When the dielectric constant is measured vs. temperature, is the change in size taken into account? In other words, is the tempco of the dielectric measured as an intrinsic property of the material or does it also include the change in size?
-- Rick C
Digikey doesn't show any surface-mount NTC caps. All their NTCs are giant high-voltage things. We use an 18pF N750, custom order from Murata.
Looks like 2pF N3300 sorts of things don't exist.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
That might work but it would annoy production.
I wonder if it's reasonable/affordable to get a PCB house to mill out a region of the board from below, like this,
to reduce the FR4 capacitance. I'd want to go most of the way through the board, leaving 10 mils or so. I know that people mill out most of some vias from below, to reduce capacitance. Same idea.
I'll find out.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
What sorta %change/ C is N3300? The capax data sheet shows a max of ~1%/C at 2.2pF. That seems like a lot, but I don't know. (If they don't like email get on the horn.)
George h.
-3300 PPM is -0.33% per degree C.
Which part is that? That would be N10000. Some of the bad dielectrics, Z5U types, have huge TCs, but not guaranteed and not very linear.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Squeek!
Curve "P" (page 2) looks to be about 10% in 30 C, they make that in 0603 packages from 1 pF to 470 pF.
Isn't that what you wanted?
George H.
Oops I read the curve wrong... Curve "T" is ~N3300. (20 C per division.. duh.)
GH.
Interesting. They don't show any data on voltage coefficient, though, and there's no tolerance on the tempco. Most N750s are -750 +- 120 ppm/K like those Kemet ones Win linked to. That's fine for one-offs, where you can cut and try a few times, but in production it only gets you a factor of 6 improvement with no tweaks.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
Those would work great, if I can get them. They never answered my email.
At 1 or 2 pF, I'd buy the biggest NTC that they can make. I'd pad it with an NPO in series to tune the net oscillator TC to zero. The rough-estimated actual fix is 1 pF at N2500, but more NTC would give me room to tweak the actual PC board. Curve V is N4700, so 1 pF would give me lots of margin to pad down.
I'll try them again.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Some vendors respond better to the phone. I guess someday I'll run across a vendor who likes Facebook and I'll have to go there.
George H.
I called them. They remembered my email but hadn't got around to answering.
They can do 2 pF at N3300 in 0805 or 0603, a few weeks lead time, 4K reel at a reasonable price. I can pad that with a theoretical 3pF NPO and compensate my oscillator without pulling it too far.
Simultaneously, two PCB houses are willing to mill out the board from below and leave about 10 mils of FR4 under the parts, which might cut my PCB capacitance by maybe 2:1. We could use some em simulation software to estimate the actual capacitance; maybe ATLC or Sonnet Lite.
Cutting away FR4 is interesting in other apps, like maybe making low-loss microstrips. "Suspended substrate" sort of.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Wouldn't a Rogers microwave substrate be cheaper, easier and more predictable?
seem to be low loss, while
seem to be FR4 compatible.
They won't be as cheap as FR4 and your printed circuit sub-contractor will probably have to buy a square metre or so of the substrate, but anything you produce in volume would eat that up.
-- Bill Sloman, Sydney
So...
Excuse me for rocking the boat, with an infuriatingly obviously-wrong suggestion, but:
Why not put the oscillator section on a teensy daughter card section.
Make it out of pricey Rogers or whatever. Put a heater on it if you like. Put a cover over it to reduce air currents and improve close-in / 1/f properties.
The entire rest of the board can be cheap FR4. Maybe even crappy two layer construction. (Or maybe not. But the method can always be extended to as many subcircuits as needed...)
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Contract Design Website: http://seventransistorlabs.com
For high volume, connectors are expensive. For prototypes, the daughterboard always gets in the way. ;-) I suppose the daughterboard could be a LGA sort of thing but you'd have to make sure both are really flat.
For optical reasons I'm doing a baby board suspended on edge-clip leadframe pins. Tyco makes SIP-style ones, but the only folks that I've found who make DIP-style ones are Batten & Allen.
They're reasonably cheap as well--certainly a lot cheaper than a C-grid header.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics 160 North State Road #203 Briarcliff Manor NY 10510 hobbs at electrooptical dot net http://electrooptical.net
That has been suggested here. All that would be expensive, and take more engineering, than compensating the oscillator on the main board. The NTC cap and an NPO series padder will cost a total of 15 cents and our pick-and-place can slam two caps down in about 1 second.
I have done too many ovenized subassemblies, and the challenge is to NOT do it again.
The main board will be roughly 8x10 inches, 8 layers, ENIG, 4 mil rules, optimistically under 1000 parts. That won't be cheap either.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.