HP 8656B repair attempt... ongoing

[Swainster]

OK, this thread is a tale of 2 HP 8656B generators. I actually came by them 'by accident' as the the first one was bundled in with a power supply that I wanted. It was non working but complete - for the sake of clarity I shall call this #1. A quick inspection revealed a 'popped' inlet filter and the typical RIFA stank. The inlet filters are still available (at a rather salty price), so one new Schaffner FN370-2-22 and the instrument was up and running - or so I thought... Looking at a few random frequencies on a spectrum analyzer showed that the level was down a few dB, but other than that basically following the attenuator settings, and the frequency was pretty close. The fact that my work has a fairly new Siglent SA for EMC pre-testing was turning out to be pretty handy.

Anyway, I happened to send some pics of the working generator back to the guy I bought it off, and he came back to say that he was moving office, and had an incomplete 8656B spares unit, which I could have (as I recall) for the equivalent of about USD30. Sounded good to me so I ended up with another, somewhat more sorry looking patient, unit #2. This one was missing the power transformer, and once again, the inlet filter had vomited its insides.

Some time later, in the course of another thread here on EEVBlog, https://www.eevblog.com/forum/testgear/fool-for-the-8656a-sig-gen/, user Brumby mentioned that he was breaking a 'B for spares. Hmmm, should I do the same to my 'spares' 'B, or fix it up? Well, TEA is a harsh mistress, so fix it up was the way to go. Do I need 2 RF sig gens? No... I didn't even need the first one at the time, although now I'm starting to get into this 'RF stuff'.

First job was to check on the state of #2 - easy to do as I just swapped over the rear panel from #1, which contains the transformer and inlet filter. Sure enough, #2 powered up and levels and frequencies were ok. Well, other than being 6dB down below 100MHz, and 2dB down on the rest of the range. The latter could be put down to measurement set up (i.e. cables or whatever).

So to conclude, at this point I had 2 (hopefully) mostly functional HP 8656B signal generators, but only 1 working power supply between them.

Edit: Just for fun, added a pic of unit #2 at 990MHz as measured by an older relative (reading on the HP 5245L is .95GHz on the dial + 40000.019KHz on the display)

[Swainster]

And here is the new transformer, courtesy of EEVBlog member Brumby - cheers! And with it, a new inlet filter courtesy of Radio Spares (RS Components). Once the nasty crispy bits were scraped off the back panel then the new parts could be easily fitted - nothing much to report here. A quick check on an adjustable AC power supply, making use of the current limit to avoid any accidents, and we are good to go... Right?

[Swainster]

Hmm, actually a couple of issues have popped up.
1) The fan is not running - could be a faulty fan, or I screwed up the wiring
2) I've lost, or never had, the cable assemblies connecting the REF in and REF out to the back panel.
3) There is a problem with the RF output. The RF PLL seems to regularly fail to lock on certain frequencies, and then basically sweeps up and down the range. This happens in ranges 2, 3, and 4, however range 1, the low frequency heterodyned range, seems stable.

Note that at this point, unit #2 has the back panel assembly of unit #1 fitted, whereas unit #1 has the repaired back panel from unit #2, which is missing the REF IN/OUT cables and has a non-working fan. The RF issue is with unit #1 which previously tested out OK. So... could the back panel cause the PLL unlock, or perhaps this a newly developed problem? If the former, it could be due to the lack of the REF cable assemblies. Although they are not attached to anything other than a BNC jack on the back panel, perhaps their lack may be allowing noise into a sensitive area. So that's where I am now - about to swap the back panels back to their original owners.

Edit:
I just thought I'd at that at the frequency shown in the pic below, it will unlock maybe 4 out of 5 times, but perhaps 1 out of 5 times it will succesfully lock to 130.1MHz

[Tony_G]

Looking forward to following this thread with interest - Good luck Swainster!

TonyG

[Swainster]

Thanks TonyG! In the end I didn't get around to swapping the back panels before packing up for the day - I got distracted with putting the finishing touches to an Atari Punk noise maker, that will hopefully be a good distraction for my little one, as she is growing up to be a right little knob twiddler like her old man. Hmmm, not sure about that last phrase... Naturally I put the volume control inside the casing, so it's not user accessible 

[Tony_G]

Naturally I put the volume control inside the casing, so it's not user accessible 

I also look forward to you finding some of your screwdrivers missing...

I can't imagine that the 10MHz ref out cable would have any impact (not saying that it couldn't but I just find it hard to believe). I'm more interested in hearing what you find out about the basic band, seeing that you have "good output" in band 1 but not band 4 - Seems strange to me.

TonyG

[Swainster]

Well, I swapped the rear assemblies back to their original owners and as TonyG suggested, the issue followed the main unit, not the back panel. Some good news though - the uncooperative fan started working. I suspect that it was just jammed up by a stray cable.

It looks like the unlock issue is getting worse, with acquiring lock getting less and less likely. This is potentially a good thing as acute problems are often easier to troubleshoot. As far as the heterodyned range 1 goes, it still locks on all my randomly selected frequencies. However, you can see that it spends half a second or so before locking. The good unit locks instantly, or at least without any perceptible delay.

Anyway, looks like this thread might end up a bit more interesting than a simple transformer change.

[jrharley]

They don't call 'em rabbit holes for nuthin'..

Great to see the 8656 information continue to flow, thanks for posting up swainster! It's really interesting to see the B version. The back transformer/power section is particularly interesting to see. HP was really changing and - theoretically, improving these kits year over year. I wonder what the thinking was about bolting the TO-3 packages directly to the finish though. I suppose they reckoned that the guys shouldn't get too hot. That detail would not escape HP for sure. At any rate, looking forward to following the effort, best of luck!

JRH

[Tony_G]

I wonder what the thinking was about bolting the TO-3 packages directly to the finish though. I suppose they reckoned that the guys shouldn't get too hot. That detail would not escape HP for sure.

JRH - If you were to nail me down and say give me one reason they didn't care about the heat, I would say that they expected that these things would be used in racks that were in heavily cooled installations so the heat output wasn't critical (though I'd probably also vacillate that with it was the way things were done at the time).

I'd probably start looking at this by tracing the 50MHz reference signal and seeing what I see - I'd then probably look at the VCO and make sure that it was putting out a good signal.

Actually thinking about it a bit - Given that the basic band is horked as well - I might actually start looking at the VCO and working out from there - Maybe a bit of both (really helpful I know).

Anyway, I did buy the manual from Artek for the unit (sad news, Dave who ran Artek and was an absolute stand-up guy, passed away over Easter, so I don't know when I'll get the manual) so I look forward to offering my $0.02 in future posts.

TonyG

[Swainster]

Yes, the overall cooling on the HP 8656B seems to me to be a bit inefficient. They have that big exhaust fan, and air ports from the rear compartment into the main compartment, but as far as I can tell, the only air inlet is the tiny gap between the top and bottom covers and the front panel assembly. They could have used a much smaller fan if they had designed a more free flowing air path. I guess that they had to compromise with the RF shielding.

I think that one of the biggest improvement going from 'A to 'B was the attenuator - reading around, it sounds like the reliability of this block on the 'A was a real problem from the launch of the instrument (probably why your unit had parts missing), and may have led to some design qualification process changes at HP.

Regarding my 'patient', I think I will do a quick check of power supply voltages, then stop to do a thorough read through of the relevant sections of the service manual before proceding.

I just heard about Dave from Artek in another thread. I remember reading on this forum that there was some concern over his health but I thought that it had been reported as a false alarm. Such a shame that he is no longer with us.

[jrharley]

No doubt the attenuator is the Achilles Heal of the A version swainster, it's infamous. The B version is superior in many ways from what I can tell. I know you'll get 'er up and running right.

To Tony_G's point on checking the VCO, if I learned anything from my efforts on the A, it's that even if the power supply looks good at first, check that sucker well, including the devices regulating it out of circuit. It's a good overall confidence/maintenance effort in any case. I found it fascinating that if I had just performed the simple task of testing the regulators and cleaning them up FIRST, I would have been farther along much quicker. Just sayin'.......

Sorry to hear of any fellow fan boy passing on, appreciate the heads up on that.

Looking forward to the progress.

JRH

[Brumby]

Glad the transformer made the trip intact and that it is headed for a new life.   

[ekoloski]

Worth mentioning, use care when replacing the back panel with that transformer. There's a sheet metal housing around the GPIB connector that easily catches the Ref IN/OUT coax. My unit unfortunately crimped these cables and damaged them when someone before me had it open...its had a rough life but I am getting it back into shape.

I do not believe having those two wires missing would cause the locking issue, but who knows. What seems strange is that it is stable in the other ranges. I've been doing some work around the fractional-n asic on my 8656B, due to a failed A3U28. I'll dig into the pll side and see if I can offer any advice. The good news is, you've got a good one to compare with! Glad to see both getting fixed up!

As for issues with the output power. There's an AGC circuit that samples the output and applies a correction voltage to a string of pin diodes. Mine needed some help there, the 500 ohm bias potentiometer had failed and caused intermittent issues with the output level. Is your output power consistent?

[Swainster]

I'd say the output levels are stable, except for the offset between the heterodyned and non heterodyned ranges. Interestingly, the offset is very similar on both generators - perhaps there is some predictable ageing mechanism which is causing this effect, as they only about a year different in age - one is an '88 model, the other is an '89, going by the chip markings.

[Swainster]

Not much to report yet, but got some more pics to share.

As I have 2 generators I decided to do a board substitution for the more easily accessible RF board I.e. the large outer one with the 50MHz reference. Part of the reason I did this is that I realised that I don't have any SMC cables (other than the REF in/out cables on the generator itself) to connect the board test points to my spectrum analyser.

This was the first time I've had a poke around in the RF guts of these instruments, so I took the opportunity to have a 'peak under the covers'... so very many covers 

Anyway, to cut to the chase, the problem doesn't appear to be in the 50MHz ref board, so most likely resides in the RF magic living in the compartments beneath.

[Swainster]

About time I updated this repair... it is still ongoing so no change there.

Anyway, first had a few problems to overcome. The first one was lack of SMC cables or adapters - this was solved by borrowing the REF IN/OUT cables. Annoying to have to remove the back panel yet again to retrieve them, but well worth it, given my lack of alternatives (SMC adapters on order). Secondly was the lack of the proprietary HP test adapter. This appears to be just 2 pins (breadboard size pins i.e. component leg offcuts) on the end of a cable. I wanted to make this up from RG316, but didn't have any suitable fittings, so I ended up using RG58 as I've got plenty of cable mount BNCs which match.

I then managed to dig out my EMC troubleshooting tools, so pressed them into use as well.

[Swainster]

Here are my latest findings:

1) The 50MHz reference level is running a bit low on both units, being just under the called out limits at just over +15dBm (limits are +16 -> +19dBm). For now I'm giving this the benefit of the doubt as the shortfall could be related to my measurement setup. That said, a number of measurements were coming out below the recommended levels in both the units (working and faulty), so for the moment I am concentrating on where the 2 units differ.

2) The first significant discrepancy is with the A8 frequency multiplier output level. At TP3 on A8, i.e. after the multiplier, the level on the good unit is 10dBm higher than the bad unit. Using an RF E-field probe suggests that the output of the final multiplier transistor is within 1 or 2 dBm on the 2 units, and the difference comes in after the 800MHz bandpass filter. This filter is almost all PCB trace RF magic, so I'd have to guess that the issue is due to excessive load on the multiplier?

Between TP3 and buffer amplifier #1 is a 237ohm resistor which I guess should isolate the amplifier from the multiplier to some extent... you may notice a lot of guesswork going as I've mostly avoided RF circuits until now. Anyway, I also tried removing this resistor one the faulty unit to further isolate the multiplier. This did not got well as I cracked the covering the resistor. This hasn't changed its value so I'll assume its ok for now, and swap it out later. Anyway, while Q3 of amplifier #1 does seem to have less gain on the faulty generator (going by the base current), removing the 237R didn't improve the level at TP3 so I think that may be a separate issue. The 2 generators actually have different parts fitted for Q3, which may explain the difference.

The other circuit connected to the multiplier output is the low band heterodyne mixer. It's not really practical to disconnect this circuit, plus it is all passive until the mixer itself. After the heterodyne mixer is the low band output amplifier. The 2 transistors in this section seem to be OK, at least going by the DC bias voltages.

Incidentally, the freely available service manual is quite low resolution and really giving me a headache. I'm trying to order a CDROM manual from Artek, but as you may know, they are not yet fully up and running since the sad loss of Dave. In particular, soft copy manuals are not properly catalogued so CD ROMS are the all that's available. Meanwhile, I don't suppose anyone has a better scan of the schematics that they can share? I'm mainly interested in SS1 - 5 at the moment.

Anyway, I'd appreciate any advice on this issue, i.e. what to measure next etc. One thing I did think of was to use the working generator to feed in an alternative 50MHz reference. Am I right in assuming that increasing the input level to the multiplier will also increase the output level?

Edited for clarity

[Swainster]

Today, I've just got a sulk to report  - I bought a handful of RF transistors from the local electronics components retailers (they carry a lot of old stock), which should be have somewhat similar to the BRF91As used in the circuits that I'm currently investigating, and despite buying from 2 different stores, they all appear to be faulty, according to my component tester (and also when substituted into the board). Equivalent of USD$10 down the drain... though to be honest, I love searching through the old stock at these kind of places so I probably got a few dollars worth of entertainment. Now checking with a local private source (actually the guy who sold me the sig gens), plus also lined up some (hopefully) actual BRF91A transistors from eBay (US).

[jrharley]

I feel your pain swainster!

Can't be much help with the circuit issue at the moment, I'm still trying to catch up, but I can empathize with the increasing scarcity/reliability/etc. of certain parts.

Enjoying the process.

Later,

JRH

[jrharley]

Hey swainster, I've had a little time to study your descriptions of the discrepancies between the units at the freq. multiplier outputs. Really interesting stuff, appreciate the post and the pics.

I'm not a digital Wizard by any measure, so I'm pretty worthless when it comes to circuit insight here. As it is, I am confined to simple, fundamental analog @h!t. I hope it's not insulting to ask if the ins and outs, connectors etc, have been cleaned, re-seated, micro-examined and anything else of mechanical interest considered. 10db seems like a lot to me, but I DO know that circuit problems in gear of all types, are so often a connection problem of one sort or another, that it's a box to check every time. I've been surprised in my various journeys how just the act of "tidying up" can reveal the issue. After all, HP didn't use those semi-rigid SMA connections for nuthin'.

Anyway, I know thats probably not very helpful, but it came to mind as I looked at the pics.

As always, I'm enjoying the process and wish you the best in your efforts.

Later,


JRH

[Tony_G]

Hey Swainster - Did you ever get a good copy of the manual? My website order is hanging in the backlog (I'm sure they'll get to it at some point) but as far as I can tell there isn't a CDROM version available.

TonyG

[Swainster]

Hey swainster, I've had a little time to study your descriptions of the discrepancies between the units at the freq. multiplier outputs. Really interesting stuff, appreciate the post and the pics.

I'm not a digital Wizard by any measure, so I'm pretty worthless when it comes to circuit insight here. As it is, I am confined to simple, fundamental analog @h!t. I hope it's not insulting to ask if the ins and outs, connectors etc, have been cleaned, re-seated, micro-examined and anything else of mechanical interest considered. 10db seems like a lot to me, but I DO know that circuit problems in gear of all types, are so often a connection problem of one sort or another, that it's a box to check every time. I've been surprised in my various journeys how just the act of "tidying up" can reveal the issue. After all, HP didn't use those semi-rigid SMA connections for nuthin'.

Anyway, I know thats probably not very helpful, but it came to mind as I looked at the pics.

As always, I'm enjoying the process and wish you the best in your efforts.

Later,


JRH

Hi JRH,

All suggestions are very much appreciated! However as far as connectors go, there aren't that many, and they are all SMC so pretty secure. Between RF sections, the semi-rigid cable is almost too secure - it's part of why i'm having to take it so slow. I don't really fancy undoing all the semi rigid as it is soldered down at multiple points to the PCB ground, and even to adjacent semi-rigid cable runs. There is one tiny wire jumper in the circuit, although I think it's between the buffer amp and the mixer, and the signal is already low before it even gets to the amp.

Before I start trying to disconnect the coax runs, I've got one more scheme to try, which is to see if my nanoVNA can shed any light on the situation. It's a new toy, and I'm just starting to explore how I can make use of it. Anyway, will get back to it at the end of the weekend, or early next week.

As a distraction, my 'fancy' Keysight bench DMM randomly died. I turned it on to do some 4 wire resistance measurement, left it to warm up, and when I came back to it the open kelvin clips were reading 0ohms. According to self-test it's suffering from ASIC failure. Rather like my sig gen, I've got a spares unit, but that one seems to have similar symptoms so the spares unit may not be much use. I guess I might have to start another "ongoing" thread.

[Swainster]

Hey Swainster - Did you ever get a good copy of the manual? My website order is hanging in the backlog (I'm sure they'll get to it at some point) but as far as I can tell there isn't a CDROM version available.

TonyG

I had some correspondence with Dave's widow via eBay and I'm afraid that it doesn't look like the website will back in action any time soon. I get the impression that the plan is to convert all the files to CD ROM format and concentrate on that method of distribution, however there is no timeline for this yet. It's only been about a month since Dave passed so I didn't feel like pushing for too much detail.

In the meantime, i've purchased an original copy of the service manual. It will take a while for it to wing its way over to me, but I should have it before the end of the month. Hopefully it will be in good condition. It will be my second service manual - the only other one that i have is for a HP model that I don't even own  (HP 5300A)

[Swainster]

OK, some good news! I've had a bit of a breakthrough, especially with regards to understanding what is going on here.

First of all, I hooked up the nanoVNA to TP3, the output of the frequency multiplier, i.e. the first place where there is a noticeable discrepancy in signal level between the working and non working units. The nanoVNA revealed a broadly similar shape for the return loss, however there was some difference in the location of the dips. With the benefit of hindsight, I probably should have set up the sweep to concentrate around 800MHz.

Anyway, the important thing is that the ability to 'see' something of what was going on in-circuit gave me the confidence to twiddle the trimmers at the input to the heterodyne section. This was the jackpot - it turns out that these trimmers affect the loading of the previous section (obvious in hindsight), and also how the power is shared between the heterodyne and high frequency mixers. It also seems to be a case of one fault hiding another fault, as the 'working unit' is actually diverting more power to the high frequency mixer than it should, meaning that it can successfully lock, however leaving the low band a few dBm below its neighbour (on the working unit, frequencies below 120MHz (i.e. heterodyne mode) are about 6dB low, whereas only 3dB low above 120MHz). In both units, the 50MHz reference being at (or just below) the low end of the allowable range is probably the true root cause of the unlock issue.

With some semi-random twiddling of the 2 trimmers on the non-working unit, I was able to achieve stable output on all the ranges, and a good level match between the heterodyne and non-heterodyne ranges - basically 3dB below nominal through the whole range. To complete the fix, I think I need to bring up the 50MHz reference by 3dB, but that will have to wait till I get some more spare time to look at it.

Semi related, I was trying to decide whether it was likely that my spectrum analyzer was 3dB high or the sig gens were both 3dB low. Current assumption is that my relatively new analyzer is accurate, but like the guy with 2 clocks, there is no way to tell for sure.

[Tony_G]

Good to hear - What happened when you tried adjusting A3C21? Was it already maxed out?

TonyG