LC584AL arrived today but channel 4 is not working :(

[analogRF]

I finally received this beauty today. I bought it as-is with 4 AP020 FET probes for a good price hoping it will all work!
Everything including the 4 probes work perfectly EXCEPT one issue and that is
the channel 4 is out of whack.

The other 3 are excellent but ch4  gives an OVERLOAD message and turns to Grounded mode as soon as I connect it to anything
with any probe (perhaps to protect itself) and says "clear the overload" but of course there is no overload.
The gain and offset on that channel are totally out of order.
I have not opened the beast yet and I don't have the service manual or schematics.

Now I have 2 questions:

can I rely on LC684 or LC574 schematics for this issue? My guess is that the input stage(s)
on these scopes should be the same? any thoughts?

any thoughts on how to approach this repair job? any recommendation is greatly appreciated.
any past experience on a similar issue is a great help

thank you so much

p.s. I have posted this message on lecroy owners group as well. Hope you don't mind!

[tautech]

#1 suspect is input FET is damaged probably from an overvoltage event.
As you have 3 working channels it shouldn't be hard at all to troubleshoot.....follow the signal path until it doesn't match the other channels.

[JohnG]

At my last job, I had an LC574AM with exactly the same problem on Ch3 and Ch4. It was fixed under warranty the first time. It happened again a few months out of warranty, and a very helpful sales guy finagled me a new scope. He told me that there had been problems with relays on the inputs, and that he had seen it a number of times.

Hope this helps.

John

[analogRF]

ok, last night I did lot more tests (until this morning) and unfortunately the problem is deeper than what I thought at first.
I connected my Fluke 6060B signal generator (at -10dBm and -3dBm and 0dBm and various frequencies) to the scope via 20cm
coaxial cables (RG142) and set the scope on 50Ohm input coupling.
At first all was OK and I could play with the scope gain (V/div) and time base and the readings were consistent (now even channel 4 worked!!) but
sometimes I could hear the relays (those damn relays) sporadically stop clicking for a few seconds when they had to
and the scope became momentarily unresponsive and then come back to life again. After an hour or so, however,
the V/div started to behave erratically on all channels. For some V/div (like 50mV and 200mV) the trace would disappear
on the top of screen and I could not bring it back but it would appear correctly for the next V/div step (and disappear on the next one and so on...)
and it happened in the SAME WAY for all channels now. Note that in the beginning it was all ok  but after a while
maybe 1 hour or so the erratic behavior started.  It seems to me that the damn relays don't respond properly.

[Ice-Tea]

If you have one shorted out relay coil it may pull the the supply rail for them low? Better crack het open and start looking 

[analogRF]

If you have one shorted out relay coil it may pull the the supply rail for them low? Better crack het open and start looking 

Yes, you might be right but why it does not happen all the time? You see, right after I received the box I played with it for an hour or so
and the only problem was the channel 4 was not working (OVERLOAD error etc...) but everything else was OK. At that time I only used the AP020 active probes.
Then I came back home at night and stared again this time with 50Ohm port setting and using RF signal generator and again all was ok and even channel 4 worked! However, after an hour or so the erratic behavior started....

[analogRF]

can a firmware upgrade do any help?
the firmware is at 7.9.0. I cannot find the 9.3.0 for this model although I have seen it for LC574 I guess. Aren't they the same firmware?

[nctnico]

ok, last night I did lot more tests (until this morning) and unfortunately the problem is deeper than what I thought at first.
I connected my Fluke 6060B signal generator (at -10dBm and -3dBm and 0dBm and various frequencies) to the scope via 20cm
coaxial cables (RG142) and set the scope on 50Ohm input coupling.
At first all was OK and I could play with the scope gain (V/div) and time base and the readings were consistent (now even channel 4 worked!!) but
sometimes I could hear the relays (those damn relays) sporadically stop clicking for a few seconds when they had to
and the scope became momentarily unresponsive and then come back to life again. After an hour or so, however,
the V/div started to behave erratically on all channels. For some V/div (like 50mV and 200mV) the trace would disappear
on the top of screen and I could not bring it back but it would appear correctly for the next V/div step (and disappear on the next one and so on...)
and it happened in the SAME WAY for all channels now. Note that in the beginning it was all ok  but after a while
maybe 1 hour or so the erratic behavior started.  It seems to me that the damn relays don't respond properly.

That sounds like a contact problem to me. Take it apart and reseat every connector. Maybe clean card-edge connectors with alcohol. Under no circumstance use contact cleaner. These contain acids which will make things worse in the long run!

[tautech]

At first all was OK and I could play with the scope gain (V/div) and time base and the readings were consistent (now even channel 4 worked!!) but
sometimes I could hear the relays (those damn relays) sporadically stop clicking for a few seconds when they had to
and the scope became momentarily unresponsive and then come back to life again.

This sounds like an auto Self Cal procedure and yes after some years it may have worn out the relays.

In some units it's possible to turn this auto cal feature off as it can interrupt work flow, let alone infuriate the user when the scope becomes unresponsive.
The manual should have more about the auto self cal.

[snoopy]

ok, last night I did lot more tests (until this morning) and unfortunately the problem is deeper than what I thought at first.
I connected my Fluke 6060B signal generator (at -10dBm and -3dBm and 0dBm and various frequencies) to the scope via 20cm
coaxial cables (RG142) and set the scope on 50Ohm input coupling.
At first all was OK and I could play with the scope gain (V/div) and time base and the readings were consistent (now even channel 4 worked!!) but
sometimes I could hear the relays (those damn relays) sporadically stop clicking for a few seconds when they had to
and the scope became momentarily unresponsive and then come back to life again. After an hour or so, however,
the V/div started to behave erratically on all channels. For some V/div (like 50mV and 200mV) the trace would disappear
on the top of screen and I could not bring it back but it would appear correctly for the next V/div step (and disappear on the next one and so on...)
and it happened in the SAME WAY for all channels now. Note that in the beginning it was all ok  but after a while
maybe 1 hour or so the erratic behavior started.  It seems to me that the damn relays don't respond properly.

That sounds like a contact problem to me. Take it apart and reseat every connector. Maybe clean card-edge connectors with alcohol. Under no circumstance use contact cleaner. These contain acids which will make things worse in the long run!

Don't know about you but I have had great success with CRC contact cleaner for the last 3 decades as well as CRC 2-26. In fact when I fixed a older Tektronix TDS DSO quite a few controls had become dicky and I was only able to revive them by squirting CRC contact cleaner down a very small hole in each control otherwise it was cactus.

[nctnico]

The problems come later when the acids eat away the copper and other metals. I've seen several pieces of equipment become FUBAR because of contact cleaner. Contact cleaner is great stuff if you want to get an extra day from a piece of equipment but after that you better bin it.

[snoopy]

The problems come later when the acids eat away the copper and other metals. I've seen several pieces of equipment become FUBAR because of contact cleaner. Contact cleaner is great stuff if you want to get an extra day from a piece of equipment but after that you better bin it.

It may not be the contact cleaner doing this but rather it dissolving the crud into a solution that may cause the issues. Any chemists here could throw some light on this ?

[analogRF]

thank you guys for all your valuable inputs.
I figured how to disable the auto self cal as tautech mentioned and those relays are now much quieter and the unresponsive periods are gone.

However, to my surprise, channels 1-3 are working perfectly again for some reason
I only did a manual reset (button 1+atuto setup+return keys pushed together) and all the erratic behavior
of the other night seems to be disappeared. Even if I turn the auto self cal on, they work but since the relays keep
clicking all the time, I turned it off. I have worked with it for a couple of hours at a time (2-3 times) 
and they seem to be OK, at least for now. I did the impedance measurements and port leakage measurements according to the
service manual of LC684 and they are all very precisely within specs for channel 1-3. Also I tried to check the analog bandwidth
and although my test equipment and method was not highly accurate, to be useful for calibration for example,
but I can say with high confidence that the BW on those 3 channels is definitely larger than 1GHz (I cannot go beyond 1050MHz with my Fluke 6060A)
on all V/div ranges.

Now, if I assume the problems with channel 1-3 are gone (which baffles me), channel 4 is stuck to the ground and
no matter what I cannot change the coupling to 50ohm, or AC 1M or DC 1M. It jumps back to ground even though I don't see
any OVERLOAD error anymore. It just doesn't work

I will crack it open as soon as I get the time and report back. I will try cleaning and/or reseating all connectors and jumpers
and maybe the front panel rotary encoders if possible (with alcohol) but the issue with ch.4 is deeper than that I guess.

How can I test voltages on the main board? any idea? It seems to me by the time I can get access to the main board
I have to disassemble the whole thing so there will not be any power.

[tautech]

Does the manual go to schematic level ?
Can you provide a link for it ?

Now, if I assume the problems with channel 1-3 are gone (which baffles me), channel 4 is stuck to the ground and no matter what I cannot change the coupling to 50ohm, or AC 1M or DC 1M. It jumps back to ground even though I don't see any OVERLOAD error anymore. It just doesn't work

Which suggests (to me) there's either a supply issue or the signal path is open....dead enable relay or transistor/FET etc.

Good progress. 

[analogRF]

Does the manual go to schematic level ?
Can you provide a link for it ?

Now, if I assume the problems with channel 1-3 are gone (which baffles me), channel 4 is stuck to the ground and no matter what I cannot change the coupling to 50ohm, or AC 1M or DC 1M. It jumps back to ground even though I don't see any OVERLOAD error anymore. It just doesn't work

Which suggests (to me) there's either a supply issue or the signal path is open....dead enable relay or transistor/FET etc.

Good progress. 

I can only find LC684 manuals here:
http://www.ko4bb.com/manuals/99.254.215.1/LeCroy_LC664_LC684_Digital_Storage_Oscilloscope_Service_Manual.zip
and people say it's close to LC584, so I will have to rely on that 
Yes, it has the schematics of the input stages and many other parts.
However, I doubt this is useful because LC684 is a 1.5GHz scope, so the input stage must surely
be different. Maybe the power supply or CPU boards are the same...

LC574 service manual is also available and I believe that must be closer to 584. At least both are 1GHz but different samplig speed etc...

I hope I can measure voltages on the main board. At this point, by reviewing the disassembly instructions, it seems that
there is no way to get access to the main board without taking everything apart.

[JPortici]

You should contact lecroy and ask them, they are always willing to help people with obsolete scopes with manuals, software and whatnot

[tautech]

LC574 service manual is also available and I believe that must be closer to 584. At least both are 1GHz but different samplig speed etc...

From the Readme in the schematic folder:

Note the 554/564 series are very similar to the 684 series (except for bandwidth) so those schematics
can be used for troubleshooting purposes.

Aslo the 684 series manual includes schematics of the power supply that is used in several models.

You worry unduly IMO, this manual should get you through. 
While the schematic might not match exactly the layout and signal path might be very similar and as you have 3 other working channels to check measurements against I say rip it completely apart and lay it all out on the bench.....you might even be able to connect it up frankenstein style. 

Pics when you do please. 

[grumpydoc]

.....you might even be able to connect it up frankenstein style. 

Pics when you do please.

If you do make sure that there is a fan aimed at the board - this series of 'scopes needs good airflow or they will fry themselves.

[tautech]

For a look at a channel input see Schematics/9615-3/F615-3 Schematic a.pdf on P20 from the ko4bb link.
There's 3 relays in the signal path, any of which could break the signal path if they were crook or their driver circuitry wasn't working.

[analogRF]

ok guys, I gathered all my strength and opened this beast of oscilloscopes on Saturday morning and I have not slept for 2 days
so please bear with me 
I have taken lots of photos. I am posting a few of them here which show you the complete picture.

First of all, as I had predicted, LC684 schematics are totally useless with this scope as there is no resemblance between the two.
The front end circuitry is completely different but everybody here and on Lecroy Yahoo group said otherwise.
So using LC684 schematics was a no go.
However, as I had mentioned myself, LC574 schematics was much closer to LC584. In fact I found perfect match between the two in most
parts except the values of components (resistors) were different. Sure, there were some differences between the circuits, too, but it was minimal.
Their digital and processing parts could be very different though, I did not check.

The worst thing is that when you open up this thing, there is pretty much NO WAY to power it up outside the box and measure voltages.
Sure, I can imagine building a test gig with tons of jumper wires to connect them all up and put a big fan to cool it down but
I could not go that way as it is too scary and any slightest mistake would fry something. Besides I didn't have so many jumper wires to
connect all those connectors (see the pictures. The CPU board and power supply sit directly on top of the main board and their connectors
mate directly on board with no ribbon connectors)
So I am only restricted to DC OHM measurements and some DIODE measurements on board. Even this is scary and I am afraid that might
damage something but that is the only thing I can do for now. My multimeter is an Extech MM-570.

I have done hundreds (literally) of ohm measurements on board (only on the input front end sections) and the 4 channels looked all the same
everywhere except for one sub-circuit that I could notice a difference between ch.4 and the others. I have re-drawn that sub-circuit the way I extracted from the board and as you can see it is pretty much identical to the one from LC574 except for one small difference
at point C (the potentiometer does not exist in LC584).
Interestingly this sub-circuit goes on to make the OVERLOAD signal! However, note that it's all DC path and there is no SIGNAL coming in.
In fact the input signal from the BNC seems to have nothing to do with making the overload signal directly. In my opinion (correct me if I am wrong)
it is all based on temperature sensing of some sort. It is the same with LC684 albeit with a different circuitry. The BAV99 (CR4003) is
positioned very close to the input BNC and has a fairly large pad on board and as you can see in the pictures it is buried underneath
a gel/silicon type of material that I cannot figure what it is. The BAV99 and several components around it are covered with this gel/silicon
thing and the input relay (RL4000) is on the other side of the board beneath the BAV99 (CR4003). So my guess is that BAV99 is used as some
kind of temp sensor here. I think when it warms up it is supposed to generate something other than zero at pin 2 of AD705 (A4010) (point C) but
it could go either way (pos/neg), no?
One more thing is C4075, it is connected to ground on both sides! On board, there is a 4-5cm distance between the AD705 (A4010) which on
the other side of the board (underneath) and the BAV99 which is on top side and close to the BNC. You can see them all in pictures. Note that
there is another AD705 (A4001) on the top side of the board but that belongs to a different path. By the way, the board seems to be 2 layer (3 metal layers) since I cannot find some of the traces. They must be buried in the middle layer I guess.
Now, I have measured DC resistance between various points on this channel and compared with other channels and here are the
values (refer to my drawing of the schematics in lecroy1.pdf):

RAB=1.85K on ch.1,2,3 / 1.994K on ch.4
RAC, RBC=0.994K,1K on ch.4 / 0.961K,0.961K on others
RCD=14.25K on ch.4 / 7.28K on others
RDG=0.931K on all chnnels
RCG=13.45K on ch4 / 6.41K on others
RKB=10.42K on ch4 / 6.35K on others
RLA=14.46K on ch4 / 6.29K on others

DIODE measurements on BAV99 (I pushed my very sharp multimeter probes slightly into that silicon cover to touch the BAV99 legs)
left diode (E to ground) VF=0.57V and VR=1.3V on all channels
right diode (F to ground) VF=0.57V on all channels but VR=1.49V on ch4 and 1.29V on others ! ??

So my guess so far is that one diode in that BAV99 is faulty to some extent and therefore an OVERLOAD signal is created all the time.
In other words at point C there is an imbalance all the time due to asymmetry created by the schottky. what do you think? Or is it
the AD705 (A4010)??

On the other hand, I might be barking at the wrong tree here.... as I mentioned just based on DC OHM measurements at many many
points on the front ends I could not find any difference between the channels except the one above. However, the relays are all set to
one position and I cannot evaluate how they actually work. The current state of relays is exactly as shown in the schematics of LC574 and
the contacts that are closed right now are very good (<0.15ohm)

with that silicon thingy that covers the elements around the schottky (BAV99) I cannot even take it out and I am afraid peeling that
thing off the board....

here are the links to photos:
https://www.4shared.com/s/fvTYBKnI-ei
https://www.4shared.com/s/fu6OuiBGxei
https://www.4shared.com/s/fFoDdoemPei
https://www.4shared.com/s/f6o9hpmDPei
https://www.4shared.com/s/fcobbqHR5ei
https://www.4shared.com/s/fn5KSmAMica
https://www.4shared.com/s/fLyntxaqRca
https://www.4shared.com/s/ffqxR2ggtca
https://www.4shared.com/s/fsPtP9L1eei
https://www.4shared.com/s/ffqtw41Gdca
https://www.4shared.com/s/fxr3y4EHtca
https://www.4shared.com/s/fGS6d9p8Kei
https://www.4shared.com/s/fOa6fMVmFei
https://www.4shared.com/s/fp0-Ma9K1ca
https://www.4shared.com/s/fi88cI8lUei
https://www.4shared.com/s/fBqC9yKQRei
https://www.4shared.com/s/fdCxumxGyei
https://www.4shared.com/s/fIKY_-nlfca
https://www.4shared.com/s/fMB9mSCUWei
https://www.4shared.com/s/f9fTVNixVca
https://www.4shared.com/s/fA1QxfiXMei
https://www.4shared.com/s/fGUeQMYfpei

[analogRF]

....I just hope when/if I put all these back together, the thing works at least as good as before   
I think I should be happy if only the other 3 channels work as before 

[tautech]

DIODE measurements on BAV99 (I pushed my very sharp multimeter probes slightly into that silicon cover to touch the BAV99 legs)
left diode (E to ground) VF=0.57V and VR=1.3V on all channels
right diode (F to ground) VF=0.57V on all channels but VR=1.49V on ch4 and 1.29V on others ! ??

So my guess so far is that one diode in that BAV99 is faulty to some extent and therefore an OVERLOAD signal is created all the time.
In other words at point C there is an imbalance all the time due to asymmetry created by the schottky. what do you think? Or is it the AD705 (A4010)??
On the other hand, I might be barking at the wrong tree here.... as I mentioned just based on DC OHM measurements at many many points on the front ends I could not find any difference between the channels except the one above. However, the relays are all set to one position and I cannot evaluate how they actually work. The current state of relays is exactly as shown in the schematics of LC574 and the contacts that are closed right now are very good (<0.15ohm)

with that silicon thingy that covers the elements around the schottky (BAV99) I cannot even take it out and I am afraid peeling that
thing off the board....

I think you are correct, temp sensing.
The silicone (BTW will be neutral cure type, NOT acid cure) will be to heat insulate the diode from heat from other sources. A sharp tool can remove it, just be sure to add some more later but first measure:

If you can measure points E and F in relation to ground when powered on, they should be exactly the same.

[analogRF]

DIODE measurements on BAV99 (I pushed my very sharp multimeter probes slightly into that silicon cover to touch the BAV99 legs)
left diode (E to ground) VF=0.57V and VR=1.3V on all channels
right diode (F to ground) VF=0.57V on all channels but VR=1.49V on ch4 and 1.29V on others ! ??

So my guess so far is that one diode in that BAV99 is faulty to some extent and therefore an OVERLOAD signal is created all the time.
In other words at point C there is an imbalance all the time due to asymmetry created by the schottky. what do you think? Or is it the AD705 (A4010)??
On the other hand, I might be barking at the wrong tree here.... as I mentioned just based on DC OHM measurements at many many points on the front ends I could not find any difference between the channels except the one above. However, the relays are all set to one position and I cannot evaluate how they actually work. The current state of relays is exactly as shown in the schematics of LC574 and the contacts that are closed right now are very good (<0.15ohm)

with that silicon thingy that covers the elements around the schottky (BAV99) I cannot even take it out and I am afraid peeling that
thing off the board....

I think you are correct, temp sensing.
The silicone (BTW will be neutral cure type, NOT acid cure) will be to heat insulate the diode from heat from other sources. A sharp tool can remove it, just be sure to add some more later but first measure:

If you can measure points E and F in relation to ground when powered on, they should be exactly the same.

powering it up and measuring voltages is basically out of question.
VE and VF must be precisely opposite each other (their sum must be zero when the AD705 adds them up)
but why heat will create imbalance? If the diodes are matched, their voltages will change
equally when heated and their sum will still be zero? no?
another issue is that, this design is extremely prone to any mismatch between the two diodes in the package
because AD705 adds the sum of the diode voltages with a -100 multiplication factor! So over time this design is prone to failing I guess...

about the silicone, do you know what is called exactly? where can I buy this stuff? how is it applied?
can I re-use it after I peel it off?


 

[analogRF]

on another thought, I should say AD705 is not a comparator here. It adds the two diode voltages with a factor of -100 and then there is an attenuator
and then this amplified/attenuated voltage goes to a comparator (LM339) which compares it with a reference voltage and produces a LOW/HIGH value which goes to the digital ICs...
what I cannot understand is that when the diode is heated (1) why should there be a non-zero difference between their forward voltages if the diodes are matched? (2) this voltage difference can go either way, I mean their sum can become a small positive or negative value. So how is the decision made for overload signal?

[analogRF]

DIODE measurements on BAV99 (I pushed my very sharp multimeter probes slightly into that silicon cover to touch the BAV99 legs)
left diode (E to ground) VF=0.57V and VR=1.3V on all channels
right diode (F to ground) VF=0.57V on all channels but VR=1.49V on ch4 and 1.29V on others ! ??

So my guess so far is that one diode in that BAV99 is faulty to some extent and therefore an OVERLOAD signal is created all the time.
In other words at point C there is an imbalance all the time due to asymmetry created by the schottky. what do you think? Or is it the AD705 (A4010)??
On the other hand, I might be barking at the wrong tree here.... as I mentioned just based on DC OHM measurements at many many points on the front ends I could not find any difference between the channels except the one above. However, the relays are all set to one position and I cannot evaluate how they actually work. The current state of relays is exactly as shown in the schematics of LC574 and the contacts that are closed right now are very good (<0.15ohm)

with that silicon thingy that covers the elements around the schottky (BAV99) I cannot even take it out and I am afraid peeling that
thing off the board....

I think you are correct, temp sensing.
The silicone (BTW will be neutral cure type, NOT acid cure) will be to heat insulate the diode from heat from other sources. A sharp tool can remove it, just be sure to add some more later but first measure:

If you can measure points E and F in relation to ground when powered on, they should be exactly the same.

is this the type of material that is on the board covering those diodes and surrounding area?
http://www.ebay.ca/itm/Kafuter-100g-K-5902T-transparent-silicone-industrial-adhesives-waterproof-glue-/292277339286?hash=item440d164496:g:IicAAOSwaB5XvAD3

[tautech]

is this the type of material that is on the board covering those diodes and surrounding area?
http://www.ebay.ca/itm/Kafuter-100g-K-5902T-transparent-silicone-industrial-adhesives-waterproof-glue-/292277339286?hash=item440d164496:g:IicAAOSwaB5XvAD3

If what's on the board is pliable and not hard, yes. That ^ seems ideal. 
If it is hard it may be hot melt glue.

WRT BAV99, have you looked up any spec to indicate that is indeed temp sensitive ? Maybe there's a temp graph in the datasheet.
I'm struggling too to get my head around why it's where it is and covered in silicone ? 
Yes, point C should be 0V, sorry for the previous confusion.

[analogRF]

yes, I did take a look at BAV99 specs. They are labeled as ultra-fast silicon switching diodes
which I don't know why it should be the choice in this particular circuit. They dont seem
particularly more temp sensitive than other diodes. There are a couple of graphs showing the
VF or IF vs T but they look like any other diodes to me, not particularly temp sensitive.
For example in one graph from Fairchild, Vf drops from 0.75V at 25C to 0.70V at 50C
and then to 0.60V at 110C (all at 10mA forward current which is close to the 11.3mA used in this circuit)

yes, the silicone on the board is kinda soft and I can poke into it by multimeter probes
but not that soft which can be removed by hand, for example.

[analogRF]

...but the thing is if the two diodes are matched, even if their Vf changes with temperature still their sum will be zero!

besides, is this really a good idea for protection against overload? I mean, if overload happens
it takes a while for the diodes to get warm and do their thing and by that time the damage is done!
there must be another path for overload detection, too, I guess but I cannot see it from the schematics yet...

LC684 is using the same concept as this one except it uses LM60 temp sensor instead of BAV99 diodes.

I am not even sure if juts purely resistance measurements + multimeter diode measurements on board is a reliable way
to tell me this diode pair are at fault here except that the values I get are different that the other three channels.
besides, when I did the in-circuit diode measurement, the forward bias was OK and exactly as other channels
the reverse was off but in-circuit diode testing on reverse bias is kinda meaningless and totally depends
on the surrounding elements...so concluding that the diodes are at fault here may not be correct at all...

[nctnico]

If you see different readings between working and non-working channels you are very likely in the part of the circuit where things are wrong. There may be multiple components which are defective. If you have good soldering skills you could replace all the semiconductors in the front-end (presumably some diodes and transistors/fets).

[analogRF]

I thought I'd give a final update on the repair of LC584AL:

After I found out the overload protection circuit is at fault and extracted the schematics (see post #19) I could not explain how that circuit could protect anything!
If the double diodes are perfectly matched, any change in temperature would not trip the opamp (their sum voltage remains zero). Anyways, I went ahead and simulated this circuit in NI Multisim (I have access to at work) and still could not figure how this thing can protect again overheating of the 300 ohm resistor. I think this thing does not protect against anything unless there is some mismatch, but not too much of it, between the diodes.

If they are mismatched by a lot the opamp trips at room temp which is what has happened to my scope I guess and if they are perfectly matched then nothing will happen with temperature. In LC574 there was a potentiometer to adjust this circuit but it is eliminated in LC584. Besides, even if it were to work as intended, it would take a while for the over-voltage at input to warm up the 300ohm resistor enough so the diodes voltage difference can trip the opamp. By that time the damage is probably done. So the ability of this circuit to protect the input is very low (none). And it only works at 50Ohm coupling, however, if it trips, the input becomes disconnected and does not work on 1M either until the overload is cleared.

So the solution for me was to just DISABLE this subcircuit instead of going through the painful and expensive job of removing the silicone putting and replacing the diode and re-applying the silicone. The easiest and cleanest way to do this was to just remove the 4.7K resistor (R4081 at the output of AD705 on the back of the mainboard). This way the output of LM339 comparator remains high and the input relay will not disconnect the port anymore. At the same time I have not changed the biasing conditions of any of opamp/comparator pins. So this repair costs nothing!

I did this and the scope works like a charm. All 4 channels work perfectly and I wish I had done the same on other channels when it was open.
I borrowed a GPIB cable from work and updated the firmware to 9.3 (last version) and some of the bugs I had noticed in operation are now gone.
The scope has so many software options and looks pretty much brand new :-)
The floppy also works!

thanks for your comments and help.