Marconi 2022C - Stopped working below 62.5 MHz

[Stefan_O]

Hello,
I´ve got a Marconi 2022C and from one day to the other it stopped working for frequencies below 62.5 MHz (no output, even when I set output level to 2 Volts, it just raises the noise floor), above everything works fine, including AM modulation. It does not show any kind of error message.
I´ve got a service manual, but unfortunately it´s a bad copy and the circuit diagram is barely readable.
My first guess was a contact problem at one of the relays, but here I read that a common problem is the OM345. Could that be the case with the problem I described? Where is the best point to start investigating and does anybody know where I can get a decent copy of the service manual?
Thank you very much,
Stefan

[short wave]

Welcome to the forum Stefan,
this link may be of use for your "Marconi 2022C" it covers C and E variants

http://www.ko4bb.com/manuals/index.p...i/Marconi_2022

[Sean Williams]

Will almost certainly be one or more OM345 modules that have failed - its a simple fix, modules are still available, but are getting scarce. There was a UK based Amateur that built some MMIC replacements, but I am not too sure if he is still producing them.

[Dickie]

This thread never reached a conclusion:
https://www.vintage-radio.net/forum/...d.php?p=620195

[Stefan_O]

Thank you very much. It seems that IC2 is broken. While at pin 1 I´m able a measure a decent input level of about 100 mV, output is only about 20 mV (measured with an uncalibrated TF 2603, so this values are only relative).
All others seem to be fine. Is that plausible with the problem I described?

I searched for NOS parts, all I found is the offer for 290 Rand in South Africa (not original part) and on Ebay there is a NE68519 and a MMIC CGB-1089Z listed as a OM345 "replacement project" (seller is from Cornwall). Any idea what would fit here? For IC2 the input level is already quite high compared to the other OM345.

[Radio Wrangler]

There's no unobtainium in the OM345. I seem to remember it's a single transistor with series-shunt feedback. So a few SMT resistors, a few capacitors and one modern bipolar device ought to do the business. It ought to be possible to get something a lot more similar than the available MMICs.

David.

[Stefan_O]

There is a circuit diagram in the datasheet, it's one transistor, 5 resistors and 2 capacitors. But I'm not an HF circuit designer. I've no idea what transistor to use or how to calculate the values for the parts around it.

[G0HZU_JMR]

Originally Posted by Stefan_O

Thank you very much. It seems that IC2 is broken. While at pin 1 I´m able a measure a decent input level of about 100 mV, output is only about 20 mV (measured with an uncalibrated TF 2603, so this values are only relative).
All others seem to be fine. Is that plausible with the problem I described?

IC2 looks like it acts as a buffer amp ahead of a Plessey prescaler in the 160MHz PLL section.
So if you don't see an accurate (locked) 160MHz at IC2 then I guess the level is too low to drive the prescaler. Obviously if this happens, the 160MHz PLL will go out of lock.

The 160MHz signal is used as part of the downmix to get the LF to 62.5MHz range in the generator so I guess this does stack up with your fault symptoms.

The usual drive level for the prescaler would be about 500mVpkpk although they are usually much more sensitive than this. If you are desperate you could try replacing the amp with a 1nF series capacitor because it might just lock OK even with zero gain and get the generator 'working' whilst you look for a replacement for IC2.
You don't really need the full performance of the OM345 here because it only has to buffer a 160MHz signal. So even a basic home made clone of the OM345 circuit (with a low spec transistor) would work here.

[Stefan_O]

I did some further measurement and I found out that there is actually a signal, but at a different frequency, always about 10 MHz higher than the selected frequency (I don´t know why I haven´t seen it the first time). Level seems to be correct, the frequency is slowly drifting away, it´s most likely not derived from the OCXO.
When I add a 1nF (without removing the OM345), the frequency changes a few kHz, but nowhere near it should be.
As a replacement transistor, would a BFS17 work (NPN 1 GHz wideband transistor)? It´s available for just a few cents at my local source for electronic parts. How do I choose the other parts?

[G0HZU_JMR]

Earlier this year I used Agilent Genesys to try to reverse engineer the OM345 and OM360 circuits wrt finding out the values of the biasing resistors. this was in response to the earlier thread on this subject.

Of course the best way to do this would be to simply measure the resistances of a real module. But I don't have one to take apart.

Once I had some reasonable values for the resistors I also did an EM analysis of a typical PCB layout of the schematic using Sonnet 3D EM and getting a decent flat performance above 700MHz isn't easy to achieve with a typical SOT23 transistor due to the various parasitics.

But for most of the OM345 locations in the 2022 this doesn't matter because they only operate at lower frequencies. It's the ones used on the upper ranges that will need to deliver decent performance across 500-1000MHz.

[G0HZU_JMR]

I did a basic simulation for a 160MHz circuit with a BFS17 and this came up with these values.

Note that this circuit will only work well up to a couple of hundred MHz because it uses a BFS17 and also because I optimised it to suit this device.

The biasing is reasonably stable over temperature but it will bias slightly harder at higher temperatures. The Pdiss for the BFS17 will be about 120mW so as long as the ambient keeps below 70 or 80degC it should be fine.

I did this very quickly so it might be worth someone else verifying it is OK.

You would ideally need to do it all in SMD but I guess if you are using a BFS17 you would do this anyway. It will be a tight fit to keep it a similar size to the original.

This simulation doesn't include any EM analysis because there is no PCB layout. I can add an EM analysis but it would only be relevant to my layout. However, I don't think it matters much at 160MHz. However, the layout becomes a significant issue up at 500-1000MHz.

Note: It's best to click on the link below to see a detailed image of the simulation as the thumbnail linked image is very blurry due to size restrictions etc.

http://media2.turbosport.co.uk/2013/...7OM345_160.jpg

[Radio Wrangler]

BFR540 abut 43p at Farnell in singles 500mW and 9GHz

You might want to add a bit of base spoiling resistance just to have a component to cal it if it feels a bit too frisky, but this will still have hfe around 9 at 1000MHz given the age of the hybrid, I'd think something like a BFR93A die might be in the original. But we don't want something like the original because they keep dying.

David

[Stefan_O]

I replaced the IC2 OM345 with 1nF and it works!!! Thank you very much!

Thank you very much for the calculation! The BFS17 was just an example. I could also get a BFR540 (0,55€) or a BFR93A (0,19€), all in stock at my local electronic distributor.

[G0HZU_JMR]

I had a go at manually optimising the circuit with regard to reducing the Pdiss of the device and this circuit seems slightly better than the first.

http://media2.turbosport.co.uk/2013/...OM345_160b.gif


I've changed the collector resistor from 150R to 470R and the 3300R resistor to 1800R and the emitter resistor from 4.7 to 5.6R.

It approx halves the Pdiss of the device and it also improves the bias stability wrt temperature. It also gives lower IMD in the simulator with a two tone test.

So it's probably a better bet than my first circuit

[G0HZU_JMR]

Originally Posted by Stefan_O

I replaced the IC2 OM345 with 1nF and it works!!! Thank you very much!

That's good However, I do think you have to fit an amp here at some point because the lack of buffering might cause subtle issues with spurious responses. That buffer provides a very useful degree of reverse isolation for one thing.

To show you the advantage of the better transistor (eg the BFR540 suggested by David) I ran the simulation in the later version of Genesys because it has this BFR540 model in the library.

You can see that the performance at 160MHz is about the same but you can see how much better the BFR540 is at maintaining a good input match and it should perform better at higher frequencies.

http://media2.turbosport.co.uk/2013/...OM345_160c.gif

[Stefan_O]

Thank you so much!
Three questions:
1. Are the 100nF at input/output just for the simulation or are they necessary in the replacement?
2. The datasheet of the OM345 says 75 ohm, the screenshot says 50 ohm, does that matter?
3. Would the replacement with BFR540 work up to 1 GHz, in case some other OM345 die in future?

Another question: The capacitors, do they need to be C0G/NP0 or would X7R also work? Or something completely different?

[G0HZU_JMR]

The 100nF caps aren't needed in the sig gen because there are already some blocking caps in the circuit.

In the case of IC2 I'm not sure it matters about the 75R vs 50R. It looks like it is fed into the prescaler in parallel with a 51R resistor. I think the input impedance of the prescaler is quite high so I optimised the buffer circuit for a 50R load.

In the Mi2022C you can see IC2 has a 100R resistor in series at the input and I suspect that this is a crude way to prevent IC2 from being overdriven and the series 100R at the input may also provide extra reverse isolation and it also kind of prevents IC1 from being double terminated.

The circuit I gave above really is just for IC2. It won't work very well at higher frequencies because I'm sure the performance will degrade due to the PCB layout.

When I looked at this last time I had to add inductance in a couple of places to get good performance at 500-1000MHz and I did wonder if the OM345 has some little inductors inside somewhere to help perk up the performance at the top end.

So the simple circuit I gave really is only suitable for IC2 down at 160MHz

I'll see if I can find my original simulations because I also included the three tone tests for the DIN IMD spec in the datasheet. I couldn't quite meet this spec the last time I tried but it probably doesn't matter for most of the locations in the Mi2022 because all but one appear to only have a cw signal. One of them is inside the ALC loop on the high range and this one will see AM modulation.

[Stefan_O]

Thank you so much for all that work!
I quickly designed a pcb, would that be ok (the contacts I would solder on the top, no holes in the pcb)? All C and R are size 0603 (smaller is not possible for me on a homemade pcb). Is a 0,5mm dual-layer pcb with using the other side as ground a good idea?

[G0HZU_JMR]

The PCB layout looks good enough to me for operation at 160MHz

I'd be tempted to use the BFS17 for operation at just 160MHz. I think it will be OK to put ground at the back of the board and 0.5mm board thickness should be fine although with a board this thin you will obviously need to make sure you don't flex it and stress the SMD parts. Once this happens they become intermittent time bombs for later failures.

I would add an extra ground via or two to connect the bottom ground plane to the top ground node but it probably doesn't need this with a BFS17 at 160MHz. I had a look at a few OM345 images and I think the layout is slightly different to yours but it's hard to tell because the components are covered in a coating.

Have you tried to take your broken OM345 apart? Looking at the images online I think the resistors will be printed onto the PCB substrate and the caps and the transistor will be the only conventional SMD parts used.

[G0HZU_JMR]

Quick update:

Alan (Biggles) kindly sent me a dead OM345 to dissect. I've just started scraping away at it and I'm actually quite excited

It 'does' have little printed inductors on the PCB and they look to be in the same place I put them on my EM simulation.

I was so pleased when I saw the little meandering traces on the PCB because that is exactly what I had to do on the Sonnet EM simulation to perk up the gain at 500-1000MHz.

The only downside (on my simulation) was that adding these did compromise the K factor meaning the simulation indicated the layout wasn't unconditionally stable up around 1500-2000MHz.

I also ordered some contenders for the active device from Farnell so over the next week or so I'll see if I can replicate the original circuit.

It's really difficult to chip away the coating on the OM345 and I may end up trashing this one. I'm worried I'm going to destroy the delicate film resistors.