HP 8566B - What to look for?

[KE5FX]

I'm not sure where my copies came from, I think they were originally from Agilent's own site.  At least some of the files, including the copy of 85662-90085 I clipboarded the above from, are available on the Internet Archive. 

The problem is, whoever OCR'ed them didn't GAF what they were doing, and used some seriously-inappropriate compression settings  As a result various bits and pieces of text including many numeric values were replaced with the OCR'ed text, rendered in the OCR app's best guess as to the correct font. 

This actually changes the information in some cases:



So you don't want any part of these files, believe me.  That's one of the reasons I keep paper copies on hand.  It's unfortunate if Artek's files weren't captured from a later edition, but at least they don't screw up stuff like this.

[W4PJB]

I had never even considered OCR, as I never use it when scanning PDF's. But that's an EXCELLENT point!

Thanks again!

[W4PJB]

It's always the last place you look, right?    In this case, it was literally the last board in the unit there was to check.

I found capacitor C22 was missing its magic smoke. This coincides with the approximate location of the volcanic eruption I saw from the right side of the display unit when it failed.

An equivalent seems to be available from Vishay (DigiKey 150D225X9020A2BE3) and I will probably go ahead and just replace all 3. C21 measures OK at 2.19uf but it's seen some heat damage from C22, and C20 measures around 2.8uf and has about 20% less phase rotation according to my LCR meter, so I'll guess that one on its way out next.

Once I get the display up and running then on to the 2-22 GHz problem. Thanks for everyone's help!

[KE5FX]

Yep, the neighboring tantalum caps are a good exception to the "Don't replace unless defective" advice. 

[W4PJB]

Well I got the new caps in today and viola! We have a screen!

Now onto the real work; figuring out why I have no signal on the 2-22Ghz band. Thanks for everyone's help!

I'll start with checking all of the voltages very closely, I understand the YIG can be very sensitive to even small errors here.

[W4PJB]

Voltages all looked good, except the "18V Typical" on the A18 was low at 16.2v. The -10v side of that test port was bang on, relative to the "REF GND" on the A19.

C4 and C19 on the A21 board are looking a little sus, though in-circuit they seem to measure OK.  They're also over $40 EACH! 

[KE5FX]

Just use whatever's handy to replace them, same or higher voltage, same or (slightly) higher capacitance.  Might have failed shorted if they appear OK on the ESR meter while looking like that.

[EggertEnjoyer123]

Have you checked the RF switch? Had an issue with another model where the switch clicked but the contacts didn't actually make contact well.

[W4PJB]

Have you checked the RF switch? Had an issue with another model where the switch clicked but the contacts didn't actually make contact well.

Thanks for your reply. Not yet.... but it's high on the list! I've been doing through the service manual and three things are pointing me in that direction. 1: looking at common points of failure, that's one of the first in the chain for this issue. 2: I do not have any errors, such as "YTO Unlock", on the display. 3: Switching from 2-22Ghz to the low band the "click" from the relay is crisp and fast. Switching to 2-22Ghz is kind of "lazy".

But there's a bit of teardown to get to that component, and with a "real feel" of 107*F in Maine today, this evening I've switched from inspecting vintage HP instrumentation, to inspecting a nicely chilled bottle of Chianti. Meaning it's probably best I keep my hands out of electrical devices for the evening. Cheers!   

[W4PJB]

Well, it wasn't the switch. Pic of a 1MHz - 7GHz sweep attached. Ignore all of the ripple, it's most likely reflections in the cheap SMA cables.

On to the next thing! I have no errors on screen, so I doubt there are any issues with the YIG itself, but if it saw some DC it's likely the mixer diodes are fried. It's my understanding this will not cause any error messages. So perhaps I'll pick one up and try it, they're cheap enough these days.

[KE5FX]

Messing with the YTX is probably not going to be productive in this case.  It could be damaged by DC, sure, but the damage would take the form of burning out the input coupling loop rather than the diodes.  Application of enough power to damage the YTX is likely to damage the attenuator as well, which it sounds like you've ruled out.

Refer to figure 1 in the August 1979 HPJ issue and you will see what I mean.  (As with many HP instruments, HPJ actually provides a better technical introduction than anything in the official manuals.)

If you were to replace the YTX you would have to recalibrate all of the the trimpots that define its tuning-current curve.  You really don't want to go there if you can help it.  Instead, try applying a CW signal at a few GHz and tracing the resulting 321.4 MHz IF through the Figure 2 block diagram in the HPJ issue, using (e.g.) your Anritsu.  If you can't find the IF signal at all, check the current through the YTX tuning coil to make sure it's not open, and that its driver circuit is functioning at all.

[W4PJB]

Thank you. I've been working on this during my spare time in the evening, so I haven't had much time to really dive into the block diagrams yet. I have a lot more time to do research over the weekend.

Also, I always think of the archival value of these posts. Hopefully at some time in the future, someone may find the information here helpful, just as I have learned from past posts on this forum and on others!

[G0HZU]

Good to see you got the display working again!

If you get a bit of free time, could you do a quick screenshot of the phase noise at 100MHz on a 100 kHz span as in the image below?

The image below is taken from the manual and it shows the front panel 100MHz (-10dBm) calibrator test signal being measured on a 100kHz span with the RBW set to 1kHz.

Could you quickly show me what your analyser looks like with this test setup please? I'm interested to see how much cleaner the later models are in terms of phase noise. I checked the serial number of mine and it was made in 1982 so it would have been a fairly late A version converted to a B with the factory upgrade. My analyser looks very similar to the plot below. I'm interested to know if the loop bandwidth was changed at this span setting on the later models. It seems to be very narrow in the image below? On narrower spans the loop bandwidth appears to be wider on my analyser.

I think my analyser has 1989 firmware as it flashes up 1989 when booting but the actual hardware appears to have been made in 1982.  I'm not sure what the latest version of firmware is for these analysers.

[W4PJB]

Ah ha! Progress is being made! I set the 8566B to 3 GHz, 10 kHz span, and low and behold, a little 4 dB blip showed up right at 3 GHz. (Sweep rate is slow so I have time to hit "Peak Search" and grab a photo before the next sweep.)

Also, I connected my Anritsu to the 1ST LO out, and as to be expected, 3.321 400 GHz.

So the great news is, everything is working! At least somewhat, enough to know the main components of the RF section aren't dead. It just has a very, very weak signal that disappears in the noise floor unless it's set to a very narrow span. I checked the attenuator and as expected, it's working fine.

So, once again, on to the next thing!

[W4PJB]

Good to see you got the display working again!

If you get a bit of free time, could you do a quick screenshot of the phase noise at 100MHz on a 100 kHz span as in the image below?

The image below is taken from the manual and it shows the front panel 100MHz (-10dBm) calibrator test signal being measured on a 100kHz span with the RBW set to 1kHz.

Could you quickly show me what your analyser looks like with this test setup please? I'm interested to see how much cleaner the later models are in terms of phase noise. I checked the serial number of mine and it was made in 1982 so it would have been a fairly late A version converted to a B with the factory upgrade. My analyser looks very similar to the plot below. I'm interested to know if the loop bandwidth was changed at this span setting on the later models. It seems to be very narrow in the image below? On narrower spans the loop bandwidth appears to be wider on my analyser.

I think my analyser has 1989 firmware as it flashes up 1989 when booting but the actual hardware appears to have been made in 1982.  I'm not sure what the latest version of firmware is for these analysers.

My pleasure! Is this what you are looking for? Or did I miss a step?

[KE5FX]

You're set to 1 dB/division, so you can't see anything but the peak.  Try these settings:



(from 27xx serial #)

[G0HZU]

Thanks! The plot I posted up is taken from the B manual but it is a reprint from the A manual. It looks like the loop response is different on the later models. Close in phase noise is better but the phase noise at 50kHz offset is higher.
Quite surprisingly, my analyser still uses the same loop response on a 50kHz span and this really doesn't look good. Maybe there is a way to upgrade it to emulate the later versions. However, that is a topic for another day...

My pleasure! Is this what you are looking for? Or did I miss a step?

Can you set it to 10dB/div as shown by KE5FX. I think you will then see a similar plot as the one from KE5FX's analyser.

[G0HZU]

If you have access to a VNA (even a nanovna) that can cover up to 3GHz or higher, then you should be able to connect the VNA to the input port of the 8566B and then set the 8566B to (say) 2.7GHz on a 1MHz span and then set the input attenuator to 0dB. Then set the VNA to an s11 measurement. Set the VNA source power to -10dBm or less.

You should then be able to see a dip in the return loss at 2.7GHz if the YIG preselector is working and tracking correctly. It should show a return loss of >10dB across about a 20MHz? bandwidth (at a guess).

It is then also possible to press 'SHIFT' 'GHZ' and this should bring up an oncreen YIG DAC control feature. You can rotate the wheel on the front panel of the 8566B and then the YIG filter frequency can be tuned over a limited frequency range either side of 2.7GHz.

This should give an indication of the health of the YIG spheres and also the YIG tuning and control sections without having to take anything apart inside the analyser.

If you don't have a VNA then maybe make or borrow a directional coupler and use it with your Anritsu analyser + tracking gen to do the same thing.

[W4PJB]

If you have access to a VNA (even a nanovna) that can cover up to 3GHz or higher, then you should be able to connect the VNA to the input port of the 8566B and then set the 8566B to (say) 2.7GHz on a 1MHz span and then set the input attenuator to 0dB. Then set the VNA to an s11 measurement. Set the VNA source power to -10dBm or less.

A 3 GHz VNA... funny you should mention that

Thank you for your advice!

[W4PJB]

Thanks! The plot I posted up is taken from the B manual but it is a reprint from the A manual. It looks like the loop response is different on the later models. Close in phase noise is better but the phase noise at 50kHz offset is higher.
Quite surprisingly, my analyser still uses the same loop response on a 50kHz span and this really doesn't look good. Maybe there is a way to upgrade it to emulate the later versions. However, that is a topic for another day...

My pleasure! Is this what you are looking for? Or did I miss a step?

Can you set it to 10dB/div as shown by KE5FX. I think you will then see a similar plot as the one from KE5FX's analyser.

Here you go! Sorry about the first go-around; my unit is butterflied out on my bench right now, so everything is either upside down, sideways, or both, depending on how I crane my neck to see anything.

[G0HZU]

Thanks! That looks very similar to the plot from KE5FX.

Mine is probably about 1-2dB noisier than the (noisy) image I posted up from the user manual. It used to be exactly the same as the image in the manual, but I think my analyser needs a service. The phase noise has become a bit fluttery at times and I think I will have some more capacitors to replace. It has been quite reliable for the 14 years I've owned it, and it was in regular use for the first 8 or 9 years.

When you get the VNA set up, I think it's worth trying the input s11 test at 2.7GHz. You have to set the analyser to a narrow span at 2.7GHz to keep the YIG preselector in one spot. I would suggest using a 1MHz span or maybe less. Also, make sure you select 0dB attenuation and turn the VNA source power at or below 10dBm.

[W4PJB]

Glad that was helpful! I would imagine the 8566 was on for about 10 minutes or so when I took that, so perhaps it would settle in a bit more after 30 minutes or so.

I tried the VNA (that has it's own thread, now  ) and was not able to get much resolution. Just as you suggested, I set the 8566 to 2.7 GHz, 1 MHz span (I even tried 100 kHz), 0 dBm, and set the VNA to -10 dBm, 2.7 GHz, and several different sweep widths. I calibrated the VNA on each change of span. Even at 100 MHz, there was no discernable notching in the reflection anywhere across the span.

[KE5FX]

Try entering ATTEN 0 dB on the 8566.  You're probably looking at the YTX through the 10 dB attenuator (which is 20 dB once the VNA stimulus signal goes through it and comes back.)

They make you enter 0 dB explicitly to help protect the mixer from static damage.  Be very careful with ESD and signal levels while 0 dB ATTEN is in effect.

[W4PJB]

Thanks, and great point about the signal being attenuated twice. I did set the 8566 to 0dB. VNA was -10 dB, as G0HZU suggested.

Guess it's not looking too good for Uncle YIG?

[G0HZU]

Give me some time and I'll power up my 8566B and an old 8714B VNA. Both are in storage (under the staircase!) but I can dig them out and power them up together. It's very late here in the UK so I won't be able to do this until tomorrow but I'll try and produce some images of the return loss you can expect to see if the YIG is tuning OK.

As KE5FX warns, be careful of ESD, especially if you start taking things apart and checking stuff with a meter or other test gear.

I think you should see a dip in s11 of at least 10dB at 2.7GHz and it should be about 20MHz wide. It might be a bit wider than this, but I think it will be no wider than 50MHz. You do have to select 0dB attenuation with the 8566B. With 0dB attenuation the return loss will probably only be 1-2dB outside the passband of the YIG and maybe 10dB to 15dB inside the ~20 MHz bandwidth of the YIG.

This assumes that all is well with the YIG/mixer of course.