HP 8563E Spectrum Analyser

[Julesomega]

I'm hoping someone can give me some simple advice on an 8563E which a colleague is trying to sell. The 8563E is only marginally Vintage but this series of SAs has been discussed before in these pages.

The one I thought I was buying is genuinely Vintage, an 8559A in a 182C mainframe which I bought him for his business almost 20 years ago. We meet up every year and last month he asked if I'd like to buy his SA so I made a reasonable offer knowing it was looked after well and he agreed, but when he brought it round it turned out he had retired the 8559A and rented an 8563E to work on a new product at 24GHz. This is far above my hobby requirements and worth a lot more so I said I'd try it out and see if I could find a buyer. I expect he'd settle below the commercial market value.

So to the question: the display looks fine but there is no sign of any signal, from the calibrator or an external µW source. He quoted me over the phone from the manual which mentions you have to replace a battery. I don't want to get inside the thing, but there seem to be two replaceable batteries accessible from outside. One invitingly even says "Press to release" but I can't see anything pressable, and I can't see how to get into the Memory Module without damage.

[Radio Wrangler]

Yes there is a battery.

DO NOT just try to replace it.
It holds memory alive holding calibration data. Search the internet and you'll find manuals and how to get it to display its cal data so you can write it down. Only when you've got the lot do you replace the battery. Then, if the memory dumps, you know the numbers to put back in. This is factory cal data and you'd need a lot of test equipment and software to re-generate it.

Not seeing a signal could be lots of things. There should be fault finding trees in the service manual, but they can assume you have a lot of test gear.

Rough value is that Stewart of Reading keeps selling these for £2000-2500.

CRTs are irreplaceable, so if you have a crisp, bright one, the box is probably worth fixing. They are thoroughly good instruments when working and are pretty reliable. However, like all analysers someone can always transmit straight into one, and sometimes they do fail of their own volition.

David

[G0HZU_JMR]

At work we probably still have 8-10 examples of this type of analyser but I rarely use them nowadays. They are all getting a bit tired and most of them have issues with the front panel switch membrane. Usually it is the up/down buttons and the 1 or the 0 buttons that go intermittent first. So even if you find a way to see a signal, I would recommend checking every button on the front panel to see if it is working. Once the buttons start to fail you have to press quite hard to get them to respond. I'm not sure if this is wear or dirt that causes this.

Be careful with the battery removal. Don't do this yet. I think you may lose the internal calibration constants and I think you are supposed to remove the battery with the instrument already turned on (to prevent this happening).

However, I have no experience of doing this and I'm not sure a dead backup battery would cause your symptoms anyway.

Just seen David's post above and he agrees about the battery. Don't try and remove it. I would recommend finding out if it is possible to do it with the instrument turned on. It might not lose the cal constants this way. But this might not apply to this model of analyser...

[Radio Wrangler]

There are checksums on the memory contents, so if the battery had discharged to the point of losing data, I'd expect there to be warning messages displayed bleating about it.

An E suffix box will have the 'flubber' keymat and I've heard of people cleaning the PCB and the little carbon pills attached to the silicone rubber mat to good effect. Other people have glued on new pills.

David

[G0HZU_JMR]

I think the mass memory module is only needed if you want to capture and store lots of trace data and maybe process it within the instrument. In all the years at work I'd be surprised if anyone has ever taken advantage of the features offered by the extra memory. We have a few of the newer LCD display versions at work and I think they all have the mass memory but I can't be certain. I'm not sure what happens if this battery fails but I doubt it will cause your symptoms.

Does the analyser show anything on the display when set to a centre frequency of 0.00 MHz? i.e. when set to 0.00 (zero) MHz is there a signal at zero MHz?

[Julesomega]

You mean I've got to turn it on
I had understood that the zero spur was showing normally but it isn't, and it shows Error 334 low amplitude. I'll have to get the manual off him and work through the fault procedures.
Thanks so much for the observations

[Julesomega]

After warm-up

[G0HZU_JMR]

I think error 334 is an error to do with the levelling of the first local oscillator. Are there any other errors that appear?

I think you have to recall the errors using RECALL and then go to MORE page 2 and then select RECALL ERRORS. I think you can then scroll through all system errors using the up/down keys or maybe the rotary wheel. The display can only show one error at a time so you have to manually scroll through them all. You might find that there are errors to do with the YIG oscillator and its lock status. The YIG oscillator is a known weak point on these analysers.

Scroll through all the errors and write them all down. Some of them might be old (stored) errors that are no longer relevant but it's best to write them all down.

[G0HZU_JMR]

The noise floor level on your analyser looks reasonable so this points towards a fault near the front end. If the first LO (YIG?) is faulty then this would give similar symptoms to what you are seeing with a reasonable noise floor but no signals and no spur at 0.00MHz.

[Radio Wrangler]

The 'signal' you should see centred on zero hertz is the local oscillator = LO = the YIG tuning across the IF frequency and leaking into the IF. It is usually a big 'signal'

There is nothing showing at the left edge. Try entering a centre freq of 0Hz on the same span. If you see nothing in the middle, the LO is either dead or far off frequency.

I don't see any lock and roll fault messages so maybe the LO is running and the distribution amp (I think that model has) is duff?

You'll need to check voltages on the cable going to the YIG, and also put a microwave power meter on the YIG RF output.

David

[John Caswell]

Hi Julian,
Sorry no experience of this spec an, however I have had a great deal of success cleaning this type of rubber membrane switch by carefully dismantling it and cleaning the individual pads and "lands" with IPA on a cotton bud and rubbing gently, leaving it to dry then reassembling.
Of course in this case "sods law" says you will fail

John

[Radio Wrangler]

That is almost two spectrum analysers in the same box, sharing some common stuff.
If you tune between 2GHz and 3GHz and vice-versa, you should hear a solid thunk inside as a solenoid switch changes over the front-end in use. Up and down transitions happen at different frequencies so you don't get any blind spots. BUT the analyser will refuse to sweep across the changeover point so you can't leave it sitting there thunk thunk thunk life-testing the solenoid switch.

In the low band, where you've been trying it. I'm trying to remember the exact IF of that model, but I was involved a lot with later versions and I've got them in my head. the YIG oscillator is used to mix the incoming (0 to 2-point-something GHz) up to an IF at (for example) 3.9214GHz. Then a 100MHz crystal oscillator and multiplier chain (100MHz to 300MHz to 600MHz to 1800MHz to 3.6GHz) is used to mix 3.9214GHz down to 321.4MHz. From that crystal-multiplier chain 300MHz is used to mix the 321.4MHz IF down to 21.4MHz. Most of the IF filter bandwidths are done at 21.4MHz and the logarithmic detector array runs at this frequency. The YIG has to tune 3.9214GHz up to almost 6GHz to cover this band. A 3GHz lowpass filter before the mixer takes care of images.

THat's the simple front-end!

When you tune above 2 and a bit GHz the solenoid switch thunks over and the 3.9214GHz IF is shut down. A different mixer (a packaged job with SMA connections) is brought into play. This front end is a down-converter mixing the signal down to the 321.4MGz IF directly. So now we have an image problem, the mixer will be equally sensitive 2*321.4 MHz = 642.8MHz above the wanted frequency. To suppress this image a tracking filter is needed. This filter is also YIG based and employs four resonant balls of YIG (was 3 in some models) to make a 4th order bandpass filter which is tuned by applying a magnetic field. Analogue circuitry and adjusters set the YIG filter coil current appropriate to the analyser's current frequency.

The YIG oscillator and the YIG filter are tuned upwards to scan inputs in the 2-point something GHz to about 7GHz frequency range, but that's limited by the tuning range of the YIG oscillator. The YIG filter is designed to go right to the top 2.5 to 22GHz and that top end takes a lot of coil current. The YIG filter is quite something, technology-wise.

The YIG oscillator jumps down to about half the frequency it had reached and the mixer continues the conversion of higher frequency signals using the second harmonic of the YIG oscillator. This gets us to about 14GHz when the YIG oscillator again hits the end-stop and jumps back down while the YIG filter continues upwards and the mixer runs with the third harmonic of the YIG oscillator to eventually reach 22GHz.

So, if you have a working example of one of these analysers, and set it scanning 3-22GHz, you'll see the noise floor jump up where the mixer changes overtone mode This is a quick indication that most of the microwave front end is working and has an LO signal.

If you have it tuning in the below 0 to below 2 point something GHz range, you see the 'zero hertz blip' if things are normal. This is the YIG oscillator crossing the first IF frequency making the analyser plot the selectivity curve of its own IF filters. THis is a handy check. And your machine is showing trouble here.

The YIG oscillator has several things to drive. It must supply LO to both the UHF and microwave front end mixers, it must supply a signal to the frequency control system so that it can be phase locked for the lock-and-roll system it uses. It must also supply a signal to an SMA connector on the front panel, which is there to allow the use of external microwave mixers up into the 100GHz region.

These are very capable analysers and there's a lot going on in them. You feel the weight when you carry one.

So from what I've seen so far the screenshot you posted says that the low band mixer doesn't seem to be getting a local oscillator signal OR it's getting one at a wrong frequency which never crosses the 3.9214MHz IF. There could be a disconnection in the 3.9214GHz IF path or one channel of the solenoid band switch.

A YIG far off frequency could mean a fault in the YIG oscillator coil driver board and the YIG is just sitting at some other frequency, usually below 3.9214GHz. You can check currents by probing across the sense resistor on the board.

No LO signal could be a dead transistor in the YIG oscillator itself (Needs a new YIG) or a missing power supply rail to the YIG

The YIG could be running but its signal might not be making it to the mixer, due to a fault with the LO distribution amplifier.

Another possibility is that someone keyed up a transmitter straight into the input and blew up your low band mixer. There may be LO activity, but no mixing going on. People before now have fixed low band mixer boards with a new pair of diodes.

Well that's a brain-dump on the things. I'm working purely from memory as I don't have the manuals in front of me, and my memory will be confused with all the other variants on that same theme.

You can trust the 100MHz....300...600...1800 MHz multiplier series. I'm wondering if this variety used 10.7 not 21.4MHz IF.

some used 3600-321.4MHz. but you should get the picture of how it works, and where the fault may lie.

Tests go along the lines

Can you get the zero hertz blip? seemingly not, but do scan across 0Hz from say -100 to +100MHz

Scan 3-22GHz and do you see the steps in the noise floor?

Aftr this you need to check YIG coil currents, try to measure RF power coming out of the front panel SMA connector to see if the YIG is alive.

All of these things will narrow the area to be searched.

David

[G0HZU_JMR]

I'm wondering if this variety used 10.7 not 21.4MHz IF.

Yes, the frequency plan is slightly different on these analysers compared to the classic HP8566B.I've been using the HP8560E and HP8563E(C) analysers at work since the 1990s but not so much in recent years so I am a bit rusty these days.

I think the IF ranges for the HP8563E on range 1 are 3.9107GHz, 310.7MHz and 10.7MHz.

Unlike the HP8566B the HP8563E can continuously sweep across from range 1 to the higher ranges. If it is preset it will sweep 0-26.5GHz and it can sweep this range in about 500ms. So it will happily switch ranges during the sweep, unlike the classic HP8566B which is limited by its mechanical range switch.

The close in phase noise on range 1 is very good, much better than the HP8566B. It also has digital RBWs for the narrowest bandwidths of 1Hz through 100Hz. Back in the 1990s this was a very useful feature for us at work compared to our existing analysers that were limited to old school (slow to sweep) analogue RBW filters. What isn't so good is the CRT display, I think it only has 600 points in the x axis compared to the 1000 for the HP8566B. This is a bit disappointing. So some care is needed when using the sample detector and narrow RBW settings or it can be more prone to display (amplitude) errors compared to the HP8566B.

The IP3 and 2HI performance is average at best so some care is needed here when measuring distortion, especially on range 1.

The HP8563E (and the 2.9GHz HP8560E) also have an annoying habit of pausing the sweep or producing a slow and stuttery sweep rate now and again and I assume it is doing some internal alignment/cal checks when this happens. But this can happen a lot and it is one feature I don't like about these analysers especially when looking for rapid changes in a signal. It often hesitates and slows down the sweep just at the wrong moment!

The CRT display can get quite grubby and tired looking over time and these analysers are very dated now. Our examples at work have been absolutely hammered in terms of demand/use since the 1990s and I'm amazed we have so many of them survive this long.

[Radio Wrangler]

Ah, so it is the 10.7 one. A later generation of analysers, project named 'Mosquito', went back to 21.4MHz, but the digital narrow filter trick was retained though watch out for reduced dynamic range, there weren't really enough bits of width. They used the ADC that normally digitised the output from the logger in Mosquito, while the next top-end model went to full DSP (Chipset ADC and DSP called Boris and Natasha). By this time R&S had got well ahead in analysers and HP/Agilent never caught up, they'd just kept starting new R&D projects then canning them. None ever lasted long enough to bring out a product, but the money was spent and time went by. It was a frustrating time, Rohnert Park would start a new project to get us back to the market position we were in in the days of the 8566, and then a year or so later it was cancelled.

David

[Julesomega]

While waiting for my friend to bring the manual I've followed up a couple of the suggestions.
The error history shows just Error 334 LO amplitude
There's no sound of a relay, neither tuning 2-4GHz nor during a Full Span so I expect the switches are solid state.
Full Sweep does show steps in noise floor so the switching must be working.

When I get the manual I'll know how to get the covers off so I'll be able to take a few measurements later.

[G0HZU_JMR]

On one of your images it is also showing error 335 and this is something to do with an unlocked oscillator.

Do your up and down buttons work OK? i.e. can you reliably change the span using the up and down buttons for example?

If these buttons are tired and a bit deaf you might not be able to scroll through all the errors because the analyser doesn't realise you are trying to press these buttons. You could try using the rotary wheel instead but I'm not sure it will be tied to this function. The other way to do it would be via GPIB but not many people will have the hardware for connecting GPIB to a PC.

[G0HZU_JMR]

This might be your service manual but you will need to check the serial number.

http://sa.support.keysight.com/8560/...8563-90214.pdf

You might be able to find some CLIP information online if you search hard enough. The CLIP files will cover the internal modules in more detail and they may include circuit diagrams. I think the relevant CLIP number is 5967-8582 if that helps.

[Julesomega]

Thanks again Jeremy. I hadn't clocked that you have to scroll through the error memory.
Doing that reveals
356 sweep data problem finding "Bucket 1" of the span accuracy calibration sweep
319 YTO coarse tune DAC is near its limits
317 main coil coarse DAC at limits
353 no acceptable YTO DAC value found
351 YTO error voltage not settling

I measured the LO o-p (none), calibrator (-10dBm) and 10MHz o-p (+3dBm). This all points to the YIG osc or distribution amplifier

Thanks for finding me the service guide. There's nothing at component level but at least it tells me how to take the case off so I hope to make the relevant measurements

[G0HZU_JMR]

If it helps I did a quick google for HP8563E YIG repair. I was hoping to find a common failure that could be fixed 'at home' if your YIG oscillator module turns out to be faulty.

Here's some images of the insides of this type of HP YIG:

https://www.eevblog.com/forum/repair...-85628563a-sa/

Also, there appears to be an online company that offers to repair the YIGs (with high success rate) so I suspect that the failure could be a fractured wire bond or a semiconductor failure. It will probably require specialist equipment to work with parts this small. However, a broken YIG may be fixable instead of it becoming an exotic paperweight!

[Radio Wrangler]

They might be repairing a crack in a substrate by bonding bridge wires across it.

Some YIGs were made by Microsource of Santa Rosa. I suspect there was some connection, given the address.

David