Heatsinked RF transistor in oscillator caused frequency drift

Hi I have made a power oscillator for HF using an RF transistor that is mounted onto a metal enclusure that is used as a heatsink using mica insulator to insulate it from the enclosure (fin connected to collector)
The emitter of the transistor is connected to ground and the metal enclosure is not connected to ground. When I touch the metal enclosure with my hand frequency of oscillator changes a bit.
Why does it happen?
Is it some kind of capacity that my hand adds? But the transistor is insulated from the metal enclosure.

By the way, if I do connect the metal enclosure to GND, the oscillator signal is a bit more distorted. The only logical explanation I can give in this second case, is that the mica forms a capacitor from collector to emitter (ground).

It's likely due to parasitic capacitance, but could possibly be due to temperature change as well. Try blowing on the heatsink and see if it produces the same drift.

If it's capacitance, then the best solution may be to put a thin RF screen between the transistor and the heatsink (with insulation on both sides), and connect the screen to circuit ground. This will detune the circuit somewhat, but the effect will be much more consistent and you can adjust the tuning back where you want.

It is also good to earth all bits of metal or at least decouple them so they are at RF earth. One way round your problem is to rearrange the circuit so the collector and heat sink are earthed, this would mean running in a -ve Vcc.
frank

chuckey said:

It is also good to earth all bits of metal or at least decouple them so they are at RF earth. One way round your problem is to rearrange the circuit so the collector and heat sink are earthed, this would mean running in a -ve Vcc.
frank

Click to expand...

I could use 2sc3133 instead of 2sc2166, that has the fin connected to emitter, which is connected to ground to my circuit. However the 2sc2166 has more desirable results in my circuit.
I just blown a 2sc3133, so I won't be able to test this before a new one arrives.

- - - Updated - - -

mtwieg said:

It's likely due to parasitic capacitance, but could possibly be due to temperature change as well. Try blowing on the heatsink and see if it produces the same drift.

If it's capacitance, then the best solution may be to put a thin RF screen between the transistor and the heatsink (with insulation on both sides), and connect the screen to circuit ground. This will detune the circuit somewhat, but the effect will be much more consistent and you can adjust the tuning back where you want.

Click to expand...

I think it is capacitance because the change is instant.
Can you provide me more information of how to make this RF shield? Now I just use a mica insulator. Maybe these rubber insulators have less capacitance?

neazoi said:

I could use 2sc3133 instead of 2sc2166, that has the fin connected to emitter, which is connected to ground to my circuit. However the 2sc2166 has more desirable results in my circuit.
I just blown a 2sc3133, so I won't be able to test this before a new one arrives.

- - - Updated - - -



I think it is capacitance because the change is instant.
Can you provide me more information of how to make this RF shield? Now I just use a mica insulator. Maybe these rubber insulators have less capacitance?

Click to expand...

What you observe is a typical amateur poor design fact. If frequency stability is important, never try to use a power oscillator. Instead build a low-power ( like 10 mW) oscillator with a high-Q LC resonator (possibly use a quartz), then a buffer amplifier feeding a power amplifier.
Good oscillators are built in a good metal enclosure and a good filters are needed for DC power input.

With any open structure the frequency is affected by touching components or moving wires around.

Totally agree. The rule of thumb designing ANY oscillator is to have minimum power possible. This would help for all of the oscillator characteristics, as frequency stability, phase noise, harmonics, etc.
After that, you can amplify the oscillator signal however you want and get any power you need.

... which takes us full circle to the start of Neazoi's other thread about RF power oscillators. Short answer is don't do it!

I did see an interesting article a while ago (I've lost the reference to it) which used an optocoupler as the load for a single transistor VCO then keyed the power to a single transistor power amplifier stage which was biased through the optocoupler output. It was quite a neat design as the capacitive coupling between stages was virtualy non-existent so barring power line fluctuation, it would be rock solid frequency wise. I think it was for frequencies around 20MHz maximum.

Brian.

betwixt said:

... which takes us full circle to the start of Neazoi's other thread about RF power oscillators. Short answer is don't do it!

Click to expand...

My best design so far is **broken link removed** and a bit more relaxed (but broadband) the **broken link removed** So I have an idea of all the things you insist and the warnings about power oscillators you refer to. Indeed poor designs no doubt!

However, with just 14 components (including the 4 components of the LPF) I am able to generate a full 8W of carrier power and the stability is adequate at least for manual morse. That is a dirty cheap design that can work many countries of the globe in good conditions. It is really the philosophy of QRP, do more with less and a lot of fun.
Having a power oscillator is not a good thing and I think in some countries it is prohibited by the radio amateur rules.
But I barely can achieve such high powers with so less components, easy to build and so low cost. Why not?

Saying so, if I could "convert" the power oscillator to a power amplifier following just a small oscillator I would do it. But I am not convinced that keying the power amplifier would lead to a stable frequency too. usually buffer amplifiers are needed, which would make the design even more complex.

The purpose of this thread relates to the power oscillator I posted before, but it addresses the parasitic capacitance issue and if something can be done at all.