it oscillates!

That's called a 'snivet'.

Cheers

Phil Hobbs

There are a bunch of the 13 GHz version on eBay for about $22k, probably not unrelated to Win7 being out of support.

I might get one if my clients don't start paying late.

Cheers

Phil Hobbs

They followed the grand Windows traditions: it's really hard to find things. After you find something, it's hard to find it again. And HELP isn't much.

I beginning to like it, but they sure don't make it easy to love.

I did discover that my two Lascar bench power supplies were increasing the jitter about 10:1. And I just discovered that the frequency is radically sensitive to the -5 volt supply.

Such a simple circuit!

That's 4.5 ps RMS jitter after almost 200 cycles of oscillation. Gotta track down the -5 thing.

That I believe. Cap multipliers are your friend when building sensitive discrete circuitry. Negative PSR can easily be negative in dB as well, and is never very high.

Cheers

Phil Hobbs

I'm going to see of the emitter current of the Colpitts transistor is super sensitive, or maybe it's something else.

This is fun: if the guard pour is grounded, the frequency tempco is negative. If it's connected to the emitter, it's positive. So some capacitor value between the emitter and the guard should make it have a zero TC.

You might be able to get rid of the varactor!

Cheers

Phil Hobbs

Phil, why not take the capacitor -ve to the output and make a gyrator, instead of just a capacitance multiplier?

Question asked for a mate, who's used gyrators like to make clean power in every RF gadget he's designed in the last 30 years :)

Clifford Heath.

Thanks!

CH

You only get one pole per section, and there's a sneak path via the series RC that trashes the high frequency rejection. Plus you need an additional (and much bigger) cap to ground for the same corner frequency because you lose the effect of beta.

My most commonly-used circuit is a two-pole cap multiplier with an RC lowpass in series with the collector, which pretty well eliminates feedthrough due to interelectrode capacitance and Early effect.

That'll get you 140 dB or more rejection if the layout is good, and since you don't need the big bypass on the output, it'll probably wind up being cheaper and smaller than the gyrator approach.

The gyrator approach is good for making quiet currents, provided you come out of the collector. I use a modified version for driving diode lasers. You can use 2 poles there, too, usually with a slow op amp loop keeping the DC constant.

Cheers

Phil Hobbs

I need frequency stability as well as low phase noise. Our DPLL will discipline the Colpitts at a maybe a 250 KHz bandwidth, so low frequency modulation, including thermals, are the big issue. If the system sits around for a while untriggered, and the upcoming oscillator frequency has drifted, the next trigger+lock event will be ugly.

We are being compulsive about jitter here, past what is needed to sell this product, but then we do have time on our hands, so we can learn stuff for future use. [1]

I added a biggish resistor to the emitter, to a hv power supply, so I can trim the transistor current up/down without affecting impedances much. More current seems to have a diminishing effect on frequency. Why are a few parts so complex?

My uncle Sheldon, who had a TV repair shop and taught me to solder when I was three, called it "squegging." The horizontal sweep tube would burst oscillate (possibly Barkhausen-Kurz oscillation), and make RF that the tuner would pick up. The result was vertical bars on the screen. The fix was usually to change the HO tube.

AFAIK 'squegging' usually refers to spontaneous rectangular AM caused by an oscillator whose bias time constant is too slow. It starts up, and when it goes into Class C the bias starts to shift until the average gain drops below unity, it stops. Then the bias has to recover till it's unstable in the small-signal regime, at which point it starts up again.

Cheers

Phil Hobbs

If you're only letting it free-run for a couple of microseconds, there's lots of room to wrap an op amp round the cap multiplier to keep the voltage stable (or apply temperature compensation, even).

Well, better than meeting the spec is nailing it to the floor.

Probably due to the f_T getting high enough not to matter.

Well, your oscillator probably has 10**22 atoms in it. ;)

I should be more disciplined about that.

Cheers

Phil Hobbs

I need the varactor for the DPLL. What I can get rid of is the N7400 cap and its series padder. Especially if the tempco is already positive.

We now have 132 Z-folders in the J:\Protos\ collection. There is a procedure and a text index file that people use to log an experiment. I figure that whenever someone learns something, they should share and preserve it. Some are actual PCB layouts, some are breadboards, some just a graph or whiteboard photo or something. Anything interesting.

Our parts database also saves data sheets and notes about parts, and references a Zxxx folder when appropriate.

That's the super-regen mechanism, but Sheldon didn't know that.

Superregens are really interesting. Chaotic too.

I had to solder for Uncle Sheldon because he always had a cigarette in one hand and a Dixie beer in the other.

Well, it depends on how large a range you get from the power supply. (*)

Cheers

Phil Hobbs

(*) Not seriously suggesting this for a product, but I've done lots hackier things in protos. ;)

The FPGA triggerd the laser via user pressing the trigger button.Target designator for laser guided bombs...Complex stuff... energy and pulse width is variable for Diode array.Qswitch window is variable.Better not say much more than that..

Yeah, it's neat being able to make a single tube superregen. External quench has its advantages, but when you start piling on parts, the superregen starts to lose its charm. Still, being able to amplify thermal noise up to headphone volume with a single triode is pretty amazing. Edwin Armstrong was a very smart guy.

Agreed. I've mentioned a truly wonderful book on superregens here before: "The Superregenerative Receiver" by J. R. Whitehead. I have a hard copy, but you can also get an OCRed scan:

Really a good read for radio buffs. I've often wanted to do an optical version, but so far it's never been the right solution to the problem at hand.

Dunno about that. Since a superregen's oscillations build up from noise on every cycle, it seems like it should be fully deterministic apart from the noise statistics.

Well, he didn't have grad students available. ;)

Cheers

Phil Hobbs