Quick repair project - Marconi 2024

[grumpydoc]

A while back there was some (slight) interest expressed in posting repair projects. The Rapco GPS conditioned oscillator repair which I described a few weeks ago seemed useful to a couple of people so I though I'd post details of another project.

In this case the item is arguably more interesting than the repair - a Marconi 2024 9kHz - 2.4GHz signal generator. This is basically the same signal generator as Dave's teardown in EEVBlog #261 but it has twice the frequency range. I picked it up on fleabay in need of a little TLC - the seller said it had "a weak display". This was his picture.



It looked like poor connections on the zebra strip in the LCD, an easy fix (I thought!) so I bought it.

Ok, got it home and powered it up. The original description was certainly correct - the display was not happy. However it worked well enough for me to do a quick check of the generator and all seemed fine apart from the LCD which was good news. The generator was fitted with options 04 (OCXO) and 11 (pulse modulation and high +25dBm output).

After a while the display seemed to improve but it was still a bit difficult to read.

From here down click on the picture for a large version.


OK - time to crack it open



The first problem is that someone has been here first and not put the unit back together correctly - if you watch Dave's video there should be an aluminium plate screening the PSU. Ho Hum, at least they left the screws for me. Time to have a look for any more damage.

Hmmm, broken mains switch....



... and someone's "had a go" at the back of the N connector.....



... and there are only three screws holding the display in....



...but otherwise it doesn't look too bad.

The main board is almost exactly as in Dave's teardown except for the OCXO in place of the TCXO in his 2023. Option 11 swaps out the attenuator unit for one with a higher power RF amplifier but that's hidden in the RF section. I'm not planning on opening that unless there's a clear need.



A closer look at the OCXO, it's an NDK unit but I can't find a datasheet anywhere for it.



Looking at the date stamps on the various ICs and the stored calibration date it would appear to have been built in Feb 1998.

Moving on, to the LCD and I wasn't quite as happy as I had been. The problem seems to be the row drive connections as whole rows are not displayed or are not displaying correctly but, although there's a zebra strip (the grey block in the bottom right half of the display) it's only for the column connections. There's no zebra block for the row connectors but these use a flexible PCB strip.



Here's a close-up of the connector. It's bonded to the PCB and to the LCD itself but it's clear that whatever I do is not going to improve the connections. I did try a bit of pressure against the PCB end up and down the actual contacts but it didn't make any difference.



At this point I was a bit hacked off with myself for buying the sig gen - I've looked for replacement LCDs without much luck for previous items, so I was a bit doubtful whether I could locate one for the Marconi. All too often equipment has a custom panel, however this looked generic - an Optrex DMF 5010 so I ploughed on into Google.

On the positive side I found a data sheet fairly easily. On the less good side there are few hits from suppliers - not even hits from the "we'll quote you anything" brokers. The only place I found with stock had a MOQ of 60 (which appeared to be the entire stock) - I might have considered it at the right price but £25 each before shipping and VAT wasn't the right price Even ebay couldn't help

More Googling and things brightened up a bit. The Optrex board has a T6963 controller so I figured if I could find a panel with the same resolution/display area and controller it might work - in the end I found one quite close to home at Rapid Online and not all that expensive.

It's not an exact match though, it has an LED backlight, not CCFL and it's smaller at 180mm rather than 200 and about 8mm less deep, but it has a compatible controller (the RA6963) and the same pin out so I though it was worth buying one.

The first problem was to get it connected - the Winstar panel has a 2x10 pin connector, the Optrex one a 1x20. I didn't want to alter the original ribbon cable - generally if non-original parts need to be fitted I try to make things as "reversible" as possible so the plan was to use a piece of single row PCB header strip with ribbon cable soldered to make a plug to fit the cable IDC socket.



This didn't quite work as well as I'd hoped - the strip isn't really designed to have a cable soldered to it and even keeping soldering time to a minimum and the iron temperature as low as practical the pins almost immediately became loose. You can see in the photo that they're uneven. Fortunately it worked well enough without having to come up with another solution.

Next was the LED backlight supply. This turns out to be very easy as the CCFL inverter board is powered from 5V. The Winstar data sheet says that the LED has a V_F of 3.5V and a forward current of 80mA, so with a 1.5V drop a 18ohm resistor should do nicely.

Connected it up and....




Bingo It works!!

The next thing was to fix it in securely. I found some tiny aluminium plates, about 10x15mm in the junk box - absolutely no idea whether they had a purpose in life or were offcuts from something else but they were perfect to make small brackets to carry the new display. Two holes drilled 8.5mm apart to make up the horizontal difference and a 6mm threaded spacer to offset the panel a bit - here's what the result looks like.



and here they are attached to the display ready to go into the front panel



Finally here is the display mounted in the front panel and a shot of the new display in place.



The CCFL inverter board also carries the supply for the fan, decoupled with two 470uF electrolytics and two 1mH inductors so I still need to make up a small board to reproduce that and carry the current limiting resistor for the backlight LED. I also need to fix the mains switch but have some in the spares box so that's not a problem. All in all I'm very happy with the result and it's another piece of rescued test equipment.

It's slightly disappointing that it's already impossible to find components for an item that's not even 15 years old. Looking at some of the pictures in ebay auctions for 2023's and 2024's I note that quite a few have missing display lines, so this might be a common problem. It's nice to know that a newer LCD module can be adapted fairly easily to fit.

One other piece of good news regarding the 2023/2024 is that the service manual has just been uploaded to KO4BB's Manuals Repository. It's still in "Recent Uploads" but should be available for download fairly soon.
 

[RCMR]

A tip -- when soldering cables to pin-headers in the way you did, plug the pinheader strip into a matching socket.  This will help sink the heat away and preserve the alignment (straightness) of the pins while soldering.

[SeanB]

Tip for the mains switch is to look for an old XT tower case ( I know, really old but still a toss item) and use the power switch from it. Comes pre insulated, and with a 5 core cable with earth connector via a tag at both ends and insulated spade connectors on the PSU end.

[jeremy]

This was cool! thanks for sharing

[muvideo]

Interesting, thank you for documenting it.

As for the hotseal connector I've found that many times the problem is
where the connector is glued on the pcb. In the past I've "repaired" a few of them,
used a clean and round tipped soldering iron at 160°C, making a little pressure
on the adhesive connector where it is glued to pcb, moving the soldering
iron tip along the conductive strip, if you do this while the lcd is powered
you can see the line affected going on and off when you touch it.
But there is the risk is to damage the tape and once perforated no hope
to repair it anymore...

Fabio.

[djsb]

Nice work.

David

[grumpydoc]

A tip -- when soldering cables to pin-headers in the way you did, plug the pinheader strip into a matching socket

Thanks - now you mention it that seems the most obvious thing to do and I'm kicking myself for not thinking of it.

In the past I've "repaired" a few of them, used a clean and round tipped soldering iron at 160°C, making a little pressure on the adhesive connector where it is glued to pcb

I'd applied a little heat (hot air gun set at 150°C) to the strip to see if I could get the adhesive to flow but, as I said, this didn't alter much at all. Because of the design of the display it's not possible to apply heat and monitor changes while it's powered up - the frame covers everything and needs to be closed for the zebra strip to make contact.

As for the LED supply I had intended to replace the whole of the original board but that would really need a PCB making up. Also I didn't have any high current 1mH inductors so I just took the CFL inverter module off and replaced it with a bit of stripboard with a connector and two 36ohm resistors in parallel - I wanted a bit more power rating than 1/4W as the dissipation was about 0.12W and a single 1/4W resistor was getting slightly too warm.

This is the original CCFL inverter/Fan PSU board


An post modification it now looks like this.


If necessary the inverter can easily be put back together with the connector.

[w2aew]

Nice retrofit repair!

[reagle]

Nice to see things put back in service with a bit of elbow grease

[grumpydoc]

Thanks for all the positive responses.

It turns out that the display wasn't quite the end of the story. I mentioned that I'd had a quick check to see if it worked but that only really covered whether it could generate a carrier. I hadn't tried getting a modulated carrier, when I did I was rewarded with less than comforting error messages such as "Fractional N loop too high" and "Amplitude modulator unlevelled" and no modulation.

A bit of fairly straightforward fault finding followed: there was no modulation but trying AM caused a fixed increase in output amplitude and in FM a fixed shift in frequency so it looked as though the modulation circuits were being fed DC but were, themselves, working. Trying external modulation worked so it looked as though the LF generator was the cause of the problem.

The modulation signal is generated by a DSP (overkill?) then goes through a 20kHz low pass filter and a couple more op-amp buffers, this is the circuit (click on image for full-sized version)



Everything is together in this area of the board by the OXCO, the DSP is a Analog devices ADSP-2105 with an AD1856 serial DAC, the op amps are a couple of SOIC TL074's



It was fairly easy to follow the signal through to find that IC207d was hosed, signal present on the input but nothing on the output. A quick replacement:



Not completely happy with the soldering but I'm still getting the hang of these SMD things (yes, I know SOIC hardly counts

Finally checking the output shows all is well.



Hopefully that's it, but the 2024 is complex enough that I still might find a bit that doesn't work, hopefully it will be not too difficult to fix if I do.

All I need to do now is figure out what the "level 2" password is so I can calibrate it - someone has kindly changed it from the default!

[mianchen]

It's very impressive work! Well done.

[comprotudo]

Hello,

I have also a 2024, witch a "little" problem.
When switch it on and activate the control knob to tune the frequency, level etc. he does not work, even I can see the under line, the values does not change.
After waiting a little I switch the generator on/off and everything work fine.
Does anyone have the same problem?
Any clue about that??

Regards
Rodrigo

[grumpydoc]

Maybe the rotary encoder is unreliable.

One of the menus has an option to do a keyboard and encoder test, try that to make sure the encoder is working correctly.

[comprotudo]

I have checked that. When it runs ok I'm able to check the rotary encoder.
When it fails at beginning, the rotary check fail.
Probably the generator makes a initial check and some times he can't "see" the encoder.
I need some one can confirm this fact.

Rodrigo

Maybe the rotary encoder is unreliable.

One of the menus has an option to do a keyboard and encoder test, try that to make sure the encoder is working correctly.

[grumpydoc]

To continue an old thread..........

The Marconi had been behaving itself nicely for quite some time but it looks like it wants to attract some attention to itself.

I was using it for some bandwidth/frequency response checks on a 'scope I've just repaired and tried to dial up a 20MHz output only to be greeted with



and no output either - rats! 

Some fiddling later I noticed that it's only lost output from 18.75MHz through to 37.5MHz, everything else is fine, so checking the service manual I noticed this in the description of the error




OK, that looks promising in terms of narrowing it down, lets have a look at the schematic. The band pass filters seem to use steering diodes to switch themselves in and out of the signal path.





Click on the image above for the full circuit diagram of this section.

So it looks like SW9 is the switching voltage point - that comes from here





I had to reduce that a bit so it's slightly hard to read - again you can click on the image for the full sized one. Basically two 74HC138's are used to form a one of 16 decoder and then three LM324's are used as comparators/level shifters to drive the filter sections. One thing which does seem odd is that SW10 has two comparators driving it (IC604a and IC606b) as does SW1 (IC604c & d) but the others just have one comparator. Absolutely no idea why they did that

I'm really hoping that one of the 74HC138's outputs has died - or the comparator. If not I'm stuck, the only other active components are the steering diodes (which I suppose it could be). After that I can't see any of the inductors or caps in the filter section suddenly keeling over.

It's a bit late to open up the RF section and start looking so that will have to wait until tomorrow but I'll report back. In the meantime - thoughts anyone?

[Jay_Diddy_B]

Hi,

The LM324 have been put in parallel for the SW10 drive to increase the current to diode switches.

The LM324 simple shifts the 0-5V output from the 74HC138 to +5V or -11V for the pin diode switches.

SW10 should negative if the 18.75 to 37.5 MHz band is selected.

How to troubleshoot will depend if you can get good access to this board with unit running. If you can access the components you can follow the signal through the filter assembly with a scope.

Regards,

Jay_Diddy_B

[G0HZU]

All I need to do now is figure out what the "level 2" password is so I can calibrate it - someone has kindly changed it from the default!

I bought one of these generators a few years ago and had the same issue with the level 2 password. I overcame this by contacting IFR UK and they gave me contact details for someone in the USA.

I phoned him up at a prearranged time and gave him the S/N of the sig gen and he quickly gave me the 'hard' unlock code for level 2 for my generator.

i.e. there are two level 2 password codes. One you can change yourself and one that is hard coded (but unique to each generator). So once I had the hard code I could unlock and then set my level 2 code back to the default. Hope this helps

[grumpydoc]

I bought one of these generators a few years ago and had the same issue with the level 2 password. I overcame this by contacting IFR UK and they gave me contact details for someone in the USA.

I phoned him up at a prearranged time and gave him the S/N of the sig gen and he quickly gave me the 'hard' unlock code for level 2 for my generator.

i.e. there are two level 2 password codes. One you can change yourself and one that is hard coded (but unique to each generator). So once I had the hard code I could unlock and then set my level 2 code back to the default. Hope this helps

Yes, I heard the same and tried contacting Aeroflex to see if I could get the code but no luck. In the end I wound up sniffing the reads from the EEPROM to figure it out.

The LM324 have been put in parallel for the SW10 drive to increase the current to diode switches.

Yes, I wondered if that was the case, but why on only two of the switches? It's only SW10 and SW1 which have the doubled up drive.

The LM324 simple shifts the 0-5V output from the 74HC138 to +5V or -11V for the pin diode switches.

Yes, that was fairly obvious - I mentioned it in my post.

SW10 should negative if the 18.75 to 37.5 MHz band is selected.

I'm pretty certain it's SW9. In another table in the manual it confirms the select codes for the various low pass filters which are:

Table 1-4: Frequency band selection

BS3	BS3	BS2	BS0	Selected Band
0	0	0	0	420 - 600 MHz
0	0	0	1	300 - 420 MHz
0	0	1	0	150 - 300 MHz
0	1	0	0	75 - 150 MHz
0	1	1	0	37.5 - 75 MHz
1	0	0	0	18.75 - 37.5 MHz
1	0	1	0	10 - 18.75 MHz
1	1	0	0	1.2 - 2.51 GHz
1	1	1	0	0.84 -1.2 GHz
1	1	1	1	600 - 840 MHz

The code for the 18.75 - 37.5MHz band is 8, the description in the manual mentions that the codes are chosen so that the same control bits are used to control the division ratio of IC602. If you follow the 8th select line (Y0 on the 2nd HC138 - IC603), highlighted in yellow on the snippet of schematic that I posted, it drives SW9 which also looks to be the right switching point in the LPF section. I'll check, though, when I get it open later today.

How to troubleshoot will depend if you can get good access to this board with unit running. If you can access the components you can follow the signal through the filter assembly with a scope.

Yes - actually it looks pretty easy. Dave did a teardown of a 2023 which is basically identical to the 2024, if you look at around 27:02 you can see the frequency generator section and the two HC138's and three LM324's clearly, just above the divide-by-n chip which Dave points out. Access for testing should be fine, depending on whether it is IC604 or IC606 I might have to remove the diecast internal screening to get a soldering iron/hot air in.

https://www.youtube.com/watch?feature=player_detailpage&v=Gmryy5KWm_g#t=1622

[G0HZU]

I'd be tempted to (initially) troubleshoot this circuit by looking at the bias voltages and currents  rather than to try and trace an RF signal through it.

If any series component in the chosen RF path fails open circuit then the bias chain will be broken and it should be easy to work out where the break is.

Something doesn't look right around D647,  D619 and D618 so I'm not sure I trust the circuit here. Is there something missing from the circuit here? eg a bias feed? Or is D647 drawn backwards?

If the bottom series diode path is unbroken then I think there should be something like 4V dropped across R603. However, these are just guesses based on the circuit diagram in the service manual.

It looks like the top series diodes are only weakly biased at maybe 600uA but the bottom section of series diodes is biased at maybe 8mA. Quite a difference? Is there a mistake in the schematic around D647 as mentioned above?

The reason SW1 is doubled up is because it has to drive three parallel diodes in forward bias. If Vcc = 5V and there's 1.5V dropped in the LM324 then the bias current for each opamp will be (5-1.5-0.7)/274R = 10mA each. So the total forward bias into the diodes will be 10+10 = 20mA (about 7mA each diode?).

I'm not sure why SW10 is doubled up. Maybe they did it because it was there as an unused opamp in the quad LM324 package? So they just used it as it was free?

I did spot a typo in the 37.5MHz LPF. L635 should be 330n (or maybe even 270n) and not 330m.

If I had to guess where the fault was I'd say it was a poor soldered connection under L633, L634, L635 or L636.

[G0HZU]

I'm pretty sure D647 is drawn backwards and if this is corrected then the whole bias system makes sense because the bias current in the top section will be several mA (biased from R602 up near the top LH corner of the full schematic)

So the whole circuit will be more balanced in terms of bias at around 5mA for both top an bottom

Doubling up SW10 on the 18.75MHz range will increase this bias by a couple of mA. Not sure why they are doing this but maybe this reduces the RF loss slightly. Seems odd to do this on the lowest frequency range...

[grumpydoc]

Thanks for your analysis - it far exceeds my understanding of what's going on!.

I'd be tempted to (initially) troubleshoot this circuit by looking at the bias voltages and currents  rather than to try and trace an RF signal through it.

Agree, there's really no point tracing the RF if the switching voltages aren't present andit's easy enough to chek these.

If I had to guess where the fault was I'd say it was a poor soldered connection under L633, L634, L635 or L636.

I'd be slightly surprised if this were the case - I'd expect a dodgy soldered joint to be intermittent before (even if) it failed completely.

Anyway - I've now had the chance to have a look. Unlike Dave's I just get standard mild steel torx screws, not sexy black ones. There are a lot of them though
Click on any pic for a larger version.



I almost wonder whether the screws in Dave's were the correct ones  - he mentioned his puzzlement at the fact that the cutouts in the aluminium above each screw weren't quite big enough to let the screw out unless he slid the whole tray out a bit. The cutouts are fine with the smaller head of the torx.

OK, we're in  Looks like the rubber gaskets in mine want to stay stuck to the aluminium plate, not in the slots that they are supposed to go in.The area of interest is the lower left corner.


Visual inspection is OK except I'm not sure I like what's going on around IC605, that looks a bit odd.


Taking a closer look reveals that I'm not the first to come by here - it looks as though IC605 has already been replaced and no real attempt to clear the flux away has been made.


I can clean that later, for now it only took a couple of quick measurements to confirm that the Y0 output of IC603 isn't pulling down to logic 0 when I select a frequency between 18.75 and 37.5MHZ - it sort of makes a limp attempt but only gets from just under 5V to about 3.

Whether that's a problem with the HC138 or the LM324 I don't yet know but I'm pretty certain that I don't have any 324's lying around and equally sure I should be able to find an HC138

Off to hunt in the parts bin.

[grumpydoc]

I'm pretty sure D647 is drawn backwards and if this is corrected then the whole bias system makes sense because the bias current in the top section will be several mA (biased from R602 up near the top LH corner of the full schematic)

So the whole circuit will be more balanced in terms of bias at around 5mA for both top an bottom

Doubling up SW10 on the 18.75MHz range will increase this bias by a couple of mA. Not sure why they are doing this but maybe this reduces the RF loss slightly. Seems odd to do this on the lowest frequency range...

I think you're right, here's the PCB:


D618 and D619 are the two vertical diodes in the centre. The cathodes are marked for most diodes on the silkscreen and these two are joined anode to anode as on the schematic.

D647 must be the unlabelled horizontal diode in the centre of the picture  - its cathode is connected to the junction of D618 and D619, not its anode as shown in the diagram so I think that you're spot on in your diagnosis.

[peter-h]

This has been a very interesting read - especially as I am about to buy one of these (2023 or 2024). I need something fairly portable and at about 8kg these units are unusual. Many thanks for documenting it.

That circuit with the 3 diodes was obviously wrong because there is no way to get any current to flow through the middle node!

The other thing which is readily apparent is that if you buy one of these and it stops working, it is rather likely that it will have to be scrapped unless the service manuals are available. I have an old HP 3314A low frequency signal generator which uses a load of reed switches and I've had to replace a number of them, but (a) I did get the service manual with it; (b) usually it is obvious that one of them has gone; (c) they are easy to fix. The microprocessor control side of it (TMS9900!) would be impossible to fix unless it is something really obvious.

How does one get the service manual?

The other Q I have is: where can one buy handles? The units I see for sale (at a reasonable price) don't come with them. I can machine some up but it would be easier to buy something...

[grumpydoc]

How does one get the service manual?

Ko4bb has them for download - link in reply #14

As to handles - I've seen a few units for sale with them fitted but haven't seen any handles for sale by themselves.

[G0HZU]

That circuit with the 3 diodes was obviously wrong because there is no way to get any current to flow through the middle node!

Agreed. However, I have learned from bitter experience to avoid saying (on a forum) that a manufacturer has 'definitely' made a mistake in their documentation. People reading the thread can get upset or annoyed and I end up getting slated for pretending to know better than the manufacturer.

As for the handles, I would not bother buying any. They do provide a degree of protection for the front panel but they often break. I think what happens is that they age and weaken due to knocks and stress and eventually someone will pick up the generator by just one of these handles and the handle will snap. I think this is one reason why these generators get damaged internally as they suffer impact damage.

At work we have dozens of them and quite a few of them are missing one or both handles. Sometimes there's half a handle still fitted after it has broken off.