tuning an LC

A big series C does the same thing, and the voltage-variable element sees a much smaller swing, leading to much less nonlinearity. It's only about 3 lines of algebra.

Cheers

Phil Hobbs

Because they are discussing varying the DC bias on a Y5V ceramic cap to change its capacitance, essentially make a high-value varicap out of it.

Which will give you harmonic content like crazy... much like a pumped varactor... which is how I got usable output power on 2 meters with the available semiconductors of the early '60's. ...Jim Thompson

We're talking about using two in series, though, which mostly cancels the nonlinearity--ideally, only the odd orders survive. It would be interesting theoretically to look at whether there are more general arrangements that would extend the cancellation to higher order. Of course given the loose tolerances of Y5V capacitors in general, schemes like that would be unlikely to make things much better.

Cheers

Phil Hobbs

It would be fun to play with, but maybe too tricky for production. It might not be repeatable, and tempcos could be bad.

We have noticed, recently, that a lot of high-value ceramic caps, from big-name makers, arrive out of spec, generally low.

Yeah, I'm not pushing it too hard for your particular application. OTOH a reel of Y5Vs is pretty cheap, and after characterization, it would probably be enough to keep production going for the next 100 years. It would certainly need some sort of phase- or frequency-locked loop to make it reliable in the field.

Cheers

Phil Hobbs

When out of the box becomes stupid, I'll stay in the box.

This tuned circuit is a tiny part of the project. The signal processing will be a 14 bit ADC, a ZYNQ fpga (with two hard ARM cores) and a 14-bit DAC, with the ADC and DAC clocked at 48 MHz.

We did do an active floating resistor simulator, which turned out to be a bunch of work.

What box? ;)

Cheers

Phil Hobbs

mosfet

around 50

shielded

trimmable

the caps

tuning steps.

uniform.

in

opamp

drift

it's 60

Goodness, you usually are the one that likes to think out-of-the-box. A mere 100 W worth of power supply and a couple power opamps can't be too challenging for you can it?

?-)

Ceramics lose C as voltage goes up. Some lose half, or even 2/3, at their rated voltage.

Except it cost you a large capacitor for no actual return on investment. A voltage coefficient on an AC signal leads to harmonic distortion, not an error in the center frequency. It is the center frequency we are trimming.

The issue in trim is monotonicity with predictability. Going back to thermometer coding, which is probably how it would be implemented on a chip, you would measure the initial frequency, cut one trim tab, measure the change, predict the number of tabs to cut, then do the cut.

Repeat for the next million parts or until the chip is deleted from the catalog.

Correct, but it is the center frequency of a filter that is being trimmed.

Yep, it is all about duplicating the manufacturers test conditions.

I never worked in "the cap factory", but semiconductors get 100% testing at room temperature. I assume this is done in passives as well. Chips most likely get a 100% test at hot and QA testing at cold. I suspect caps would just get QA testing per batch over temperature.

Now with caps, there is also binning. If you order a part with a high tolerance, it is likely the binning removed all the parts that were close to nominal.

Manufacturers should have a traceability on the entire test flow, including the accuracy of the instrumentation, plus some sort of statistical process control to insure the process itself isn't drifting.

Then again, if you are getting parts from the 3rd world......

Jim is asking the key question. More often than not, the customer is doing it wrong.

On some custom chips, I have managed to throw away half the customer's circuitry and achieve better results. I got a high five from some engineers on one project because I finally shut up one of their pain in the ass designers by pointing out the right way versus his way, which was more complicated and wouldn't work.

standard ferrite less, but still significant.

I would really like to know what kind they are when as you state they dont drift?

For oscillators, you tend not to flog the inductor, so I don't see it being a target for drift. It isn't like a switcher.

There are applications where a VCO needs to withstand more shock than a crystal can handle, and LC resonant based VCOs are used. You can abuse passive components a lot of you don't give them significant electrical stress.

But back to this circuit, it does look like the L will take some punishment.

Of course not. Have you ever looked at V/C curves for ceramic capacitors. No, I suppose you haven't.

Have you noticed nobody cares what you think nor do they listen to you?

We aren't talking about an IC design, we're talking about an application circuit problem, a small part of a larger design. Should all our boards wait till the Omniscient Chip Gods decide to take pity on us? We'd be waiting quite awhile for this one, I expect.

If you weren't so determinedly parochial, you might be willing to learn something from folks who work in a different domain, but of course your oft-rehearsed disdain for stupid customers would tend to work against you there. Did you see how many boards John was talking about? IIRC he said about thirty. A whole reel of 2.2 uF Y5V caps cost about $40. (See e.g.

The benefit of using a Y5V 'varactor' is the same as using any other VCO--you can control it electronically, using a PLL or FLL. I agree that they aren't the most stable things in the world, but neither are silicon varactors. So what? It's inside a feedback loop. Of course, one has to maintain enough design margin to guard against eventualities, but that's SOP.

And I haven't claimed that it's the only, or even the best, solution to John's particular problem. It is, however, one very cheap, simple, and probably effective solution. I didn't invent it, but I think it's pretty.

Cheers

Phil Hobbs