HP 8566B - What to look for?

[W4PJB]

I'm going to look at an 8566B tomorrow. I can power it up and inspect it, but I don't think I'll be bringing my 8665B to test it as it's a little large to carry

The unit supposedly powers up and displays no errors, anything else i should look for? If i set it for "full span", is there anything obvious on the screen I should see if it's got a fault? I know the noise floor will shift a bit every few gHz (and that's a good sign). Anything else I can test?

I do have an Anritsu MB2721B with TG I could use to create a signal up to 7gHz.

[G0HZU]

Connect the 100MHz test signal (BNC) to the main RF input and see if it adjusts OK when doing the RECALL 8 and RECALL 9 tests in the manual.

You can also try running the automated error correction routine by connecting the 100MHz test signal (BNC) to the main RF input and pressing the blue SHIFT button and then press the big FREQUENCY SPAN button. This runs an automated self correction routine that aims to improve performance by correcting errors between the RBW filters etc.

This takes about a minute to run.

Also put it on a 10MHz span and look at the 100MHz test signal. There will be some wobble seen on the signal when a 10MHz span is selected. A poor unit will show more wobble than a good unit.

Make sure the 0-70dB attenuator works on all settings. Note that you have to press ATTEN 0 dB to get it to select 0dB attenuation. It won't let you select 0dB with the down arrow key.

Probably the best advice I can offer you is to not buy the 8566B unless you really do need the extra frequency range it offers. It's unlikely to be very reliable in the coming years so you can expect to have to take it apart to replace tired capacitors from time to time.
Also, the RBW filters may need to be aligned/adjusted as per the service manual. It is also very big and very heavy and the fan noise is very loud. They can also smell a bit musty if operated in a small room as the fans will spread the musty smell into the room.

I've still got my 8566B but I rarely use it. I upgraded to something more modern about 5 years ago. Better performance, less noise and much easier to move about!

[KE5FX]

One thing you should look at is the CRT condition.  All display elements -- text, trace, graticule lines -- should be razor-sharp at all reasonable brightness levels.

The suggestion to hook up the calibrator and do a recall-8, recall-9, blue shift-W cycle, followed by a blue-shift-w command to inspect the results, is a good one.  Take the time to read about that in the manual beforehand.  It is unlikely that you will need to tinker with the RBW filters but if there is a problem, this procedure has a good chance of indicating it.  Watch the shift-W cycle while it runs, keeping an eye out for gross errors in filter skirt shapes.

Agreed that these units are likely to become more "maintenance-prone" going forward.  But at least you can work on them.  Mine will leave my bench shortly after I do, feet first.

[G0HZU]

If at all possible, try powering it on from a cold start to see how it behaves. Normally, you are meant to leave it powered in standby for maybe 15-20 minutes before powering it on. This allows the 10MHz ref oven to warm up and also other circuits need to be warmed up before it will be able to achieve reliable control of the swept LO system.

However, if you deliberately power it on from cold it should report REF UNLOCK and OVEN COLD on the CRT display. If the analyser is in good health, this probably won't stop it from being basically functional on wide spans at least.

However, when I first purchased my 8566B about 14 years ago, it would misbehave badly if given a cold start. It needed about 30 minutes to warm up and be functional. It showed all kinds of strange behaviour onscreen when started from cold. It would drift badly and in its attempts to try and sweep correctly it even wildly changed the centre frequency (to negative frequencies!) and basically went crazy.

Over time this got worse and worse (almost certainly why the rental company sold it to me so cheaply) and it eventually refused to lock even when warmed up.

This was traced to several failed capacitors on boards involved in the YTO (YIG) tuned LO system. Once these were replaced, it was basically useable from a cold start.

It will also report the condition of the battery when you first power it on. If there is an issue with the battery it will show BATTERY on the CRT display.

There are also a couple of red status LEDs along the lowest edge of the front panel. These should turn off once the analyser is booted and running correctly. If one or both of them stay lit then the analyser has problems somewhere.

[W4PJB]

Thanks everyone for your replies, they are all very helpful! I certainly appreciate the warnings about it's age. That is, of course, a downside. But it's also part of the hobby. I'm the type that learns by doing; I need a problem to solve, I can't just sit and learn from a book. So if happens to kick down the road, I'm ok with that... I'll learn something! Whether I can fix it or not. Either way, the 8566 is just such a cool piece of gear I'd be kicking myself i missed the chance to own one for a few hundred bucks.

I'll be sure to let everyone know how it goes!

[nctnico]

If the 8566B costs a couple of hundred bucks, I'd save a bit longer and buy a much newer spectrum analyser. The ones from Advantest are typically not sure expensive. I'd avoid anything with a CRT screen.

[KE5FX]

But the Advantest lacks service documentation, "tribal knowledge," and widespread parts availability.  It has a substantially-inferior display, a less-expedient UI, all the soul of a Subaru, and worst of all, you miss out on the comforting roar that will remind you of your summer job at the airport as a teenager... 

One nice thing about the 8566B, especially with an LCD, is that it's up and running almost by the time your fingers come off the power switch.  Seems like a goofy hill to defend but the fact that it's always ready when I need it is a big part of why it's still on my bench.  Of all the spectrum analyzers around here, only the tinySA matches it in that respect.  What typically happens is that the 8566B is used to look for signals, while a newer FFT-based analyzer (usually an FSP or E4406A) is then used when necessary to drill down and look at the signal. 

Admittedly, if I only had room for one, it wouldn't be the 8566B just because life is too short for swept analog filters.  But everything with a modern signal-processing back end either has lower coverage limits like the OP's MS2721B, is in a higher price bracket like the FSP, or is intentionally compromised like the E4406A.

[W4PJB]

Well, the good news is, the 8566B has a new home! I was able to run the self cal's at the sellers house and everything seemed to run, at least well enough to make an offer. The screen was bright and crisp. All controls responded well. YTO Unlock was on the screen briefly, but quickly disappeared after a minute or two of warmup. The other good news is, I only paid $400 for it.

The bad news is, once I got it home and was able to connect it to my 8665B, I learned the 2.5-22 gHz range has no signal. Then, after about 10 minutes of operation, the screen went to a blurry dot, and the very familiar smell of capacitor filled the room along with a gentle puff of smoke. Not daunted on that, it's likely a display capacitor, as this poster experienced: https://www.eevblog.com/forum/repair/hp-8566b-display-problem/

The fun part will be diving into the 2.5 gHz section once the display is fixed. Might as well get a few capacitors on order while I'm on DigiKey. Does anyone have a list of common caps to replace in these?

[KE5FX]

Looks like a late-model unit that will clean up nicely.  

You'll want to get an ESR meter if you don't already have one, and then shotgun-check every axial tantalum cap on the PC boards under the plastic covers in the RF section.  Any good-quality electrolytic (Ta or Al) with similar voltage and capacitance ratings can be used to replace most/all of them that are found to be defective.  No need to replace caps that aren't actually defective.

[W4PJB]

Looks like a late-model unit that will clean up nicely.  

You'll want to get an ESR meter if you don't already have one, and then shotgun-check every axial tantalum cap on the PC boards under the plastic covers in the RF section.  Any good-quality electrolytic (Ta or Al) with similar voltage and capacitance ratings can be used to replace most/all of them that are found to be defective.  No need to replace caps that aren't actually defective.

Thanks! I was thinking it was a later unit, based on the encoder knob being the plastic / dimpled one, and not the aluminum / plastic one. Is there a way to date an HP unit by serial? Or only get a general idea based on sequence?

My Fluke 123 Scopemeter can measure capacitance, and I also have one of those cheap transistor / diode / capacitor / resistor meters that sell for $20 on the big websites. Guessing the Fluke should be up to the task.

Do you think it's likely my non-working 2.5-22 gHz section could be as simple as a bad cap? The only error that didn't go away in the first few minutes of warming up was "REF UNLOCK". It seemed to be blinking roughly in pace with the scan rate. But the cap in the CRT power supply went poof before I could let it warm up much.

[G0HZU]

It does look to be in nice condition. I bought mine as an ex-rental about 14 years ago for just under £1000. At the time this was a really low price for an 8566B. It turned out to be an A version that had had the factory upgrade to a B. So it is older than I was expecting. I bought it from one of the main rental companies here in the UK (Microlease) so I had to trust them that it worked OK (it wasn't OK).

Apart from the YTO instability, the CRT was also very fuzzy even when the focus control was optimised. I managed to fix the fuzzy CRT by servicing the Z axis board. I also went around all of the RF boards with an ESR meter and replaced several dead caps and this fixed the YTO issues.

The other problem it had was a very poor frequency response across the 2-22GHz range. Playing with the YIG PRESEL button just made it worse in various parts of the 2-22GHz range.

I'd recommend that you try pressing the blue shift key followed by the PRESEL PEAK button when on the 2-22GHz range. This should restore the factory control of the YIG preselector. In my case this fully restored the frequency response on the 2-22GHz range. There were then no more deep holes in the frequency response up to the 21GHz limit of my old 83752A sig gen.

Hopefully, the high band mixer is OK on your unit. This would probably be expensive to replace assuming you can find a replacement.
There is also a changeover relay that might have failed in some way. This relay switches across to the high band range.

Is the high band range completely dead or is the displayed level just really low?

[KE5FX]

Thanks! I was thinking it was a later unit, based on the encoder knob being the plastic / dimpled one, and not the aluminum / plastic one. Is there a way to date an HP unit by serial? Or only get a general idea based on sequence?

Adding 60 to the first two digits of the serial # will give you the date of the last design/manufacturing revision, which is a lower bound on the manufacturing date.  E.g., if it starts with 27xx it will be 1987 production or later.  Then there are the date codes that will be present on many of the caps and ICs. 

My Fluke 123 Scopemeter can measure capacitance, and I also have one of those cheap transistor / diode / capacitor / resistor meters that sell for $20 on the big websites. Guessing the Fluke should be up to the task.

No,  you want an ESR meter.  You can't reliably measure capacitance in-circuit, but you can read the ESR.  It will be obvious which ones are open (and there will likely be a couple.)  The axial tantalum electrolytics fail open much more often than they fail short.

Do you think it's likely my non-working 2.5-22 gHz section could be as simple as a bad cap? The only error that didn't go away in the first few minutes of warming up was "REF UNLOCK". It seemed to be blinking roughly in pace with the scan rate. But the cap in the CRT power supply went poof before I could let it warm up much.

REF UNLOCK is normal during warmup, before the OCXO comes up to temperature.  If it stays on, make sure the reference jumper and switch are set correctly on the back panel.  YTO UNLOCK and other error messages are never normal, except perhaps BATTERY.  There will be a Tadiran primary lithium battery on the CPU board that may need to be replaced if that message shows up.

The 2.5-22 GHz section could fail for any number of reasons, most of them not too painful to find and fix.  You'll probably need to go through the manual to track that issue down.  Bad capacitors can cause almost any symptom including that one, but they are far from the only potential trouble source.  Fortunately the high-band mixer is rarely bad, as the diode is well-protected by the YIG coupling loop.

Artek Media has a good set of manuals (the free ones are pretty crappy.)  I like to keep both paper and .PDF copies on hand, but the paper manuals take up almost as much space as the 8566 itself...

[W4PJB]

Thanks again for your reply!

The first two digits are 31, so that should make my unit a '91 or later model. Great!

I will look into a better LCR/ESR meter for this. Never hurts to have one on the bench anyway. The DE-5000 seems to have a good following here for the money.

Also great info on the REF UNLOCK error. Unfortunately I was not able to keep it in warm-up long enough to see if it went away. YTO ULOCK did come up for a minute or so when I powered it up at the sellers house, but it quickly went away. It did not come on again when I powered it up at home, just REF UNLOCK. They got the SA from an estate sale, and are not electronics savvy, so I unfortunately have zero history on the unit.

Once I get an ESR meter I'll check all the caps in the RF section as well, hopefully that does the trick for the high band. I plan to replace the large caps in the power supply too, while I'm in there. I'll be sure to keep this thread updated as I progress, but now we're entering the waiting phase to get test kit, parts, and most importantly TIME to mess with it.

Thanks for the tip about Artek Media. The whole set of 5 manuals is only $15. I'll be placing that order right now!

[KE5FX]

Thanks again for your reply!

The first two digits are 31, so that should make my unit a '91 or later model. Great!

I will look into a better LCR/ESR meter for this. Never hurts to have one on the bench anyway. The DE-5000 seems to have a good following here for the money.

Also great info on the REF UNLOCK error. Unfortunately I was not able to keep it in warm-up long enough to see if it went away. YTO ULOCK did come up for a minute or so when I powered it up at the sellers house, but it quickly went away. It did not come on again when I powered it up at home, just REF UNLOCK. They got the SA from an estate sale, and are not electronics savvy, so I unfortunately have zero history on the unit.

Once I get an ESR meter I'll check all the caps in the RF section as well, hopefully that does the trick for the high band. I plan to replace the large caps in the power supply too, while I'm in there. I'll be sure to keep this thread updated as I progress, but now we're entering the waiting phase to get test kit, parts, and most importantly TIME to mess with it.

Thanks for the tip about Artek Media. The whole set of 5 manuals is only $15. I'll be placing that order right now!

  Definitely a later model.  If you had $61.5K to burn in 1991, you could have bought this 8566B or a new Porsche 911.

I would absolutely not recommend replacing the large power supply capacitors (or any others) unless they are actually defective.  They should be checked, of course, but the truth is that the newer capacitors you replace them with will not necessarily be any better or last any longer.  (Plus, have you checked the price of new 'computer-grade' electrolytics lately?)

[W4PJB]

I'm continually blown away, and simultaneously humbled, by the brain trust on this forum. Thank you!

I took the liberty of popping a few panels off tonight, just to see if I could find an obvious "smoking gun", since, well a pretty significant jet of electrolytic smoke emanated from the side of it.

From a quick visual nothing obvious, and on a positive note, the unit was SUPER clean inside. Not a speck of dust. This must have been in a clean environment, rarely used, or both. Judging by the quality of the display (when it worked), and the softness of all the wire jacketing, I'm guessing this may have pretty low hours on it.

One thing I DID find, when pulling one of the small side panels off of the RF unit, was a few tell-tale sunflower seed shells. So perhaps I need to be on the alert for critter damage as I go through the unit.

Also, how on earth did humans design this thing? In the 1970's, no less? If I can't get this fixed, I'm pulling the covers off and hanging it on the wall. What a masterpiece of technology.

Apologies for the weird color cast in right side of the photo, it's an artifact from the rolling shutter in my iPhone and the crappy fluorescent light above my bench. Everything inside is as clean and silvery as it should be

[W4PJB]

Is the high band range completely dead or is the displayed level just really low?

I only had a few minutes until the display quit, but from what I could tell, dead. I had a source going in at 0dbm, through a 30db pad, and 10db of attenuation in the 8566, and I got absolutely nothing on the graph. The low range peak was exactly where expected. As soon as the relay clicked into the 2-22 gHz range, I couldn't get a reading anywhere. I have a 6 gHz source and tried at 2.5, 3.0, 4.0, 5.0, 5.5, and 6 gHz, and nothing registered on the screen.

The high band noise floor looked like the one in this photo (not mine, it's from the web).

Thanks!

[KE5FX]

Yeah, that's a low-hours unit.  The CRT HV lead is the real telltale, and it looks like new.  Should be well worth fixing up!

Trivia: they started work on the 8566A in the early 1970s, and it didn't ship until 1979.  Its code name was "Doomsday."  The earliest drawing I've seen is from 1971 (and never mind where it came from):

[W4PJB]

That's a very fascinating site, thank you! Not surprised it was code named "Doomsday". It doesn't take a vivid imagination to figure out what kind of projects would have motivated development of a cost is no object, 22 GHz spectrum analyzer in the 1970's.

[nctnico]

Apologies for the weird color cast in right side of the photo, it's an artifact from the rolling shutter in my iPhone and the crappy fluorescent light above my bench. Everything inside is as clean and silvery as it should be

This looks like new indeed. I'll admit that looking at the photo of the internals does put a smile on my face 

[G0HZU]

Yes, it does look to be as new inside. I recently acquired a late model HP 8405A vector voltmeter that looks inside like it was made yesterday. It's absolutely sparkly clean inside.

To give some contrast, I know the history of my 1500MHz HP 8568B analyser back to about 1987 and it was made in 1986. It was donated (as part of a contract) to the company I work for back in about 1987.

I joined in 1990 and this analyser was used for >8 hours every working day for about 15 years since 1987. It was never used in an ATE rack so was switched off every evening. The company gave it to me (I made a small donation to a charity to secure it) in about 2005 after it had fallen out of use at work for a few years.

When I took it apart, I was surprised at how much dust there was around the CPU board. It was thickly coated in dust to the point it wasn't possible to make out the ribbon cables. The display unit showed quite a lot of black soot around the HT areas.

I'm not sure why it was so dusty and sooty after maybe 30,000 - 50,000 hours operation time. It may be that the filters in the fans weren't working well for many years or maybe there was more dust in the air than I realised in the RF labs at work!

My HP 8566B is a few years older but the display unit was very clean inside. Not nearly as sparkly clean as yours but I assumed that it had been internally cleaned and serviced at some point because it showed none of the soot and dust that I found inside the 8568B. I suspect that my 8566B has seen a lot of use as it was an ex-rental analyser from Microlease.

[W4PJB]

Well, mine has a .gov tag on it, so it was probably used twice, put in a storage facility for 30 years, then went to auction for pennies on the dollar   

[G0HZU]

My HP 8568B was also a govt owned analyser. The govt donated it to the company when it was about 1 year old.
They also donated a nearly new Advantest TR4172 analyser a few years later and this would have cost them about $60k due to the high cost of the Yen in the late 1980s.

I also now own that Advantest analyser as the company gave it to me about 20 years ago. I had to make a charitable donation for it though. It was a worthy competitor to the HP 8568B. It wasn't as polished as the 8568B in terms of all round performance, but it outclassed it for spurious free dynamic range. That's why it was donated to the company. The 8568B was good, but in terms of SFDR it wasn't good enough for the stuff we were doing for govt back then. The TR4172 also had really low phase noise at 100kHz offset. IIRC it was typically about -128dBc/Hz. It also had a tracking generator and could be used as a crude VNA with an external RL bridge.

If you can get that 8566B fully working, you could probably sell it to a collector for a lot of money as it looks to be in fantastic condition inside.

[W4PJB]

Since it looks like I'll be working on the CRT power supply section, any advice on how to ~not~ kill myself? 

Just kidding... mostly. I'm very aware of the voltage levels in CRT circuits, even potentially when unplugged. Do the HP CRT boards typically have bleeder resistors to discharge the capacitors when unplugged? I see nothing in the service manual instructing to manually discharge anything, or wait "x" time before servicing; it simply states to not work on this area of the unit with the power cable connected. I could always cobble together a drain circuit to discharge it in a controlled manner, but it isn't exactly easy to get to HV lead on the CRT on these units.

Thanks! 

[KE5FX]

Since it looks like I'll be working on the CRT power supply section, any advice on how to ~not~ kill myself? 

Just kidding... mostly. I'm very aware of the voltage levels in CRT circuits, even potentially when unplugged. Do the HP CRT boards typically have bleeder resistors to discharge the capacitors when unplugged? I see nothing in the service manual instructing to manually discharge anything, or wait "x" time before servicing; it simply states to not work on this area of the unit with the power cable connected. I could always cobble together a drain circuit to discharge it in a controlled manner, but it isn't exactly easy to get to HV lead on the CRT on these units.

Thanks!

It is indeed described in the manual:



It's worth following this procedure, as the CRT really will hold a charge for quite some time.  Contact with this voltage probably will not kill you but it will hurt.  

Be sure the CRT shield ground wire is in place.  How I know this is a long and boring story, but trust me. 

All of that being said, whatever went 'poof' and smoked is probably not in the HV section.

[W4PJB]

Edit: Now that I knew what to look for, I found the procedure in a different manual from EMCTEST. Thank you for sharing this!

Also, the manual from EMCTEST is a 1990 printing, so likely more representative of my unit, rather than the 1984 printing from Artek, so added bonus.

-----------------------------------------------------------------------------

Thank you for your reply. Two things I find interesting here:

1: My manual is completely different, and does not list this procedure at all.
2: HP engineers advise using the wire on a screwdriver trick on a sensitive $60,000 piece of RF equipment  I would have expected something a bit more.... controlled.

I'll post below the only two pages in my manual that reference a service procedure for the A1A3 or the CRT. The only place I see a reference to a discharge procedure for a high voltage in the A1 section is in the A1A8 section, where it advises, "The time constant of the filter capacitor circuit is about 25 seconds. Before servicing, allow adequate time for discharge after line power cable is removed.". But this is referencing mains voltage levels, not the HV circuit. Even the CRT removal procedure does not have any flags about HV.

Was your manual sourced online? I'm using a manual from Artek, and the page numbers for the A1A3 stop at "A1A3 9/10", which is the schematic diagram for this board. I don't see it in the manual on Xdevs, either.