CGW glass-dielectric capacitors

Yes I still have boxes of them!

--
Best Regards:
                     Baron.

Several people make mos capacitors, up to a couple hundred pF. They are amazing, near zero da, huge q's, low tc, low voltage coeff.

Metelics, Vishay, maybe others.

John

Winfield Hill wrote in news:1194130811.919667.138700 @v3g2000hsg.googlegroups.com:

AVX makes glass capacitors, up to 2400 pF. Possibly the best capacitors made, within their capacitance range.

Not cheap.

--Damon

Corning Glass Works capacitor division was sold to AVX in 1987 according to this article.

[8~{} Uncle Monster

I fondly remember them, and Corning Glass Works for some unknown reason stopped making them. Far better than most precision, selected Micas.

But those damn things are eXpen$ive!

Hey, Baron, do you want to sell any of those?

Er ! What do you have in mind ? These have been stored in my loft for a number of years now. They were originally purchased when I was doing development work.

--
Best Regards:
                     Baron.

Have you checked the dielectric absorption? I understand that's pretty bad.

Do you happen to know the temperature coefficient?

Cheers, Tom

Well, quality costs. More significantly, they're really hard to mount. They are usually either leadless parallel-plate things, or the Vishay parts have bump pads on the bottom, impossible to solder by hand.

Oh, sorta off the topic, but this is cute:

One of my products uses this to coarse-tune an LC oscillator at every powerup, to get within varicap range. That eliminates a trimmer cap and associated labor, and takes care of longterm drift. It was a pita to program.

John

OK, I just answered at least part of the questions I just posted, via the AVX website. I suppose with a high (+140ppm/C) but stable temperature coefficient, they could be useful in thermal compensation, but unfortunately, it's more often that a negative temperature coefficient is appropriate for that (esp. in an LC circuit).

I'm scratching my head, wondering why I'd want to use a glass cap instead of a C0G ceramic. Though I don't see a voltage coefficient listed in the AVX data sheet for C0G caps, my distortion measurements tell me that it cannot be very large; I suppose even in the glass caps, it's not truly zero, just vanishingly small.

Cheers, Tom

Consider a 7-pole 40 MHz lowpass filter, 50 ohms or so, running at levels up to maybe 2 volts RMS. Using C0G caps, do you have any guess as to how much THD the caps might generate on their own?

We're doing a new DDS synthesizer/arb: lookup table, 14-bit diff-current-out dac, diffamp, lowpass filter, output amp. It's clocked at 128 MHz and max output will be half Nyquist, 32 MHz.

We're seeing complex patterns of harmonic distortion versus frequency and amplitude, but I hadn't thought about the caps as culprits. We'd like to hold -60 dB to 10 volts p-p, which is optimistic. Commercial RF signal generators and arbs seem to have distortion specs above

-40dB, some as bad as -20.

John

Thank You, I had forgotten Corning Glass Works capacitors.

Datasheet says low, about 0.012% per MIL-C-19978, similar to polystyrene, but thanks, I'll measure that.

The datasheet says about +125ppm from -55 to +45C, increasing to +165ppm at +125C. They say it's VERY stable at these values, with 5ppm retraceability.

One big difference: "glass capacitors exhibit zero piezoelectric noise." (Quoting from the AVX spec page.)

Piezoelectric effects are easy to miss with conventional capacitor-testing equipment, like I have, because they show up very strongly at discrete resonance frequencies. Yes my hp4192 has a 0.1 to 1ppm frequency resolution, good for measuring crystal parameters, etc., but one could take forever looking for capacitor resonances that way. I imagine a new test jig with a sweep generator is required.

Or one stays away from C0G parts in critical signal applications like John's filters (next post). But, they sure are inexpensive!

So, I can tell you that I'm doing tests at +20dBm much of the time, from about 1MHz to 100MHz. I use common SMT and axial-leaded C0G MLCCs in various filters; some are filters I'm testing, and some are filters I use to clean the generator output or kill the fundamental. They range from 5th order to 11th order, typically, and generally have zeros at specific frequencies; though I've also tested common Butterworth and Chebychev designs too. If I assume simple second and third order models, I can assure you that these capacitors won't even come close to giving you trouble at your +30dBm level, down to well below -60dBc. As far as I know, nothing goes suddenly bad between

+20dBm and +30dBm with them.

Cheers, Tom

Hmmm... I'm interested in distortion levels well below what John's talking about, at levels up to +20dBm (only 10dB lower than his), and I've never seen a hint of problem from C0G's.

Based on the low Q (relative to a crystal resonator) of ceramic filters, I'd be really surprised if MLCCs showed all that narrow a mechnical resonance bandwidth. Could be wrong, but it would be a serious surprise. Actually it should be easy to see if it's there. I can excite with random noise and look at 20kHz to 45MHz or so easily, with 300Hz RBW or lower, and invoke a whole lot of averaging. I'm well aware of high-K dielectrics exhibiting piezo effects, but my impression is that C0G shows very low piezo.

Cheers, Tom

Why not try a step function?

That's good to know. We have waveform table quantization errors (with digital interpolation!), DAC errors, diffamp errors [1], then the filter, and then the output amps. Distortion versus frequency and amplitude is looking very complex, but at least the filter's caps shouldn't be a big contributor.

John

[1] so everybody is making great "TX" dacs, 12 to 16 bits, clocks in the hundreds of MHz, great stuff. But the outputs are differential currents. So why don't they make compatible diff-to-single ended amps? They make lots of them for diff-in adc's, but none for diff-out dacs.

If you make your own out of opamps, it gets messy.

Do they assume that everybody will just use transformers?

John