inductor tempco

These look great, very high Q. I'm thinking about a 50 or maybe 100 MHz Colpitts oscillator PLL.

I wonder if the thermal expansion tempco of the FR4 board will stretch the coil and change its native tempco. FR4 is variously cited as being around +5 to +17 ppm/K, which isn't bad. That might even reduce the tempco of the inductor.

I also wonder how a PCB ground plane effects L and Q. I guess I'll order a kit and try it.

I meant "affects" of course.

If you do, please post your findings.

Colpitts with a JFET is a good way to go. Low distortion.

It depends on whether the individual turns of the coils are stuck together.

If they were wound with self-bonding wire, movement in the FR4 would bend t he mounting leads rather than changing the length of the coil. To get a max imum SRF in a cylindrical air coil, you are supposed to gap the turns by th eir thickness, which isn't the way the data sheet shows them

but the wires may have a thicker layer of insulation than you might expect.

If the wire's wasn't circular to start with, life could get even more inter esting.

The stray field from a cylindrical coil mounted parallel to a ground plane would induce some current in the ground plane, and lower the inductance and the Q, but if you think about the currents induced in the ground plane, it can't be much.

I'm thinking about...

It's a clock generator, so distortion doesn't matter much. Low tempco and low phase noise are good.

I guess I'll build one and try it.

Good idea. I LTSpiced it and it looks okay, but the amplitude is only

Yup. It will drive a comparator to become a clock.

The amplitude is limited by the c-b conduction of the transistor. With a high-Q inductor, it takes surprisingly little emitter current, like

We'll coarse tune it with a digital capacitor, and fine-tune/phaselock it with a varicap. Varicaps are awful, so it's best to minimize the varicap pull range.

It will be slightly microphonic though not as much as if the turns of the coil were not bonded together. Still an inductor deposited on ceramic or silicon would be more mechanically stable and have less thermal expansion. I suspect that if you are building a VCO, the varactors and other semiconductor junctions will dominate the temperature coefficient of frequency.

It is difficult to get consistently low phase noise (and reliable oscillation, and predictable tuning sensitivity) if the amplitude is not controlled by some sort of feedback system. On cellphone chips there is pretty much always some sort of amplitude control. This has to be done carefully as it may introduce its own noise, but is worthwhile.

If you need good phase noise then you would also want a quiet supply regulator. On cellphone chips it is common use one substantially better than 20nV/root-Hz. (For some reason people don't sell those as stand-alone regulators, you have to buy a cellphone radio with it.) At high frequencies you can get similar noise by low-pass filtering, but then the cost, size and startup time are orders of magnitude worse. If you want low phase noise close to the carrier (e.g. at 1kHz offset) then the supply needs to be quiet at frequencies similarly far from DC (e.g.

It is also much less practical to design quiet oscillators without something like SpectreRF which can not only simulate phase noise of free-running oscillators (which you could also measure on a prototype) but also tell you which components the noise is coming from (which a prototype is not able to tell you).

You might find that an integrated circuit microwave VCO plus a divider is better than building one from scratch, though I don't know if you need any special features that those would not offer as standard products.

If you do build your own oscillator, and if you find that the inductor is the main cause of temperature coefficient, then you could use a ceramic coil former (or buy a coilcraft inductor that is wound on a block of ceramic). Provided the wire is tightly wound, I expect that its diameter will vary as does that of the ceramic, which would be more stable than copper or FR4.

Chris

In the circuit that I posted, amplitude is limited by conduction of the transistor c-b junction. That probably has costs.

It's kind of a weird application, so I have to build my own oscillator.

On a previous generation of this sort of PLL, I used a Coilcraft inductor that's wound on a plastic form. They must wind it tight, because the plastic cold flows and the inductance creeps over time. We fixed that by baking them before assembly, to relieve the stress.

Standard surface-mount inductors, 1206's and such, are handy and stable, but Q tends to be low. Coilcraft has an 1812 part with Q=80 at

I haven't used these springform inductors before, so I don't know if they will be stable. I'll try it I guess.

I'm simulating 1.7 volts p-p at the tank, and 170 mV p-p at the emitter. I'm not sure what might be the optimum capacitor ratio. 10:1 sounds unusually high, but it seems to work fine.

I'm not an RF guy! Sinewaves are my least favorite waveform.

Why a Colpitts? If you wanted a clean and reasonably fast tuneable sine wav e oscillator you could use a pair of Analog Devices fast multipliers - the AD834, AD835 or ADL5391 are all fast enough for a 100MHz oscillator.

One multiplier would provide the adjustable in-phase feedback to keep the a mplitude where you wanted it, and the other would provide adjustable (posit ive or negative) quadrature input to allow you to pull the frequency up or down. At 100MHz, half a metre of coax could provide the quadrature componen t to be fed into the second multiplier. A coax delay line isn't a broad-ban d solution, but sufficiently broad-band for something you might otherwise f ine-tune with a varicap.

It's a more complicated solution than you'd come up with on your own, but i t would be easier to explain to customers than the traditional approach.

It might be quite noisy though. Multipliers tend to be noisy because the "LO port" is in small-signal operation all the time, and therefore the devices contribute noise. If you replace the multipliers in your scheme with hard-driven mixers then it would be less noisy as the switching devices in the mixer core don't contribute to the current noise when they are fully on or fully off. The harmonic content in the mixer output current should not matter as the tank will filter it out, and even if it didn't, filtering out 3rd and higher harmonics is not very difficult unless the tuning range approaches an octave.

Chris

Adding a one-transistor ALC loop may fix that. As in:

piglet

wave oscillator you could use a pair of Analog Devices fast multipliers - the AD834, AD835 or ADL5391 are all fast enough for a 100MHz oscillator.

he amplitude where you wanted it, and the other would provide adjustable (p ositive or negative) quadrature input to allow you to pull the frequency up or down. At 100MHz, half a metre of coax could provide the quadrature comp onent to be fed into the second multiplier. A coax delay line isn't a broad

-band solution, but sufficiently broad-band for something you might otherwi se fine-tune with a varicap.

ut it would be easier to explain to customers than the traditional approach .

The multipliers are injecting small adjusting currents, not the whole signa l, and the scheme can - in principle - generate a pretty clean sine wave.

The Colpitts oscillator is a class-C oscillator so the current that sustain s the oscillation is essentially a Dirac spike, with every harmonic up to t he limit imposed by the width of the spike. A feedback stabilised oscillato r can use linear amplifier to supply the bulk of the power required to sust ain the oscillation, which isn't much with a high-Q inductor, and the ampli tude control multiplier is just a tweak around that.

The frequency adjusting multiplier has to inject (or subtract) an appropria te proportion of the current circulating in the tank circuit, which would g et nasty for large frequency shifts, but not as nasty as Class-C spike.

Sure does--collector limiting will hurt the phase noise. Cutoff is way cleaner than saturation. Since the gain is so high, you'll be loading down the tank, too, which will not do good things to the Q.

Cheers

Phil Hobbs

OP

I don't understand your orignal question.

Is the coil for the PLL VCO?

Most modern PLL ICs have on chip VCOs.

Mark

John is usually doing something weird like instant-on VCOs that are phase locked eventually. His stuff is pretty much all time-domain. You could instant-on that Colpitts by putting a Schottky across the top capacitor and swinging the supply from +5 to -5.

Cheers

Phil Hobbs

Yes, it is an instant-start oscillator that is eventually phase locked to an OCXO. So any slow loop, like an ALC, or any inherent slow amplitude limiting mechanism isn't good. I want steady-state as soon as possible. It's a messy crossover between time domain and RF-type issues.

The inductor is a big part of the problem. Nasty things, inductors.

A BFT25 costs us 36 cents.