Yo! RF dudes!

That's LTspice. It doesn't need much to run. It won't crash or do bad things to your files. It runs on any Windows version from Win98 on.

Sure. It might take a while. I'm working on an ozone machine to kill mold spores in fabrics. These make me very ill, so it is my top priority right now.

I feel the same way about 813's.

Regards,

Mike Monett

We have one product that uses a 600 MHz gated oscillator, using a coaxial ceramic resonator and a mmic.

Speaking of MMICS, Mini-Circuits has switched vendors again, and, one again, the ERA series is screwed.

John

I used the gated oscillator I posted earlier in an octave-bandwidth vco running from 350MHz to 700MHz to measure hard disk error rate. It went directly from SPICE to a production pcb. Since the oscillator phase noise is part of the measurement, a low value is desirable. I believe it was about 54ps rms. The oscillator jitter remains the same when it is divided down, but then it is much smaller part of a cycle so the vco noise became insignificant compared to the readback signal.

How long does the ceramic resonator take to come up to amplitude? If it's too slow, can you goose it? That might be the solution to your 50Mhz gated oscillator problem. Run the oscillator at a high frequency with a high Q element, then divide down. That should give much better phase noise than running at 50Mhz with a low Q surface-mount inductor. You should be able to pull the ceramic resonator enough to phase lock with your reference. That could solve your tempco problem and give better performance.

I've heard of people starting a crystal oscillator in one cycle. I don't know how they do it without damaging the crystal, but it might be worth looking at.

Regards,

Mike Monett

It starts in a couple of ns.

The LC-based design we have now is fine, and has low power consumption compared to a ceramic resonator osc. It's been highly evolved over a decade or more. The oscillator starts in about 4 ns and gives a nearly perfect 50 MHz square wave clock out after that, and can be killed and restarted in about 40 ns. Jitter is single-digits picoseconds. It's just that the tempco has got weird on us and we don't know why.

Barring some revelation, we'll just program the trim dac voltage to be a polynomial on temperature, which will flatten the oscillator tc curve. I hate to do stuff like this when I really don't understand the physics.

HP did that in one of their old delay generators. It's basically terrible. For a lot of work you wind up with a bad crystal oscillator. The worst part is that you can't quench it very well for the next burst... the quartz keeps ringing for a very long time. You can damp an LC tank to near zero residual oscillation in a single cycle.

John

Whew, luckily I'll escape that since I am still rolling my own amps. Somehow I never got to like MMICs much. A few months ago I almost used MMIC but the largish gain tolerances blew that one.

Isn't there that trick of drilling a hole into the quartz can and reduce the Q?

I have a friend who is running AT-cut shear mode quartz crystals under water, no can. The seem to work fine.

John

Interesting! I guess now someone has to try how they'd run on beer. Reminds me of a Kenyan who used beer instead of (expensive) real brake fluid for his car. Nigerian beer though. Asked about why Nigerian beer and not a brew from Kenya. "It brakes better".

Is that a normal oscillator without special starting circuitry? I thought the unloaded Q was in the hundreds to thousands. How long does it take to settle to full amplitude?

That sounds a bit like the circuit I just posted. I didn't show the stop waveform, but it dies pretty quickly.

That's the advantage of working at a single frequency. Most of my stuff is octave bandwidth to allow binary division to any lower frequency. The jitter is pretty bad in comparison, but it's a lot better than DDS. I haven't tried the latest ADI chips, so the jitter might have improved.

That's what mine does also. But I'd be careful about any residual energy in the tank. A small amount could cause significant phase error on restart if it happened to be 90 degrees out of phase with the previous timing.

I ran into that problem on an old data separator for hard disk drives. It was the first time an LC oscillator was used for this application, and it solved a lot of problems with previous multivibrator vco's. That is patent 3,810,234 on my web site at

The start waveform for the oscillator shows a 180 degree phase difference. That turned out the be the easiest to handle.

Regards,

Mike Monett

I tried stopping the oscillator I just posted at different points on the cycle.

The original stopping point just happened to be at the perfect location to kill the oscillator instantly.

Moving the stop point slightly had a very dramatic effect. The oscillator took many more cycles to die.

So stopping in a single cycle is easy if you can pick the right point. Otherwise, it might be quite difficult. Regards,

Mike Monett

Of course it has special starting circuits. It starts at zero crossing, makes a perfect full-amplitude sine cycle, and keeps doing that until we shut it off. The LC loaded Q is probably in the ballpark of 30 (1206 inductor!), and the coaxial resonator version is likely in the Q=400 ballpark. The sub-GHz coax resonators have relatively low Qs.

I couldn't get LTSpice to display that. Can you post a pic?

I think DDS has fundamental spur problems. We usually see jitter around 1/20000 of the sinewave period, not very good.

If you shunt a ringing LC with the optimum resistor you can kill it pretty dead in one cycle. I think a resistor of around 0.8 of the inductor's reactance works pretty well.

John

I notice that you don't mention me any more in your Binary Sampler pages.

John

Heh - Sorry. I had a disk crash in the middle of updating the site and tried to reconstruct the files from pieces I found on old backups. Then I stopped using Sympatico many years ago, but they didn't erase the web site. I can't update it, but I figured I might as well leave it there since it showed up in google pretty good.

I'm still recovering from the disastrous effects of mold toxins. I'm trying to build an ozone machine to kill the spores, and will probably begin moving and updating everything as soon as I get my health back. Regards,

Mike Monett

John Larkin wrote:

OK, back to square 1.

I know about the 1206 50MHz oscillator. That's fine. I was asking about the ceramic coaxial resonator oscillator, the one with Q=400 ballpark.

How long does it take to come up to amplitude?

If it comes up in a cycle or so, can you use it in place of the 50MHz LC oscillator that is giving you problems with temco? That should give much better jitter and solve the tempco problem forever.

But if it comes up to full amplitude in a cycle or so, you must be using some special startup circuitry. Yes?

Even better. Here is an updated ASC file where I tried to show two different timings on shutdown. It's not perfect, but you can definitely see a difference depending on where in the cycle the shutdown occurs. The same PLT file as before should work.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Version 4 SHEET 1 880 708 WIRE -48 -128 -160 -128 WIRE 96 -128 -48 -128 WIRE 128 -128 96 -128 WIRE 240 -128 192 -128 WIRE 240 -112 240 -128 WIRE -48 -96 -48 -128 WIRE 96 -96 96 -128 WIRE -160 -80 -160 -128 WIRE -448 -32 -496 -32 WIRE -288 -32 -368 -32 WIRE -224 -32 -288 -32 WIRE -288 -16 -288 -32 WIRE -48 -16 -48 -32 WIRE 240 -16 240 -32 WIRE 96 0 96 -16 WIRE -496 32 -496 -32 WIRE -288 64 -288 48 WIRE -160 64 -160 16 WIRE -160 64 -288 64 WIRE -128 64 -160 64 WIRE -112 64 -128 64 WIRE -288 96 -288 64 WIRE -496 128 -496 112 WIRE -160 144 -160 64 WIRE -288 192 -288 160 WIRE -224 192 -288 192 WIRE -288 208 -288 192 WIRE -160 256 -160 240 WIRE -112 256 -160 256 WIRE -96 256 -112 256 WIRE -160 272 -160 256 WIRE -160 368 -160 352 WIRE -160 464 -160 448 FLAG -160 464 0 FLAG -128 64 Vosc FLAG -496 128 0 FLAG 240 -16 0 FLAG -112 256 Vem FLAG -288 208 0 FLAG 96 0 0 FLAG -160 -128 Q1C FLAG -288 -32 Q1B FLAG -48 -16 0 SYMBOL voltage 240 -128 R0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 WINDOW 0 13 11 Left 0 WINDOW 3 -20 137 Left 0 SYMATTR InstName V1 SYMATTR Value PULSE(0 5 100n 2n 1n 515n 575n 2) SYMBOL npn -224 -80 R0 SYMATTR InstName Q1 SYMATTR Value 2N2369 SYMBOL cap -304 -16 R0 WINDOW 0 35 18 Left 0 WINDOW 3 38 47 Left 0 SYMATTR InstName C1 SYMATTR Value 2.546E-10 SYMBOL res -176 256 R0 SYMATTR InstName R2 SYMATTR Value 1.2k SYMBOL cap -304 96 R0 WINDOW 0 41 17 Left 0 WINDOW 3 37 45 Left 0 SYMATTR InstName C2 SYMATTR Value 8.488E-11 SYMBOL ind -352 -16 M270 WINDOW 0 32 56 VTop 0 WINDOW 3 5 56 VBottom 0 WINDOW 39 -16 53 VBottom 0 SYMATTR InstName L1 SYMATTR Value 1.5915E-07 SYMATTR SpiceLine Rser=1.66 SYMBOL npn -224 144 R0 SYMATTR InstName Q2 SYMATTR Value 2N2369 SYMBOL voltage -160 464 R180 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 WINDOW 0 8 11 Left 0 WINDOW 3 -7 58 Left 0 SYMATTR InstName V2 SYMATTR Value 2v SYMBOL voltage -496 16 R0 WINDOW 123 0 0 Left 0 WINDOW 39 0 0 Left 0 WINDOW 0 8 11 Left 0 WINDOW 3 -14 59 Left 0 SYMATTR InstName V3 SYMATTR Value 2.5V SYMBOL schottky 192 -144 R90 WINDOW 0 0 32 VBottom 0 WINDOW 3 32 32 VTop 0 SYMATTR InstName D1 SYMATTR Value 1N5817 SYMATTR Description Diode SYMATTR Type diode SYMBOL res 80 -112 R0 SYMATTR InstName R1 SYMATTR Value 3.3k SYMBOL cap -64 -96 R0 WINDOW 0 35 18 Left 0 WINDOW 3 38 47 Left 0 SYMATTR InstName C3 SYMATTR Value 10pf TEXT 216 40 Left 0 !.tran 0 1.5u 0 200p TEXT -408 -168 Left 0 ;'SED 50MHz StartStop Colpitts

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ > John

Regards,

Mike Monett

Same as the LC; it starts instantly at full amplitude. Just poke the proper initial conditions and then let go. It will think it's been oscillating forever.

John

why is ozone better than, say, chlorine? chlorine is easy to make.

Hey, since you're on the subject, how well does the mmics transistors hold up in a strong EMF ? At work we have a board that is mounted near a large unit that periodically pulses a strong EMF, and now and then, one of the transistors will simply fail on this board. We can't move the board any where else! It has to be there. I know the mmics are made of slightly different materials than the usual BJTs.

The slower (1 GHz) mmics are silicon, but the fast stuff (3-10 GHz) is usually InGaP. Sirenza has some very quiet SiGe parts. And they are all a bit tender, especially the SiGe. They are also very small chips run at high current densities, so operating junction temps are high. But they seem quite reliable once they ate running and not abused.

John

Ozone doesn?t hang around long?