Electronics > Repair
Series defect on agilent 167xx boards?
ahakman:
Yes, the comparators are right down by the pod cable connectors on the external side of the card, but I think the reference voltages come from a DAC that's way up close to the backplane connector. They run right under where one of the plastic runners was, and the trace was broken there. I had already bridged that trace with a wire, but it was completely disconnected from one side, either from cleaning the board with q-tips and acetone, or it just didn't solder well the first time.
For others reading this thread, just because it says "comparator failure" and the comparators are down on the external connector side of the board doesn't necessarily mean that the problem is there. Always focus on the areas with the plastic runners and the areas around where they were.
And if you have memory data bus errors, focus on all of the 33 ohm resistor packs. Especially focus on any signs of corrosion where the pads turn into traces right at the edge of the solder mask opening for the pads.
These are the kind of self test numbers I like to see:
MarkL:
--- Quote from: ahakman on February 26, 2023, 10:24:50 pm ---Yes, the comparators are right down by the pod cable connectors on the external side of the card, but I think the reference voltages come from a DAC that's way up close to the backplane connector. They run right under where one of the plastic runners was, and the trace was broken there. I had already bridged that trace with a wire, but it was completely disconnected from one side, either from cleaning the board with q-tips and acetone, or it just didn't solder well the first time.
For others reading this thread, just because it says "comparator failure" and the comparators are down on the external connector side of the board doesn't necessarily mean that the problem is there. Always focus on the areas with the plastic runners and the areas around where they were.
...
--- End quote ---
100% agree. It's the path between the DAC and the comparators that can get severed which causes these problems. It's much more rare that it's an actual chip failure (although it has happened).
On the 16752A, the DAC is U39 (AD7841ASZ), and as you point out is near the backplane connector on the top. The traces to the comparators run on the bottom of the card on the very outer edge on the POD1/2 connector side. And as you say, they pass right under one of the runners. The crossing is adjacent to U90, and is a favorite corrosion spot.
There is a similar looking set of traces on the opposite edge, also on the bottom, and if corroded these can create board ID errors.
There's a mapping of DAC output pins to comparator inputs earlier in this thread for anyone needing to do the end-to-end continuity check:
https://www.eevblog.com/forum/repair/series-defect-on-agilent-167xx-boards/msg2720304/#msg2720304
fpgaarcade:
Bit of an odd request.
I've been hunting on ebay for a while for a dead 167xx or similar board - anything with the modern low density probe connector. Most of them are in the US and the shipping costs a fortune.
I am producing a new high end FPGA board for retro gaming, and it has a daughterboard slot. I'm thinking about using the HP front end comparator and circuit around it, connected to the FPGA.
I should be able to get speeds of @1.6Gb per channel.
I see the pinout of the comparator is quite well understood but has anybody drawn a complete schematic yet?
Does it vary much with the highest speed boards, say the 2GHz 16751a?
I could do with the cable as well, I have some probes to play with..
I'm in Sweden and happy to pay for parts - although I do need at least one intact front end with all the parts present.
One reason is to make a modern logic analyser with huge depth, but the other is for real time debug/emulation of chips in arcade boards.
Thanks for reading.
Mike.
www.fpgaarcade.com
mike@fpgaarcade.com
MarkL:
--- Quote from: fpgaarcade on March 22, 2023, 10:30:10 am ---Bit of an odd request.
I've been hunting on ebay for a while for a dead 167xx or similar board - anything with the modern low density probe connector. Most of them are in the US and the shipping costs a fortune.
I am producing a new high end FPGA board for retro gaming, and it has a daughterboard slot. I'm thinking about using the HP front end comparator and circuit around it, connected to the FPGA.
I should be able to get speeds of @1.6Gb per channel.
I see the pinout of the comparator is quite well understood but has anybody drawn a complete schematic yet?
Does it vary much with the highest speed boards, say the 2GHz 16751a?
I could do with the cable as well, I have some probes to play with..
I'm in Sweden and happy to pay for parts - although I do need at least one intact front end with all the parts present.
One reason is to make a modern logic analyser with huge depth, but the other is for real time debug/emulation of chips in arcade boards.
...
--- End quote ---
Not an odd request at all; an interesting idea! The probing is not a trivial piece to get right. Why not take advantage of a probing system that's already out there and readily available.
The 16715A, 16716A, 16717A, 16718A, 16719A, 16740A, 16741A, 16742A, 16750A/B, 16751A/B, and 16752A/B all use the same front end design, including the DAC. In fact, many of the cards are identical and only differ by model setting resistors. The 16753A, 16754A, 16755A, 16756A, 16760A use a different front-end design and comparator.
The former all support 2GHz Timing Zoom (except the 16715A which has unpopulated areas for it), but that's just the sample clock rate. If you're shooting for 1.6Gbps, you should take note that the max state capture is 400MHz in the 16750/1/2 cards, and channel-to-channel skew is only specified as <1.0ns. Besides the acquisition ASICs, the front-end could be contributing to those limits. When you get a board, you might want to measure the actual switching characteristics of the comparators in their natural habitat before proceeding with a design.
Unfortunately schematics don't exist. Some of the passives connected to the incoming data lines are unlabeled and would need to be measured with appropriate high-frequency gear. The easiest approach would probably be to duplicate their layout exactly and lift all the front-end components from the board. Length-matched traces may include some post-comparator delay compensation, so you may need to tweak lengths in your final design.
I'm in the US, so I'm unfortunately in the category of "costs a fortune" shipping.
fpgaarcade:
Thanks for the detailed response.
I can get hold of a working 1680A for a bit which looks to use the same front end? and I can probe around that. I've got access to a decent 'scope at work.
The test mode feature of the comparator is interesting, I should be able to use that to compensate for delays between the front end and the FPGA.
I doubt I'll get the layout quite as good as the original, but hopefully sufficient.
The MPSoC Xilinx device I am using is quite a beast, with Ethernet, USB and a couple of built in ARMs. It will be able to stream the captured data direct to the connected DDR4 memory.
Navigation
[0] Message Index
[#] Next page
[*] Previous page
Go to full version