I was thinking that a phemt follower could be made to oscillate by coupling its source into a tapered transmission line whose skinny end feeds back to the gate, to make a high-Q resonator with a bit of voltage gain. The impedances sort of work.
Problem is, I'd probably have to make the tx line from a pcb trace, with all the associated losses and tempcos that pcb's have.
It needn't be an actual width taper. Just relieving some ground under a microstrip would taper the impedance.
Then there's the bootstrapped transmission line, another animal.
--
John Larkin Highland Technology, Inc
The best designs are necessarily accidental.
Not the only way; you could 3D-print the sucker in a suitably low-loss polymer, and fill a couple of tracks with conductive paint , then sandpaper off the conductor from the embankments that define the tracks.
Heck, you could electroplate a styrofoam replica from a 3D-print mold, making an air-like transmission line good to really high frequencies.
A thermostat can get rid of the tempco problem easier than a materials choice, and probably more accurately.
You don't have to make the printed circuit board out of epoxy-glass. Rogers make a bunch of better high-frequency substrates. If you made your transmi ssion line as a buried strip-line, rather than as microstrip on the surface of the board, it wouldn't be dispersive.
Buried strip-line would be even better, and you could cut slots in two grou nd planes.
A sort of transmission line transformer? You could vary the coupling coeffi cient along a pair of buried strip lines, but that would probably need a si x layer board.
An electronic thread in the Covid-19 era! Sadly John won't take it seriousl y - he won't feel flattered by it.
3Dprinting of a small object isn't expensive, and resistive losses in a silver-paint transmission line depend on the length of the line; how fast IS the oscillator? And, what impedance is the line? I'm not seeing a problem in the numbers.
Sure, that would probably work--it's a modified Colpitts. The gain mechanism is interesting, and might be fun to use with some two-ended resonator such as a SAW device, where the Colpitts approach doesn't work well.
I've run bootstraps over (short) transmission lines, but this sounds like something else. Do tell.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com
I need to start and stop my oscillators instantly (otherwise I'd just buy them) so anything mechanical is out. Colpitts works. The OLD HP time synthesizer and 5370 counter used ecl transmission-line oscillators, which can be started and stopped fairly fast.
I've used coaxial ceramic resonators, basically slow velocity transmission lines, but they are low impedance and don't come below about 600 MHz.
I was thinking about plopping one microstrip on top of another, probably in a circular pattern. That might get me voltage gain and would get my frequency down too. That, and tapered lines, are tricky to simulate.
Just musing. I'll probably stick with my Colpitts, but maybe use a SAV551 next time instead of a bipolar.
My first 'bootstrap' was a driven shield (coax) to reduce C of wire down a ~18" LN2 probe. (measuring noise of 10k- 1M R's) max freq ~1 MHz. I never thought of it as a transmission line... The bottom of the coax was floating.
The temptation with a driven shield is to add a bit of gain... that leads to squirrellies. :^)
s. Rogers make a bunch of better high-frequency substrates. If you made you r transmission line as a buried strip-line, rather than as microstrip on th e surface of the board, it wouldn't be dispersive.
nder
two ground planes.
ng coefficient along a pair of buried strip lines, but that would probably need a six layer board.
seriously - he won't feel flattered by it.
ever-ending pissing contest...
out electronics. If he did he'd be aware that you can't get into a pissing contest with John Larkin - if an element of a thread isn't going to make hi m look good, he ignores it.
some comment about John
Not exactly everything. The bulk of what I posted was squarely on topic. O ne line predicted that you wouldn't react to the electronic content - which you haven't.
You have reacted to a personality nonsense reaction my post, which makes yo u a hen-clucking, my-oh-my, personality nonsense practitioner.
And you do a lot of clucking, and take any excuse to avoid talking about el ectronics. You want to use s.e.d. as a platform where you can boast about y our electronics, but don't seem to be interested in discussing how your ide as might be improved.
You chose to do that. I don't happen to think that this is a good idea, and when I had to deal with a similar problem - back in 1988-91 - we looked at a continuously running oscillator twice - once at the start of the period to be timed and once again at the end and interpolated twice. It worked.
ansmission-line oscillators, which can be started and stopped fairly fast.
The HP designs go back a long way, and now that it's a lot cheaper to do di gital processing at ECLinPS speeds, their approach isn't as attractive as i t was.
on lines, but they are low impedance and don't come below about 600 MHz.
We used an 800MHz local oscillator, phase-locked to a good 50 MHz crystal.
One could go appreciably faster now - I was limited by the Gigabit Logic Ga As four bit synchronous counter - which formed part of a rather longer sync hronous counter. ECLinPS was just marketing promises at that time, and really fast programma ble logic was barely on the horizon.
The tapered line has to be electrically long, so it isn't as good a candidate for instant start as a lumped-element design.
I'd be interested to hear how its 1/f noise affects the jitter.
Cheers
Phil Hobbs
--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510
http://electrooptical.net
http://hobbs-eo.com
Probably so. The HP version ran at about 100 MHz and I think the delay line was some sort of ceramic hybrid. 5 or 10 ns is a lot of PCB trace. In those days, HP would throw a man-year or two on details.
I'm thinking that the major issue is circuit Q, as opposed to device noise. I typically swing maybe 1.5 volts p-p, so device noise should be insignificant. Base current could be a q killer as a bipolar runs out of Ft. We do phase lock after a microsecond or two, so the PLL dynamics dominates after that. I just want the first ~20 cycles to be really good.
Amplitude limiting is always an issue in an oscillator, doubly so for a triggered oscillator that ideally makes constant-amplitude swings from the very beginning and can't have any sort of slow amplitude-limiting loop. The more obvious fast amplitude limiters look like Q killers and wobulators to me.
A comparator-based (infinite gain limiting amplifier) gain element is another approach to making an LC oscillator. The HP txline oscillators used an ECL gate.
I think the Tek 11801 series scopes use a triggered txline oscillator as their timebase. Not phase locked. Some newer sampling scopes seem to phase lock a triggered oscillator.
I have enough dead 11801s that I could take one apart.
--
John Larkin Highland Technology, Inc
The best designs are necessarily accidental.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.