2GHz oscillator not working - faulty step recovery diode?

[Radio Wrangler]

It looks like things are coming together, AND the matching simulation is the cherry on the top.

Rather a good present for the new year.

Faults are always simple after you've found them!

Happy Hogmanay

David

[G0HZU_JMR]

Yes, that is good news!

It took a bit of tracking down but you got there in the end. Time to box it all up and try it in the analyser!

Happy New Year to all...

[orbanp1]

Hello Everyone.

I put that 200MHz driver stage together with the SRD/helical filter assembly, and the best I could coax out of them was -8dBm power at 2GHz. That is with adjusting the various trimmer caps and the filter.

I do not think I am out of the wood yet!

I am not sure of that 2dBm LO-power requirement either, the part numbers of the mixers (1st, 2nd) in my SAs are different from those in the documentation of the A-version.
If anything, I would expect more drive requirement than that 2dBm!
Having a peek inside the mixers I could see a quad diode package and transmission line transformers using thin parallel wires.

I guess I could excite the SRD a bit harder!
Power from that 200MHz circuit can effectively be controlled by the value of the R15 resistor.
That resistor was 330 Ohm originally, in that board that is the only cc resistor among those that I measured whose value decreased and not increased! It reads now 299 Ohm!

I did put in a diode in my simulation to see how much current is pumped through that by that oscillator circuit.
I used a modified 1N914 model. I increased the series resistor in the model to match the forward voltage drop at 10mA that I measured on my diode.
While such a model would not show the harmonic generation, I would expect it to show the loading on the diode with reasonable accuracy.

Any misconception in the above theory?

According to simulations I could go up to 390 Ohm with that resistor and the peak current would stay below the 50mA diode limit, and with an RMS current of 26mA.
The spec sheet never mentions if the SRD current limit is the peak current or average current.
I have to order such resistors.

Unfortunately I can not measure the 2nd LO-module from the working SA for comparison, that is the only equipment I have to make such a measurement.

Regards, Peter

[G0HZU_JMR]

A really rough estimate of the conversion efficiency of a typical x10 diode multiplier like this would be 10-20%.

So if we go with 10% efficiency and maybe 200mW from the 2N3866 you might get 20mW (+13dBm) at the diode at 2GHz. Then there would be some loss in the filter that follows the SRD and this could be several dB so you might end up with 10mW (+10dBm) LO power at 2GHz with a typical LO design like this?

[orbanp1]

Before I hooked up the SRD and filter I did put back into the circuit the original R15 resistor with that reduced 300 Ohm value.
When I have seen the diode current in the simulation I was worried that high output current might kill the SRD!
And the designers probably had a good reason to use 330 Ohm and not 1K!

The "original" simulations were done with R15 1 kOhm, that is the one that gave the 6.5Vpp value at the output.
Using the 300 Ohm resistor in the simulation gave 3.4Vpp output, measurement gave 3.56Vpp.
That is roughly half the voltage, quarter the power.
If that thinking holds, and using the estimate for the SDR conversion, that would still be just 6dBm increase, still not enough, if the 1k resistor were used!
The manual mentions 2dB - 3dB insertion loss for the 2GHz helical filter when tuned up.

Even if I can not measure the 2nd LO output of the working SA, I guess I can measure the output from the working 200MHz exciter circuit, and the one I repaired, those ones I can measure and compare.

Regards, Peter

[G0HZU_JMR]

I'd be tempted to leave the bias resistor at the original value. If you are brave, you could try breaking the signal path at C15 and then drive the 2N3866 from your HP8640B sig gen at 200MHz at a drive level of a few mW. Experiment with sig gen drive levels up to +12dBm to start with and see what Vpkpk you see in the diode matching section after tweaking the matching trimmers. I think a healthy HP8640B should be able to deliver >+18dBm so be careful not to overdrive the 2N3866. You could put a 6dB attenuator inline with the sig gen to prevent the risk of overdriving the 2N3866 and damaging it or the SRD.

[orbanp1]

Hi Jeremy,

Thanks for the suggestion, sounds very interesting and useful, I will probably do it!

One more point on that R15 resistor.
With that 330 Ohm value it is clearly stressed to its limit!
The dissipation on it, depending on how is it measured, AC or DC, is 200mW to 250mW. (I know, there is only one value, but the simulator also reports the DC dissipation after the initial DC operating point calculation.)
The resistor size is 6mm x 2mm dia, present day such CC resistors have a 1/4W rating. I am sure that the rating of such a size resistor was not more forty years ago, it was possibly less!
Just for comparison, that 100 Ohm resistor in the collector, R11, that dissipates 12mW or 78mW, depending on the value of R15, and has a rating of 2W!

Clearly that 330 Ohm value as requirement for R15 must have came up after the original design, and was not thoroughly thought through!
Probably that is the reason that now it has a value of only 299 Ohm.

Thanks, Peter

[G0HZU_JMR]

It's late and I haven't done any simulations or design checking but I'm a bit confused why that 330R resistor will be stressed. Everything below is just a ballpark guess at what might happen after the 2N3866.

If the multiplier is passive and the 2N3866 can produce (say) 300mW RF at 200MHz then that 300mW is the total power (heating) budget available for everything that comes after the 2N3866.

I doubt that the 2W 50R resistor R14 is double terminated at 50R so maybe 100mW of the 300mW budget gets burned up in that resistor. But it depends on how the matching components are adjusted. R14 is probably rated higher than this to allow for maladjustment of the matching sections where a lot of the RF power from the 2N3866 ends up in R14 in error. But a 2W package does seem a bit overkill here.

I'd expect that the SRD will eat up another 100mW so that just leaves 100mW for the 330R resistor and any other passive losses due to the component Q of the trimmer caps.

Just for comparison, that 100 Ohm resistor in the collector, R11, that dissipates 12mW or 78mW, depending on the value of R15, and has a rating of 2W!

Again, I'm only guessing but I would expect the doubler to be quite inefficient and if it has to produce (say) 50mW then I'd expect the DC input power for this stage to be quite high if the stage was working correctly.

So a late night stab at guessing the total Pdiss in R11 might be something like 300mW if the extra current draw for the doubler action is included in the Pdiss calculation for R11. This is where a decent service manual really would help to remove a lot of the guesswork.

[Radio Wrangler]

If this is a unit you obtained in a dead state, is it possible that it was known to be faulty and attempts to repair it had been made?

That SRD will make a broadband comb, so filtering is needed before it get used at the mixer. So I'd expect at least one stage of cavity filter - likely with that diode/SMA assembly dipping into it. Is it possible that someone has already tried tweaking a tuning screw or two in order to see if they can compensate for the early stages of that dry joint?

If it's built this way, the 2GHz signal is created only at the field in the cavity, so it's not feasible to sample and measure.

Basic diode quad mixers with Schottky diodes are usually specified for about +7dBm LO drive. In early spectrum analysers you sometimes find single-balanced mixers to reduce the LO requirement a bit.

Just musing a bit.

David

[orbanp1]

Hello Everyone,

Made some comparative measurements, and it looks like that the drive power from that repaired 200MHz unit is down, compared to the unit from the working SA.

I guess my next step is to replace that 2N3866 PA transistor!

Took out that 2nd LO from the working SA, removed the SRD-2GHz filter filter assembly from it and hooked up the exciter unit to another SA.
The working SA's exciter indicated +14dBm at 200MHz, while that repaired unit only indicated +4dBm.
I call it indication as the output impedance of those 200MHz units is nominal 330 Ohm, the DC-bias resistor for the SRD.
The SRD in this B-version is part of the filter assembly, it is the input coupling loop.
As the 200MHz exciter unit has an SMC connector at its output I just hooked up a coax-cable and the SA.

There seems to be some confusion about those 330 Ohm resistor, there are two such resistor in the B-version circuit.
One is at the output, it is the DC-bias resistor (the A-version does not have such a DC-bias resistor for the SRD), the other 330 Ohm resistor is R15, at the collector of the doubler stage transistor (Q3), that is the one that was 1 kOhm in the A-version of this circuit.
Simulation showed that the value of this resistor controls the output power from this exciter module.
I changed it to 1 kOhm, like in the A-version.
The power output was exactly the same!
That resistor loads and reduces the collector voltage on the doubler transistor.
The working unit also had 330 Ohm in it.

On the plus side of things, when I put that working SA together again, it still worked!

Both SAs were acquired as non working units. I managed to fix the other one.
From the look of it, none of them were touched before, it seems that no repairs were attempted on them.

Regards, Peter

[G0HZU_JMR]

Sorry for the delay in reply, I've had a busy few days and not much free time.

There seems to be some confusion about those 330 Ohm resistor, there are two such resistor in the B-version circuit.

Yes, sorry, I thought you were referring to the 330R bias resistor for the SRD. I now see that R15 (330R) is in the doubler section and this is the resistor you were discussing.

I guess my next step is to replace that 2N3866 PA transistor!

It's a long time since I played with a 2N3866 but I do recall they can fail with a leaky collector and this means they can draw current of a few mA or more even with no base bias. I can't remember how much this fault typically affects a basic B-E and B-C junction test or an HFE test but it might still pass them.

[orbanp1]

Hi Jeremy,

Thanks for the reply!

I did measure that 2N3866 with one of those Chinese clone component testers.
The tester did not show any transistor failure, though I would have to read the manual what does the "tester say" about excessive leakages, if they are present. (There is a very detailed technical manual available on the German web-site of the original tester project.)

I see if I can find such a transistor to harvest from some of the older boards that I have stashed away for exactly such purposes, though I do not have that many older RF boards to scavenge from.
I will order it from Digikey if I do not find one, but it might take some time (to make the order).

Regards, Peter

[orbanp1]

Just a quick question about the replacement for the 2N3866 transistor.

I did find a few 2N4427 transistors (made by TRW) to scavenge.
From the datasheet it looks like it is the lower voltage version of the 2N3866.

Would I be pushing my luck using that transistor as a substitute or should I wait and order a "genuine" 2N3866?
Though the datasheet mentions the 2N4427's use as a PA at 175MHz, and higher, with 12V Vce supply (same as the SA's 2nd-LO exciter board uses).

Thanks, Peter

[orbanp1]

Hello Everyone,

There was lots of very valuable and appreciated help from the members of this forum!

Just to recap, I did find a faulty solder joint in a coil tap, but that did not seem to provide the output power as compared to the same module from the other working SA.
As a last resort, I replaced the final 2N3866 transistor with a 2N4427, that is a lower voltage version of the 3866 transistor (40V vs. 55V Vcb0, that was all I had).
The output still did not match the output of the working module.
That was when I set the repair project aside in frustration.

Fast forward to yesterday, when I thought I would try the module again, and see if I increased the 12V supply voltage a bit (to 15V) would I get more output.
This time I fully assembled the module, buttoned up all the shields, and added the step recovery diode module with the helical filters.
Lo and behold, when tuning up all the stages, I did get about +6dBm output at 2GHz!
That should be enough to drive the second DBM mixer!
And that is with 12V supply voltage.
Interestingly, increasing the supply voltage barely increased the power output.

The bad news is that when putting back the LO into the SA the SA still does not work ;-(

Peter