Keyed power oscillator frequency drift, why and how?

[neazoi]

One way would be to compensate the drop in frequency, with a varactor diode that increases the frequency over 5 seconds. One way would be to feed the keyed Vcc via a 1M and a 10K to the cathode of a varactor who's anode is earthed. Feed the cathode to the base via a 50 PF capacitor. Decouple the junction of the two resistors with a 4 MF capacitor. The theory is that when keyed the varactor has no volts on it so it has maximum capacitance, after a couple of seconds the volts across it rises and its capacitance falls, so hopefully raising the crystal frequency to offset the fall in frequency caused by your present circuit.
Frank

Very interesting idea!
You use the capacitor charge on the varactor to compensate for the frequency drift. I have noticed that with different crystals, the amount and time of drift is mostly the same (which makes me skeptical about if the crystal is the source of the problem), so I suspect if the capacitor is carefully chosen this will do the trick. I also believe this would compensate just right, because the capacitor has an exponential charging curve and the frequency change is exponential as well.

The only thing that I do not understand is this "Feed the cathode to the base via a 50 PF capacitor". Will that change the frequency? I thought that this must be done in a point in the feedback resonant circuit, not directly on the transistor base?

 

[FvM]

One way would be to compensate the drop in frequency, with a varactor diode that increases the frequency over 5 seconds.

I believe it's possible somehow. Related to the problem of designing a simple power oscillator, it sounds suspiciously like inventing a square wheel...

 

[neazoi]

I believe it's possible somehow. Related to the problem of designing a simple power oscillator, it sounds suspiciously like inventing a square wheel...

Yes, but it would be an interesting experiment to see if this can be done.
As said, keyed oscillators are not good designs, we take that for granted, but they are simple. How can we make such a circuit better, that is the idea of experimenting.

Again, the relay/dummy load seems a simpler approach that uses less components and more straight forward as the varactor and capacitor curves do not play any role, but the resistor required for the dummy load is huge
I wonder if I could use a ceramic wire power resistor for the dummy load. I know it is inductive but who cares in that application?We are not doing any measurements on it we just want to dissipate power. Am I right?

 

[volker@muehlhaus]

Again, the relay/dummy load seems a simpler approach

At the switching moment, there is no load on the oscillator, because the relay is open between the contacts. I think this will cause the frequency to move around (chirp).

I agree to FvM that it's a square wheel design.

 

[neazoi]

At the switching moment, there is no load on the oscillator, because the relay is open between the contacts. I think this will cause the frequency to move around (chirp).

This is a point. I will have to experiment on that to verify this, I will do so.
Does it matter too much in this application if the resistor is wirewound?

 

[volker@muehlhaus]

Does it matter too much in this application if the resistor is wirewound?

That will add some inductance, likely not a problem at a few MHz. Check the VSWR and you will see if it's good enough.

 

[chuckey]

Square wheel?, I think its a historical hang over. in the 1930s, you could get a single valve and your expensive crystal to generate up to 25W of CW. To add a oscillator, would involve a bigger chassis, more holes to be drilled, trying to keep the whole thing stable.. it was a big deal. Some one has just updated the cheopo arrangement with a power transistor. The addition of a continuously running oscillator would cost peanuts and occupy a few square inches of PCB.
Frank

 

[biff44]

thermal effects in the transistor and varactor. Use a power transistor with a big flange, to get small thermal resistance, and just throttle it back like it was a small signal transistor. The varactor, decouple it as much as you can.

Why not leave it all DC powered, but just snub the RF signal so it stops oscillating.

 

[neazoi]

thermal effects in the transistor and varactor. Use a power transistor with a big flange, to get small thermal resistance, and just throttle it back like it was a small signal transistor. The varactor, decouple it as much as you can.

Why not leave it all DC powered, but just snub the RF signal so it stops oscillating.

I did so. I used the 2sc2166, which is quite powerfull. However I run it harder (4W output).

I also left the oscillator continuously running and added a relay to switch the oscillator output between the antenna (50R load) and a second 50R dummy load.
The instability described in post #1 dissapeared (as it was expected). However at the time there the switching is done, a quick frequency drift can be noticed, as predicted on post #24. This drift is much quicker in duration (<0.5 seconds or so), but it's frequency is larger and so an unpleasant tone is heard everytime the relay is closed.

Not a good solution as initially thought it to be...

Now I am flerting with the idea of the varactor described previously. However this will require 5 more components, which could otherwise make a small oscillator prior to the power transistor.

It would be interesting to see is a powerful QRP transmitter with good stability, could be implemented using so few components.

 

[BradtheRad]

This drift is much quicker in duration (<0.5 seconds or so), but it's frequency is larger and so an unpleasant tone is heard everytime the relay is closed.

Somewhere I read this is how morse enthusiasts recognize a keyer with a homebrew power supply. The sudden current draw causes a dip in supply voltage, which in turn causes the frequency to drop. It is distracting and causes difficulty in interpreting the dots and dashes.

The article said tighter voltage regulation to the oscillator is one way to cure it.

Possibly separate power supplies, one to the oscillator and one to the transmitter.

 

[neazoi]

Somewhere I read this is how morse enthusiasts recognize a keyer with a homebrew power supply. The sudden current draw causes a dip in supply voltage, which in turn causes the frequency to drop. It is distracting and causes difficulty in interpreting the dots and dashes.

The article said tighter voltage regulation to the oscillator is one way to cure it.

Possibly separate power supplies, one to the oscillator and one to the transmitter.

The oscillator IS the transmitter itself. See post #1.
I suspect you might be correct though about the power supply.
I will try a separate power supply, eg an lm58xx to see how it goes and let you know.

- - - Updated - - -

The oscillator IS the transmitter itself. See post #1.
I suspect you might be correct though about the power supply.
I will try a separate power supply, eg an lm58xx to see how it goes and let you know.

I just tested it. Not it is not the PSU.

 

[neazoi]

Yes, something like this.

Consider that the transistor power disspation may be different without oscillations, the bias circuit may need to be switched, too.

Hi, how can I get more power out of this oscillator?
The greater power so far is 4W at 22v with a 2sc2166.
I have also tried the much more powerful 2sc3133 but I cannot get more than 4W again.
Any ideas?

 

[chuckey]

The absolute power output is limited by the Vcc and the reflected load impedance. i.e. If the oscillator transistor is switching hard between ON and OFF, then changing the transistor type will not increase the power output. So more Vcc or a lower load impedance for the transistor (lets it take more current).
Frank

 

[neazoi]

The absolute power output is limited by the Vcc and the reflected load impedance. i.e. If the oscillator transistor is switching hard between ON and OFF, then changing the transistor type will not increase the power output. So more Vcc or a lower load impedance for the transistor (lets it take more current).
Frank

Wouldn't it be able to take more current by changing the base resistor too?

Saying so, I have tried 8.2k and 12k instead of the nominal 10k for the base resistor, but I did not notice any significant improvement in output power, just a few 10s of mW. Maybe it is that bias diode that limits this?

Also I have tried to increase the voltage from 22v to 24v with no signifficant improvement in output power as well.

Any advice to try would be appreciated!

 

[neazoi]

Hey, I have tried to increase the voltage from 24v to 29v. Th output is boosted to 8W at 50R. I have an LPF at the output so this measured output is pretty much the power of the carrier. This has been tested with a 2sc2166.
I wonder how much can I increase the voltage on this transistor more?
Also what about the more powerful 2sc3133?

 

[D.A.(Tony)Stewart]

The tempco of resistance of copper is approximately +3930ppm/K at room temperature, depending on how pure it is and upon annealing.

But I found the inductance has a negative tempco as I recall, raising the f with T.

Thus large positive ceramic caps are required, which don't exist. My guess you have large negative tempco caps causing your problem or you are heating your crystal will may have a -50 ppm swing with rising temp up to ~60'C then go positive.

Most ceramic caps have a large negative Tempco while few are more than P200 (+200ppm/'C) and zero tempco caps are recommended or NPzero (NP0) often mislabelled as NP-oh.

What cap types are you using? NP0 or general purpose? NP0 are also called C0G type material

Xtals will be damaged with more than their rated 50 uW of power. So it wont ever be hot, but should be insulated from heat. ( move heat far away or use a Foam sleeve)

So what is your PPM/deg'C slope ... make a guess then measure it.

 

[neazoi]

The tempco of resistance of copper is approximately +3930ppm/K at room temperature, depending on how pure it is and upon annealing.

But I found the inductance has a negative tempco as I recall, raising the f with T.

Thus large positive ceramic caps are required, which don't exist. My guess you have large negative tempco caps causing your problem or you are heating your crystal will may have a -50 ppm swing with rising temp up to ~60'C then go positive.

Most ceramic caps have a large negative Tempco while few are more than P200 (+200ppm/'C) and zero tempco caps are recommended or NPzero (NP0) often mislabelled as NP-oh.

What cap types are you using? NP0 or general purpose? NP0 are also called C0G type material

Xtals will be damaged with more than their rated 50 uW of power. So it wont ever be hot, but should be insulated from heat. ( move heat far away or use a Foam sleeve)

So what is your PPM/deg'C slope ... make a guess then measure it.

All the capacitors I have used in the prototype are Silver mica 1%. As far as I know silver mica caps excibit a small positive temperature coefficient, but this is quite small.

NP0 caps have theoretically zero tempco, but they have also larger losses and higher Q than the silver micas, except if expensive hi-Q ceramics are used (which are even harder than silver micas to find).

The cores I have used have a very low (~32ppm/C) positive tempco.
The crystal does not to get heated at all, at least it's metal case body. I do not know what is hapenning inside the crystal chip, but the body seems cool even at 8W of output power. The tuned LC at the feedback plays a role on that, in the sense that it limits the feedback signal through the crystal. However I would expect the crystal to heat up (even the case) at 8W, but surprizingly, this is not the case.

Because I suspect the frequency drifting at the beginning of each transmission is a thermal issue, the problem with temperature compensation, is that the compensating components must heat up to affect compensation. But we cannot remove the case of the crystal neither the transistor, and put a compensating component onto the chips directly.

 

[chuckey]

when the transistor is fully on the current it takes is Vcc /ZL where Zl is the load impedance. Changing the base bias, just increases the DC current and not the RF current.
Frank

 

[neazoi]

when the transistor is fully on the current it takes is Vcc /ZL where Zl is the load impedance. Changing the base bias, just increases the DC current and not the RF current.
Frank

Thanks, this is useful to know that it does not depent on the base resistor.

 

[chuckey]

Its based on tuning old fashioned valve PAs. You first resonate the anode circuit, which reduces the anode current, you then increase the coupling to the load, which increases the anode current. This might have effected the tuning of the anode circuit, so you dip the current with the tuning, then raise it again with loading control until you get to a maximum (or required) power output. (anode~ collector )
Frank