colpits AGC stability problem

In article , Jim Thompson wrote: [....]

You could also just move the cats whiskers a bit futher apart too. :>

Another method worth a look would be to vary the voltage on the second gate of a dual gate device.

Thats only if you have 2 emitters you can hook them between.

simplified circuit as ASCII art:

Vcc ! ( ( L1 ( +---------- ! ! !/ --- Vb--! --- C1 !\\e Y1 ! ! !! ! +---[!!]---+ ! !! ! \\ --- C2 / --- \\ ! ! GND GND

L1,C1 and C2 form the frequency selection circuit. The C2/C1 ratio is largish. Y1 is running in series mode and sees low impedance on both of its ends..

Thanks, its a grounded emitter for the sake of minimum parasitics, a reasonably high voltage is needed to get it to oscillate over a wide range and keep Ccb low. Im not sure about the squeging I dont think ive seen any since its behaved.

In effect the fairly quick inacurate one is monitoring the collector current. reducing the RF gain with the 200R resistor in the feedback was the key to success in taming the positive feedback, but ive not seen this used in the countless oscillator circuits ive looked at.

Im not sure what the swing on the base is, exept that it is high when uncontrolled. measuring such signals is problomatic. It should be very small when it is controlled as it goes through another gain stage before it is level controlled to .5vp-p

The non linearity doesnt seem to be a problem now, its said this improves the phase noise anyway. The problem before was uncontrolled limiting in subsequent stages cuasing awfull harmonics wich exceeded the fundamental in the final stage, now the output looks at worst like very rounded squarewave wich is fine to drive my laser.

Colin =^.^=

I had assumed:

! !/ ---! !\\e ! +---- ! ! \\ --- / --- \\ ! ! GND GND

Yes, assuming the current rises when it oscillates that will do it. The emitter resistor I had assumed effectively does this same thing.

If you need wide tuning you may not be able to do so but normally I would suggest that changing the capacitive divider in the feedback path be used to lower the gain.

I was expecting a calculated estimate. You could do a spice model of the circuit. If everything else is accurate, the base swing is likely to be also. In the startup condition, you want to make sure you can't get into breaking down the EB junction.

Personally I have my doubts about the non-linearity making phase noise better. The non-linear action makes the stage operate as a mixer it takes noise from near harmonics and mixes it down to near the carrier. This seems like a bad idea to me.

You didn't respond to the question. Show me a by-hand mathematical analysis. Otherwise I relegate you to blow-hard status.

...Jim Thompson

I was trying to keep parasitic inductance as low as practicably possible, as it considerably lowers the range of tuning, conveniently these transistors have 2 emitter leads.

Yes thats true, the tuning is done by 2 varactor as a capacitor divider, I was unsure as to the necessary ratio of this divider as many circuits had quite different ratios,

It should work well enough for now to test the limit of the laser/detector anyway.

IME the accuracy of the spice simulation would be entirley dependant on my estimate of the parasitics wich isnt very good, Ive usualy added parasitics to the sim till it works as the real thing, sometimes far more than would seem realistic. I think the main problem is the model of the varactor diode, the series resistance varies drasticaly over the tuning range.

Seems to make it considerably better judging by the figures in a few reports, at least for close in phase noise. The maths suposedly shows the tank is least susceptable to a noise impulse during the peak voltage. I cant say ive measured it for myself though.

I remember I had a similar discussion about this some time ago, I ended up being more confused than before. Its also interesting that the collector current monitoring transistor suposedly negates some of the RF transistors 1/f noise.

Colin =^.^=

Sounds pretty duff to me.

- YD.

I had seen it for decades and noted it worked well and replicated it, but that post made the LED turn on.

From a DC point of view, the ground plane of the oscillator circuit can be above the emitter resistor. This can let you use the "opposed emitter" as you normally would by taking them straight to the plain. The bypass capacitor on the emitter resistor will include the capacitance from the oscillator's ground plain to the overall PCB ground plane.

You could just assume that the EB junction has 1/2 the swing of the collector and see if that would break down the transistor. The loading of the input drive will ensure that the signal is less than that.

In terms of phase shift caused by an impulse, the effect varies as the sine of the angle from the peak. As you go further into class-C, the angle at which there are noise impulses decreases. On the other hand, the amplitude (or number of) the noise impulses is increased. This makes it sound like it may matter what the noise source's spectrum looks like.

Yes, moving the ground plane leeds to many possibilities just like all the different colpits configurations. Ive used a 100R collector load instead of the inductor/33R wich probably has the same effect, and the base bias is now derived from this wich otherwise is reasonably conventional bias circuit, although it is further controlled through a pnp.

I chose common emitter for the colpits becuase it means the capacitances from each end of the inductor to ground are effectivly in series wich means the effective parasitic parallel capacitance is half that compared to one end of the inductor being grounded, wich has less efect on the tuning range, also the blurb about noise advantage of dual emiter being conected directly to grnd was in the back of my mind.

When it was uncontrolled I did put my RF sniffer probe on the base and found about 1.7v p-p wich surprisingly was higher than at the collector, this probably means it was exceeding the 1.5veb on the spec sheet, the impedance probably limited any damage.

When it is controlled to low level the waveform looks quite sinusoidal wich means the base drive must be small ( angle at which there are noise impulses decreases. On the other hand, the

The noise spectrum would also be dynamic, probably the biggest part of noise is related to sqrt of collector current, therefore a higher peak current would mean lower shot noise contribution for the same average current.

Colin =^.^=

In message , dated Fri, 1 Sep

Actually, you also chose common-base and common-collector. The oscillator doesn't 'know' where its circuit is grounded.

Yes but the parasitic capacitances to ground, biasing components and also the output connection all have diferent influences as they are referenced to wichever transistor conection is grounded.

Colin =^.^=

In article , colin wrote: [....]

But but but..... when you go further into class C you include noise components from higher and higher frequencies. The noise increases as the sqrt() of the bandwidth so there is another sqrt() term to multiply the sqrt(I) by removing the advantage.