Colpitts crystal oscillator in LTSPICE

I know a low-power IC designer who relies on kick-start schemes initiated by a uP... but no check for success. I think he's an idiot

...Jim Thompson

You can do a lot of fun stuff if you don't mind using 70 transistors to do it.

Sometimes an IC data sheet will have a transistor count. The numbers can be amazing.

Sounds like it. There are enough non-obvious ways to fail that there's no reason to include the obvious ones.

Cheers

Phil Hobbs

Within an ace of saturation, like the oscillator. ;)

Cheers

Phil Hobbs

It's really important to avoid saturating BJT oscillators. The bias-dependent time delay alone is enough to ruin the stability and phase noise, not counting limit cycles and so on.

Cheers

Phil Hobbs

I'm pretty skilled with Numpy and SciPy. Do you have an interesting set of equations I can set it to work grinding on?

Not right handy--I haven't done a crystal oscillator in years.

Cheers

Phil Hobbs

That's why the gold-doped 2N2369's.

My chip versions use level measurement and current source control... much better spectral purity. ...Jim Thompson

It's really easier to buy oscillators. The phase noise and suchlike specs are right there already. A pretty good OCXO can be cheaper than buying just an SC-cut crystal.

Yup. I mostly build oscillators in protos these days, or if I need to do something fancy, such as tuning it faster than the RC network on the varactor wants to let me, or getting more than an octave of tuning. (Both from recent examples--LC not crystal, obviously.)

Cheers

Phil Hobbs

VCOs tend to have miserable modulation bandwidths, if specified at all.

I'd like to make or buy a VCO that runs at 1.5 or 2 GHz maybe, with

A lot of them have time constants in the 10-100 microsecond range, even for a UHF VCO. Pathetic.

There are a bunch of parametric effects that you have to worry about in a case like that. What do you want it for?

Cheers

Phil Hobbs

I want to transport some analog signals out of a high EMI environment, over fiber. The customer wants at least 500 MHz bandwidth. Think fiber-optic scope probe.

I got some fiber-coupled LEDs, which will modulate (baseband, pure analog) nicely without mode jumps or interference wobbles, but they are too slow. I was thinking PWM or FM with a laser, but the FM thing will be difficult, as you note. Blindingly fast PWM is a possibility. The budget doesn't include lithium niobate modulators.

Near a CO2 laser?

One approach is to current-tune a laser, and demodulate it using a delay discriminator. You can use the delay discrim for AFC as well, just to maintain an operating point. The AFC can be really slow, so you can filter the daylights out of it. The delay discriminator will get you a full range signal for a small tuning input, but of course it'll be sinusoidal, so it'll need calibrating.

That can all be done easily with connectorized fibre parts--you just need two 50:50 couplers with a few centimetres of path difference.

Another approach would be to put AM on a subcarrier, and modulate the diode with that.

Cheers

Phil Hobbs

No. I don't know the actual application, other than the EMI environment is extreme. I assume that battery power should be an option. But we get occasional requests for this sort of thing, so I'm playing with it. Slower stuff we could just digitize.

We might AM modulate a laser directly, if we can avoid mode jumps and interferance effects. FM or PWM would make gain independent of fiber and connector losses, which would be nice.

You can't do that over a very wide amplitude range, though, whereas with FM you get a zero-centred output basically for free, and the tuning range is narrow enough (some fraction of a wave number, maybe 10 GHz p-p) that you don't have to worry about mode hops much.

Connectorized fibre couplers are pretty cheap if you get them from China, and they'll make you the unbalanced Mach-Zehnder as a single small assembly, complete with two photodiodes, for a reasonable price in quantities of > 10 or so. (Check out Photop, who are the ones I've used. Great outfit.)

Cheers

Phil Hobbs

Run the VCO at say 4.5-5 GHz, mix it down with a fixed 3 GHz oscillator to get 1.5-2 GHz. It should be fairly simple to filter out the sum (7.5-8 GHz). The 3 GHz is outside both the desired frequency range as well as the actual VCO range.

What did they used to send analog CATV signals over long distances using fibre ? Apparently they just amplitude modulated the light using the full CATV downlink frequency range.

I think they generally use lithium niobate Mach-Zender (interferance based) modulators, which tend to be expensive.