Hewlett Packard HP8640A


 Here's a view of my latest HP acquisition. Not exactly as it seems because the range knob dial is just resting on the panel and the left handle is snapped off at the bottom. Hopefully I can sort these things out and clean up the corrosion on the alloy fittings. At least the panel looks clean and tidy. It remains to be seen whether the various switches and mechanism parts are in good order. The tuning is a little stiff but may just be lack of lubrication or a build up of surface corrosion.

I've shown below the differences in panel layout between A and B examples. Clearly the A version has no built-in counter but I understand its spectral purity is superior to that of the B version.



 One of the first tasks will be to find a suitable knob for the range control and to fit back the dial then remove the panels and see how the thing has fared over the years and make a judgement on refurbishment. I quickly found a suitable knob for the range switch and fitted it but the thing failed to turn and because of the complex plastic gear train behind the panel I didn't force it to rotate until I'd investigated further. Initially I detached the top cover which needed lots of persuasion due to heavy oxide deposits but it freed up to reveal a light covering of yet more oxide. An old electric toothbrush soon removed most of this as well as some of the lettering on the largest panel. Oddly, although the top and bottom seals were intact, whoever had last been inside had fitted the panel covering the innards back to front (easy to spot because arrows should point to the power supply pcbs at the rear of the chassis). I cleaned up everything including the fan blades which were smothered in aluminium oxide. To complete this area I had to detach the metal filter. Once happy with the results I refitted a cleaned up lid and turned the chassis over so I could detach the lower cover. This covers amongst other things the plastic gears which I felt were probably going to be pretty rough.



 What's left of the range knob removed. Now, I must signal a warning to anyone contemplating working on an HP8640 signal generator, which is to purchase a set of suitable Allen Keys. Many knobs are fastened with 0.05 inch grub screws and these may well be seized and can only be removed with a good quality 0.05 inch key.


 The rear panel shows that this equipment indeed is fitted with Option 2, an internal doubler extending the range to over 1GHz.

Voltage selection is accomplished by pulling out the fuse assembly and refitting in the alternative mains setting position.

Nice to have an IEC connector.




Below is a picture of the underside of the top cover.




 Above is what was revealed under the top cover. Not too bad and amazingly you'll see the extender board for testing pcbs and a set of spare fuses in the plastic cup lower left (and below). Thankfully the tuning dial cord (below left) looks intact and in perfect condition...





 Above is the pcb under the back-to-front lid. Apart from a trace of rust on one gold-flashed chip the thing looks perfect and after detaching the lower lid everything looks perfect apart from a thin coating of dust. All the plastic gears look intact although they seem to be stuck from dried grease. All the other gears such as those attached to the attenuator etc also look perfect.



 Under the air filter was a mass of debris and aluminium oxide which readily cleaned up..


 Above is the view once the lower lid had been detached.. even the spanner (centre right) was present..

I'm really pleased to see yaxley switches instead of the abortionate spring-clip devices used in the HP8640B.

Now to look at the gears etc and see if they need lubricating.

All the range gears were OK and I added some copper-ease grease to these and the low pass filter operating mechanism, however I spotted a couple of tuning gears with cracked centre bushes. I could superglue these later to preserve them. The reason for the cracking was probably due to unequal expansion of the plastic and brass bushing. Unfortunately the smaller centre gear of the main tuning gear has a missing tooth. This hardly affects operation as it stands but I did find the tuning pointer doesn't move to the extreme left. In fact at the furthest point about an inch from the end, reversing tuning causes the dial cord to unwind and slip from its bush. I managed to refit this using a pair of tiny crochet hooks but this problem is the likely reason the equipment was set aside.

Clearly the tuning needs to be resynchronised after making good the missing tooth. The tuning for the cavity oscillator has a saeries of disks which allow for a pair of mechanical end stops, but if the gears connecting to the cavity gear slip out of position I can envisage a tooth breaking off due to excess pressure used in rotating the knob. In fact this area was found to be faulty when I refurbished my HP8640B.

Below are some documents which comprise Service Sheet C in the HP8640A manual. Click to see full size.








 In fact. although I read these pages I hadn't appreciated the clever design of the tuning and wavechange display. Removing a few screws enables the entire assembly to be merely withdrawn from the equipment. The dial cord arrangements in their entirety are then exposed just as shown in the pictures above. But why did I need to remove the display assembly? Well, basically something was jamming up. A previous knob twiddler had broken off a tooth on a gear and anyway the pointer didn't reach the extreme left side of the dial. If an extra turn to overcome resistance was applied then the dial cord slipped off the small drum on which it's wound.

Once the assembly had been detached it was clear that all was not well. Sometime in the dim distant past a tiny pulley must have come loose and this had been refitted, not with a simple wire axle but with a steel pin (another term for a nail). My guess is two things resulted... firstly, superglue had been used to fix the pin in the plastic moulding and this had travelled up the pin and seized the pulley and secondly, the pin had rusted. The result was a completely seized pulley and this had affected smooth running of the dial pointer. Whether this alone had messed up the mechanism remains to be seen.

As I haven't come across any other description of this repair I've provided the following set of pictures.




You can see there are two independent dial cords. That for the range switch rotates the drum and needs investigating due to stiff operation whilst the second, for pointer movement, is in a mess.

 Note the cord has slipped off the RH pulley resulting in extreme slackness in the dial cord, and it unwinding from the black part as soon as the tuning gear was moved.




 The pulley block now detached from the housing revealing a stain at the RH pulley centre. The copper grease on the range pulley is from earlier attempts to free this part of the mechanism.


Note also the slightly odd design of the black part on which is wound the dial cord. A screw is provided to fine tune the amount of take-up of the dial cord at one end of the dial.. a sort of fiddle factor.


The metal pulley was tricky to detach from its metal shaft as any slip might have wrecked the plastic parts essential for operation. I had to smooth the shaft with emery paper to refit the pulley.



Below, the tiny seized pulley removed with its steel pin.

To part the axle from the pulley wheel (ie. break the superglue bond) I aligned the head of the pin over the hole in a square dexion style nut and tapped the end of the axle with a hammer. Fortunately this worked and I was able to clean up the axle and refit it in the pulley then the plastic T-piece. No more superglue used..



 The next step is to reassemble the parts on the dial assembly and check for smooth operation of both the range and the tuning mechanisms before refitting in the equipment. To align the tuning I guess I'll need to power the equipment and adjust it without the dial to a specific frequency then set the dial to match this and refit it.

Reassembly of the dial assembly was easy enough and after using copper-ease on the plastic range ident and its ball bearing the switch changed from ropey and stiff to perfect. The tuning dial was also perfect without any glitches. Turning to the front panel, before refitting the dial, I found the cavity tuner was pretty stiff compared with the one in my 8640B. WD40 on the front bearing and the bearing on the front of the cavity had no real effect.

There are two potentiometers driven by plastic gears from the cavity tuner. One of these gears is cracked but the crack closes up in operation with only a slight glitch. I might be able to glue this but I also need to fix a missing tooth on the cavity gearwheel and this would mean detaching all three gears. This looks tricky without removing the whole front panel. In fact I just detached the cavity tuner and moved it back abut an inch. After a lot of experimentation I discovered the circlip holding the large plastic gea wasn't fitted into a slot but just gripping the shaft so I was able to lever it off, removed the large gear and then the two smaller gears after slackening their pairs of 0.05 inch hex screws.


 The worst gear has a split between a pair of teeth so I pressed out the brass bush and turned this down so it would slide back into place after gluing the cracks. This was only partly successful because the crack opened slightly but I can refit this gear to its pot in such a way as to avoid the slightly wider spacing beteen two of the teeth. This is because the pot does not turn a full 360 degrees.

The second smaller gear had numerous cracks but not extending to the teeth so I glued a metal washer on the bush to revent further damage. I also fitted a similar washer to the first gear.

The most difficult task will be to fit a new tooth to the larger gear as this missing a tooth. I reckon I might be able to repair this by gluing a sliver of plastic in its place. The new tooth will measure 6.5mm x 1.5mm x 1.1mm thick and will need to be profiled to match the mating gearwheel teeth.



 A sliver of plastic glued in place. I found a flexible plant marker which was 1.1mm thick and cut a tiny section using a scalpel to match the teeth but allowing some excess so that the material could be scraped to match the other teeth. The key check was using a vernier caliper to measure the diameter across opposite teeth. Once this was within 0.1mm by scraping and filing, I contoured the edges of the plastic with the scalpel until it roughly matched the mating gearwheel without dragging.

 During the above repair, whilst the glue was curing, I checked to see if other options were possible and found that one could add baking powder to superglue to fashion a hard material which could be roughly contoured using the mating gear whilst still flexible. A third option is to use something like a JB two-part epoxy and a miniature hand tool fitted with a tiny cutting disc and a dentist-style drill.

Refitting the set of three gears proved the repairs were successful. One pot goes from near fully clockwise to near fully anti-clockwise whilst the other is opposite going from anti-clock to clockwise. I'm not sure whether the position of the two pots is critical but if it is I can reposition the relevant gear on the pot.

Maybe new gears are available? I counted 80 outer teeth and 12 inner teeth, but the diameter is critical because the outer teeth mate with a steel gear on the cavity tuner and the inner drives the twin pot gears. Maybe an interesting option is to use a 3D printer?

Anyway.. moving on to the next issue. Revisiting the display.. below is the space where it fits.


 I fitted the display box back in place after sorting out a seized pulley and rethreading the cord. Tuning worked almost to the RH end where the pointer jammed followed by a loud click from the display box. Tuning backwards revealed something had gone wrong so off with the box and a second investigation began. After several failures I put a 6 inch ruler against the inner edge of the front panel and found it was slightly bowed.. maybe by only 1 to 2mm but that together with corrosion was causing the pointer to jam. I scraped and filed away corrosion and tried again. This time it was better but the pointer still jammed. I found the outer surface of the pointer had a lump of dried glue sticking out so I filed this off. Of course we're dealing here with maybe 0.1mm or so and the front panel bowing is at least 1mm at the point of jamming, but finally the pointer moved freely.

At this point I needed to fix the range display. I'd already applied grease to the mechanical bits and this allowed the range mechanism to switch.. albeit with some resistance. My 8640B was much the same and to reduce the resistance I'd modified a metal indent in the filter box but this time I didn't bother to do this.

One practical problem is fitting a range switch knob. Because of its years of use, ending in a (major?) problem the original range switch knob was now merely a brass bush which I'd removed. Looking in my junk box I found a suitable knob but it slipped on the shaft, so I had two options.. file a flat on the quarter inch shaft or find a decent knob. The diameter was key because the range marking scale needed to be visible or I'd have to make a paper copy of larger diameter. Fortunately a third rummage revealed a correct sized knob having large Allen screws so I glued the scale to this and put it in place as you can see below.



 The original range knob has a lever (much like that on the left) which gives a decent mechanical advantage to overcome switch stiffness.

 The display box range switch worked fine on the bench but once refitted into place it just jammed solid and the dial cord flew off one of the pulleys.

Already fed up with getting the tuning dial sorted there was no option but to virtually repeat the whole process to fix the range drum mechanism (without upsetting the working tuning dial!!)

On the bench everything worked smoothly.... the green plastic gearwheel rotated and flipped the drum to all its settings without a hitch but as soon as the box was fitted and the four fixing screws were tightened the thing jammed solid.

On the right is the left end underside of the display box with the two outer holes for very long securing screws and that centre hole.. for what?

 Well, the end of the range drum locates over a plastic end cap with a tiny pip to ensure correct positioning. The drum has a pulley around which is fed dial cord and the pulley connects to an ident switch whose position is fixed by a ball bearing held by a strong spring. The end cap is therefore under some considerable stress and to fix it in place there are several screws. Two of these screws locate the end cap securely in position whilst the two long screws holding the box to the chassis also have a measure of control over the end cap position. I found the problem resulting in jamming was the end cap was being skewed by the box fixing screws because that centre screw was loose. Using a light and magnifier I found the plastic into which the fixing screw threads had a piece missing. After puzzling over this I realised that someone had fitted an oversize screw and broken the plastic. The end cap wasn't held in place properly and the drum was jamming. Can I drill a couple of small holes either side and use new screws to do the job? In fact this wasn't necessary because I found the real problem (or had I?).




 I began to remove screws from the complete but totally seized display assembly. Starting with removing the ident spring just in case for some odd reason the ball and spring were seized.. but they weren't and detaching the spring had no effect. To give you some idea of the problem, attempting to rotate the end pulley or the eighth inch axle at the other end (above right) proved impossible. Even rotating the drum itself was impossible so could the axle bearing or the bearing at the rear of the pulley be seized?

I removed the front bezel and then the two countersunk screws holding the black metal extrusion to the end plates and suddenly the drum freed up completely. What's happening is akin to operating a drum brake in a car. Above right you can see a gap between the end plate and the extrusion. If the gap is present... no problem, but closing the gap locks the drum. Basically any screw that closes the gap locks the drum in place and this includes the pair of very long screws holding the display box to the chassis. A straight edge along the drum confirms that it's not distorted, neither is the black metal extrusion.


 I looked at the driven end of the drum to see if that had the same problem but it didn't. The end plate was snug against the extrusion and the drum didn't bind. Note that the spring between those two upper lugs has been temporarily detached to permit freer drum movement during testing.

You can see everything is not perfect here as the plastic pulley has cracked. Without removing the cord I can't superglue the crack so it's a problem for the future. Copper grease has been used on the detent mechanism.

The fix for the jamming needs to be thought out. First I'll check no parts are distorted, and if everything is OK I think I'll put down the problem to plastic shrinkage or maybe more likely, slight distortion in the drum. The fix might be to add shims to prevent the gap between the end plate and the extrusion from closing up.



 This picture shows the underside of the display box. When the drum is free to rotate the extrusion is proud of the underside of the end plate by 1.56mm or a sixteenth of an inch. That means I need to add a sixteenth of an inch shim under the end plate to prevent the gap closing and the drum to lock up.


Because of the way the gear wheels driving the range drum and the tuning pointer mesh with mating gears under the chassis, the effect of this shim needs to be tested otherwise an alternative solution needs to be found. However, as the underside of the extrusion will still be flat against the chassis I think the shim under the end plate should be fine.

 I started to test the hypothesis by first of all checking to see if there was a way to easily fit the new shim and found that by making it about 3 and a half inches long and 11mm wide (don't say anything) a shim would bolt into place at the extreme edge of the chassis. I removed various screws in the display box and refitted them to check no other problems existed and the fault was repeatable.

At about the same time I'd discovered that, with everything in place, the extrusion was no longer proud of the endplate, AND with the drum rotating freely, and wondering why this was so... something fell to the floor. I picked it up and saw it was a small piece of black plastic with part of a thread visible. Without pondering this I looked at the end plate and found no gap... I also noticed the end of the axle was dead centre in its hole and the drum turned perfectly freely... then the penny dropped. The piece of broken plastic had been jammed behind the drum and sure enough when I looked at the drum closely there were now slight score marks visible where the plastic bit had squeezed out when I was jiggling the drum to check for clearances. Below.. you can guess where the plastic part had been stuck. Behind, where the gap was widest, between the black metal extrusion and the back of the drum.




 After umpteen attempts the end plate suddenly fitted perfectly. After and before >>>



 Here's the tiny piece of plastic that gave all the trouble. It was probably responsible, not only for the drum jamming but also for the pointer sticking because the clearance between the drum and bezel at the RH end was lessened. No doubt the seller or the previous owner had broken the range knob, broken a tooth off the tuning pulley and these faults had ended the life of the 8640A.. at least temporarily as it was probably working electrically at the time.

The thread gave up when an oversize US screw had been forced into it.

 I need to refit the display box to the chassis.. check everything works properly mechanically, then plug it in and see what happens. This went fine with both tuning and range working smoothly.


 Initially the RF meter read close to zero but the three meter select buttons weren't latching. I poked at these a few times and the RF meter suddenly came to life with 0dBm indicated. This was maintained as the range switch was moved to different frequency bands.

I used my Tiny SA (spectrum analyser signal generator) to confirm everything was OK.. as above for 100MHz and below at 200MHz and 800MHz (this example includes the 500-1000MHz range).



 I now need to use a spot of switch cleaner to free up the meter switches and the modulation switches etc. then finish the cosmetic work, mainly filing the scabby aluminium to make it look half presentable. Then check everything works as per spec...

It's unfortunate that aluminium alloy corrodes over time when in damp conditions. Salt present in the water makes things very much worse and I remember way back in the late 1960s seeing this for the first time. Our factory in Liverpool made computers. This was donkeys years before tiny PCs, in the days a computer was pretty big. Ours were around seven feet high and built in stove enamelled cabinets with plenty of chrome bits and pieces. A typical computer had maybe five or more cabinets and in terms of real time operation were actually quite powerful. Our boss decided he'd sell some of our computers to Russia. I'm pretty sure he wouldn't have been allowed to actually follow through because of the cold war, but nevertheless he arranged for a computer to be shipped to a Moscow trade fair. Because computers were notoriously tricky to get working lots of test equipment went with it. I guess things went tolerably well but alas the stuff was shipped back to Liverpool on the deck of a boat and the Tektronix and HP oscilloscopes were ruined. We never did sell a computer to Russia but we did sell an air defence system to Burma using those very same XL9 computers.

 Initially I found the output RF indication on the meter was intermittent and sometimes zero so detached the pair of circuit boards adjacent to the meter and freed up the set of three push-button switches which were reluctant to lock into position, but replacing them still hadn't fixed the intermittency. I soon found this was associated with the attenuator rotary switch, most likely the rear-mounted parts. After rotating a few times and with the application of switch cleaner the intermitency was resolved.

The next task was to confirm the tuning dial corresponded to the RF output frequency. The dial read something like 101MHz at 100MHz output so I tuned the output to precisely 100MHz then loosened the two screws at the end of the dial box. After clearing the metal gears from meshing (cavity tuner and dial box) I turned the dial to 100MHz and refitted the two screws. The dial now reads the output frequency. The design doesn't include anti-backlash gears and, because the cavity tuner is deliberately made stiff (which I understand is to make it stable), tuning to zero beat on a test receiver is a bit messy, but drift is negligible and mechanical stability excellent. All the testing in this these tasks was completed using an SDR Play which tunes from a few KHz to several GHz.

I also tested the RF output level and the SDR dBm readings followed the settings on the attenuator. At higher output levels I needed to add a couple of in-line attenuators 20dB and 10dB to overcome receiver saturation. During tuning and modulation tests I used AM as this gave an easier way to check tuning. Switching to FM gave rather strange results so I need to look at the user manual and perhaps track down any faults. For example, when I set the FM dial to say 20KHz deviation the signal on the test receiver jumped from say 100MHz to 100.02MHz. In the 8640 FM is achieved in rather an odd way and I suspect there's a faulty component lurking.

A couple of annoying things showed up. One was modulation knobs binding and the other was the attenuator dial was offset and not all the marked settings were possible. The latter might be to do with linkages not aligned with the main attenuator switch or even the wrong dial fitted to the atteuator. Again a study of the user manual will help.

During testing the rear-mounted fan was very noisy so I squirted it with switch cleaner and after an hour it ran silently. I checked a random electrolytic capacitor to see if there's any clue to these failing but it read the exact capacitance with a very low ESR so hopefully none will need swapping.


 Reading the manual I noted the power supply rail voltages and their test points and adjusters so I initially removed the boards as you can see above. All the rails use two capacitors, one for smoothing and a second in the control circuit. I checked all of these and found every one was in a very poor state so removed them and fitted new.

Upper left supplies neg 5.2 volts, upper right plus 5.2V and plus 44.6V. Left supplies neg 20V and plus 20V.





 Replaced with

 33uF x 10V

 37uF x >40 ohms

 47uF x 16V

 33uF x 10V

 33uF x 20 ohms

 47uF x 16V

 33uF x 10V

 33uF x >40 ohms

 47uF x 16V

 33uF x 10V

 33uF x >40 ohms

 47uF x 16V

 33uF x 10V

 33uF x >40 ohms

 47uF x 16V

 33uF x 75V

 33uF x >40 ohms

 47uF x 63V

 220uF x 9V

 208uF x >40 ohms

 220uF x 25V

 300uF x 10V

 104uF x >40 ohms

 470uF x 50V

 300uF x 25V

 62uF x 21 ohms

 470uF x 50V

 300uF x 25V

 61uF x >40 ohms

 470uF x 50V

 As you can see all the old capacitors have very large ESR values although oddly some have maintained their marked capacity.

As an example the larger values should have an ESR of less than one ohm whilst the smaller something like up to 5 or 6 ohms.

 Power supply capacitors 


 During testing I'd noticed the fan was intermittent and soon it needed a little push to get it going but, after replacing the capacitors in the three power supply boards, nothing would persuade the thing to run. I imagined it was in need of lubrication because earlier on it would make a rough start. Getting it out of the chassis looks tricky but I worked out it could be done fairly easily. The horizontal mounting bracket is in two halves held together by two small screws. The bracket is also screwed to the rear metal casting and these two screws can be removed using a short pozidrive tool. Then the two screws holding the horizontal bracket are removed. Note that these are non-magnetic stainless steel so don't drop them into the chassis. once the top bracket is detached the fan body can be slid backwards so the fan blade is clear of the rear casting, and the electrical connector just lifts off the regulator board.

The blade is held in place by a grubscrew which can be slackened with a 5/64th Allen key (no other size should be used). The key may need jiggling if oxide is present. To remove the blade you might need to lever it off using pliers to grip the shaft for fear of damaging the bearing. Once off you can lubricate the shaft. The other end of the shaft is located by a small screw which can be removed for lubricating this end of the shaft.



 While the fan is detached you can clean and then align the blades by rotating it against a table edge to check each blade position.

Why HP use such a complicated fan circuit puzzles me but perhaps I can be enlightened?

After lubricating the fan, but before refitting it, I checked to see if it was OK. Unfortunately not one jot of improvement so either the thing is kaput or there's a drive circuit issue.

The user manual suggests testing the various transistors etc.


  The fan circuitry is obviously on the board into which it's plugged so I pulled this out and checked the two rows of TO39 and TO18 transistors.

I immediately found Q9 had a short circuit base-collector junction and Q12 was open circuit base-emitter plus base-collector.

Removing the Q9 proved it was faulty. The HP code marking is an 1853-050 made by Motorola (the manual has 1853-007 listed as 2N2351). I fitted a BC261B because this was my nearest equivalent.

I unsoldered Q12 and found only one leg remained. As you can see below two legs have rotted through. I jiggled the adjacent transistors and sure enough Q10 suddenly had an open circuit base-emitter, so I fitted a pair of 2N4036 transistors in their place. The HP transistor is an 1853-027 (a custom device made by Fairchild).



 Plugging back the board and re-installing the fan proved it was now working perfectly.

I have come across this type of transistor leg corrosion a few times before, but why the TO18 device shorted is puzzling.




 Now the original fan is refitted and working I've found it's unduly noisy. The complexity of the fan circuit (not to mention it's unreliability) makes me think that back in 1970 or 1971 a group of Hewlett Packard guys were looking for something interesting to do rather than come up with something simple and just as effective. Read this and see what you think.

Fan circuit is below...

 Anyway I'm considering just removing it and fitting an ex-PC cooling fan. Of course, as these are usually 12 volt I'll need to check available supply voltages and see how it performs perhaps using reduced voltage.

 Return to Reception