faulty yet irreplacable IC

[merseyless]

Hi guys,
I managed to get my hands on an advance digital multimeter dmm2 (nixie tube readout). Which I immediately broke .
I've traced it to this unusual IC that I have linked below.

Does anyone know how to (or if you can) perform a repair on an IC like this or where I can get a replacement.
I'm half hopeful that it is an intermittent connection on the IC as shock causes it to stop working, and when left alone it will eventually start working again. for a period. Power supply is stable and all the inputs are doing their job as far as I can see.

Schematic attached. The IC is IC3 on schematic.
The fault in particular is that the P and Q outputs stop functioning, usually the P output. This causes only one nixie and the indicator lights to be usable while in this mode. Otherwise it is either non-functioning or working flawlessly. Both oscillators that feed into IC3 are working. I'm out of ideas on how to fix this.

PDF's pulled from here: http://www.vintage-radio.net/forum/showthread.php?t=18969

[KJDS]

I'd check the soldering, but if that looks ok then as it's working intermittently it's likely to be a wirebond inside the package.

If you have access to a wirebonding machine in a decent cleanroom and the skills to use it then it may be possible to remove the old bond and put a new one in, however unless you have can do this yourself or have a friend that can do it for you then it will cost more than a new meter.

[Ian.M]

IC3 is on the sheet DMM2002 7.pdf.

It looks like it only performs logic functions and has a 5V supply. It contains a counter and output latches for three BCD nibbles and an extra status light nibble + a 2 bit binary output on P and Q for display multiplexing.   

If you can work out its functionality, I am sure it could be duplicated in discrete logic or even a PIC or AVR, tested on breadboard and then made up as a daughter board to plug into a turned pin socket in place of IC3. 

Of course this would hardly be economic, but if you love Nixies . . . . .

[SL4P]

What did you do initially to 'break' it?
That may help identify possible repair options.

[PaulAm]

Is there a part number on it?  Manufacturer's mark?  Anything?  It seems unlikely they ran custom chips for something like that.

If you can track down data on the chip, it wouldn't be too hard to engineer a replacement.  From the circuit, it looks like there a few bcd counters and some multiplex circuitry.  Doesn't look like anything really unusual; even if you can't find data you could come up with an equivalent functional circuit.

[android]

The fault in particular is that the P and Q outputs stop functioning, usually the P output.

I notice IC3 has a RESET input (TP9). Maybe the chip is ok but it is being intermittently reset? Are all the supply voltage ok? If shock is involved it could be a dry joint somewhere.

As an aside, I wonder what the -12V on pin 5 is for...doesn't appear to be purely 5V logic.

[merseyless]

Just managed to get my hands on a computer again.
I had it on top of a stack of radios at wyong field day on the back seat of my car and I drove off without properly securing it all.
It fell forwards and down so a bit of shock but nothing drastic.

I checked the 5v supply, ground, reset pin, load pin, strobe pin and count pin. All are within tolerance. No ripple on the supply lines.
I did not check the -12v supply or the carry, I might just go over the solder joints and reflow them

What I would love to know is how this chip is getting the numbers to be displayed, I can't see any single point where the data is added, unless they are doing something crazy with the multivibrator on the count pin. If so that may take the cake for the most insane way to input data that I've seen.

Sorry if I was unclear, An image of the IC in question is here:

the markings are as follows:
GIM-1
29011
D3035A

I'll have another poke at it today. See if I'm missing something obvious.

[Ian.M]

As its an extremely old chunk of logic, it probably uses the -12V as a bias rail for the memory cells in its latches.   I said it is 5V logic because all its outputs and nearly all its inputs appear to be digital and go to/come from circuits with 5V & 0V supply rails.  The exception is the capacitively coupled LOAD input on pin 14, which is a -ve going puse, idling at 5V and being pulled down by the circuit C30 - D45 -TR13 to the -12V rail.   I am reasonably confident this could be handled with an appropriate potential divider hanging off the 5V rail or may be a series resistor + clamp.

As its working intermittently, getting good scope traces of all its input and output waveforms and their relative timing, then hooking up a logic analyser and sweeping the meter's input voltage over its full range should be a high priority for any reverse engineering effort.  This *MUST* be done before any resoldering of any suspected dry joints on IC3 is attempted in case the fault is internal and the thermal stress makes it permanent.

There are four test points associated with it and it would be worth searching all available documentation to find any waveforms or test procedures involving them that may be given.

[amyk]

The -12V and +5V, along with the inverted logic ("+ve = 0 0V = 1") suggests this is a PMOS part. "GIM" suggests GI Microelectronics (now Microchip.)

Digging deeper,  it appears to be similar to the AY-5-4057 mentioned on page 193 of this old GIM data catalog - that has pins for +5V, -12V, ground, BCD, and "Q" and "P" outputs.

[android]

Well done amyk
I have to admit to spending half an hour doing a fruitless image search for the manufacturer's logo on the chip. It seemed to me to be a rather heavily stylised "CEA".

[Rog520]

This company lists one AY-5-4057 in stock:

http://www.4starelectronics.com/part_detail/AY54057.html

There's also a seller on eBay (from Italy) that has a handful of them:

http://www.ebay.com/itm/321534728809

Assuming this is a functional replacement, you'd have to adapt for the different pinout.

[amyk]

That logo is not unlike the Advance logo on the multimeter itself, so probably a custom-produced part for them. AY-5-4057 is not a replacement as it's a 4-digit counter while this IC is only 3-digit.

You may try asking GI (Microchip) what it is, it's worth a try...

[Ian.M]

If you look closely at the DMM2002 7.pdf schematic, there are some neons in place of the fourth digit and it can apparently display 0, 1 or OL which is connected to the '2' cathode.  A four digit counter may be exactly what's needed.

However its not a cheap replacement and will require an adapter board so my original suggestion to use a PIC or AVR stands.

[SL4P]

What is interesting, is the mechanical nature of the cause.
As someone already suggested, a bonding wire, but the mass of those wires is tiny.

Have you scoped the pins and surrounding tracks?
I wonder if the problem could be a microscopic hairline crack on the ic pin or nearby...?
You mention that it comes and goes.  This may still be an option.

[android]

Someone on UK Vintage Radio Repair forum was kind enough to post the entire Advance Instruments DMM2 manual, which unfortunately lists IC3 as simply "LSI 4-Decade Type 1".

[amyk]

...so it is a 4-digit counter after all, but a custom part.

The CuL960 (which is similarly rare), is from the 1960s according to this link

[macboy]

I'd suggest trying one of those AY-5-4057 parts. It isn't a sure thing but it looks like a very good bet based on the original circuit schematic and the IC description in the catalog. Of course you will need to deal with the different pinout, but you would need to do that if trying to use a microcontroller as well. Consider the time spent writing the program to emulate the functionality of the circuit. It isn't terribly complex, but you would still spend some hours I would think. That assumes of course that you are set up with the tools and experience to bang it out in a timely and inexpensive manner. Is your time worth more than the $25 or so to buy the IC?

[Ian.M]

I'd want to see all the signals round that chip before going down that route.  There may be something nonstandard about how it handles gating and reset or similar, or the fourth digit may not be a full decade ,otherwise why did the manufacturer  pay extra for a custom chip?

[macboy]

I'd want to see all the signals round that chip before going down that route.  There may be something nonstandard about how it handles gating and reset or similar, or the fourth digit may not be a full decade ,otherwise why did the manufacturer  pay extra for a custom chip?

Why custom indeed, and yet they do it all the time.
I have a nice HP distortion analyzer here. Full of all manner of op-amps, marked with part numbers like 1826-0783 instead of NE5534. Not only are they custom made for HP with HP part numbers, but they are in fancy ceramic or metal can TO-99 packages. Even a TL072 is ceramic and HP numbered. All very expensive. Why? Maybe they like not being copied so easily. Maybe they just like feeling special. Special parts = special gear.

[merseyless]

Thought I might revive this thread with an update.
A mate of mine was able to lend me his Logic16 for a bit, from which I was able to capture some information.

Turns out that this thing takes the input voltage, uses it to charge a capacitor for 1000 cycles of one of the multivibrators, then counts how many cycles it takes for it to discharge. The number of cycles taken to discharge is equal to the input voltage, with the decimal point placement done in hardware. Very bloody clever.

Shouldn't be too hard to implement in software either, would be nice though to find something that fits the pinout.
I've attached the logic analyzer data and the table I did up of its relevant inputs and outputs to determine its operation.

Edit. When I did the table, only one of the nixies was lighting (center). So I changed the input voltage to get it to go through the entire range of values and matched up the logic outputted to the nixie driver to the numbers that appeared. Further to the right, I recorded the oscillator count and noted the value on the display from when the counter IC decided to co-operate.