Repair of a 8201 function/pulse generator

[theHWcave]

I got this Global Specifications 8201 Function/Pulse Generator as a “display works no output” unit and took it on as a challenge. 
This type of generator was also sold under other names like Kontron in Germany.

This is a beast of a machine from the mid 80’ties, containing no less than 135 ICs (mostly opamps) and 176 transistors/FETs and hundreds of diodes, more than 250 capacitors, all spread out over one large horizontal main board and 4 additional vertical boards.

There is an ancient 8031 microcontroller with separate EPROM and RAM chips but the actual signal generation is done the old fashioned way. A VCO produces a selectable charge/discharge current into selectable capacitors to produce a triangular wave from 0.002Hz to 20 MHz, which is then shaped to create a sine as well as a rectangular wave. The rectangular wave goes then into a complex circuitry to change pulse width, delays and, if an option is installed, rise and fall times of the slopes. This option is present in mine so this is actually an 8241 model.  In either variant, there is an additional frequency counter capability and it looks as if the CPU is borrowing this function internally to check if the selected frequency is actually generated. At the end is an amplifier board with selectable DC offset and selectable gain or rather attenuation for a 15V peak output into 50 Ohm.

In all the above, “selectable” means that the microcontroller sends a clock and serial bit patterns to lots of 8-bit serial to parallel shift registers of type CD4094 and a strobe to convert the currently clocked-in pattern to the 8 outputs, which then change things, mostly by using FETs to switch resistors/capacitors or activate relays.  Some shift registers are chained to get 16 or 24 bit words.

The first challenge was to get some documentation and schematics for which I have to thank fellow EEVBlog member TonyBe who has a Kontron-badged 8201.

I documented the whole fault-finding, repair and final tests and adjustments in two YouTube videos.

Part 1: https://youtu.be/LqNo7llgOw8
Part 2: https://youtu.be/axHzIyAz6N8

The videos are quite long since they follows the process as it happened, including me going down the wrong path occasionally… 

There are two hard learned lessons for me out of this: 

• Don’t believe the schematics are 100% correct. If a vital signal on an IC pin is marked “not connected” on the diagram, check the PCB if that is indeed the case. 
• Read the complete theory of operation section from first to last page, in case the authors casually mention a vital fact about function X in a section that has absolutely nothing to do with X, while the description of X doesn't mention it at all

Spoiler alert, stop reading now if you’d rather watch the video

One tantalum capacitor on the main board had shortened the -15V rail to ground and while replacing all capacitors, I found another tantalum capacitor with burn marks that had failed open. Two AD7533 chips (these are multiplying 10bit DACs) had failed, one in the circuit adding offset to the signal and one in the current generator. One LM311 comparator had failed which was responsible to engage the first 10dB signal attenuator in the final amplifier. Finally, the fan, an 115V AC type, was beginning to fail but lubricating the bearings fixed that, at least for now.

Yes, maybe apart from the beefier output amplifier, and the capability to change pulse rise/fall times, any half decent modern function generator can probably do all of what the 8201 can do and much more, on multiple channels, and more accurately ... , but as a lesson in circuit design this thing is great and I learned a lot. I am glad I rescued it from the skip and will keep it for a while, probably giving it away when I run out of space.