Philips PM6654 high resolution counter
A Teardown and Repair Log
Fine (Ice-)TEA from Belgium
A (not so) mystery box showed up at my doorstep. Vaguely remembering having had some chat with
a Belgian eevblog inhabitant, some shadows of memory came back: This must be the defective PM6654.
Indeed, opening the package revealed the truth - it's a defective Philips PM6654 high resolution counter.
So the normal eevblog member would follow the "don't turn it on, take it apart" workflow. For some
particular reasons (basically because I'm curious), I decided to turn it on before I take it apart.
Nothing happened, the display was blank, some LEDs were lit at the front panel. Randomly pushing and
turning buttons and knobs revealed this:
The A and B inputs apparently worked, since I could turn on and off the trigger LEDs by changing the
trigger level. With an input signal applied, these LEDs start blinking, showing the input is working to
So one might think now, at least some analog supply is alive, operating the inputs, but maybe the digital
supply is dead. The next step shall prove this assumption wrong:
Pushing the reset button brought some more life into the unit, the display showed a row of zeros and
more buttons operated something else but the button was intended to. E.g. one of the numeric keypad buttons
changed the mode. Normally one is intended to change this setting by the arrow up/down buttons.
OK, We've at least power to the digital components, and the processor does something though not what it's
supposed to. Maybe there's a faulty EPROM?
Now I've tried to hold the reset button while turning on the unit. This apparently works, the display
shows "E4" now. Pressing the reset button again now brings the unit to a somewhat better state - still no
counting or measurement working, but all the buttons work as expected, and one can read out e.g. the trigger level settings.
So the power on reset circuitry most probably is toast in some way.
I had to look up "E4" in the manual (didn't read that until now). It's an internal self test, indicating
"Measurement Logic Fail". First step in the fault finding flow is to check the int/ext ref source switch (so I did,
was OK), next is to swap the "COAC" ASIC. Can't do that, since this chip most probably is unobtanium now.
So my next thought was: That darn COAC (counter on a chip) ASIC might be dead, could one replace that by some CPLD or
FPGA stuff if neccessary? For sure this would be an interesting task!
Reading the manual revealed some more self testing procedures, this showed "E4" as the only failure. So RAM and EPROM
and other stuff is considered OK now.
More playing around shows the counter can display the "Hold Off" setting (time), though the knob doesn't work.
knob is stuck, one cannot turn or pull, obviously another point to repair. According to the manual, a special internal
path to the mighty "COAC" is used to display the "Hold Off" setting. So for now I concluded the COAC is OK as
it can measure this setting. Phew.
Now, with Hold Off turned on, the counter actually shows plausible frequency measurements up to around 200kHz. Above
This is another hint the COAC most probably is fine, but the CEX board might be faulty, as this acts as a prescaler and high
resolution extension to the COAC.
Fault finding #1
The CEX board has some means to generate 100MHz and 500MHz reference frequencies from the main 10MHz ref.
So I decided to check them. Of course, the real schematic is a bit more complex than the above block diagram,
and the main CPU can turn off the 500MHz multiplier if it isn't needed. Switching of the 500MHz is controlled
by IC712, Pin 6 (the Signal is labelled K12). The CEX board uses a wild variety of logic circuits, ranging
from CMOS to ECL. Interfacing between the logic levels is done by various methods, often just resistors.
So the K12 signal finds its way into the ECL gates (IC704), effectively disabling the input signal to TS719.
The PM6654 must be set up in a way the 500MHz is enabled, the manual provides some instructions. If you
choose to ignore them, basically turning on and off "Hold Off" switches the 500MHz on and off in "Freq A" mode.
This matches my observations above.
The manual specifies an amplitude of 500...800mVpp at the 500MHz test node (R742, R718, R766 ... junction).
Using a FET probe, I found the amplitude at about 200mVpp.
This clearly is on the low side. The SP8635 datasheet specifies 400...800mVpp for this chip to operate,
though there's a diagram showing that it just might work anyway. The SP8635 is the first decade of the "TIME COUNTER",
where the 500MHz is routed to.
I've used the signal generator to couple a 500MHz signal into the circuit node,
and adjusted its amplitude to provide 800mVpp. But all she said was: "Okay, so you're a rocket scientist -
That don't impress me much". OK, let's carry on.
Now take a look at the schematic diagram again:
Pin 6 of IC702 outputs a 100MHz signal that is coupled through C747 into Pin 14 of IC703 (said SP8635).
Depending on some control signals, there's nothing, 500MHz from the multiplier circuitry or
100MHz from IC702 at Pin 14 of IC703. Signal "K12" that enables 100MHz to Pin 10 of IC702 when the 500MHz
path is disabled. One can find this 100MHz signal here. It never showed up at the output (Pin 6). This is
because one of the other inputs (9, 11) is at a constant "HIGH" level.
This made me believe there's something wrong within that convoluted logic circuitry. It's just logic gates
and a few D-type registers, but the sheer amount of them (look dad, it's half a metric ton of ECL gates) and
nets drawn in a rather confusing way (someone must have had fun bringing the whole circuit onto one page)
doesn't make it easier to follow the signals.
Anyway, the Philips engineers were nice people and provided the unit with a special test stimulus mode and
a whole set of pulse diagrams of what to expect where. Read the manual for details. This is a real useful
feature for debugging this kind of circuitry.
Basically, this test mode provides the inputs of the logic network with a repetitive stimulus pattern on
most of its inputs and specifies resulting pulse patterns for many nodes that one can verify with a scope.
This way it's way more easy to follow the signals through the circuit and locate faults.
So I hooked up the scope according to the instructions and checked the first signals. Two out of four were
a fail, so yes something must be bad here within the logic circuitry. The manual recommends to check all
the "blue" signals now, these provide the input for the feedback circuit that should be checked first.
I continued checking signals and found something weird at IC715.
Pin 11 receives a pulse train that is a common reference for the whole signature test. It's the yellow trace,
the other trace shows the input signal at Pin 10. Some interesting side note: The outputs of IC723 (9) and
IC715(14) are just tied together and fed into the next gate. With ECL logic, the result is a wired-OR of these
A properly working "NOR" gate now should pass the inverted pulse train to its output while Pin 10 is "LOW" and output
"LOW" while Pin 10 is "HIGH". Here, the output signal just looked like the inverted input. So one would think,
IC723 is defective. As 10k ECL gates are a rather rare species today, I don't have one to simply swap that IC.
But there's an open input at Pin 13 of the same IC, while Pin 12 receives the same pulse train. As expected,
Pin 15 outputs a copy of the input. By tying Pin 13 to Pin 10, this gate should work as the other one is expected
but doesn't. But ... this gate doesn't work either. WTF?
If one examines the pink trace closer, one might notice the "HIGH" level of the signal doesn't reach the
ECL "HIGH" level. It switches between near 0V and around 3.5V, aligned with the "LOW" level of the yellow trace.
So this input never reaches ECL "HIGH" and why should the gate gate the signal then?
Measuring the supply voltage of IC708 and IC712 (that's where the signal mentioned above originates) revealed
a rather lowish 3.6V. Looking at the schematic diagram, it's supposed to be 4.3V provided by one diode drop
from the 5V supply.
Providing an external 4.3V supply here changed the signals to this:
Better, isn't it?
Now reverting to normal mode, the counter showed a plausible result for a 100kHz input signal at
its full resolution ("Hold Off" turned off). First bug found, but the self test still shows "E4", so there's a bit more left to do.
Replaced that diode (GR704) and carried on. The diode had failed open, so one could expect
IC708 and IC712 to receive no supply at all, but there was a 3.6V high level signal at
its outputs. This is quite common with CMOS ICs, the chip receives its supply through
the ESD protection diodes from inputs at "high" level, or outputs with attached pull-up
Fault finding #2
Next fault turned out to be intermittent. Sometimes the unit started working fully
OK, sometimes showed "E4" again but worked then and sometimes it couldn't even be bothered to display
a measurement with "Hold Off" turned on.
A few days later, to my luck the intermittent fault turned into a permanent fault. So
I started again to check all the CEX board signals using the above signature method.
So I quickly discovered IC712 having all its outputs static while some of them were supposed
to toggle. Next check would be its input signals (would be OK, since IC708 did put out
the correct signatures) and the /C2 signal at P11. Bingo, that signal was just missing.
Further inspection showed one pin of this connector was contaminated with some sticky
whatever. Cleaning the pin and the connector fixed the fault. Just to be sure, I've cleaned
the other connector too. When I first openened this unit, afair the calibrations stickers
looked not broken, but there were clear signs of someone has been into the counter before:
Some screws fixing the IEC and CEX boards were missing. So maybe one had touched that
connector with sticky fingers or whatever.
Fault finding #3
The power on reset still wasn't working as desired. The Philips engineers couldn't be
bothered to describe the operation of the state machine controlling the counters brains
(that's what I've diagnosed with that signature analysis above). Instead they provided
you with well-made instructions for fault finding. For other circuits, they've put
a nice description of how it is supposed to work into the manual. So they did for
the reset circuitry. Having that and the schematic:
it was a rather easy job to find the fault. Turned out Pin 6 of that 555 (IC208) wasn't
connected to the rest of the circuit. Further inspection showed "there's someone been
here before" and most probably de-soldered IC210, IC208 and some of the components around.
Looks like he'd just missed (or caused) that broken trace from GR231/R290 to IC208 Pin 6.
One cannot see that in the photo, most probably the trace is broken at one of the pads.
Fixed that, and the reset circuitry worked as desired.
The "HOLD OFF" knob is stuck. One can neither turn nor pull that knob. For now, I couldn't
be bothered to take the front panel apart to fix that. It's not an important function to
me, so I might leave it as is.
Some late night testing:
A 10 MHz input signal and an A/C ratio measurement 1GHz / 10MHz input.
As this counter doesn't have any high quality reference frequency option, I've used
the external reference input to synchronize the signal generators and the counter
to a single 10MHz (GPSDO) reference frequency. This way, the counter shows exactly
the correct number, slighly flickering by 1 digit.
Goto here for some proper teardown pictures.
Some components datasheets
Thanks to eevblog member Ice-Tea who sent me this unit on a "if it breaks it breaks" basis.