[Updated June 2013, see end of article.]
[Also see: newer article on IBM 8060A/AA Refurb]
The Fluke 8060A is another early-80s vintage handheld DMM. With a 4½ digit 20,000-count display, RMS-responding AC bandwidth to 100kHz, and direct readout in dB, it is still in use today and a favorite of those working with audio gear.
This eBay acquisition was described as “powers up, but I can’t test it.” I assumed it would be non-functional, and when it arrived, that suspicion was confirmed. The display flickers, is sometimes very faint, but mostly shows a fixed arbitrary number, regardless of what function or range is selected. It is also in sad cosmetic shape, being very dirty and grimy. Repairs will be required to get this unit working again.
Disturbingly, the area around the buttons shows rust-colored deposits, looking as if there has been some water intrusion.
After opening the unit up, the internals do not look as bad as the outside. The components all seem to be in good shape, with only a slight layer of dust. The bottom of the PCB looks pretty good, no obvious signs of damage, missing parts, or any previous repair attempts.
The LCD assembly contains a daughter board with a QFP surface-mount microprocessor under the display. Hidden under this is a 40-pin DIP analog processor on the main board, in a nice purple ceramic package with gold-plated pins. Fluke calls this the Measurement Acquisition Chip (MAC).
But I usually refer to this as the analog processor or AP chip throughout this document. [The official name is MAC, so I've standardized to avoid confusion.] Most of the parts have 1982 date codes, confirming the vintage of this unit.
The case parts and buttons get a good scrubbing with a strong solution of basic household cleaner. A rag, a toothbrush, and a little muscle removes most of the grime. The clean-up efforts make a marked improvement in the appearance. Other than a previous owner’s name engraved in the top side, there are only a few minor permanent dings and scratches, so the make-over is successful.
The switch assemblies make it hard to fully clean the circuit board with IPA (isopropyl alcohol), which is usually step #1 in any restoration project. So after the shield is removed, the accessible areas are dusted and given a cursory cleaning. The microprocessor daughter board has two elastomeric connectors, one for the LCD, and one back to the main board. Cleaning these improved the look of the LCD segments, but did not otherwise improve the functionality.
The Fluke manual for this multimeter is downloadable from Fluke’s manual archive and contains full schematics and service information. There are two versions of the manual available (1000V and 300V), and unfortunately neither one seems to include the serial number of my unit. A little research indicates that Fluke manufactured and sold this model well into the 90s, and these service manuals likely refer to the later revisions. The 1000V version refers to the lowest serial numbers, and going through it reveals a few minor differences, but it seems mostly accurate. The chief difference is the addition of an RMS converter sub-assembly. My unit does not have that sub-assembly, having instead a 14-pin DIP RMS converter directly on the main PCB.
In the absence of more specific clues, the best place to start trouble-shooting for just about any equipment is usually the power supply. The service manual has a list of test points, including those for the power supply.
TP1, the +5.2V supply measures 4.96V, which is outside the ±0.12V range specified in the manual. Close, but not quite right. TP2, the negative power supply rail measures -0.89V, but is supposed to be -5.1V. Obviously this is a serious problem, so let’s tackle that first. Looking at the schematic, we can see that an 8-pin DIP ICL7660 voltage converter IC (U4) is responsible for creating this supply, using an external 10μF charge pump capacitor (C21). The output of this device (pin 5) is connected to TP2 through a 12Ω resistor (R35) and goes on to provide the negative supply for the analog processor (Vss).
Examining C21 closely reveals a rather dull crusty-looking solder joint on the bottom side of the PCB, which is indicative of the capacitor having leaked corrosive electrolyte at some point. When removed, it’s obvious the cap has spewed its juice. It has an ESR of 69Ω, and actual capacitance measures 4.5μF. Both measurements are not good, and at this point, this cap appears to be the cause of the faulty negative supply.
After finding the first cap that has leaked electrolyte, all other electrolytic caps are automatically suspect. Visual inspection reveals that two other capacitors have leaked significantly. This includes C1 and C12, both of which are very near the 40-pin socket for the MAC chip. Lifting the device part of the way out reveals significant corrosion has wicked its way up some of the socket pins. All other capacitors look and measure OK, but a complete re-cap will be in order.
Removing the MAC chip completely reveals the extent of the damage. Pin 32 near C12 has weakened to the point of failure, as one side has broken completely off. Pin 21 near C1 is cracked and will no longer make contact with the chip leg. The socket will definitely have to be replaced, but waiting on a parts order puts quite a drag on the trouble-shooting process, so the socket gets scrubbed thoroughly with IPA and is allowed to dry. A slight outward bend on pins 21 and 32 will hopefully provide enough contact for further testing.
The bad capacitors are temporarily replaced with whatever’s on hand and the analog processor is carefully re-inserted into the dubious socket for further testing. Part of its function is regulation of the 5.2V positive supply rail, which is now looking good at 5.16V (TP1). The bad caps and corrosion on the socket contacts probably caused the original low voltage reading here. Unfortunately, the negative supply is still only managing about -1.1V on TP2.
Across the ICL7660 output is a reservoir capacitor (C23) that is integral to the charge pump operation. Even though it measured OK, it gets replaced. Doesn’t help, however.
I removed the ICL7660 and tested it on a breadboard with a DC power supply providing a 5.2V positive supply. It still only outputs -1.1V, so it is definitely bad. Maybe it was damaged by the nearby leaking charge pump cap. A new 7660, a new 40-pin socket, and enough capacitors to re-do the entire board are placed on order. Now waiting for the postman…
A small DigiKey box has arrived in the mail, so back to work. The new 7660 chip is tested in the same breadboard setup as before, and it definitely generates a negative supply voltage like it should. Confidence is high.
The new 7660 is soldered in. And after much careful de-soldering work with a solder-sucker and some wick, the old corroded 40-pin DIP socket is removed. The 4-layer PCB is somewhat difficult to work with, but fortunately it is quite sturdy. No pads or traces were damaged in this process. The new socket goes in easily. It has machined pin contacts, and should be much better than the old one.
A number of pins on the MAC are very high-impedance inputs, as evidenced by the guard traces around some. Soldering flux residue has to be cleaned, so that no current leakage paths are created.
The MAC is now settled in its new socket and all the suspect capacitors have been replaced with good-quality Nichicons. All major DC test points are now within acceptable limits.
The meter boots up and passes its self-test! The LCD segments are solid. All functions seem to basically work. But with no input, or even a shorted input, the display never seems to rest at or near zero, like it should. A test input of 1.9V DC in the 2V range shows a pretty good reading. But higher ranges are considerably worse. Resistance measurements are 10% or more off. Something is not right.
With open inputs the 8060A will eventually settle on a reading of around -0.0150, sometimes more, sometimes less. This is not good. Another multimeter bridged across the input jacks seems to load this offset down some and bring the reading closer to zero. Interestingly, the other meter reads essentially the same as the 8060A. All this suggests the fault is a real, measurable (but tiny) offset current flowing somewhere it shouldn’t be, and not some digital counting error. But where?
R20 is a 220KΩ resistor that provides the sole path to couple the input voltage divider section into the analog-to-digital (A/D) converter inside the MAC (pins 6 and 7). Removing it decouples the entire front end from the MAC. While it’s out, an inspection reveals that the electrolyte leakage from the charge pump capacitor had some effect on one of the leads of this resistor, but not enough to cause a problem.
The front end is essentially a passive voltage divider configured by the range switches, and can be tested by injecting a voltage into the front panel jacks and measuring the result on the upstream side of R20′s location. This is done with the meter powered off, and is possible because this meter is manual-ranging. The divider is configured solely by the switches, and not by a microcontroller.
The divide ratios for each of the 5 DCV ranges are shown in this table from the service manual. So for example, setting the range to 20V and injecting 10.00V, the result is 0.100V because the divide ratio is 1/100 on the 20VDC range. Checking all the ranges reveals that the divider is spot-on, and probably is not the cause of any error.
In retrospect, the previous exercise was mostly unnecessary because of two things:
- The residual offset voltage measured at the 8060A’s input jacks by another DMM (with the 8060A powered on) disappeared when R20 was removed. The current producing the offset was coming from the MAC side of R20.
- The residual offset voltage was negative. Pulling the positive input terminal of the 8060A to a voltage below ground requires a path to the negative supply voltage Vss. There is no such path in the input circuitry except for the RMS converter, which is disconnected when the switches are set to DCV.
All of these clues point to the MAC chip itself as the source of the fault current. Of course I had rather find any fault other than the custom Fluke device, so I looked a bit too hard for any other explanation. For good measure though, C22 was removed and checked and it tested fine for capacitance and current leakage. Plus a few other dead ends I went down that I won’t embarrass myself by documenting here. That really rules out just about everything but some kind of serious PCB contamination that has escaped notice, or the analog processor.
It is fortunate that the MAC is socketed and can be removed and replaced easily. Also fortunate, is that a derelict Fluke 8062A model appears on eBay for cheap and no one else bid on it. Checking the 8062A documentation, it shows that the Fluke part number 612713 for its MAC analog processor is the same as the one in this 8060A.
When the parts donor meter arrives, the MAC inside does indeed have the correct part number. Instead of the nice ceramic package, it has a plastic package. After installing this ‘new’ analog processor chip and performing a cursory calibration procedure, we finally have a near-zero idle display, and good readings on every range.
This 8060A will join an 8020B of similar vintage in my small personal ‘museum’ collection.
DMMCheck calibration check results for this Fluke 8060A:
|Ohms||100||100.04||REL feature used|
Important Update: Some time after this article was written, I decided to find out if the 8062A mentioned above that donated the ‘new’ MAC was repairable. It now had the MAC from the 8060A that I deemed faulty for the reasons explained above. After repairing some unrelated minor problems, this 8062A then started to exhibit exactly the same symptoms as the 8060A originally did. Makes sense, since it really was the MAC causing the problems.
Knowing that leakage current was to blame, it occurred to me that I had never actually cleaned the chip itself. So I dunked it in IPA and scrubbed it with a toothbrush, even though I couldn’t visually spot any contamination. At first it seemed even more flakey, but after several more hours of drying time, it began to work perfectly! As a matter of fact, it has continued to work, so I swapped it back into the 8060A, and it’s still working today (June 2013).