Repairing an HP 8593E spectrum analyzer

[giosif]

Hi,

I was fortunate enough to stumble, one day, upon a HP 8593E SA left in a dumper. 
First thing I noticed was it wouldn't keep the time across power cycles and would always start with default settings.
That was easy to troubleshoot and fix - the CR2477 battery solered on the memory board had died and needed replacement.

Following that, I tried to perform a frequency and amplitude self-calibration, which failed with no calibration signal found.
Checking the CAL OUT output, the signal is there, it's at roughly 300 MHz (as it should be), and at the right power level (harmonics too).
Still, when connecting CAL OUT to the INPUT, I see absolutely nothing at or around 300 MHz.
Going through the service guide and test procedures, I found that the +10V and -10V reference detector tests fail (I get the same looking ramp for both +10V and -10V, instead of what the guide indicates there should be).
Also, the FM coil driver test for SPAN below or at 10MHz fails as well (I get a horizontal line, instead of the ramp in the guide).
The other tests I was able to run were all ok.

Now, where I got stuck is the verification of the +10V and -10V references.
According to the manuals, those are being generated on the A7 analog interface board.
However, looking through the CLIP manual, at the schematic for the A7 board, in the area called Voltage References, I can't figure out where those reference outputs reside.
Could someone please confirm whether they are the outputs of U32 and U29, respectively?
If so, what is the reference ground for those?
Is it the chassis ground or something else?

Thanks,
George

[uncle_bob]

Hi

The "self calibration" on these analyzers is very limited. The real calibration data was all stored in battery backed up RAM. I have several of their cousins and they all run the same way. If you get the beast back up and running, it still will have some significant accuracy issues. The only way to correct them is to send the unit back to HP. They put it on a factory only test set and shoot new data into the RAM. The gotcha with that approach is that the test set is long gone and HP no longer supports that line of analyzers (as in support has been gone for *many* years). There is no documentation on the data they put in RAM and it is very specific to an individual instrument.

Everything I have is sitting in storage so getting one out to start probing points is not very practical. The ground reference is inevitably chassis ground.

Bob

[CJay]

Hi,

I was fortunate enough to stumble, one day, upon a HP 8593E SA left in a dumper. 
First thing I noticed was it wouldn't keep the time across power cycles and would always start with default settings.
That was easy to troubleshoot and fix - the CR2477 battery solered on the memory board had died and needed replacement.

Thanks,
George

OK, I needd to know where this dumpster is, I will travel 

[giosif]

Hi

The "self calibration" on these analyzers is very limited. The real calibration data was all stored in battery backed up RAM. I have several of their cousins and they all run the same way. If you get the beast back up and running, it still will have some significant accuracy issues. The only way to correct them is to send the unit back to HP. They put it on a factory only test set and shoot new data into the RAM. The gotcha with that approach is that the test set is long gone and HP no longer supports that line of analyzers (as in support has been gone for *many* years). There is no documentation on the data they put in RAM and it is very specific to an individual instrument.

Everything I have is sitting in storage so getting one out to start probing points is not very practical. The ground reference is inevitably chassis ground.

Bob

Thank you for the details, Bob!
At this stage, I am not too fussed about high accuracy - I'd be happy if I can make it work at all.
I can confirm the ground reference is chassis ground.
And I checked the +10V and -10V references and they all look good to me.
I think the failed reference tests might be a red herring and the issue is somewhere else.
I tested with a wireless AP and I see no signal in either 2.4 GHz or 5 GHz frequency bands.
Also, the CAL OUT signal which is being generated and sits at 300 MHz is not seen either, when connected to the main input.
I'll be looking at the block diagrams next and see if I understand the signal path.

Regards,
George

[giosif]

OK, I needd to know where this dumpster is, I will travel 

Sorry, you need to get your own magic dumpster. 

[Kilrah]

There is no documentation on the data they put in RAM and it is very specific to an individual instrument.

It's actually quite simple, the service manual lists the procedure to access the cal data, and what to back up with a form to fill to avoid running a recal when changing processor board. Just a few correction constants for timebase, attenuator steps and a bunch of them all over the instrument's range.
I've got a 8593A and backed up my constants in case the battery decided to die unexpectedly.

Not helping him much, BUT given the simplicity of the cal data I believe it would be feasible to put it back in reasonably good state by comparison with a known good instrument...

[SaabFAN]

Isn't there a procedure to recalibrate the instrument by yourself when you have a good RF-Generator?
If the cal-memory just stores corrections then it should be just a matter of sweeping the whole frequency range at a known level in sync with the Sweep of the Analyzer and then put data into the RAM until it shows a flat line on the screen.

[MarkL]

...
Going through the service guide and test procedures, I found that the +10V and -10V reference detector tests fail (I get the same looking ramp for both +10V and -10V, instead of what the guide indicates there should be).
...

If you're referring to the Service Diag menus, you should have a flat line for both +10V and -10V reference (not a ramp).  Make sure you've locked the analyzer to the 0-2.9GHz band before running that test.


Some ideas for you:

Since it lost its cal data and config, you can try DEFAULT CONFIG under the CAL menu, and then DEFAULT CAL DATA also under CAL menu.  For the latter, set freq to -2001Hz before pressing CAL.  This serves as a "passcode" (see pg 234 and 582 in Assembly Level Repair for details). 

You could also try setting the passcode -37Hz and doing a CAL FREQ (pg 209 and 509).

Does trying sample (as opposed to peak) detector setting help?

Can you see the LO peak at 0Hz?

When you connect the CAL OUT to the INPUT, can you see the signal at the output of the internal attenuator?

[giosif]

If you're referring to the Service Diag menus, you should have a flat line for both +10V and -10V reference (not a ramp).  Make sure you've locked the analyzer to the 0-2.9GHz band before running that test.

Hmm... I don't think I had locked the SA to that band before the +10V and -10V references tests.
And I don't remember the manual stating that.
I'll give it a try next time I get to troubleshoot this beast (weighs a ton as well).
In any case, I don't think these are an issue or the issue at this point in time.

Some ideas for you:

Since it lost its cal data and config, you can try DEFAULT CONFIG under the CAL menu, and then DEFAULT CAL DATA also under CAL menu.  For the latter, set freq to -2001Hz before pressing CAL.  This serves as a "passcode" (see pg 234 and 582 in Assembly Level Repair for details). 

You could also try setting the passcode -37Hz and doing a CAL FREQ (pg 209 and 509).

Does trying sample (as opposed to peak) detector setting help?

These things were mentioned in the manuals as well and I think I've tried most if not all of them (maybe not exactly as you describe them).
I'll give them another shot, but only after I troubleshoot the last point below.

Can you see the LO peak at 0Hz?

Yes.

When you connect the CAL OUT to the INPUT, can you see the signal at the output of the internal attenuator?

Ah, you see, that's the thing: there is no CAL OUT signal displayed when I connect it to INPUT.
And I know the CAL OUT does output the right signal - I was able to confirm it using another SA and a counter and the signal is there at 300MHz.
However, on this SA, there is nothing being shown at or around 300MHz.
Also, not having a signal generator going that high in frequency, I used a dual band wireless access point to see if I get a signal at either 2.4GHz or 5GHz.
However, there was absolutely nothing shown on the display, apart from the noise floor.
I am now planning to follow the signal path starting from INPUT and testing at different points where there are cables and connectors.
However, I don't know what to do when I would need to confirm signal at >3GHz frequency (e.g. between the dual band mixer and the second converter), as I don't have equipment that can measure stuff that high (my other SA can go only up to 1.5GHz, and the frequency counter can do no more than 1.3 GHz).
I also have one analog scope and a digital one but, again, entry level stuff: 60MHz or 100MHz.

[MarkL]

When you connect the CAL OUT to the INPUT, can you see the signal at the output of the internal attenuator?

Ah, you see, that's the thing: there is no CAL OUT signal displayed when I connect it to INPUT.
And I know the CAL OUT does output the right signal - I was able to confirm it using another SA and a counter and the signal is there at 300MHz.
However, on this SA, there is nothing being shown at or around 300MHz.
...

Understood, but I'm asking if you could see the attenuated 300MHz signal on the output of the internal attenuator.  I've had a couple of 859x attenuators with bad contacts.

It's the first step in the signal path and can be tested with the equipment you have since the signal is still 300MHz.  A little disassembly will be needed to get to the connector.

Does your unit have the tracking gen option?

What is the other SA?  Does it have an LO or tracking gen output?

[cncjerry]

How close is your 300Mhz cal signal?  Lot of analyzers multiply the cal signal up and/or mix it with something prior to injecting it into one of the many mixers.  So if it is way off, that could be your issue.  Otherwise, if all you see is noise, you probably have a blown attenuator, first mixer or even an input switch.  You need to trace the signal through the input plumbing.  That is the easiest thing to test.  Check your 300Mhz and fix that first.  Then use that to plumb the input up to the mixers. Usually you can isolate the problem to a module pretty quickly by just looking at the block diagram.  I think you'll find something is way off somewhere.

[MarkL]

Without taking anything apart, you could jump ahead a step or two and check for the 300MHz signal at the output of the A3A3 2.9GHz LPF.  This filter sits on top of the front end assembly and is accessible.  It's the longer one, and the output side is facing the front of the unit.

If you get a signal here, it would rule out the attenuator and RF switch (as cncjerry mentions).

[giosif]

Thanks everyone for the suggestions!

I was able to spend some time with the unit and found the following:
1. The +10V and -10V references are ok.
    MarkL was right: you need to lock the analyzer to the 0-2.9GHz band before running these tests.
    Funny that the manual does not mention this.
2. I ran some tests following the block diagram for the RF section, starting from the input and it appears that the A3A2 RF switch is not working.
    I did some quick checks on supply voltages and all looks ok, so I am presuming the switch is dead.

Two follow up questions based on point 2 above:
a. Would anyone happen to have such an RF switch as spare and willing to sell?
b. Does anyone know where I could find the instructions to take the whole RF section out of the main unit chassis?
    The Service Guide has a section for that, but it mentions it applies to 8590L and 8591E spectrum analyzers only.

Thanks,
George

[MarkL]

Here is the Assembly Level service manual for the 8693E:

  http://www.keysight.com/upload/cmc_upload/All/08590-90316.pdf

In particular, take a look at the pinout for the RF switch in table 6-5 (pg 289).  Make sure the RF switch is getting all the right switching signals (and power, but you've already checked that) before declaring it dead.

As a test to see if the rest of the unit functions, you can go around the switch and connect the attenuator output to the LPF input.  You should have a working band 0 (to 2.9GHz) at that point if nothing else is wrong.  You can lock the analyzer to band 0 to speed up the sweep.

Edit (more info): I don't have an 8593E to look at this directly, but if I'm reading the schematic correctly you should see switching on the following pins when the analyzer is sweeping all bands:

  J2-1: -15V to 0V
  J2-2: 0V to -15V
  J2-6: +12V to -12V

These signals are all derived from HBSB (guess: High Band Select Bar?) that is input to U110 (74LS06) pins 3 and 9.

[MarkL]

More info for you... Attached is the A7 schematic page that has the RF switch control from the 8593E Component Level Information Packet (CLIP).  You'll find this useful to verify the RF switch signals described in the previous post.

This is a scan from my paper version.  I can't find the full CLIP in the wild anywhere, but you can purchase a PDF version here (their part #5963-2951):

  http://artekmanuals.com/

They do excellent scans and everything is OCR'd.  My paper version is about 3 inches thick.  It's well worth the $25.

[giosif]

Many thanks for the very helpful info, MarkL!

I did more rigurous measurements today and found that:
1. The A7 Analog Interface board is producing the expected voltages across the J2 connector when no cable is connected to it (for both low band and high band settings).
2. The A3A2 RF switch is definitely dead and so looks to be the A3A6 dual band mixer.
    I think this because, when connecting cable W13 to J2 on the A7 Analog Interface board, the measurements over the J2 pins start to get weird: J2-1 (which is normally sitting at -15V, by default) is at -10V, J2-2 (which is normally sitting at 0V, by default) is at +(yes, plus)15V, and J2-6 (which is normally sitting at -12V, by default) is at -10V.

Also, as instructed, I bypassed the RF switch and checked for the CAL signal to see if detected (of course, with a cable between CAL and INPUT), but no signal was shown on the screen at or around 300 MHz.
So, something else is broken, in addition to the RF switch.

Oh well, need to go on a hunt for an RF switch and a dual band mixer now... 

I wonder what else is broken...
Which brings me to my other question: someone has any idea on the procedure to take the RF section out of the chassis?


Finally, with regards to the manuals from artekmanuals, I have already bought them and, indeed, they are very good scans.
Now, if only I could read them properly...

Regards,
George

[MarkL]

I agree, the A3A2 RF switch doesn't sound too happy.

FWIW, here's the voltage readings on a 8595E.  It also has a switch controlled by J2 (but it's integrated into a YIG filter), and it has the same model A3A6 dual band mixer as the 8593E:


      Low Band    High Band
      --------    ---------
J2-1:     0V       -9.92V     W13 out
          0V       -9.88V     W13 in

J2-2:   -9.91V       0V       W13 out
        -9.82V       0V       W13 in

J2-5:     0V       +3.95V     W13 out
          0V       +3.95V     W13 in

J2-6:  +14.23V    -13.25V     W13 out
        +9.63V     -9.47V     W13 in


It's going to cost several hundred $$$ to obtain those A3A2 and A3A6.  I would make sure the output drivers on A7 are working by loading the above outputs with a 5k or 10k resistor and checking the output voltages again.  The +15V on J2-2 sounds bad and you want to make sure it's not floating up to +15V because the driver has gone high impedance.

Can you see the 300MHz signal before it goes into the A3A6 mixer with your other SA?

Which brings me to my other question: someone has any idea on the procedure to take the RF section out of the chassis?

It's in the Assembly Level service manual, "Replacing Major Assemblies - A3 Front-End Assembly" (pg 311).  You have to take the front panel off, and be sure to remove the SMA connector going to the front panel N-connector before you start pulling on it.

It's not too bad getting it out.  The harder trick is stuffing it all back in.  Take pictures as you go so you remember how everything is routed.

And once you get it out, it's possible to partially put the front panel back on so you can power it up and work on the Front End assembly hanging out.  Kind-of unwieldy, but it can be done.

[giosif]

FWIW, here's the voltage readings on a 8595E.  It also has a switch controlled by J2 (but it's integrated into a YIG filter), and it has the same model A3A6 dual band mixer as the 8593E:


      Low Band    High Band
      --------    ---------
J2-1:     0V       -9.92V     W13 out
          0V       -9.88V     W13 in
[...]

Thank you for taking those measurements for me!
I know that getting inside these beasts is not a quick job (I have to take the cover off and put it back on every time I'm working on the unit, as kids would for sure start making "toys" out of its internals, otherwise).
I need to take some more measurements to build a complete table like yours, just to be clear on what values I get.

It's going to cost several hundred $$$ to obtain those A3A2 and A3A6.  I would make sure the output drivers on A7 are working by loading the above outputs with a 5k or 10k resistor and checking the output voltages again.  The +15V on J2-2 sounds bad and you want to make sure it's not floating up to +15V because the driver has gone high impedance.

Good idea with the resistors, as I was wondering the same but didn't know how to check.
I don't think J2-2 is bad, actually: I forgot to mention this in my previous post but, with W13 out, I am getting the expected -15V (for Low Band) and 0V (for High Band) outputs.
The +15V is only with W13 in.

Can you see the 300MHz signal before it goes into the A3A6 mixer with your other SA?

You mean after I bypassed the RF switch?
And is that to check that A3A3 LPF is ok?

It's in the Assembly Level service manual, "Replacing Major Assemblies - A3 Front-End Assembly" (pg 311).  You have to take the front panel off, and be sure to remove the SMA connector going to the front panel N-connector before you start pulling on it.
It's not too bad getting it out.  The harder trick is stuffing it all back in.  Take pictures as you go so you remember how everything is routed.
And once you get it out, it's possible to partially put the front panel back on so you can power it up and work on the Front End assembly hanging out.  Kind-of unwieldy, but it can be done.

And I remembered seeing some details on disassembling this section; I just could not find it when I looked for it.
I need to improve my search skills.
Ok, so this is serious disassembly that can't be done and undone in one night, if I am to also check things.
It will then need to wait until I move to a new place (planned to happen soon, if all goes well), where I'm supposed to have my own (geek) area (finally).
Just need to implement secure access to the room: wonder if fingerprint + iris + retina scans are going to be enough to keep the rascals away... 

[iXod]

You *are* sharing this with the Yahoo group for HP/Agilent fixers, yes? ::

https://groups.yahoo.com/neo/groups/hp_agilent_equipment/info

Serious set of experience and other resources (firmware, etc.) there...

[MarkL]

Can you see the 300MHz signal before it goes into the A3A6 mixer with your other SA?

You mean after I bypassed the RF switch?
And is that to check that A3A3 LPF is ok?

Sorry - that was a little out of context.

Yes, with the relay bypassed, just to make sure you had signal going into the A3A6 mixer.  I was curious what made you conclude the mixer was dead too.

Another thing to try is bypassing the switch to the high side and see if you can see your WiFi AP at 5GHz in the higher band.  This uses the other half of the A3A6 mixer.  You could also try bypassing the A3A8 YTF and go from the attenuator output to the A3A6 mixer high side input.

These are expensive assemblies and I'd hate to see you make a significant investment in the wrong place.  It's difficult to make conclusions without an SA that can see the LO, mixed signals, etc.  Any chance of borrowing one?  You'd only need it for a few hours.

If you can find one for cheap, one hack is to use a satellite down converter to go from C-band (3.4 - 4.2GHz) to L-band (950 - 1750MHz).  You should be able to verify the LO and the 3.9214GHz IF with your other SA.

[Bicurico]

Hello,

After 2-3 years of having one of my HP8594E sitting in the corner due to it being unable to perform the CAL, I somehow found out today that I could recover the default CAL using the passcode. I don't know why I did not see that before.

The FREQ CAL now passes and frequencies are correct again (they were 100MHz off!).

However, the AMPL CAL fails with the error:

CAL: NBW width failed
    Indicates that one of the resolution bandwidths is not the correct width for the calibration routine.

Can someone point me into the right directionto repair this?

Thanks,
Vitor

[Bicurico]

Just to add: it was the backup battery that seem to have failed.

After switching the power off and on, a message briefly appears EMPTY DLP MEMORY and I am back to an uncalibrated device.

At least I know how to get the frequency CAL back. And yes, I know how to swap the battery. It is odd, though, because I did swap the battery as soon as i got the device.

Also, the device was always connected to the outlet - I would have imagined that this would keep the memory alive, which doesn't seem to be the case.

Woiuld be great to fix the NBW width error, too.

Regards,
Vitor

[vishaldotgupta]

i have hp8593e which doesnt shows very high attenuation above 3 Ghz.  i suspect the RF switch is not switching to high band.  the control voltages are not there.  now two questions.

1my version A7 board is slightly different from CLIP.  it has monostable multivibrator, deriving these signals.  board part no is 08592-60110.  CLIP manual doesnt have schematic for it.  it seems the multivibrator is driven from HSB signal only

2.  if analyser is made to sweep from less than 3 Ghz to more than 3 Ghz, should there be switching with each sweep?  i mean should HSB should toggle in each sweep?

[vishaldotgupta]

i read some where that you need to select 75 ohm input impedance before attemption CAL amptd