HP 3457A Opto-coupler/opto-isolator replacement

[SvanGool]

Found an HP3457A for 95$ on Ebay "As-is", the unit had a problem which was described as "Adrs 22 Failed".
I had a suspicion and bought the unit. On checking the unit, it gave immediately FAILED after showing the GPIB address.

After some online research and checking with the scope, the unit had no communication between the digital board and the analog board, this is a bi-directional serial signal of app. 190 Kbit between processors on each board. At power on, the digital board probes the analog board but there was no signal on the analog side of the opto-isolator. Conclusion: opto-isolator broken.

The problem is that this opto-isolator is unobtainable. The HP3457A analog board is fully isolated from chassis ground and digital board, there for the opto-isolators and the powering by separate transformer windings.

After quite some prototyping/experimenting I found a good "drop-in"-ish replacement, with good isolation voltage and decided to replace both opto-couplers. The replacement of the original KT3172/HP#1990-1075 (A2 U509 and U510) is the SMT opto-isolator FOD-8160. This one can be glued upside-down on top of the "opto-gap" (see picture) and then the output can be wired 1:1 (top view) and for the LED-side cross-wise (top view).

The unit works great after that (last calibration 2014).

 



Schematic:



(correction still to be done on above circuit:  +/- input symbols of output Opto-Coupler OpAmps are reversed, no impact to the above)




 

[Ordinaryman1971]

Hi, I have the same problem with mine. Mine is failing right after the GPIB test. It gives the "Fail" message on the display.
Can I ask you to post better screenshots of the fix. I can't see clearly, they are out of focus.
I've also tried to search for the opto-isolator you've picked but I can't find it.
Would any optoisolator work in this solution or you need some conditions to be meet for it to work.
Also, I think the picture and the schematic you've posted are 180 degrees reversed.
I've seen that on the picture the U509 is on the top when on the schematic you have have it on the bottom... am I right?
Good fix, thank you for posting.

[SvanGool]

Sorry for the late reply, was not tracking the thread.

Seek for "FOD8160" at mouser.com or digikey.com and there you go, they have more than you need.
I would certainly not use "any" opto-isolator because of the speed of the serial signal and the isolation voltage, most opto-isolators/-couplers will be no good.

I have no better pictures and dis-assembling the 3457A, at PCB-level, is not something I am eager to repeat.

There are different versions of the schematic of the 3457A (dependent of the serial number) but if you want to make this repair, you already have dissected (work anti-static) your 3457A and you can verify the schematic and the above pictures, I have supplied. For my 3457A everything is correct, it might be that things are switched on the top or bottom of the schematic (a schematic is most of the time not according the topology of the circuit) vs pictures, it should not matter for the repair.

The only thing important is that the FOD8160 are mounted "dead-bug" style (component on its back) and some wires were crossing.
I think the phrase "the output can be wired 1:1 (top view) and for the LED-side cross-wise (top view)" about the opto-coupler, says it all. If you use the FOD8160, there is nothing more to it.

Success !

[almarghi]

For those interested, TLP2362 from Toshiba also works fine.
(some wires crossed)

[peterkos2]

Hello I also have a problem after startup-- ADDRESS-22. then FAILURE How can I measure whether the communication between the boards is taking place? well thank you

[almarghi]

I visualized the communication signals with an oscilloscope and they were there, so first I suspected the processor on the analog board. (There is a very good thread on it on this forum.)
But then, on further examination, I saw that the signals were not high enough, so I replaced the optocouplers and it worked.

[peterkos2]

For those interested, TLP2362 from Toshiba also works fine.
(some wires crossed)

Good day. Which are crossed-GND and Vcc?

[almarghi]

It depends if you mount it upside down (like me) or normal. Either way you always have to cross 2 pins on one side.
I mounted it upside down, so I crossed the supply pins on the receiver side.

[peterkos2]

Great, thank you, it worked, now the multimeter works. When I go to measure resistance and connect the measuring cable in AUTO mode, is it possible that it takes almost 5 seconds and shows the result?

[almarghi]

Yes, it takes some time, especially on Auto. When I sort resistors, I lock the range and it is somehow faster, but don't expect miracles...