Marconi 2022E RF signal generator repair and refurb

[bd139]

Thought I should probably write this up in a dedicated thread rather than let it be lost in the noise of the TEA thread.

Bought a Marconi 2022E from scotttrade12 on eBay for £100 delivered. This was advertised as dead, which it indeed was. Well packaged, accurately described. Good job.



Has NO STD on some masking tape on the top. More to follow...

[bd139]

After some dissasembling and testing, the root cause was found. Someone nicked the OCXO / TCXO module out of it 

That's what NO STD means clearly.

According to the service manual this drives the synthesizer and CPU.

[bd139]

Some reverse engineering later, the missing OCXO pins were identified and a 10MHz TTL level signal was injected into the synthesizer board 



Temporary bench set up 



It boots and everything appears to work panel-wise. No idea about the RF situation yet.



Next steps are to build a replacement OCXO module. There is a pot on the rear of the unit which gives an adjustment range on the ADJ pin noted on the board and there are +5V and +12V rail supplies on the OCXO footprint. I will ID which of these is supposed to supply the oven power and go from there. I've measured the footprint out and ordered 5x CTI OCXO modules from Aliexpress. I will produce an adapter PCB for it in Kicad and order from JLCPCB.

Aliexpress link: https://www.aliexpress.com/item/1005003277492150.html

Stay tuned... in about a month and a half 

[Zenith]

STD, an unfortunate abbreviation which has another meaning.

I thought you were being unduly pessimistic about the possible state of this. It has a nice sharp display.

If this uses the standard Macaroni OCXO of that era, they are a dodgy number, especially at 30 years old. If it had come with one, it wouldn't have been as cheap and you would probably have had to change it anyway.

A right result so far. I shall follow the thread with interest.

[bd139]

STD may be appropriate for some hardware I’ve used (Telequipment I’m looking at you).

Better to have disappointment elevated than happiness destroyed

Designing PCB now. I have a Kicad footprint for the OCXO module after building another one for an HP unit a couple of years back.

Oh investigating the images suggests that the unit is 1989 vintage but the synthesiser board is 1994 vintage which is weird. Possibly factory replacement. The old synth board has a very naff crystal oscillator installed usually apparently. There aren’t even footprints for this on the board.

For the money I’m very happy. I prefer to buy broken stuff cheap and fix it.

[artag]

Isn't there an external std input on the back ? The INT EXT button switches to it. Should have made testing easy.

While you've got it in bits, consider replacing the front panel buttons. They seem to die early and require heavy pushing, or bounce excessively. Replacements are available on ebay. I don't know whether it's worth replacing the lot, or just any that are faulty. You'll be kicking yourself if you get the osc working, put it back together, and then find the buttons are dodgy when you're using it in anger.

I think you got a bargain - these can go for quite a lot of money, as do all RF generators. I've also had an item or two from that source. He seems willing to sell broken devices for properly low prices, which, as someone who likes to repair things both for the entertainment and the money-saving aspects, suits me very well.

I will replace the buttons on mine (a 2022, no C, D or E) sometime and would also like to add an HPIB interface.

You may find https://groups.io/g/Marconi-Test-Instruments useful if you need information.

[bd139]

I tried that originally. Unfortunately the CPU doesn't run if the internal oscillator is dead or missing so you can't select the external reference. And it doesn't actually use that for the computer side of things anyway, only the frequency synthesizer.

Front panel buttons seem reliable. I will replace them in the future once I've checked it out.

Thanks for the groups.io link - didn't know that existed. Already member of the HPAK and Tek ones 

Did some sleuthing and managed to construct a schematic for the original OCXO module that was fitted to the 2022 series. This had the same part number as their frequency counters and it appeared in the manual. Had to unfuck the Mauritron scan which I include here for reference:



This suggests that the OCXO heater was run from the 12V rail and the oscillator from the 5V rail. I will look at options here as the OCXO replacement I have is 5V supplied.

[artag]

Unfortunately you can't use the famous HP10811 as it wants a 20V oven supply. And is possibly too big anyway.

Gerry Sweeny describes some comparable OCXO options here : https://gerrysweeney.com/diy-hpagilent-53131a-010-high-stability-timebase-option

Apparently another point of failure are the small RF modules, OM345 and OM360. See https://www.vintage-radio.net/forum/printthread.php?t=96747
They can also be found on ebay - at present there seem to be some original OM360 and a pcb to substitute the OM345 with a MMIC. Mine haven't failed (yet).

[bd139]

Yes definitely too large. Done the OCXO rodeo a few times now. Think I've got through about 6-7 of them now including 2x HP ones, couple of Blileys and a box of cheap CTI ones.

But this is one different little box. The thing is phase locked to any external standard you plug in to avoid noise issues.

The CTI 10MHz ones are 5x for £10.56 delivered from China and will drop in here. However they are a little hungry at 600mA heat and 250mA run currents. If I supply at 12V I've got to blow off 4.2 watts on a 7805 during heating and 1.75W steady state. If I supply at 5V I've got to pull 600mA off the 5V line while it's heating and 250mA when active. The 12V OCXO consumes max 400mA.

Did the thermal calculations on sticking it on the 12V with a 7805 in there and it'd exceed junction temperature on the heat cycle with a heatsink that would fit in the space so that option is eliminated.

VCTCXO is possibly an option phase locked to an external standard with the CTI in it. Those are cheaper to run...

[xrunner]

Looks like you're on a good path forward. I found the thread on the one I had, dated back to 2015. I didn't remember what all I did to it, but I did find this -

I looked in the register that contains the total amount of hours the unit has been on and it's 108156 which is equal to 12.3 years.

The reason I wanted to know it was because the fan was making a lot of noise and I had to replace it. Be interested in how many hours your unit has (I don't remember how to check the register but it's in the manual).

[bd139]

I will check it when I bring it up next time 

[bd139]

After some research and discussion elsewhere this may be a Frankenstein box. The synthesiser board may be from another 2022 series box.

Thus another option is now appearing which is to build a replacement crystal oscillator module to restore the original 2022E oscillator and phase lock it to an external OCXO. This is a reasonable solution and allows me to build a GPSDO and distribution gubbins which goes nicely with the top floor balcony of my new apartment 

[bd139]

Reverse engineered power supply. 5V limits at 2.7 amps. Investigating idle power draw next. Convenient 0.1 ohm sampling resistor in circuit

Pushing it. 5V option discarded. 12V option discarded. Final option is build internal crystal osc and discipline it externally.

[AVGresponding]

Is there space for something like an LM323K on a small heatsink, instead of a 7805?

[bd139]

It's not so much the space but the fact that it has to go somewhere or the junction temperature will be too high resulting in kapoof of the regulator. Also you have to drop 7 volts at 600mA which is about 5 additional watts during heating and then 1.75 additional watts when warm which is all being dumped into a sensitive box full of RF synthesizer

After doing a KT of all the options, actually building a replacement crystal oscillator for it and phase locking that to an external GPSDO via the ref input seems to be the most sensible solution here with the least number of risks or surprises. I can do this for about £4 too.



Plus as mentioned it give me the excuse to futz around and build a GPSDO 

I can't find any other differences between the AA1/1 and AA1/3 board doing a full inspection of both other than some additional decoupling and the missing crystal oscillator.

[AVGresponding]

Looks simple enough I guess. Is the pcb layout available?

[bd139]

It's in the 2022E SM but I will just dead bug one on some FR4. I may do a proper PCB if I can be bothered with it down the line but I want to prove the idea works first and check the entire unit over properly before I throw too much money and time into it.

I need to order an N to BNC adapter anyway and Rapid seems to actually have all the bits I need for once. Will recycle some of the parts out of that HP create you pointed me at as well 

Ordered - Rapid:

10x 120pf NP0 ceramics
10x 22pF ceramics
2x ACT 10 MHz crystals (30pf loading to match the original)
1x RF solutions N to BNC adapter
100x 100nF X7R ceramics (for stock)
100x 1N4148 (for stock)

Ordered - Langrex:

3x BB809 varactors

Bang goes another 20 quid 

[AVGresponding]

Glad you have some 150MHz transistors in stock, and didn't throw them out with the recycling!   

[bd139]

Don't even go there 

Any generic NPN will do this job so usual suspects 2n3904 / 2222 / BC547 etc are fine. I will reuse the Motorola 2N2222A's extracted from the 3776A for this only because I quite like metal can transistors

[Fraser]

I own a three versions of the venerable Marconi 2022 RF sig Gen.

These Sig Gens are still sought after due to their compact size and decent performance. The most common fault on them
is failure of one of the Philips OM series gain block hybrid amplifiers. Those hybrid amplifiers were common and relatively inexpensive in the 1980's. They were cheap enough to be used in TV aerial booster amplifiers as a gain building block component. Sadly their relative rarity in the modern world has made them somewhat more expensive. I have a 2022 that has suffered failure of many front panel button switches and that is a project that I will get around to doing when time permits.

For information, the 2022E model was specified by the UK MOD who needed a tactical field use signal generator which did not have a long settling time before producing a stable output frequency. The OCXO was ditched and a TCXO employed to meet that fast set-up and use requirement. For the MOD application precise frequency accuracy was less important than the speed with which the unit became frequency stable. I personally like my 2022E as it is portable, fast to use and its accuracy is more than enough for most tasks that I have for it. If not, I connect a small Quartzlock 2A off-air frequency standard to it that uses the BBC Radio 4 198KHz carrier as its reference. 

From memory I own two 2022's, one 2022C and one 2022E. I was going to sell one or two but could not bring myself to do it as I have used these a lot in my career and I have a soft spot for them. I sold my bulky Marconi 2019 RF signal Generator instead.
I believe that I may also have the service manual for the 2022E model, but sadly I am many thousands of miles away from home for a while so cannot check.

Good luck with the restoration 

Fraser

[bd139]

Thanks for the info - very interesting. I'm actually writing up all the notes people are providing and will publish a web page about it. There seem to be lots of little bits of information but nothing in one place so far.

As for the original oscillator here I can see the design trade off that was made and it makes sense. I supposed the MoD were using it to work on typical FM stuff at the time so frequency precision isn't as important as say narrow bandwidth. And yes the OCXOs take a while to warm up. The previous set up I had was a 5334B with OCXO module and you had to leave that running for at least 30 minutes for it to be usable.

So I got ahead of myself this morning and designed a PCB for the basic oscillator module.

[Zenith]

This is interesting because I have a Marconi 2019 which I use a lot. When it came the frequency it produced was way below what it indicated and the adjustment was at its upper limit.

I found that someone on the WWW had opened the OCXO and traced the circuit. I recall there's was in a 2955 receiver test set. The circuit is similar to your OCXO, but with different transistors and varicap at least. I opened the OCXO and tweaked the preset cap quite a way, maybe 1/6th turn, and tried it. The adjustment multiturn preset could now bring it onto the correct frequency.

It's accurate to less than 1ppm and drifts slightly over time. I don't know what the spec is off hand, but that seems somewhat disappointing for an OCXO. It takes a fair time to settle down properly. Realistically, it's more than accurate enough for what I use it for after it's been on for five minutes. I suppose we all have Frequency Counter and GPSDOitis these days.

Since the 2019 works, I'm not inclined to start messing with it for no good reason. However it does occur that the OCXO is a dodgy number; it must have a lot of hours on it and it's likely been pulled well beyond its design limits. I wouldn't be surprised if it died or became unusable and I had to replace it. Finding one out of a scrapper wouldn't be easy and it would probably have the same problems, so making one up from new parts would make more sense.

[xrunner]

So I got ahead of myself this morning and designed a PCB for the basic oscillator module.

That looks nice. Come to think of it (I wish I still had mine but I sold it to get my Agilent 8648A) one of the problems I had at first was the ext. ref input - when I tried to use it nothing happened. It just wouldn't act like it had any stimulus from the ext. ref. As I said reflowing solder joints fixed that issue too. I followed the ext. ref. input and found where it stopped and I re-flowed solder connections in the area. That fixed it. I got off easier than you are doing. 

[bd139]

This is interesting because I have a Marconi 2019 which I use a lot. When it came the frequency it produced was way below what it indicated and the adjustment was at its upper limit.

I found that someone on the WWW had opened the OCXO and traced the circuit. I recall there's was in a 2955 receiver test set. The circuit is similar to your OCXO, but with different transistors and varicap at least. I opened the OCXO and tweaked the preset cap quite a way, maybe 1/6th turn, and tried it. The adjustment multiturn preset could now bring it onto the correct frequency.

It's accurate to less than 1ppm and drifts slightly over time. I don't know what the spec is off hand, but that seems somewhat disappointing for an OCXO. It takes a fair time to settle down properly. Realistically, it's more than accurate enough for what I use it for after it's been on for five minutes. I suppose we all have Frequency Counter and GPSDOitis these days.

Since the 2019 works, I'm not inclined to start messing with it for no good reason. However it does occur that the OCXO is a dodgy number; it must have a lot of hours on it and it's likely been pulled well beyond its design limits. I wouldn't be surprised if it died or became unusable and I had to replace it. Finding one out of a scrapper wouldn't be easy and it would probably have the same problems, so making one up from new parts would make more sense.

I had a 2019 for a while a few years back. Very nice bit of kit if a little large. Paid £5 for it from some ex EE who bought a sports car and was running off to the south of France with some floozie and just wanted rid of it! (slightly jealous).  Bargain of the century that was. I'm not going to get into the demise of that unit here because that's a whole thread on its own but I will say that the previous owner was a bodger and it lead to more than 5V landing on the 5V supply which basically machine gunned everything in the box connected to the rail 

All OCXOs have a lifespan unfortunately. The crystals and/or other components will drift out of specification over time. I had that with a Racal standard a few years back with a 9444/12 in a 9915. I built a replacement for it and bypassed the frequency multiplier circuitry. Details here including photos of insides of the Racal unit:

https://www.eevblog.com/forum/testgear/repairing-off-frequency-racal-9442-ocxo/

The ex cell tower OCXOs on aliexpress etc have plenty of adjustment room yet. Plus also 5 for £10 it's probably worth just buying some to leave around as ornaments 

Actually I did a PCB for the Marconi as well and one of these OCXO's. That might be usable for your 2019 if it's the same standard. Not quite as elegant as the VCXO board and will require some pin trimming and spacers on the build but should work.



It replaces the original standard which looks like this:

[bd139]

So I got ahead of myself this morning and designed a PCB for the basic oscillator module.

That looks nice. Come to think of it (I wish I still had mine but I sold it to get my Agilent 8648A) one of the problems I had at first was the ext. ref input - when I tried to use it nothing happened. It just wouldn't act like it had any stimulus from the ext. ref. As I said reflowing solder joints fixed that issue too. I followed the ext. ref. input and found where it stopped and I re-flowed solder connections in the area. That fixed it. I got off easier than you are doing. 

I'd rather have an 8648A but they don't turn up for £100 and some futzing here

I was expecting a fight with this one. It's not as bad as it could have been. At least it looks like it's recoverable