Marconi 2022E RF signal generator repair and refurb

[Zenith]

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 

Thanks. I'll bag some of those. I've got enough stuff that uses OCXOs and I think it's only a matter of time before the one in the 2019 cashes its chips in.

[bd139]

Non trivial update finally.

At lunch today I threw together the prototype crystal oscillator replacement module based on the original schematic. This was tested with 0-5V which gave a pull range of 9.999800 to 10.000100 MHz. The pull range is +/-12V



This was crudely tacked into the 2022E. No prizes for mounting here.



Booted straight up and the CPU runs fine. The phase lock is a bit dicky so might need some more debugging but we at least have a working frequency reference.

When you poke EXT it complains about error 11 (bad reference) and occasionally error 12 (cant lock). Which is a lie because it's clearly working as shown here:



The game is however not over. It does output a signal up to 250MHz but at 10dBm output it's about 15mV and on top of that it's well off frequency. That would possibly suggest that there's a problem with an OM345 somewhere in it as it's clearly not kicking out enough wumph to drive the output amplifier and/or synthesizer divider. Will wait until I've got the final oscillator board in for that.

Found a nice debugging feature though. 2ND+3 then enter register number and you can shift binary values. I used this to test the attenuator drivers and relays to make sure that they were operating independently which they were.

[bd139]

Going to need some OM345's for this so have designed a board. This uses the MiniCircuits MAR-7SM+. Unlike other boards there will be a different mounting method which will involve inserting a pre-trimmed 5-way header pin into the edge of the castellated board. The pins can then be soldered in while still on the header plastic which is then removed. This is then inserted into the board and soldered underneath. This avoids some of the mounting issues identified with other approaches. Disclaimer: might be worse . Also the module is the same size as the original one.

[xrunner]

When I bought mine I researched that issue before it got to me, but luckily I didn't have to replace any. 

Does the board manufacturer populate your board for you? I noticed some of them will do that.

[bd139]

I will order the PCBs from JLCPCB. They don’t sell the mini circuits part so will have to order those from mouser and hand solder. Will get the other two SM parts at the same time.

I need to do a trial fit of the pins first which I will do when they turn up …. 

Did some debugging earlier and confirmed that OM345 IC10 is indeed dead. You can ID the dead ones because the input pin one goes to 0V and the output to 12V as the feedback network is broken.

I’m actually going to order 30 boards as that’s the cheapest JLCPCB increment band and sell the rest on eBay 

[Fraser]

You may find this page helpful…

https://ea4eoz.blogspot.com/2017/09/repairing-om345-hybrid-module.html

Do remember that the OM345 is a 75 Ohm input and Output impedance and the MAR series MMIC’s are 50 Ohms.

Fraser

[bd139]

Yeah have been through that already. I did a calculation on 75 ohm loss and VSWR and it’s trivial at 50 ohms. It’s actually better than the original OM345 input VSWR quoted at 1:2.0

The gain, noise figure are slightly better. Not sure about OIP3. I’ve seen that quoted but no data to back it up. Doing feedback amp calculations against transistor circuit with same topology I couldn’t get better than 20dB which is about as good as the MAR-7SM+.

RF isn’t an exact science or magic

[djsb]

I have one of these as well. Are you going to share your PCB files or the link to where you're selling any PCBs? I'd also be interested in any webpage/blog with additional info you may create. Thanks.

[bd139]

I will write up and share everything once I'm sure it's all tested and working

Edit: just ordered enough parts to build 11x replacement OM345 modules. I'll replace my broken ones, keep some in reserve for to replace the rest of mine when they blow up and stuff the rest on ebay as kits. I can get 30x PCBs for the same as 5x so will over order and sell the remaining PCBs too. Will be sold at cost. Don't need the profit!

Edit 2: perhaps I should lose the Ferengi avatar 

[bd139]

Meant to post this. RF board photo:

[Fraser]

bd139,

Are you taking Gold Pressed Latinum in payment for your boards 

Fraser

[bd139]

As long as it’s not worthless gold 

[Fraser]

 

[bd139]

Redoing the OM345 PCB without castellated edges as i didn't know they were an additional cost option 

[factory]

Would lead frame connections be better than the pin header connectors?
Example: https://www.digikey.co.uk/en/products/detail/te-connectivity-amp-connectors/1544210-2/4731536

David

[bd139]

That’s a a great idea actually. Thanks for the suggestion. Those also won’t fall out when soldering!

[George Edmonds]

Hi

Just a quick word.

The ones used on the AA4 board in a 2955 are critical for overall dimensions due to the limited space available.

PLEASE use soft pins for the legs, the last ones I used could not be cut with small side cutters, I had to use piano wire cutters to cut them, not a good idea.

G Edmonds

[G0HZU]

Note that the reverse isolation of a typical MMIC won't be quite as good as the original OM345. This might not matter in some locations in the 2022 sig gen but it probably does matter for the IC1 location at least.

The OM345 reverse isolation is typically 22dB at about 100MHz and the MAR7 manages just 18dB here. The 22dB reverse isolation of the OM345 may already not be enough to prevent some FM modulation occurring on the AM for frequencies up to about 62.5MHz. IC1 acts as an isolation buffer between the AM modulator and the 160MHz VCO. I've checked several 2022 sig gens and they all suffer from a bit of unwanted FM when AM is selected below 62.5MHz. This includes the ones at work and these all have the original OM345 fitted. There's a good chance the MMIC replacement for IC1 will increase the amount of unwanted FM that occurs here.

I designed a high reverse isolation alternative for IC1 but never got around to trying it out. This had about 50dB reverse isolation at 160MHz.

[factory]

That’s a a great idea actually. Thanks for the suggestion. Those also won’t fall out when soldering!

They were used in stuff at work, no-one knew what they were and some boards we had made (supposed to be an exact copy) they fitted right angle pin headers, they were a pain to remove, after we found the boards didn't work (eventually I found the problem).

Many many years later, some of the leadframe connections turned up in the TEA thread and I asked to find out what they were. Seems there is lots of choice for, pitch, size & board thickness.

David

[bd139]

Hi

Just a quick word.

The ones used on the AA4 board in a 2955 are critical for overall dimensions due to the limited space available.

PLEASE use soft pins for the legs, the last ones I used could not be cut with small side cutters, I had to use piano wire cutters to cut them, not a good idea.

G Edmonds

Thanks for the heads up. I located a picture of AA4 in the 2955 and it looks like my replacement module would fit in there.

As for the soft pins, agreed. See my later reply from user factory which I haven't written yet as I post this

[bd139]

Note that the reverse isolation of a typical MMIC won't be quite as good as the original OM345. This might not matter in some locations in the 2022 sig gen but it probably does matter for the IC1 location at least.

The OM345 reverse isolation is typically 22dB at about 100MHz and the MAR7 manages just 18dB here. The 22dB reverse isolation of the OM345 may already not be enough to prevent some FM modulation occurring on the AM for frequencies up to about 62.5MHz. IC1 acts as an isolation buffer between the AM modulator and the 160MHz VCO. I've checked several 2022 sig gens and they all suffer from a bit of unwanted FM when AM is selected below 62.5MHz. This includes the ones at work and these all have the original OM345 fitted. There's a good chance the MMIC replacement for IC1 will increase the amount of unwanted FM that occurs here.

I designed a high reverse isolation alternative for IC1 but never got around to trying it out. This had about 50dB reverse isolation at 160MHz.

Interesting. I will work out how to measure that when it comes to it. The 2022E is not a particularly wonderful generator specification-wise so I'm not overly bothered about some side effects as long as I am aware of them. I'm not going to be using this as a modulation source however so the modulation generators will be off all the time. I'm intending using it for very basic CW RF measurements such as MDS, return loss and occasionally VHF/UHF filter measurement. My AWG already makes a fairly decent RF generator with some exceptions.

I saw your post on the designs on vintage radio forum as well.

I actually looked at the W7ZOI termination insensitive amplifier as a design basis of another variant as they have excellent isolation and matching at both ends. This is a similar feedback amplifier topology. I shoved a BFR182W model into LTspice and got reasonable flatness up to 900MHz substituting all the transistors out for those. Obviously no idea on the noise figure. Would have to measure a prototype. I may persist with that design as well. Of course I need an RF generator and detector that is actually usable past 100MHz first. So I need to repair this with whatever I can.

[bd139]

That’s a a great idea actually. Thanks for the suggestion. Those also won’t fall out when soldering!

They were used in stuff at work, no-one knew what they were and some boards we had made (supposed to be an exact copy) they fitted right angle pin headers, they were a pain to remove, after we found the boards didn't work (eventually I found the problem).

Many many years later, some of the leadframe connections turned up in the TEA thread and I asked to find out what they were. Seems there is lots of choice for, pitch, size & board thickness.

David

Interesting bit of history. I have seen them before but didn't put two and two together on this one.

I've actually ordered two sets of boards here to try them out (as the boards are so cheap!). I have ones with leadframe suitable pads on the edges and some that can just use 0.6mm wire (resistor legs) soldered in as that might be easier to fit for people. Will see what comes out of it!

Starting to think I have a PCB problem at the moment 

[George Edmonds]

Hi

If it helps I have attached two photographs of a 2955 AA4 board showing my concerns.  The last OM345 replacements on eBay require C37 to be moved to one side, not a good idea.

G Edmonds

[bd139]

Excellent. Thanks for the photos. My design is considerably smaller than those ones from what I can see. They should fit in the same courtyard as the original OM345.  The height is also close to the original.

I redesigned the board slightly before they shipped from the above. The two sets of boards were ordered with 0.6mm thick substrate and the transmission lines approximated for that board thickness. The thin traces another poster mentioned earlier were a mistake on my part. I rechecked again and again before ordering

They’re still being manufactured so we’ll see in a few days.

[bd139]

Mouser parts order for the final crystal oscillator, the OM345 replacements have arrived as have the CTI OCXO modules.

All the CTI OCXOs are checked out and have been characterised for lowest current. I'll load up the 2022E power supply with an equivalent load and take some measurements when I get some time.

Still waiting for PCBs which is the blocking issue at the moment.