Fluke 867B dead - no internal clock - any advice? [FIXED]

[mahi]

Fluke 867B Revision L (manufacturing date 1998)

My Fluke 867B died. I bought it used 10-15 years ago, so I don't know its history, but it was like new. No signs of abuse. I don't use it regularly, but I like to 'rotate' between the multimeters I've collected over the years. I've used it last around December and it was put back in storage in working order (batteries removed).

Yesterday I picked it up again, installed freshly charged batteries (6x Eneloop AA), and switched it on: The LCD backlight came on, but that's all. No graphics/text on the display. No clicking from the buzzer or relay. None of the positions on the mode dial have any effect, nor does pressing any of the buttons.

It's not just the LCD display or driver that is dead. I know the multimeter is supposed to make clicking noises at startup and between certain mode dial positions. Also when I dial it to Ohms, I measure no voltage on the sockets. So, there's a lot more wrong than just the display...

Luckily the Fluke 867B dates from the days public service manuals were still a thing:

Fluke 863/867B Service Manual (4.5 MB)

Chapter 2-2 covers the start-up sequence:

The GMM sequences through the following steps when power is applied:
• The 3.3V power supply comes up.
• The reset pin on the microprocessor (U25-1) goes high.
• The microprocessor (U25) begins executing the program stored in EEPROMs U11 and U19.
• LCD controller (U13) data is initialized.
• VEE is turned on (-20V dc), and the display comes on.

First I checked the power supplies: VCC/VAD (3.27V), VDD (5.2V), VSS (-5.2V). All are fine and well within tolerances with regards to voltage and ripple (chapter 2-6). Power consumption is normal.

PS: If you power an 867B with a lab power supply (7.2V at the battery terminals), make sure to set the current limit to at least 700 mA. Even though the multimeter draws 'only' ~200-250 mA under normal operation, the startup surge current is a lot higher. When the current limit is set too low, the 867B power supplies do not start properly. It looks as if there's a short somewhere pulling down the output voltages. I wasted a couple of hours chasing ghosts.

Next I checked the reset pin on microprocessor U25 (Hitachi H8/510): About 130 ms after switching on, the pin goes high (and stays high until powering off).

After this the microprocessor should start executing the program in the EPROMs. I checked this by monitoring the address lines on the U11 and U19 EPROMs. And here's a problem: There is no activity at all on the address lines.

So, the microprocessor does not run. Why? I checked pin 98 (EXTAL) on microprocessor U25 and there is no clock signal while there should be a 9.6 MHz signal.

According to the service manual the clock signal is provided by gate array U24 (chapter 2-49). The chip has an external 19.2 MHz crystal and divides this clock to generate different clocks for different parts of the hardware (this is just one of the things U24 does - it's quite an important chip in the circuit).

I confirmed U24 gets power, but there's no activity around the quartz crystal. From as soon as VCC is available, U24 pin 31 (XTAL1) goes high (3.27V) and pin 32 (XTAL2) stays low (0V). There are no oscillations.

U24 appears to be a custom chip made by NEC for for Fluke (Fluke part number 928671). There is no datasheet available.

And this is where I am currently stuck. Is it likely that U24 is dead? Or might the crystal be bad (do crystals even die - I never dropped the meter)? Does anyone have suggestions for further troubleshooting?

[shakalnokturn]

Have you also confirmed U24 is not stuck in reset? (That's if it's a different reset from CPU, I haven't checked full schematic.)

I have seen crystals go bad or intermittent on a couple of occasions. It should be quick enough to inject a clock at U24-32 to check.

I've had a couple of Fluke's (of which a 867) with not so obvious faults that were brought back to life by re-soldering the QFP IC's.

[bdunham7]

I have fixed things, including multimeters (not an 867) that had clock crystals that failed.  As mentioned, just attach a 10x scope probe to a signal generator set to put out something close to the clock signal at a low level and just probe the circuit with it.  It may jump start the crystal and then it may work for a while.

[mahi]

Thank you for your replies! I was not aware that crystals could go bad without shocks or vibrations.

I don't think U24 has a reset pin. At least the diagrams do not hint at it. It's not a microcontroller.

I tried jump starting the crystal with a low level 19.2 MHz signal, but had no luck. Maybe the signal was too low due to the 10:1 probe...

I desoldered resistor R65 and capacitor C37 to isolate the crystal from the circuit. I applied a sine to the crystal on one leg and probed the other leg with the oscilloscope. I could clearly see massive resonance around 19.2 MHz (see attached image). I don't think the crystal is bad.

With R65 and C37 still removed I applied a 3.2 Vpp 19.2 MHz sine directly to U24 pin 32. I switched on the meter, and yes, it made the familiar clicking noise! The display/keyboard is not connected at the moment, so I only have the noises to go by, but when I rotate the mode dial the relay clicks and in the ohms and diode positions I can measure a voltage on the sockets. I'm pretty confident that means the meter is fully working .

So, the problem is just the oscillator. Why won't it start with the crystal on the PCB? Could the amplifier/inverter inside U24 be bad?

I found one source that sells surplus Fluke 928671 (U24), but it's rather pricey (£47 + £37 shipping = ~100 EUR/USD). And since everything seems to work with an external oscillator, I might be better off with an external SMD clock oscillator chip. In fact, judging by the schematic and PCB, Fluke considered this option as well (U28).

shakalnokturn: When you write "re-soldering the QFP IC's", do you mean reflowing with hot air, or actually applying fresh solder and drag soldering?

[shakalnokturn]

I don't think U24 has a reset pin. At least the diagrams do not hint at it. It's not a microcontroller.

shakalnokturn: When you write "re-soldering the QFP IC's", do you mean reflowing with hot air, or actually applying fresh solder and drag soldering?

I read "DIC RST" at pin89, didn't check if that was an input or output though...

For the resoldering the bare minimum I do is apply flux and rework with the soldering iron, not just hot air.

As for the oscillator section, well you know it's at least usable on input pin, if you want to take the testing a little further hook your scope to 31/32 feed pin 32 a triangle input and check toggle point voltages and high / low state voltages at pin 31.

[mahi]

In the meantime I resoldered U24, but without success. The issue remains the same: The crystal oscillator won't start.

Thus I ordered a CMOS oscillator which arrived today. Sadly I could not find an oscillator in stock with the proper 7x5mm footprint to fit the U28 solder pads, so I settled for a 3.2x2.5mm ECS-3225SMV-192-FPT. I've soldered it in place with bodge wires. Not my best job, but it's for testing. I'll design a proper adapter PCB later. The CMOS oscillator is 10ppm whereas the original crystal is 50ppm according to the service manual.

I removed R65, R111, C37, and C42 but kept the crystal on the PCB as it is isolated by removing the passives around it.

With the display/keyboard still disconnected I switched on the meter and, yes, it made the familiar clicking noises! Thus the oscillator is working!

Next I reassembled the multimeter for a full test. After switching it on I was greeted by text and graphics on the display. Success! Then my heart skipped a beat: A message to start the calibration appeared. I feared that I had somehow lost calibration data, but when trying to abort the calibration I noticed none of the buttons worked. It turns out that the zebra strip between PCB and keyboard foil had become misaligned. This resulted in the calibration button being pressed permanently. No harm done.

After testing most of the functions and comparing the meter against my references and other multimeters, I can safely say that the repair was a success!

shakalnokturn: I hadn't seen your edit. I didn't test pin 31 with a triangular waveform on pin 32, sorry.

[mahi]

Last update... The adapter PCBs arrived today so I could undo my bodge wire job. The CMOS oscillator is now mounted onto the adapter PCB. Looks a lot neater! I'm happy with the result!

[BILLPOD]

Good Morning Mahi, I know you said you purchased it used, so that might eliminate any warrantee claims, but it doesn't cost anything to ask Fluke what they could do.   I've heard others say that Fluke replaced their meters without proof of being the original purchaser.

[mahi]

BILLPOD: That's a bit optimistic, I'm afraid. We're talking about a multimeter that was introduced in 1994 (mine is from '98) and was discontinued somewhere around 2001. The Fluke "lifetime warranty" ends 7 years after manufacturing stops. The 860 Series have been long out of support and Fluke no longer carries replacement parts.

[bdunham7]

Last update... The adapter PCBs arrived today so I could undo my bodge wire job. The CMOS oscillator is now mounted onto the adapter PCB. Looks a lot neater! I'm happy with the result!

You should be happy--that's a very nice repair and a good save of a nice old instrument.