Marconi 2019 Signal Generator.

[Station X]

I recently acquired a non-working Marconi 2019 signal generator. So far I have fixed about eight faults on it, but I'm stuck with what I hope is the last fault. The SG works fine for output frequencies of up to 520 MHz, but not for frequencies of 520 to 1040 MHz. It's pretty obvious that the frequency doubler, used for the latter range, is not working.

I've been doing tests with the generator set up to output 521 Mhz. As expected a frequency counter shows a 260.5 MHz signal at the base, collector and emitter of TR4 (see attached diagram). I also see this frequency at the "input" of the two diodes D55 and D56. However at the common point of the diodes and on all the leads of TR5, instead of seeing 521 MHz, all I see is a signal which flicks all over the place between about 10 and 20 MHz. Needless to say there's no signal at the doubler's output, which is on the other side of the switched filters. It's very likely I'm being mislead here, as the output from the doubler will be rich in harmonics and sub harmonics. However the only test gear I have to check the signal is a frequency counter,

The switched filters have three ranges 520-660 MHz, 660 MHz-820 MHz and 820 MHz to 1040 MHz. I have tried test frequencies within each of these ranges, but the result is the same ie no output. I have checked the data lines to the filters on all three ranges and they are being set correctly.

Test voltages are as follows:-

TR4

C 12.35V
B 3.55V
E 2.69V
Across bias diode 0.67V

TR5

C 7.07V
B 2.725V
E 1.94V
Across bias diode 0.67V

Non-inverting input IC8 0.260V
Common point of D55 and D56 0.249V

That's all I can say really. Any help gratefully received.

Thanks.

[G0HZU_JMR]

Do you have a 60MHz scope and a x10 probe?

If I only had the 2019 and a scope as the test gear I would exploit the fact that this doubler looks to be a broadband circuit suitable for use down to HF/VHF. This is desperation stuff but you could fit a (temporary) 56R load at the output of the doubler with a series DC blocking cap as the image below.

Then isolate the doubler input as below and make a coaxial connection from the 2019 front panel output and use the 2019 as your test source at (say) 20MHz to drive the doubler. Then troubleshoot it using the scope with the hope that it gives out 40MHz at the temporary 56R load resistor. The waveforms at the various parts of the doubler could be looked at on the scope because you are only looking at 40MHz maximum. It looks like the doubler is powered all the time from 15V so this approach should be OK?

Otherwise, if you don't want to do it this way you could make a simple RF detector probe to drive a regular DVM and sniff around at 260.5MHz and 521MHz. But you won't be able to see phase info as the probe just outputs a DC level according to RF level. Or you could do some DC checks to make sure the diodes are healthy and the (active) compensated bias is working OK.

An even simpler check would be to test the two diodes on a meter to see if they are OK? The compensating diode in the active bias circuit is probably OK as I'd expect to see just under 0.3V from the active bias. But the rest of the circuit should be quite rugged and reliable unless there is a poor solder connection under the PCB somewhere.

[G0HZU_JMR]

One thing to bear in mind is that the doubled output at 521MHz will be maybe 10dB lower than the 260.5MHz drive and maybe this output level at 521MHz is too low to properly drive your counter. So maybe the circuit is still OK at the doubler diodes and your counter can't see it? There won't be a lot of gain in TR5 and maybe the capacitance of your counter probe is attenuating the signal at 521MHz? So the counter can't get a reading here? A well made RF probe with a suitable detector diode and short connections should work OK here.

A cheap way to get a crude spectrum analyser would be to buy one of the £15 USB SDR dongles on ebay. This could be used as a cheap sniffer if you make a crude E field probe to go with it. It will be a crude analyser with lots of internal spurious but still could be used to look at the signal path through the whole doubler path.

[G0HZU_JMR]

Here's my old RF detector probe. It used to have a SMD diode on it but has a classic HP2800 Schottky diode on it at the moment (70V rated)

When probing 50R circuits and the little short earth spike is used (rather than the croc clip ground) this probe is flat to several hundred MHz when probing small signals at maybe 0dBm. By 521MHz the stray inductance in the tip and the earth spike next to it interact with the diode capacitance and the diode over responds by about 1dB. But still very good.

It only outputs about 120mV DC to the meter at 0dBm and the detector response is not linear of course. It also isn't temperature compensated but as a quick and dirty sniffer (used with a lookup table to linearise the readings) it can still be a very powerful tool.
If you look closely you can see the little earth spike next to the detector probe tip. Both are just made with ordinary TCW. This probe costs very little to make!

The circuit is similar to the classic 1n34A RF probe (google for it) but I have used a HP2800 Schottky diode and SMD caps and resistors as the probe body is a PCB so it is easy to fit SMD parts like this.

[Station X]

Thanks for all your advice Jeremy.

I have found that when faced with a difficult fault like this it helps to put it on one side for a while and come back to it afresh, so I'll be trying your suggestions next week.

It also occurs to me that I might be able to use a UHF scanning receiver to listen for the doubled signal?

As far as the HP diodes go I couldn't find a spec for them and was puzzled by the low forward volt drop. I disconnected the two between the two transistors and tested them using a DMM's diode test facility. They show a forward volt drop of 0.26V and have infinite reverse resistance.

I also noticed that the two resistors in the feedback loop?? around TR4 had gone about 25% high in value. However I have not replaced them as they are physically small and presumably carbon composition types having low inductance? I don't have suitable replacements to hand and would have to order some.

[Dickie]

Hi,
I don't want to butt (?) in, but if all you have to check the output above 520MHz is a counter, all you can deduce is that there is not enough RF to drive the counter above 520MHz. So the fault could lie anywhere along the doubler path, from the doubler to the RF output amp input.
Your DMM measurements suggest the doubler diodes are OK. The forward voltage drop is about right for low-barrier Schottky diodes.
What resistors have gone high? I don't think there is any feedback loop around TR4.

[Station X]

Originally Posted by Dickie

What resistors have gone high? I don't think there is any feedback loop around TR4.

I should have said TR5 This is where I demonstrate my lack of knowledge. I'm referring to the 51R and 100R resistors on either side of L70.

[karesz*]

Hi Graham,

Older carbon resistors can be 20% in tolerance, but I think Marconi selected it with 5%?
You can find i.e. Beyschlags/Vishays HF Resistors(MMA 0204 HF, MMB 0207 HF)_ they are perfect for some application (up to GHzs ), Ok; I dont know its source.
Otherwise I think Farnell/RS/Mouser.. can have it or usable thin/thick film SMDs.
Regards, Karl

[Station X]

Spec says 0.125W CC 5%. I'll change them, so they can be eliminated as a source of the problem.

[Dickie]

Ah, yes. ISTR they were made by Allen-Bradley and were chosen for their small size and low inductance. It would be a tedious bodge, but modern SM resistors would be OK.

It also occurs to me that I might be able to use a UHF scanning receiver to listen for the doubled signal?

And get the 'normal' second harmonic too.

[G0HZU_JMR]

I don't think I'm being overly cautious here when I say that I would advise against casual changes to some of the resistors in those amplifier sections. This means changes to packages or even manufacturer.

This mainly applies to any of the resistors used in the C-B feedback or in the emitter of TR5 and TR6. Also, don't swap the BFR9x transistors for modern ones bought from ebay or from 'less well controlled' supplier chains if you can avoid it. Those amplifiers have to work across 520-1040MHz and have to maintain a certain gain at 1040MHz. I think you want to remain as original as possible here especially if you don't have any test gear to check the frequency response after any changes.

I have been inside my 2019s loads of times over the years but not for maybe a decade now so I have forgotten what this stage looks like in terms of component choice. But I suspect that the resistor packages in TR5 will be fairly clunky and the frequency response of this amplifier will depend on the package inductance of those resistors as well as the package inductance of the 39nF cap used in the feedback and also the PCB layout itself. I think the amplifier will rely on the stray inductance to change the feedback beyond just 'resistive'. I think the inductance here is wanted/required in order to perk up the gain at 1040MHz for example.

The fact that they have used two gain setting resistors in the emitter means that they were fighting the package inductance effects in order to get a flattish frequency response. Changes here of a couple of nH will change the response.

[G0HZU_JMR]

If there's anything you can't read clearly in the pdf manual I have the official/original 2019A operator/service manual (it comes in an official Marconi A4 ring binder format with foldout pages for the circuits and layouts) plus all the documentation that came with a typical purchase here. I think I have similar for the basic 2019 but I can't be certain.

[G0HZU_JMR]

If the voltages check out OK on the amplifier transistors TR4 to TR6 and the diodes then I would look at the control voltages on the 520-1040MHz range select switching diodes and also check that the varactor diode 'tuning' voltages are sensible on the four varactor steering lines as the frequency is changed across 520-1040MHz. The four tuning voltages should be fairly low at 520MHz and increase steadily as the sig gen is tuned up towards 1040MHz.

It's much more likely that there is a poor connection somewhere either on this board or the board that controls/steers it. The poor connection could be a soldered joint on the PCB or something to do with the various connectors used. These sig gens aren't that reliable and it took me several years to get mine reliable as they suffered from intermittent faults as well as proper faults caused by component failure. I think it's wise to not leave them running in an enclosed space as they can run quite hot (no fans!) and this may cause some of the long term reliability problems through thermal stress across hot and cold.

[Dickie]

Hi Jeremy
You are no doubt correct in assuming that it was a struggle to get the gain and b/w right, but surely if the design is on a knife-edge and the resistors really have gone 25% high will that not severely affect performance? BTW from memory those resistors are no more than 6mm long and rather susceptible to heat damage.

[G0HZU_JMR]

Yes, it's a worry that they have changed 25%. I suppose there is the risk that the amplifier can now go unstable (possibly way up beyond 1GHz) and this would cause mayhem in the system.

But I would at least keep the originals safe and if you do have to change them then choose ones with the same package inductance. It's difficult to offer advice on the best way forward because of the restricted test gear available. When I was a student I was lucky to have a very keen tutor and in his free time he taught us to fault find using just an Avo 8 and a homemade RF detector probe and a squarewave signal injector/tracer! That's why I suggested the RF detector probe

[G0HZU_JMR]

If you look at the circuit for TR6 you can see clues that they really had to fudge things in order to get useful gain up at 1GHz from this (higher gain, higher signal handling wrt TR5) broadband amplifier. They had to put a 2pF capacitor across the C-E. This would compromise the stability up at several GHz meaning that the amplifier would be prone to instability with certain loads. But clearly a healthy 2019 manages to stay stable here with this cap fitted.

Even the package capacitance of the feedback resistors will affect the frequency response a bit. Having two resistors in series will minimise the feedback capacitance here as well as just making up the distance on the PCB.
Today, these amplifiers would be trivial to design and implement. Just go to Minicircuits or AD/Hittite and choose a £2 50R MMIC gain block IC. These tiny devices will typically be flat and well behaved up to many GHz with unconditional stability.

But in the days when this sig gen was designed, the designer would have been fighting component parasitics and PCB layout issues in order to get flat and stable gain across just 500-1000MHz with a BFR9x based design. Jelly wrestling time!

[G0HZU_JMR]

Disturb TR6 with a soldering iron as little as possible especially the little cap that connects across the C-E. I don't know what type it is (or how they actually fit it here) but it may prove fragile and it's parasitics may be important to set the gain response and also to prevent out of band instability up at several GHz. Even small/extra blobs of solder can affect the parasitics here.

The good news is that the gain doesn't have to be 'flat' in this section and it can afford a fair bit of slop in terms of absolute gain as the final ALC section further along the system should have enough open loop gain and enough gain control range to accommodate minor changes in response and gain in the doubler. But I think you need to try and replace any feedback parts with modern equivalents with similar parasitics just to be on the safe side. The small change in gain caused by the 25% change in feedback should easily be accommodated by the ALC system. The bigger fear would be instability introduced as a result of slightly higher gain in that stage. But I'd only change those parts if instability was actually happening.

[Radio Wrangler]

The split resistors around collector to base of TR5 is a method of avoiding having a stub of wire connected to either B or C and risking instability at much higher frequencies. For these resistors 0805 surface mount types with one end soldered right onto B and the other one with one end mounted right on to C ought to work quite well.

This stage has both shunt and series feedback paths with the C-B path and the emitter degeneration respectively. L70 is a little gimmick inductor to help put a bit of lift in the characteristic at the high frequency end. Similarly in the emitter network there are TWO boost R-C networks.

Shunt/series feedback like this is able to give controlled gain with controlled input/output impedances when all is well. This one looks like it's been fiddled with rather a lot. Anything you do with it will change things, but I'd be prepared to try SMT resistors without too much worry.

Q4 is a phase splitter just like the 'concertina' ones in many an audio amplifier, and is doing duty in place of the usual transformer in a diode push-push doubler. In this case the direction of the diodes makes this a pull-pull doubler.

The prime-mover oscillator has to cover an octave to make it gapless, so the low freq end of the doubler output meets up with the top end of the VFO range. This means that the filter after the doubler has to be narrower than an octave, so the filter has to be switched to cover the range in two bites.... or more bites.

David

[G0HZU_JMR]

The split resistors around collector to base of TR5 is a method of avoiding having a stub of wire connected to either B or C and risking instability at much higher frequencies. For these resistors 0805 surface mount types with one end soldered right onto B and the other one with one end mounted right on to C ought to work quite well.

I guess only the original designer will know for sure but I think they deliberately wanted a controlled amount of inductance here in the overall feedback path. The images in the manual suggest that they use carbon comp resistors in series with maybe 13mm of skinny PCB trace (microstrip?) between the feedback points. So they definitely didn't try and get a direct low inductance path. They could easily have shortened it.

L70 is a little gimmick inductor to help put a bit of lift in the characteristic at the high frequency end.

According to the manual I have here L70 is the resistor leg inductance?

There must be some compromise here I think. More inductance will perk up the gain at 1GHz but I think the stability margin will degrade. That's why I was cautious about fitting alternative resistors that might have more or less inductance than the originals. I really don't know how much inductance (or what the package capacitance) is in those original CC resistors up at 1GHz.