture of a broken one and upload it.

Aurora

[Fraser]

OK, here are the pictures of the broken 'Nylatch'

There should be three 'spurs' at the end of the hollow part but they snapped off. The problem is that the spurs latch into the hole and must be gently 'squeezed' inwards to release them. I didn't even have to squeeze them as they fell away as soon as I pulled the centre pin toggle 

If anyone knows where these can be purchased in the UK I would appreciate a steer as I am having to use 4mm nylon screws as temporary substitutes at the moment.

I attach the datasheet from Nylatch for the same/similar product to that used in the Fluke and other equipments.

Aurora

[SeanB]

RS? Should be available in the fasteners or PCB section.

[Fraser]

Thanks,

I couldn't see them last time I looked but I will search on RS and Farnell again. There are plenty of nylon releasable rivets available but I cant find some with the toggle on top for ease of release. I may have to settle for the dome top version.

Aurora

[reagle]

Thanks for the pictures!
That's what I assumed when I was trying to get mine out- they just looked like they "grew" into each other with no dividing line

You pull the 'toggle' top.

Its attached to a center pin that holds the three retaining spurs open so when you pull the toggle the spurs can move inwards.

Warning, when I removed mine for an inspection I found the plastic to be very brittle and some of the spurs broke off.

I'll take a picture of a broken one and upload it.

Aurora

[reagle]

I finally got around to check why my 8840A has a display that comes and goes.
( Just posted a brief writeup  on my blog http://kuzyatech.com/fading-display-repair-on-fluke-8840)

It turned out to be the same exact problem as what you had. Bad solder joints of the mains transformer making intermittent contacts with the filament traces. They actually looked soldered until heated with an iron. Then you can clearly see oxide film preventing solder from sticking.
In my case the unit may have been dropped, since the corner board support bracket was broken, and the board was flexing a lot. This unit also has both GPIB and Option 9 AC boards, so I am keeping it with its new rock stable but dim display.
Has anybody come up with a way to wake these screen up like the used to do with CRTs? As in a brief capacitive discharge to "shake" some stuff off the filaments? Or alternatively, do we want to make an LED or LCD  based replacement front board? That may be fun

[SeanB]

The problem is not the filament ( it is uncoated so will not lose emission much with time, it is bare tungsten wire) but the actual anodes. Try getting the anode voltage up to the max allowed, probably the supply is at the low end of the range Check for ripple on the supply there.. If you want to liven up the filament run it from a power supply ( disconnect the one transformer lead) and take it up to a dull red ( view in subdued light and while looking at the display direct with no filter in place) for a half hour to warm it up and redistribute the surface. Dull red only, bright red or white will destroy it in seconds.

[reagle]

Oh, good point! "Though shall check voltages" as Dave says

[nukie]

How long is the memory rentention of the chips in the 8840? I have one and I have very little faith in it due to old age.

tapatalk