Marconi 2022C - Stopped working below 62.5 MHz

[G0HZU_JMR]

I had another look at the dead OM345 today and soon realised that the circuit of the real OM345 is not quite the same as the schematic in the datasheet.

There is a change to the circuit (extra resistor) and when I add this resistor to the EM simulation it cures the K factor issue up at 1500-2000MHz

This is way out of band but you have to do a full stability analysis across a much wider bandwidth than just 40-860MHz.

I really need access to more dead ones to find out if there is a common failure mode because I think it may be possible to repair the OM345 module rather than try and replace it.

[af024]

Hello Jeremy,

Exciting work you've got going on there. I'm really looking to the next installment as I too was wondering why they fail.

Regards,

Andy

[Jac]

Very interesting Jeremy!

I would be very interested in the "final" replacement circuitry.
My 2022D is still working fine, but that may not be the case indefinitely...
A DIY replacement would indeed be very handy.

Thanks all for all the good work!

Jac

[G0HZU_JMR]

Thanks!

The sample that Alan (Biggles) sent me is a bit worse for wear because the solder pins were missing but I think I've managed to repair it with a new part in the schematic.

If they all fail this way then I think there is an extremely simple way to fix the module with minimal scraping of the tan coating to fit a new part. This failure mode explains why they fail in seemingly random locations in the 2022 and the 2955.

I've not tested this one at RF yet but it now biases up OK and takes the correct current now.

Don't throw away your dead OM345 just yet...

[G0HZU_JMR]

I managed to get onto the Marconi BT/Yahoo group and ask about the OM345.
There are some fuzzy images of the insides of these devices on there plus details of the resistances etc. The info agrees with what I measured on this one including the extra resistor.

However, the images on there are quite poor quality and don't show the film resistors so I intend to take some decent macro images of this one because it's possible to see much finer detail. eg I can show down to the laser cutting slots inside the film resistors. it really is quite a pretty device inside

I also think the true schematic of this device is a bit more complicated than first appears. Apart from the obvious printed inductors I've also found two extra resistors and also some printed capacitance in a couple of places. All these extra things improve the performance.

I've asked about getting more dead OM345 devices on there but they seem happy with the limited info they already have about this device which is a shame. My goal is now to do a decent failure analysis of this device because I think the device may have an inherent weakness that causes a common failure mode.

I can also explain how the device is designed and why the layout is the way it is. I really don't think that fitting MMIC amps in place of this device will be the right thing to do in all locations in the Mi2022. I'm also not convinced it is worth making clones of this module because it may be fairly trivial to repair it and make it far more reliable than it was when new.

I just need a few more samples of failed OM345s to analyse

[dave cox]

Its not unusual to omit some of the 'magic sauce' when publishing schematics - they don't want to make it too easy to clone ! I do wonder if the encapsulation was necessary for performance or just to protect the IP (intellectual property) ? I guess your simulation is suggesting that with surface mount and modern parts it would no longer be necessary to get the performance ...

Sadly, (or luckily, depend how you look at it) I don't have a broken one to send you Jeremy but I do hope you make progress before the inevitable failure happens to my 2022 !

dc

[G0HZU_JMR]

I can confirm that this one from Biggles now appears to be fully working after the repair. Does anyone know what part usually fails? It wasn't the transistor that failed on this one which makes me think there could be a cheap and easy fix for these modules.

I'll do a few performance tests on this one because someone did ask about P1dB and IP3 etc. I'll be testing it with 50R test gear but I think it's still worth doing. I could fit 50/75R matching pads but I think a simple 50R test suite will be just as useful.

I'll compare the simulation against the real results although I'll be testing the OM345 with no encapsulation as it is all scraped off

[G0HZU_JMR]

I did a couple of initial RF tests at 160MHz and got P1dB = +4.3dBm and the OIP3 was +17dBm.

I also looked at it with the thermal imager to see if it could pick out any info. The results were very good. Here's an image of the half of the OM345 that was healthy. You can see the transistor glowing a bit with heat from its dc operating conditions.

What I haven't shown is the other side of the OM345 because this is now repaired. I would really like to test a healthy OM345 on the thermal camera because I think it will show some heating in the tiny area that failed on this one.

You can see the wiggly inductor traces in the thermal image as well as the SOT23 transistor. The inductors look 'cooler' because they are acting like a mirror and are telling the camera a false temperature due to the reflection.

[Dickie]

I would really like to test a healthy OM345 on the thermal camera because I think it will show some heating in the tiny area that failed on this one.

I have a NOS OM345 date code 8313 that I have been keeping "just in case" for years you could borrow if you promise not to blow it up .

[G0HZU_JMR]

Hi Dickie thanks for the offer but it's probably not a good idea because I would want to scrape away the coating in a certain area.

Also, the most recent ebay sale of one of these I could find sold for £30

The measured results I got for P1dB and IP3 were within 1dB of the simulation. In Eagleware Genesys I also have a non linear model of the MSA 0735 MMIC and this device is is often used as a replacement. I found I had to bias it at nearly 16mA (OM345 is 11.5mA) to get a similar gain and the P1dB, the harmonics and IP3 were several dB worse than the OM345. eg the OIP3 on the simulator was only +11dBm at 16mA bias for the MSA 0735. To get the MSA 0735 to better the OM345 for OIP3 on the simulator I had to bias it at 23mA and this gave just under +19dBm OIP3 which kind of agrees with the MSA0735 datasheet.

In some locations in the 2022 this probably doesn't matter but the OM345 shows much lower distortion than the 735 MMIC even with the MSA 0735 biased 5mA harder.

[Sean Williams]

I replaced all the OM345 in my 2022 - Due to the fragile nature of the PCB I had to resort to the generally accepted practice of destroying the original module, in order to make removal simple.

I may have a couple of new modules here still, if so I would be happy to sacrifice one in the interests of science!

Will have a rummage when I am next in the shack

Sean

[G0HZU_JMR]

Thanks Sean

I've had some more feedback on the Marconi groups page and someone called Steve on there has taken several of these modules apart and also looked at a few broken ones.

He has found the same failure mode as me although I'm not sure whereabouts on the PCB he found the failures. But it was not the transistor that was failing. On my board it was one of the printed/etched sections that had failed open circuit. So it may be the case that the board has several weak areas.

My initial thought was that my unit had failed because there is a tiny area of heat generated where the PCB has a thin (resistive) trace and 'maybe' the repeated thermal expansion and contraction over the years causes the outer encapsulation coating to pull the printed part away from the board or simply cause a tiny fracture.

Or it could be that this area on the board simply fails without the coating causing the fracture. i.e. it is caused by thermal stress in the board itself. I was hoping that all boards fail in this one area because it would be very easy to repair the module.

What would be great would be if Steve could send me a stripped 'good' unit so I could look at it on the thermal camera and then return it to him. I'll try asking if he still has the modules...

[G0HZU_JMR]

If the board does fail in the way I think it may be failing then it could be the case that the sig gen will have a brief phase where the fracture will be intermittent and cause the odd dropout in signal on some ranges. But I think this would only be very brief phase before it fails for good. To me it looks like a thermal stress 'timebomb' type failure and maybe it depends on how many times the sig gen is power cycled or maybe how hot/cold it gets internally in some scenarios.

[G0HZU_JMR]

I've had some feedback from someone who has just taken a few of these OM345s apart and two out of three showed a failure of R3.

This is the same failure mode as mine so here is a quick image showing the failed 3K9 film resistor. It's the little zig zag printed shape in the PCB that the two yellow arrows point to and you can see where it has a missing section. It's supposed to connect across the big SMD capacitor alongside it.

Isn't R3 tiny and skinny!

I'd like to see one of these on a thermal camera to see how hot R3 runs but I think the main cause of failure is thermal stress from lots of hot and cold cycles over time. Possibly the resin coating pulls this little film resistor off the PCB after many temperature cycles and the resistor eventually fractures.

If this is a common failure mode there are two ways to fix it. Either scrape away the resin coating (only) above C1 to expose the metal contacts of C1 and then solder a 0603 3K9 resistor across C1.

Or, the roadside 'in situ' repair would be to fit an external 4K7 leaded resistor across pins 1 and 5 of the OM345 module.

This temporary method with the external 4K7 resistor isn't a highly recommended method to repair the module but it would work in most locations in the Mi2022 if R3 has gone open circuit and you need to use the sig gen urgently. You could leave the OM345 in place and just tack the 4K7 resistor across pins 1 and 5 (i.e. the two outer pins of the OM345)

This second method produces a slightly different circuit because the bias path for Tr1 is now outside the module. But the RF negative feedback will still be there via C1 and R2 (approx 1nF and 470R in series)

The original bias is via R3 + R2 in series = 3900 + 470 which is just under 4K7. With the external 4K7 the dc bias will be slightly softer and there will be slightly more negative feedback at RF but nothing significant so this would be a possible method to do a temporary repair if R3 has failed open circuit.

[Wendymott]

Too late. I binned my OM345 as the legs had come off on removal. If anyone out in radio land has a "duff" OM345 I will copy the repair as Jeremy describes, and hopefully get my 2022 back into service. "not holding my breath though".
Regards
Wendy G8BZY

[G1VRH-Stephen]

Hello everyone,
I've also been investigating failed OM345 modules and have had a couple of emails with Jeremy regarding them.
I'm hoping that some unused modules get sent Jeremy's way as I'm interested in seeing the thermal images of the area that we both found issues with.
On with the reason why I joined the forum to make the post.

A friend brought his broken 2022C over for diagnosis. IC8, IC9 and IC10 had all failed at the same time.
We looked at the prices that some people are asking for the OM345s and he asked if I'd mind taking a look at the modules with a view to repairing or replacing them. Initially i was going to replace with one of the numerous MMIC parts available but curiosity got the better of me.

I searched and read all i could on the om345 and that's how i came across this thread and wound up emailing Jeremy to see if he'd found the same problem area.

Of the three broken modules I had to check R3 was a common fault.

On one unit R3 was open circuit. Simply soldering a 3k9 resistor across C1 fixed it and it biased up at 10.5mA.

Second unit R3 was also open circuit and I soldered a 3k9 across C1 as before and powered the module up on a breadboard to test. This worked for a couple of seconds then failed again. Further tests showed that TR1 had failed. I fitted a BFR93a and powered up again to test. Module worked perfectly and biased up at 11mA.

Third unit R3 was measured as 600r and TR1 had failed. R3 looked as if the carbon laid down for it had smudged across the board. I fitted a BFR93a and cleaned off the R3 carbon deposits, effectively removing R3. Fitted a 3k9 across C1. Powered the module up on the breadboard and it worked biasing up at 11mA.

OM345 modules appear to be easily repairable and cost less than 50p to do so.
Hope this is useful.

Stephen G1VRH

[Colin_G4EML]

Hi Jeremy and Stephen,

I would just like to thank you for the investigative work you have done on the OM345 modules. A friend of mine has just brought his 2955 round to me with self test failures in test 1 (generator/counter loopback). He was amazed that I managed to fix it in about 20 minutes thanks to your work.

The second OM345 on the Frequency counter board had failed and I decided to try Jeremy's idea of adding an external 4K7 bias resistor between pins 1 and 5. This was easy to do on the component side so I didn't even have to remove the PCB. I can confirm that it worked a treat. The modules gain appears to be fully restored and all self tests now pass. We tested the 2955 over the full frequency range and its performance seems to be back to normal.

Had the module been in a more critical part of the circuit I might have tried the scrape and add a resistor method, but I am not sure this is really necessary and it adds the possibility of further damaging the module during removal or repair. As replacement modules are now so rare and expensive that's something that needs to be avoided. Certainly in the frequency counter the absolute performance is not in any way critical so the external resistor is the 'least risk' thing to do.

I am now a little less concerned about my own 2955A. If it fails in the future hopefully it will be as easy to repair.


Thanks.... Colin G4EML

[AC/HL]

One post moved here: https://www.vintage-radio.net/forum/...d.php?t=102918

[G0HZU_JMR]

Hi Colin
Glad you got it working OK

I guess you are the first person to try the 4K7 method! I do think it's worth finding an optimal type/package for the 4K7 resistor so I think I'll buy a few flavours of 4K7 from Farnell on my next order. I can test them as a two port device on a VNA up to 3GHz to see which one behaves best up to 3GHz.

Otherwise, some people may fit unsuitable types of 4K7 resistor and this may cause instability if the impedance of the resistor falls too low in the GHz region.

It would still be nice to get hold of some dead OM345s to investigate further. It may be possible for me send them off to be X rayed as (at my place of work) we often send failed semiconductors or PCBs for X ray. I don't know if the X ray will be able to see a fractured R3 though as it is so tiny and skinny.

I could also test an OM345 + 4K7 against a new OM345 on a test fixture and take S parameter data up to a few GHz to show how the 4K7 resistor affects the performance up at 500MHz-1GHz. I suspect the performance will tail off a bit up there but it may still be OK.

IMPORTANT:
One other thing to note:
As part of my failure analysis I looked at (external) ways to cause premature failure of R3 and one of these is to cause a temporary short on pin 1 of the OM345. This could easily happen when probing with a DMM or scope probe tip etc.

What will happen is that the transistor inside the OM345 will be biased OFF and its collector voltage will rise up towards 12V. This then means R3 will be dissipating about 25mW of heat instead of 5mW typical. So my guess is that R3 could get a bit toasty if this happened and it probably only takes a split second to give it this extra thermal stress. So it may be possible to kill an OM345 through clumsy probing around pin 1.
This is one reason I wanted donor OM345s to look at R3 on a thermal camera.

[Colin_G4EML]

Hi Jeremy,

I used a 0.125W MF12 resistor. I think Farnell stocks these. It would be interesting to see a comparison between a good module and a faulty one with the external resistor, even if its only out of curiosity.

..Colin