more Dremel fun

On Sep 7, 2016, John Larkin wrote (in article):

I?d start with using a better dielectric than FR4.

Rogers has a low tempo circuit board material that processes like FR4, RO3030. I don?t see that in their catalog any more, but Rogers has a whole line of such materials.

Joe Gwinn

The real board will be six or eight layers and pretty big, and an exotic laminate would be expensive. We could put just the oecillator on a small baby board, with better laminate, but that adds cost too.

It's hard to buy NTC ceramic caps, which would be handy. The FR4 has a positive capacitance TC. A cap very roughly 0.5 pF and N200 is what we'd need.

IIRC N750 caps (the most common type) have a TC of -600 to -900 ppm/K or thereabouts--so sloppy it's only useful in one-offs.

With those values, you could make an active cap using a noninverting amp with an RTD in its FB network. ;)

Cheers

Phil Hobbs

We have stock on an 18 pF N750, which is bigger than I need. Maybe I can "attenuate" that one somehow. We have to buy that in 5K reels, special order.

At 600 MHz, that smilie is appropriate! A dac-driven varicap might work, except that varicaps have horrible TCs, which change with bias voltage!

The other idea is to use a temp sensor and a few 1206 resistors and ovenize that region of the board.

A question that may show my ignorance of oscillator performance: in what way do the commercial oscillators not satisfy your requirements? (rather than making your own oscillator using crystals)

Actually a microwave pHEMT or BJT would probably work.

Cheers

Phil Hobbs

Do you mean, say, a c-b junction as a varicap? That would be easier, and cheaper, than buying a varicap. The TC problem would be about the same, but of course if a uP and a DAC drive the temp comp varicap, we can just add code to fix the defects.

A BFT25 is a pretty linear varicap between 2 and 8 volts. C changes about 20 fF, which is probably not enough, so I'd need a bigger transistor or diode.

Nuisance.

This experiment is a 600 MHz delay-line oscillator, which can be started or stopped in nanoseconds. I don't know of a commercial equivalent.

I meant using Miller and changing the gain via g_M, probably. Letting the gain change from -1.1 to -1.0 would give you -200 ppm over 50 degrees

Cheers

Phil Hobbs

Cheers

Phil Hobbs

I must have missed something. ISTM that you could just put an NPO in series to drop the C. Yes, it also drops the N750, but you can decide if it is still enough, yes?

Cool! Thanks

Tim

A very cool dude even wrote a JS calculator that tells you how much series and parallel capacitance you need to get there. ;-)

Tim

To get 1 pf at -100 PPM, from my 18 pF -750, it says that I need a series cap of 0.13 pF.

There is another configuration for the three caps, with C3 in parallel with C1. NPO C3 "dilutes" the NTC of C1.

Huh, I was going to suggest that, but I figured it would be to expensive. Is that cheaper than an special "baby board"?

Well and you have to run it hot for a large temperature range.

Say is there anyway to purposely add some pcb capacitance to a different part of the circuit to counteract it?

George H.

A few resistors, a temp sensor, PWM from the uP. That would cost maybe

Yes!

Probably servo to something like 60C. Maybe a couple of watts?

Not that I can think of.

You're welcome to solve the formula which leads to that solution.

Or you can find a smaller N750. :)

Tim