Cheap stable VFO design

[StillTrying]

The diode is shorted by the inductor, and the diode is shorting the inductor during half the cycle...

Hi
.... as mentioned above, you have a DC short to ground across your tuning diode. ...

FWIW, I agree, is there a mistake in the oscillator part of the diagram?

[MrSlack]

It's not shorted. The diode doesn't conduct until Vforward is above ~1.1v on a 1n4002 and it is seeing about 600mV p-p. Also the diode doesn't get turned on because it can't switch at 7MHz because it's a basic rectifier. The only effect we see here is capacitance from the size of the depletion region. Also the voltage source for the diode is reverse biasing it so it doesn't conduct.

So no the thing is right.

[uncle_bob]

It's not shorted. The diode doesn't conduct until Vforward is above ~1.1v on a 1n4002 and it is seeing about 600mV p-p. Also the diode doesn't get turned on because it can't switch at 7MHz because it's a basic rectifier. The only effect we see here is capacitance from the size of the depletion region. Also the voltage source for the diode is reverse biasing it so it doesn't conduct.

So no the thing is right.

Hi

Pardon me, but as drawn, the coil clamps the diode at a DC voltage of zero. You can not bias the diode in that configuration. You can not tune the VCO with the tuning diode shorted at DC. It can have interesting AC voltage on it, but it's DC must be zero.

Bob

[StillTrying]

Pssst. If there's no one watching I'm still with you uncle_bob.

At first I thought it was tuning by stretching the negative peaks a few mV into the diode's conduction, - in sensitive parts silicon diodes start conducting at about 330mV for me. Then I saw that these diode's revV/cap graph is quite steep near 0V. But now I think most of the tuning is done by the changing characteristics of the tuning voltage's emitter follower. But I'm not at all good with analogue RF stuff!**

But it works somehow, - LT version oscillates at least.

**I came across this thread while searching to see if anyone had used a diode to edge a cheap 100ppm or 50ppm xtal a little bit closer.

[MrSlack]

I'll pop a scope on this later and show you what it's doing. The diode doesn't behave like a diode here at all.

You couldn't put a signal diode in here - it would behave differently.

[GK]

The inductance value of that torodial inductor would be sensitive to the amount of DC current flowing through it. I have an old spectrum analyser that frequency sweeps the 2nd (LC) LO this way.

[G0HZU]

The inductance value of that torodial inductor would be sensitive to the amount of DC current flowing through it. I have an old spectrum analyser that frequency sweeps the 2nd (LC) LO this way.

In this case I don't think a few mA of current will affect the inductance of that T50-6 inductor significantly (as in enough to tune the VFO across a band) because it has a powdered iron core but it's not really a wise thing to do in a circuit like this. So the DC current should be avoided.

The overall circuit is poor.  There's a 1k resistor damping the Q of the inductor for a start. It should be possible to make a basic narrowband VFO here with phase noise in the ballpark of -155dBc/Hz at 10kHz offset.

I'd expect that design to be maybe 20dB worse than this and I agree that the varactor diode is shorted at DC by the inductor. Maybe it does still tune over a short range but this isn't a good VFO circuit IMO.

[GK]

The inductance value of that torodial inductor would be sensitive to the amount of DC current flowing through it. I have an old spectrum analyser that frequency sweeps the 2nd (LC) LO this way.

In this case I don't think a few mA of current will affect the inductance of that T50-6 inductor significantly (as in enough to tune the VFO across a band) because it has a powdered iron core

Well the "varactor" diode has no significant DC voltage impressed across it, so what do you propose? The OP should remove the diode altogether and report on the results.
 
 

[G0HZU]

The inductance value of that torodial inductor would be sensitive to the amount of DC current flowing through it. I have an old spectrum analyser that frequency sweeps the 2nd (LC) LO this way.

In this case I don't think a few mA of current will affect the inductance of that T50-6 inductor significantly (as in enough to tune the VFO across a band) because it has a powdered iron core

Well the "varactor" diode has no significant DC voltage impressed across it, so what do you propose? The OP should remove the diode altogether and report on the results.

TBH I'm not sure how/if the circuit will tune in its current state. It just looks bad from a design point of view and it kind of hurts just to look at the circuit. I wouldn't expect to see the inductance change (at 7MHz) with a few mA passing through the coil.

If it still tunes without the varactor then there are other possibilities that cause the 'tuning'. eg the loading changes caused by the emitter follower? I'd be very surprised if the inductance of that coil changes (enough to cause band tuning) with a few mA DC. If it does then maybe it isn't a genuine T50-6 core? I'd expect to see so little change in inductance it would be hard to measure.

[MrSlack]

Some notes:

1. It doesn't tune without the diode. That was the intial state of the oscillator. It sits around 7.15MHz.
2. The 1k resistor was a 22k trimmer to start with. This was reduced until the circuit oscillated. Initially it didn't, I suspect because the reverse bias wasn't low enough impedance to overcome the diode's action and allow the oscillator to start.
3. It is a genuine T50-6 core.
4. Tuning isn't very linear. At the bottom end of the tuning range ~6.95MHz it compresses. This is the price of using a 1n4002 instead of a proper varactor diode. If I could be bothered to plot capacitance vs voltage I could pick a more linear region to operate in but this will not result in the tuning range reqiured.
5. The emitter current on the transistor is stable at all frequencies so this isn't loading. This is deduced from reading the voltage across the resistor with a scope.

For ref, this is purely an experiment to ask the following questions:

1. Can you build a VFO with ridiculously cheap parts that everyone has (the T50-6 will be eliminated as well)? Apparently yes.
2. Is it stable enough to be able to tune CW? Apparently yes.
3. Would I be happy putting this near a transmitter? Hell no.

I can only ask people to review the outcome not the design. The design is horrid and makes me cringe. Purely an exercise in minimalism. It started with op amps, Infineon varactor diodes, FETs, buffers, filters, power supply regulators and evolved towards zero.

[GK]

Some notes:

1. It doesn't tune without the diode. That was the intial state of the oscillator. It sits around 7.15MHz.

If it doesn't tune without the diode then that also rules out the suggestion that "loading changes caused by the emitter follower" are responsible.

With zero DC bias the 1n4002 probably has a lot of junction capacitance; whether that changes enough to tune the oscillator over the observed range with a variation or maybe a mV or a fraction of a mV effective variable DC bias is another question.

[MrSlack]

I measured the 1n4002 with an LCR meter standalone outside the oscillator and the numbers add up. I don't have my notes handy to check this at the moment as they are at home.

I'm going to stick a decoupling capacitor from the cathode of the diode to the rest of the oscillator in line with other designs. That will eliminate the DC bias current from the picture and implicate only the diode in the tuning - how does that sound?

[G0HZU]

As a side note, you could probably tune the T50-6 core 'thermally' if you passed a lot of DC current through the winding. This core material is fairly stable over temperature but you probably only need to pass 10mA through it to heat it slightly. That would cause some frequency drift in a VFO but probably much less than 1kHz (at 7MHz) and the response time would be very slow

[uncle_bob]

I measured the 1n4002 with an LCR meter standalone outside the oscillator and the numbers add up. I don't have my notes handy to check this at the moment as they are at home.

I'm going to stick a decoupling capacitor from the cathode of the diode to the rest of the oscillator in line with other designs. That will eliminate the DC bias current from the picture and implicate only the diode in the tuning - how does that sound?

Hi

A blocking cap sounds like a real good idea.

To get reasonable Q out of the diode, you want it reverse biassed.

Bob

[MrSlack]

Will sort out this evening. Approximately 8 hours away for me

[GK]

I measured the 1n4002 with an LCR meter standalone outside the oscillator and the numbers add up. I don't have my notes handy to check this at the moment as they are at home.

I'm not sure that the numbers do stack up. Here is the capacitance versus reverse-bias plot for the 1N4002:



It's a pity the graph doesn't go all the way down to zero volts, but unless something seriously funky is going on <0.1V I'm at a loss to explain how a <=mV or so of variation could cause enough of a capacitance change to result in the frequency tuning range that you are reporting.

[uncle_bob]

I measured the 1n4002 with an LCR meter standalone outside the oscillator and the numbers add up. I don't have my notes handy to check this at the moment as they are at home.

I'm not sure that the number do stack up. Here is the capacitance versus reverse-bias plot for the 1N4002:



It's a pity the graph doesn't go all the way down to zero volts, but unless something seriously funky is going on <0.1V I'm at a loss to explain how a <=mV or so of variation could cause enough of a capacitance change to result in the frequency tuning range that you are reporting.


In any case the 1N4001 looks like a poor varactor diode for a conventional 40m oscillator. A reverse bias span of 2V to 20V only yields a 9pF variation in capacitance.

Hi

As soon as you forward bias the diode, that curve changes a lot. Since an LCR meter (or an oscillator) could be putting several volts p-p on the diode, you may start to "see" forward bias at the low reverse bias levels.

One way to look at it:

By the time the diode gets to 0.7V forward bias, it's basically a short RF wise. As it approaches the "zero ohm" point, it's capacitance goes up quite a bit.

Bob

[GK]

Hi

As soon as you forward bias the diode, that curve changes a lot. Since an LCR meter (or an oscillator) could be putting several volts p-p on the diode, you may start to "see" forward bias at the low reverse bias levels.

One way to look at it:

By the time the diode gets to 0.7V forward bias, it's basically a short RF wise. As it approaches the "zero ohm" point, it's capacitance goes up quite a bit.

Bob

The junction capacitance does typically rise at an accelerated rate as 0V DC bias is approached, but the slope in this case would have to be ridiculously steep to effect the capacitance variation required to tune this oscillator with the limited range of variable DC bias that the OP must be getting with the diode shorted by the DCR of the inductor.

[uncle_bob]

Hi

As soon as you forward bias the diode, that curve changes a lot. Since an LCR meter (or an oscillator) could be putting several volts p-p on the diode, you may start to "see" forward bias at the low reverse bias levels.

One way to look at it:

By the time the diode gets to 0.7V forward bias, it's basically a short RF wise. As it approaches the "zero ohm" point, it's capacitance goes up quite a bit.

Bob

The junction capacitance does typically rise at an accelerated rate as 0V DC bias is approached, but the slope in this case would have to be ridiculously steep to effect the capacitance variation required to tune this oscillator with the limited range of variable DC bias that the OP must be getting with the diode shorted by the DCR of the inductor.

Hi

.... which is why several of us have been pointing out the short circuit issue. Of course, there are many ways to get a circuit to change frequency.

Bob

[GK]

Hi

As soon as you forward bias the diode, that curve changes a lot. Since an LCR meter (or an oscillator) could be putting several volts p-p on the diode, you may start to "see" forward bias at the low reverse bias levels.

One way to look at it:

By the time the diode gets to 0.7V forward bias, it's basically a short RF wise. As it approaches the "zero ohm" point, it's capacitance goes up quite a bit.

Bob

The junction capacitance does typically rise at an accelerated rate as 0V DC bias is approached, but the slope in this case would have to be ridiculously steep to effect the capacitance variation required to tune this oscillator with the limited range of variable DC bias that the OP must be getting with the diode shorted by the DCR of the inductor.

Hi

.... which is why several of us have been pointing out the short circuit issue. Of course, there are many ways to get a circuit to change frequency.

Bob

You are talking about something completely different now. That there are numerous things wrong and sub-optimal in this design is beyond question. Precisely how the described circuit can function exactly as claimed is what I am trying to figure out.
     

[uncle_bob]

Hi

As soon as you forward bias the diode, that curve changes a lot. Since an LCR meter (or an oscillator) could be putting several volts p-p on the diode, you may start to "see" forward bias at the low reverse bias levels.

One way to look at it:

By the time the diode gets to 0.7V forward bias, it's basically a short RF wise. As it approaches the "zero ohm" point, it's capacitance goes up quite a bit.

Bob

The junction capacitance does typically rise at an accelerated rate as 0V DC bias is approached, but the slope in this case would have to be ridiculously steep to effect the capacitance variation required to tune this oscillator with the limited range of variable DC bias that the OP must be getting with the diode shorted by the DCR of the inductor.

Hi

.... which is why several of us have been pointing out the short circuit issue. Of course, there are many ways to get a circuit to change frequency.

Bob

You are talking about something completely different now. That there are numerous things wrong and sub-optimal in this design is beyond question. Precisely how the described circuit can function exactly as claimed is what I am trying to figure out.
   

Hi

Well, that's not exactly the point of the DC short either.

How can it work:

1) The schematic and the circuit are not the same

2) The parts on the breadboard are not what they are "supposed to be".

3) The measurements are in error.

That's a short list.

Bob

[GK]

Hi

Well, that's not exactly the point of the DC short either.

How can it work:

1) The schematic and the circuit are not the same

2) The parts on the breadboard are not what they are "supposed to be".

3) The measurements are in error.

That's a short list.

Bob

Sherlock Holmes I presume?

[StillTrying]

A few random notes from last nights simulations on the off-chance someone finds them interesting, or even useful. - I simulated(on and off) for over 2 hours so you don't have to. LOL

Despite it only going down to 0.1V, GK's 1N4002 junction capacitance/reverse voltage graph above is very close to what I 'measured' in LT. A 1Amp diode starts off at 35pf at 0V, and applying a real 1V to 8.5V tuning voltage gives only a 2:1 change in capacitance.
LT's variactor starts at ~52pF, - still with a 2:1 range.

I 'measured' the tuning diodes capacitances by probing the 5.4Mhz current through it and it's parallel 30pF, and compared the 2 pk-pk currents ratio.

In the simulation if I fix 'the short' so that the tuning voltage of 1V to 8.5V gets across the tuning diode it makes very little difference to the frequency or its full range - it's still about 5.4Mhz with about a 250Khz change.

You'll like this next one.

Now knowing the tuning diodes capacitances, if I swap out LT's variactor diode for a fixed 50pF with ESR 10R does the frequency still change with the 1 to 8.5V? - Yes! But in the opposite direction - at 8.5V it drops about 1Mhz.

Just for some colour, using LT's variactor, no short, green =output, red =tuning voltage, blue = emitter-collector.

[G0HZU]

Obviously the impedance at the emitter of the 'tuning' transistor/buffer will change with changes in bias. At very low bias voltages the impedance at the emitter will be much higher and the (potentially severe) damping impact of the 1K resistor will be less because there is a high impedance at its cold (emitter) end. So there should be more RF voltage developed across the T50-6 inductor at low tuning voltages because of the reduced damping. But I think the diode/transistor may interact as a form of crude RF clamp here with the side effect of changing the RF frequency downwards as this ALC/clamping happens. So maybe you don't 'see' this increased RF voltage because the diode is doing something in the form of ALC/limiting. Maybe the side effect of this ALC process is how it appears to tune with changes in tuning voltage (at the emitter) despite the diode being shorted at DC by the inductor? The diode will be a different animal if this clamping/limiting is happening and the limiting probably only happens strongly/significantly at low tuning voltages?

Ideally, the emitter node should be bypassed at RF so maybe some of the tuning strangeness will go away when there is a decent RF decoupling cap across the 10k resistor in the emitter? This would clamp this node at RF and this would define the damping effect of the 1k resistor across all tuning voltages and frequencies. A 1k resistor across the inductor is going to damp the loaded Q of the system quite a bit and this will degrade the phase noise and drift/stability.

The circuit just looks dodgy to me... I'd bin it and start again with a JFET and connect the tuning diode(s) differently to prevent clamping and also aim for a higher loaded Q (I.e. less damping of the main resonator!).

[uncle_bob]

A few random notes from last nights simulations on the off-chance someone finds them interesting, or even useful. - I simulated(on and off) for over 2 hours so you don't have to. LOL

Despite it only going down to 0.1V, GK's 1N4002 junction capacitance/reverse voltage graph above is very close to what I 'measured' in LT. A 1Amp diode starts off at 35pf at 0V, and applying a real 1V to 8.5V tuning voltage gives only a 2:1 change in capacitance.
LT's variactor starts at ~52pF, - still with a 2:1 range.

I 'measured' the tuning diodes capacitances by probing the 5.4Mhz current through it and it's parallel 30pF, and compared the 2 pk-pk currents ratio.

In the simulation if I fix 'the short' so that the tuning voltage of 1V to 8.5V gets across the tuning diode it makes very little difference to the frequency or its full range - it's still about 5.4Mhz with about a 250Khz change.

You'll like this next one.

Now knowing the tuning diodes capacitances, if I swap out LT's variactor diode for a fixed 50pF with ESR 10R does the frequency still change with the 1 to 8.5V? - Yes! But in the opposite direction - at 8.5V it drops about 1Mhz.

Just for some colour, using LT's variactor, no short, green =output, red =tuning voltage, blue = emitter-collector.

Hi

So what does the schematic look like now?

Bob