Automatic Level Control (ALC) drive circuit in Marconi TF2015 generator

[paulfifties]

I have a theoretical question concerning attached extract of the Marconi TF-2015 (10MHz-520MHz AM/FM signal generator):

What I don't understand is the function of the 2 Schottky diodes inside the red circle.
In addition, why each Schottky diode has in addition an additional 100pF parallel capacitor, while the Schottky's are exactly having a minimal 2 pF junction capacity, just to be able to perform very fast switching (low "charge buckets").

So the lower half of the circuit shown is from A17, the Automatic Level Control, controlling the current through the active (selected) oscillator circuit on A12.
At the top left, the carrier is fed from the selected oscillator circuit on A12 into the RF output amplifier circuit A17.

(all the remainder of the RF-amplifier, further to the right, with the BFR90 transistors, is pre-amplifying stage of the RF-amplifier, which we are no further considering at this point.)

Now, if the amplitude of the oscillator signal (taken from a Philips PM5145/01 12MHz generator) is increased, then I observe the following :

1) the DC voltages as measured at both the red arrows, start deviating more and more from each other.

2) the amplitude of the AC-components at both points start increasing also.

At this time, it is noticed that the amplitude at the cathode of D1 is increasing stronger than the amplitude at the anode of D2, which is connected to ground via C3.

So far I am still following with everything.


But now: what I do not understand is the exact function of those 2 series-Schottky's in de base circuit of the left-hand balancetransistor TR1 at the bottom.

And with each Schottky diode in addition being bridged/shunted by a ceramic 100pF.


Can anyone tell me about the exact purpose of those circled Schottky's ?
Why two in series ?
Why these 100pF capacitors across ?

PROBLEM DESCRIPTION
Just in case you want to know: the problem I am troubleshooting, is that when switching on the generator, at first an RF output is produced, but after less than a second, it is getting distorted, and finally collapses.
I noticed the A17 circuit TR3 output signal (which is a DC level) increasing steadily, until about some 10V, the RF output is getting distorted again, and when the TR3 output reaches over 10V, the RF output collapses.

When hooking up a scope probe (ATT X10) to the base of A17-TR1, things get more stable.
However, when selecting carrier ranges below 200 MHz, the RF is distorted very much. (Due to overdriven oscillator circuit I guess...).
So here is the reason why I really want to understand this ALC circuit to the bottom of it...

Many thanks!
PS: the full "Instruction Manual" for the Marconi TF 2015A can be downloaded from the web. But if desired, I can always mail it. It's a 5645 KB pdf file.

[Herald1360]

I believe those Schottkys are being used as part of the attenuator control of the ALC system. They don't need to look like diodes, rather as resistors whose value is controlled by the dc applied to them.

[Radio Wrangler]

More likely then PIN diodes than Schottkys in a current controlled attenuator

David

[jimmc101]

The answer is provided in the manual, the circled diodes provide temperature compensation for the detector diodes (arrowed).
I've attached the relevant extract.

[paulfifties]

@JIMMC101: thanks for your answer. However the manual does not mention the A17 capacitors C12 and C13 that are across the diodes D1 and D2.
So their function is still unclear...
In addition I have noticed the cathode D2 joint with the base of TR1 is very "touchy". If I hook up my oscilloscope to this point, the DC-voltage at this point increases a few volts.

I wonder if the potmeter A17R5, which is not discussed in the manual, would have to be used to adjust exact 50 mV RMS at the input of the A16 RF Output Amplifier PCB ?
Any hints for adjustments are welcome

I did replace the A17 D1 and D2 by BAT83S diodes (original types were showing HP2 835 405, so I suppose HP2835 Rectifier Diode, Schottky, 8 Volt, DO-204AH)
According to the manual, these should be HP5082-2811. I don't know whether the type of Schottky is very critical. I selected therefore a low capacitance type to be on the safe side...
But something is still wrong apparently...

[G0HZU_JMR]

I can have a go at guessing how it all works but I can't guarantee anything...

This looks to me to be a fairly basic ALC loop and (as you will know) the final transistor TR3 in A17 controls the bias to an oscillator that acts as a means to vary the RF level. Everything south of C4 on A16 in your diagram is DC or very close to it as it is all post RF detection.

Probably best to start with the RF detector section D1 and D2 up at the top in A16. This is best viewed as a voltage doubler that is riding on a 3.5V voltage pedestal set up by the potential divider on A17. Note that the detector diodes are set up to produce a falling dc voltage at D2 anode with increasing RF drive. I wouldn't expect the 3.5V reference voltage at D1 cathode to alter much with or without RF present because it looks like a reference. But I think the reference will be able to change a bit during open loop before it all settles.

IMPORTANT: The next bit of the analysis is with no RF present.

The diode detector D1 and D2 will have a tiny bias current through it and this will be a function of the 22k emitter resistor R9 in A17 and the the big divider network and the BCY71 transistors on A17. This diode bias current is going to be tiny. I'm going to guess it will be in the ballpark of 1uA but it could be less. With 1uA going through the RF detector (A16) D1 and D2 there will be a voltage drop across them both. This voltage drop is important and will be diode specific.

The design uses some classc HP Schottky detector diodes and at <= 1uA bias the voltage drop across both diodes is going to be quite small. It could be just 100mV.

So in the absence of any RF the detector output voltage at D2 anode on A16 will be about +100mV wrt the 3.5V reference at D1 cathode. I think that D2 anode is the 'real' detector output.

When you add loads of RF (far too much RF) the 3.6V detector output voltage at D2 anode will begin to fall a lot. Way below the 3.6V level that was there in the absence of any RF. But something special happens if you add just the right amount of RF such that the DC voltage at D2 anode drops by an amount similar to the Vf diode drop (100mV drop?) such that it becomes the same as the 3.5V reference voltage at D1 cathode. I think this is the condition when the ALC settles in a healthy levelled system.

Therefore, I think the Vpk of the RF waveform when the ALC settles will be a function of the Vf of the detector diodes but the detector efficiency is important too. So you ideally need to fit the same diodes with the same Vf characteristic and also the same detector efficiency if you want to get the typical (ALC levelled) 50mV RF level at A16 input. Ideally the diodes should all be matched in terms of vf vs current.

This is where it gets a bit more complicated. If you look up the Vf of these diodes vs temperature and forward current you will see that the Vf changes a lot with temperature. Not good for a system that relies on Vf to define the RF level at a settled ALC level. So this is why the two compensation diodes are there in A17. Any change in Vf due to temperature gets cancelled by the compensation.

You asked why the 100pF caps are across the compensation diodes and this may be to reduce 1/f noise in the Schottky diodes that could otherwise modulate the oscillator. At just 1uA bias these diodes look like a highly non linear resistor with a very high resistance and there will be 1/f noise at audio
frequencies that could modulate the oscillator. The alternative explanation could be to help with stability but I think the loop is heavily damped by the 22uF cap C11 on A17 next to TR3. This big cap probably helps in terms of system startup as it will take a while to charge allowing lots of time for the oscillator transistor (on the A12 module) to start up at a high bias level and settle before the ALC takes control and brings it down to the correct drive level.

I noticed the A17 circuit TR3 output signal (which is a DC level) increasing steadily, until about some 10V, the RF output is getting distorted again, and
when the TR3 output reaches over 10V, the RF output collapses.

I think this may mean that the system is open loop with too much detector output and it is trying to turn down the RF level of the oscillator on A12 using TR3. Eventually the transistor in the oscillator will be biased off so much that it with stop oscillating? If you have changed all the diodes to ones with different detector efficiency and different Vf vs current then you could have trouble closing the loop. Can you dig out the original diodes and check them on a meter? I'd refit these or if I was desperate I'd try some 1N5711 diodes.

In addition I have noticed the cathode D2 joint with the base of TR1 is very "touchy".

With 3.5V dc here and (maybe) just 1uA bias current through the diode a 10Meg scope probe will drag 0.35uA current here. I think this will load the circuit and will affect the bias balance in terms of the detector diode bias current and the bias current through the compensation diodes. I think the system will work best (over temperature) if the bias current through the detector is the same as the bias current through the compensation diodes and maybe this is why the trimmer R5 is there on A17. Maybe it acts as a fine tune to correct for subtle differences in the BCY71 transistors and the potential divider network to keep the bias the same?

I wonder if the potmeter A17R5, which is not discussed in the manual, would have to be used to adjust exact 50 mV RMS at the input of the A16 RF Output Amplifier PCB ?

I don't think it will have much impact on the RF level but maybe someone else can comment? One think that looks odd is that I would have expected R6 and R7 on A17 to have each other's value. That is, R7 to be 560R and R6 to be 390R. They do look to be select on test parts and maybe the system balances up better with them having this swap across in values?

[G0HZU_JMR]

You could also try the Marconi Yahoo group or whatever it is called now. I think they have some ex Marconi Instruments employees on there.

I had another think about the role of the trimpot R5... If the RF level really is as low as 50mV rms then this is close to the square law region so maybe a small change in detector bias will affect the detector efficiency enough to make this a means to set the RF level to 50mV rms. Otherwise R5 could be there to balance the bias current but there don't appear to be any obvious test points where this adjustment would be carried out. So it may well be the level calibration trimpot for the A16 RF input level after all.

[jimmc101]

With the values shown the offset between the junctions of R1-R2 and R6-R7 (A17) can be varied between -70 and -160mV with -120mV at mid range of R5.
This ties in quite well with the expected output from (A16) D1,D2 (which form a pk-pk detector) when fed with 50mV RMS (=141mV pk-pk), so yes I would agree that R5 is used to set 50mV RMS from the oscillator bank.

Re C12,13 (A17) I think these are there to ensure that the temperature compensation diodes are not affected by any stray RF. 100pF seems reasonable for the range of frequencies (10MHz to 520MHz).

As G0HZU_JMR has mentioned it is important for good compensation that D1 & D2 on A17 should be the same type as D1 & D2 on A16. Out of interest what are D1 & D2 on A16?


I've just re-read my original post, sorry if it came over as a bit abrupt, I posted in a hurry.

Jim

Jim

[G0HZU_JMR]

I just had a look at a better quality schematic and R3 on A17 in the potential divider is 4.7k and not 47k so all my voltages in my earlier posts are wrong.
I may have misread some of the other values too.

This means that where I said the reference would be 3.5V at D1 cathode it will be more like a 12.7V reference. So some voltages will be wrong by about 9.2V. I don't think there's much point posting detailed circuits on here. The website shrinks them down to be very fuzzy. The other thread with the stabilock circuit was just as fuzzy and caused loads of confusion because the resistors were sometimes being read in error by some people by a factor of a 1000.

I can't go back and edit it so I guess the test point voltages will have to remain incorrect for anyone who finds this thread.

Otherwise, could a mod change all the 3.5V entries in post #6 to 12.7V (there are five cases) and the two 3.6V entries to 12.8V please? it might save some confusion

[G0HZU_JMR]

Re C12,13 (A17) I think these are there to ensure that the temperature compensation diodes are not affected by any stray RF. 100pF seems reasonable for the range of frequencies (10MHz to 520MHz).

Yes, that seems like the most likely reason. I thought at first that it might have been to do with decoupling the noise from the high diode resistance that would modulate the carrier or maybe for stability but there are so many other big caps in the system to slug it already. So it is probably RF decoupling. It's a good idea to have these 100pF caps here I think.

I think that detector diode D1 cathode and D2 anode on A16 will be very close in DC voltage when the ALC settles on a healthy system. I can guess that it will be in the order of 100mV or less. Each somewhere around 12.8V DC? If the detector output voltage at D2 anode goes lower than this then I think the system will be open loop with too much RF drive into the detector forcing a very low detector voltage here. This would cause the symptoms at A17 TR3 where the voltage starts off low but it climbs up to 10V as the big 22uF cap charges and the A12 oscillator then gets biased off.

[paulfifties]

Jeremy, you have been of a fantastic help in understanding this circuit already !
As a matter of fact, I have simulated the ALC circuit on a breadboard, using the 47K value for A17 R3.
But you are absolutely right about the 4K7 value, as it appears to be listed in the parts list for Unit A17 (Chap 6 page 16) ! I also noticed it only just now that you mentioned it !

I will measure the actual DC values (hopefully this evening), both in the Marconi and on the corrected simulated circuit.

The reason that I changed the A17 D1 and D2 was that one of them showed visibly a cracked-off glass fragment at one of the ends.

I will also try to set up an additional simple test circuit to compare the voltage-current characteristics of the removed, but still ok, original diode, with the cracked diode, and also compare it with the BAT83S replacement that I have used.
Unfortunately, the A16 D1 and D2 (also listed in the parts list as HP5082-2811) are very hard to reach, in case I should need to replace those. As a matter of fact, I would not know how to best remove the A16 PCB in order to reach the downsided (component) side. I would at least have to unsolder a cap and 2 resistors to the connector towards the RF Attenuator unit... And even then I'm unsure whether the attached wires would not prevent the A16 PCB to be flipped around...

Incidentally, I just found some interesting info chap 6 page 16 : TABLE2: here the DC voltage levels are listed that A17 TR3 collector should measure, in function of the carrier range selected.

I will also measure the values of 12,7V and 12,8V that Jeremy calculated for no RF input at A17 D1 cathode and D2 anode respectively.

Paul

[G0HZU_JMR]

I think that between us all we are getting closer to the right answers

It might also be worth reverse engineering this looking backwards from the differential pair because we can make some fairly reliable assumptions here and apply some basic theory.

The A17 diff pair TR1 and TR2 has an emitter current of about (19-13.5)/22000 = 250uA. The large size of the 22k resistor R9 will look like a fairly decent current source. This 250uA total current will be shared between TR1 and TR2 depending on the status of the ALC loop.

We know that the TR3 (BC109) collector current on A17 controls the RF oscillator level and it will probably need to work with several mA of collector current when in the ALC levelled state. If we assume a decent HFE for TR3 then the TR3 base current will be maybe 20uA to achieve this. But there is also a 10k load resistor R10 in the collector of diffamp TR1 to consider as well as this will take up most of any TR1 collector current.

So the total current share in diff pair TR1 collector at a typical closed ALC loop condition would need to be about 80uA. This is because we need about 0.6V Vbe for TR3 to conduct reasonably well, so 60uA is needed in the 10k resistor to get 0.6V Vbe and 20uA is needed in TR3 base as above.

All of this means that the diff pair needs to share out its 250uA total of current to give 80uA in TR1 and 170uA in TR2 when the ALC is happy and closed. We can use some classic diff pair equations to predict what voltage difference there needs to be between the diff pair bases to give this particular current sharing. I've attached a spreadsheet below. It predicts that when the ALC is healthy and closed there will be about 20mV difference between the bases with diff pair TR2 base being just 20mV lower than diff pair TR1 base. This 20mV difference will deliver the current sharing described above.

This should be quite close to reality assuming the transistor HFE is reasonable in TR3 and I've read all the resistor values correctly in the schematic. But 20mV across the TR1 and TR2 diffpair bases seems really small! Hope it's all OK. It looks right at first glance....

[G0HZU_JMR]

The other thing to consider is the efficiency of the diode detector. The compensation diodes take out the Vf/temperature issue in the detector but I don't think a weakly biased Schottky diode detector is going to be very efficient in this circuit. That's why I think there won't be as much difference as might be expected between the DC voltage at node R1/R2 and the DC voltage at R6/R7 node when the ALC is healthy and set for 50mV rms RF. This voltage difference is set by R5?

I also think you will have to make differential voltage measurements using a decent high impedance DVM at the diffamp rather than any classic shunt measurements reference to ground. This is because a typical x10 scope probe or a typical 10Meg DMM is going to heavily load (and corrupt?)the circuit if you tried measuring at TR1 or TR2 base with a DMM. I think you would have to measure across the bases to see the 20mV difference? Don't try measuring TR1 base (to chassis) and then TR2 base to chassis and work out the difference as you will corrupt the circuit. 12.8V and 10Meg Ohm is 1.3uA of leakage/bleed to chassis which is going to be highly significant in this part of the circuit and will lead to confusing and false measurements.

[paulfifties]

Some DC voltage values measured with no RF input, i.e. A12 pin 14 desoldered:
A17 D2 anode : 11,83V
A17 D1 cathode : 12,67 V
R3-C1 : 18,98 V
A16 D2 anode : 12,741 V
A16 D1 cathode : 12,747 V --> only 6 mV over the 2 diodes...
TR1e - TR2e : 13,337 V
R9 power supply side : 18,992 V (should correspond to 257 µA through R9)
TR3c : 13,278 V
TR3b : 0,114 V

I will post the comparative measurements of different Schottky diodes (incl the original ones) in a next reply...

Paul

[G0HZU_JMR]

As a matter of fact, I have simulated the ALC circuit on a breadboard

It would be much easier to simulate this circuit on a SPICE simulator. I've just simulated the diff amp section and it agrees with the spreadsheet for a 80uA vs 170uA current share with 20mV differential voltage across the bases. You can see that with this current share, the TR3/Q3 transistor is operating at a few mA collector current as its collector is at 15V. 4V drop means 4mA Ic. See below:

I could probably simulate the whole detector and ALC system in SPICE, but I might not be able to find an accurate SPICE model for the detector diodes.

[G0HZU_JMR]

I managed to get it working using a SPICE model of a classic 1N5711 Schottky diode.

I set up the simulation such that the ALC loop levelled a JFET buffer amplifier to 50mV rms. It seems to settle correctly and the youtube video below shows all the test point voltages and some currents as well.

You can see that the 20mV differential voltage at the diffamp bases seems to be right although this depends on the HFE of the TR3 transistor to some degree. The real generator might see slightly more or less than 20mV here.

Note that I did have to speed up the ALC by drastically reducing the time constant caps in the circuit but it still settles quite stably even though it does it in just a few milliseconds. The real circuit will be much MUCH slower than this.

Sorry, there's no sound. I really need to sort out my sound card wrt Camstudio and get a proper microphone as well. Hope the video is useful even if there still may be a few bugs in this analysis because I'm still unsure I'm reading the Marconi circuits cleanly.

https://youtu.be/eeXyn8QyLkE

The video above is best watched in 720p.

[G0HZU_JMR]

To back up some of the theory and show that the ALC in the simulation really is working I did another youtube video with it being driven much harder in terms of unlevelled RF from the JFET amplifier. But this circuit still manages to level it to about 60mV rms. It's a bit higher in level than before but not much.

You can see that the RF levelling overshoots but still settles to be quite similar to before. But because there is much more RF drive to level the ALC control voltage at TR3 collector is a bit higher at 8V as it has to bias back the JFET amplifier a bit more. Also note that the differential voltage across the diffpair bases is now 25mV (was 20mV)

This change in diff voltage across the bases is because TR3 needs less drive now to generate that 8V ALC control voltage so the current share in TR1 is now down to just 70uA as you can see in the video.

So a quick play with the excel spreadsheet shows that for 70uA in TR1 the differential voltage across the bases needs to be.... 25mV So this is expected.

But you can see how much harder the ALC has to work to level in this second video? Maybe this video is better but it does show that the ALC works reasonably stiffly over a big change in drive level.

https://youtu.be/9MLWeB9xZgI

Again, best watched in 720p

[G0HZU_JMR]

Some DC voltage values measured with no RF input, i.e. A12 pin 14 desoldered:
A17 D2 anode : 11,83V
A17 D1 cathode : 12,67 V
R3-C1 : 18,98 V
A16 D2 anode : 12,741 V
A16 D1 cathode : 12,747 V --> only 6 mV over the 2 diodes...
TR1e - TR2e : 13,337 V
R9 power supply side : 18,992 V (should correspond to 257 µA through R9)
TR3c : 13,278 V
TR3b : 0,114 V

I think you have to make differential measurements if you want to work out the voltage at A17 D2 anode. i.e. measure the voltage when you connect your DMM across A17 D2 anode and A17 D1 cathode.

I'm expecting you to see ballpark 100mV differential voltage here with no RF. For example, the voltage at A17 D2 anode should be about 100mV or so higher than D1 cathode with no RF. You can then reliably measure the absolute voltage at A17 D1 cathode with respect to chassis with your DVM and then work out what voltage 'must' be at D2 anode by adding the (100mV?) differential voltage you measured. Also, use an isolated handheld DVM for the differential measurement across D2 anode to D1 cathode and not a bench DVM?

Otherwise you are going to be confused by false results. Putting a typical 10M ohm DMM from A17 D2 anode to ground will upset the circuit because there will be a bleed current to ground. If you tried to measure D2 anode voltage wrt chassis with a DMM on a healthy system I think it would cause some serious issues.

I would expect that 12V/10e6R A will flow in the DMM itself as leakage current = 1.2uA extra emitter current in TR1. With an HFE of several hundred TR1 will grab all of the diffamp current and so TR3 will be turned on enough to unbalance the system fully one way.

Besides, if you disconnect the DMM from the system, TR3 should be being turn ON by the diffamp if there is no RF. This is because it should be trying to bias the oscillator on A12 as hard as possible. But your results suggest that TR3 is OFF. So your results above look very suspicious to me.

Whatever you do don't leave one or more DVM connections in situ at A17 D2 anode or at A16 D2 anode whilst you probe elsewhere. The best advice I can offer is don't try measuring A16 D2 anode or A17 D2 anode with a DVM with respect to chassis at any time because the DMM will load the circuit and change the status of the ALC system.

[G0HZU_JMR]

A17 D2 anode : 11,83V
A17 D1 cathode : 12,67 V

I tried simulating the effect of a 10Meg Ohm DMM at A17 D2 anode on my simulation.

I get:

A17 D2 anode : 11.800V
A17 D1 cathode : 12.785 V

very similar to your results. But if I remove the 10Meg leakage path and measure with a voltage probe I get:

A17 D2 anode : 12.789V
A17 D1 cathode : 12.867 V

This is a difference of about 80mV. The voltage at the emitter of TR1 and TR2 should be about 13.4V and you measure this correctly. So for the D2 and D1 diodes to have a tiny forward bias it makes sense that the base voltage of TR1 will be about 0.6V lower than this at 12.8V. This is what the simulation shows.


One other thing to note here is that the alternative schematic I have here has Marconi's DC voltmeter test voltages on it and some of these measurements have been made with the same problem. They have used a meter that loads the circuit giving the wrong (running) voltages on the circuit. Their measurements have a black dot next to them indicating that the result is meter range sensitive.

But they get 13.4V on the emitter and just 8.4V on the base of PNP TR1 (which is the same connection as D2 anode). Obviously, this is a daft measurement as there should only be typically 0.7V across a forward biased diode. But their schematic suggests there are about 5 Volts dropped here! Totally wrong and potentially misleading.

[paulfifties]

Jeremy,

You're absolutely right regarding the A17 D2 anode (which is same as TR1 base) !
I had very little time yesterday, but I managed to put a 50 mV RMS signal of 12 MHz on the desoldered input of A16 (connection to A12 pin 14).
Actually, I soldered a little extension wire to this mini coax cable end, and connected that directly to the 50 Ohm RF output of my 12MHz Philips generator.

I noticed the following:
1) at this 50mV 12MHz RF level, all values measured were nearly identical to the values measured with no RF input, as listed previously.
2) though the DC-voltage measured with respect to GND (with a Brymen) were rock stable at A17 D1 cathode, the DC-voltage measured at A17 D2 anode was indeed drifting slowly, but definitely !
This weekend I hope to find the time to make the differential measurements, as you suggested, and come back with the results.
In addition, I was hoping to create some comparative measurements with the original Schottky diode that looks undamaged, the cracked one, and some candidate replacements (BAT83S and a different totally different Schottky diode type).
I really hope I can get away without having to replace the A16 D1 and D2, as I really don't see how remove the A16 board without having to unsolder a lot of critically short connections...
Regards, Paul