Magamp oscillator (2023 Update)

Not with an ordinary Schottky diode it isn't. With a tunnel diode, it should be possible.

Jeroen Belleman

It doesn't work. The observed double-period component is just the tank ringing from the start-up transient. It doesn't last. Look at the waveforms a bit later: There is no trace of a double-period signal to be seen.

Jeroen Belleman

Varicap diodes were once popular as microwave parametric amplifiers, when other semiconductors were too slow.

so I figure it can be a subharmonic oscillator, too.

I think it will work if I get the values right.

The paper I linked to (paywall unfortunately) talks about optimizing frequency halvers based on varactors and schottkys.

All you need is a nonlinear capacitance, which all diodes have, and low enough loss.

It should be possible to do it with a magamp too--it's a degenerate parametric oscillator.

Cheers

Phil Hobbs

Does that mean a magamp oscillator would be expected to produce frequencies of Fclk/N? Or, perhaps, *M/N for modest values thereof?

I know very little about parametric amps... All the explanations are crap, which usually suggests that it's either a stupendously complex subject, or a preposterously simple one (which is merely being presented in a confusing manner as a barrier to entry, as many otherwise-simple academic subjects tend to do).

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

The simplest parametric amp is a variable capacitor connecting an RF source to a resistive load. Change the cap and you can vary the load power. It takes a tiny amount of energy to change the capacitance of a mechanical variable cap or a varactor, so there is huge power gain.

Rectify the RF output back to DC, and you have a DC-controlled, DC output amp with enormous power gain. But it needs the AC pump.

There are some simple graphical ways to look at my (currently not working) parametric divider. And there is serious math, too.

There are varicap-based parametric amplifiers, mixers, dividers, multipliers, and oscillators. Still used in microwave apps, apparently.

It sounds like there was a telephone central-office 20 Hz ring generator that was a magamp based divide-by-3, off the 60 Hz line.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

OK, I give in. Here is a simple circuit that generates a strong and persistent f/2 from an input at frequency f.

Jeroen Belleman

======== Cut here ======== Version 4 SHEET 1 880 680 WIRE 240 80 144 80 WIRE 288 80 240 80 WIRE 144 96 144 80 WIRE 288 128 288 80 WIRE 144 208 144 160 WIRE 144 208 96 208 WIRE 144 256 144 208 WIRE 288 272 288 208 WIRE 144 352 144 336 FLAG 288 272 0 FLAG 144 352 0 FLAG 240 80 tank FLAG 96 208 pump SYMBOL varactor 128 96 R0 SYMATTR InstName D1 SYMATTR Value MV2201 SYMBOL ind 272 112 R0 SYMATTR InstName L1 SYMATTR Value 1? SYMBOL voltage 144 240 R0 WINDOW 123 0 0 Left 2 WINDOW 39 0 0 Left 2 SYMATTR InstName V1 SYMATTR Value SINE(1 1 96.08meg) TEXT 176 24 Left 2 !.tran 10u

Arm waving explanation:

In an LC paramp, you want the capacitance to be decreasing as the output goes through its +-peak values, and increasing at the zero crossings. (A varactor wired across a bridge rectifier is one example of a device that does this. Two varactors back-to-back also work.)

The voltage goes as 1/C, and the energy in the capacitor goes as CV**2, so at the peaks

dE/dt = E/C dC/dt.

The circuit will oscillate at f/2 by building up from noise, provided that the input is strong enough to make the loop gain at f/2 more than unity. The output will be loosely phase locked with the input, of course, because the gain isn't a very strong function of the relative phase.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 

hobbs at electrooptical dot net 
http://electrooptical.net

How do you wire that? An L || C only has one resonant mode, and no isolation. Would you use one resonator for the pump, one for the signal frequency, and span the "funky C" between them? And then the signal resonator then looks like a negative resistance, so you have to come up with your own hybrid / isolator / circulator / etc. to get an amplifier as such?

(The magamp equivalent would be two resonators, where the caps go to a common ground, except it's not ground, it's the magamp core.)

How does one relate the impedances of such a circuit? The resonator frequencies are defined by their LC components, plus the average capacitance (more or less, and I suppose, give or take if the pump and signal are correlated or not), and the pump magnitude is something about the (Cmin / CJO) ratio and resonator impedances and Qs.

Does it only work at f/2? That seems rather less useful, needing an oscillator at a much higher frequency all the time. (Of course, doublers and triplers, of modest efficiency, are a thing. Often using varactors again, it seems. But a stack of those gets rather wasteful.) After all, magnetrons can be modulated, and that's old fashioned, like 10GHz in the

1950s, late 40s even. Though it's not like you can phase lock a magnetron to a quartz oscillator (or better)...

I suppose that's the point, the pump can be any dirty old AC (give or take "PSRR"), while the signal is something cleaner.

As for "degenerate", I presume that's something about being able to handle DC (or most anything between DC and Fp/2)?

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
Website: http://seventransistorlabs.com

Ooh, that's nice. The varicap is the nonlinear thing and the tank capacitor simultaneously.

I had to change the L value to 1 uH; the "mu" symbol didn't survive Usenet.

--

John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Yes. The best known is possibly the Lorain Sub-cycler family:

The "decimonic" K5 produced 20Hz and 30Hz by para-amp action and then mixed to generate 40Hz and 50Hz (10Hz spacing=decimonic) by a demonic circuit shown in the download fig 8.

piglet

Is that LTSpice? It complains, Multiple instances of "Flag"

--
 Thanks, 
    - Win

Yes indeed. It doesn't complain here. OK, I declined to upgrade LTspice since almost four years. Some asc file format changes may have occurred since.

Beware of the value of L1: It should be 1uH. The 'mu' doesn't fare well on Usenet. Sorry about that. I just fix that using a text editor when it happens to me.

Here's an ASCII version of the schematic:

tank +-----------+ | | | | V MV2201 L = varicap L | L 1u | | - | / \ | | V | GND \ / | V=1+sine(97MHz) (pump) | GND

Next challenge: A circuit that produces two *different* frequencies, such that f1+f2=fpump.

Jeroen Belleman

I ran it in the current LT Spice, except that ? should be u.

It's amazing. Divides by 2 with gain!

It works here in LTSpice, after correcting the value of L1 to 1u.

--

-TV

Check your source impedance/current. Is it a fair trade?

RL

Not the point. The argument was about using parametric effects to make oscillators.

Come to think of it, I posted about another such thing, in an argument over using mains-frequency driven magnet coils to sustain a pendulum swinging at a ~1s period. That was a parametric oscillator too.

Jeroen Belleman

It's not gain.

RL