strange oscillator

Yes, decades ago I built something similar using an NPN BJT at kHz frequencies. I used two molded chokes and had initially them loose coupled but was amazed that I could separate them or re-orient so there was no mutual couplling and still got oscillation.

I think Irving Gottlieb described three types of Hartley in his Oscillator Handbook and this is one of his "type 2" Hartleys, there is no mutual coupling and the action is negative resistance.

piglet

does the impedance transformation, with no help from the transformer action of the mutual inductance.

Cheers

Phil Hobbs

I was just fooling around. What I really want is a 1.5 GHz oscillator with two antiphase outputs. A MMIC might be a good gain element.

Found the book online, p145-146 were what I was thinking of ...

piglet

Nice. It's from 1997 but includes oldies like thyratrons, spark gaps, tunnel diodes.

What about cross-coupled ECL-type oscillator

To me, this classifies as a Hartley, because it has two L's and one C as its tank circuit. The fact that the L's aren't coupled, to my mind, is only a minor point.

Jeroen Belleman

To me, a Colpitts has two C's and one L in its tank. This one doesn't. A capacitive divider is also an impedance transformer.

Jeroen Belleman

I could persuade a diff-in-out ecl gate or comparator to oscillate, but I'd have to try it. I don't have Spice models for those.

I might get lucky and just connect a diff gate into itself, inverted. and watch it oscillate. Hey good idea! Might be fun to try.

A MMIC is a great gain element, but it's basically 50 ohms in and out, the ideal tank Q killer. The best MMIC ocillator is probably delay-line, which will give me antiphase outputs. Again, gotta solder to test that.

I wish ADI would include their ADCMP parts in LT Spice.

Not without the resonance, it isn't. The input is at the tap and the output is at the top of the tank, so it isn't a voltage divider.

The math of a real Hartley is different from the Colpitts case, but with uncoupled inductors it's the same except for the sign of the reactance.

Cheers

Phil Hobbs

The concepts of input and output are somewhat slippery in oscillators. One could choose any or none of the three terminals of a transconductance gain element as common, and inputs and outputs move about as a result. My point is that two of the three reactances that make up the tank are of the same sign, and one is opposite, and the low-Z node of the gain block connects to the node where the two equal-sign reactances connect. It's easier to draw than to describe.

The signs of the reactances are to me what makes the difference between Colpitts and Hartley oscillators.

Jeroen Belleman (Whose oscillators oscillate and whose amplifiers amplify. Usually, anyway. :-) )

Gottlieb's 'Understanding Oscillators' was in its 3rd Ed in 1971.

RL

First published 1970. 4th Ed ?

RL

Why bother? An IMPATT diode will do the oscillation (negative resistance) just fine, and antiphase is pretty easy; at 1.5 GHz, a few inches of transmission line is a hundred eighty phase shift. Tunnel diode type items are the bomb for this kind of thing.

Have you built an impatt oscillator in this frequency range? Impatts are usually pulsed, high power at higher frequencies.

Does anybody still sell TDs?

Without doubt, the worst oscillator I have ever seen anyone design. You are right. You are not an analog guy.

What's wrong with it? It worked first try.

We have surface-mount inductors in stock, many values, but no tapped inductors in that range.

You don't need tapped inductors. You make the feedback with tapped capacitors. That means capacitors in series. The problems are obvious. You tried to make a colpitts with a hartley configuration. You ended up with parasitic oscillations and very poor control over the oscillator amplitude. This gives unwanted sidebands in the output and very poor phase noise.

Just for you, I looked up my old xtalosc.zip file. Start with the readme.txt file. It is intended for colpitts oscillators but will work with any lc oscillator. Just start with an inductor impedance of 50 ohms at your target frequency and use a fast enough transistor or logic device.

Sure. Colpitts has been around for about a century. This is a design group. Instant hostility to ideas is easy.