nonlinear caps

What page of AoE3?

Closest I can find is a line on a chart reference to High-K Ceramic caps on page 301 (figure 5.4).

Page 28 section 1.4.5 Not Quite Perfect... talks about memory effect, but no warnings...

Maybe this is in AoE-X?

John :-#)#

It's a bad idea. Audio oscillators are supposed to produce clean sine waves, and hi-K ceramic caps have a voltage dependent capacitance.

As a sine wave goes up to peak voltage and back again, equal changes in voltage will produce varying chunks of charge, and you won't have a sine wave of current.

We've had a couple of threads on sine wave oscillators, and you don't seem to have learned anything from any of them.

Might be. One of the interns found it in one of my books.

I just ordered three sets, Aoe3 and X-chapters, for my team. Nearly $200 per set.

Gosh. I never knew that.

Lots of people who have no ideas, are hostile to ideas.

Of course you did, but you've never bothered to think about the implications.

This should have been an obvious implication, which you don't seem to have noticed.

I have plenty of ideas, and I'm not in the least hostile to good ideas. I am hostile to bad ideas. Their proponents have been known to waste a lot of time and money following them up - not just their own, but other people's as well. Your junior engineers will have been trained not to say boo! to the goose.

I have seen voltage controlled oscillator designs published a long time ago which used kigh-K dielectrics as the tuning element. I think the main issue is not so much the non-linearity itself but the hysteresis and temperature dependence which are difficult to calibrate out. John

But you have bad ideas yourself. An example might be the sinewave oscillator circuit you produced some time ago. It had at least four times as many components as anything produced by JM and much worse performance.

You then told me in a subsequent thread that I didn't like ferrite filters. This is also ridiculous. I will use them when they are needed. You will, of course, tell me that I don't know when they are needed but that's just another fantasy of yours.

There would likely never have been any such thing as Trinitron if a lot of time and money hadn't been wasted on:

Of course I do. Winnowing out the good ideas is the tricky bit. Other people can help a lot there. JM provides an excellent example of the constructive critic.

The idea there was using an asymmetric current mirror as the gain controlling element. JM finally got it to perform rather well.

Bitching about the component count in one of the half-baked versions rather misses the point.

Seems unlikely. You don't seem to like ferrite beads/chips, which are a rather specialised sort of filter, but useful for keeping high-frequency hash out of the power rails. They are cheap and compact, even if they add to the component count.

But you don't seem to know how often they are needed in real circuits. Voltage sources in simulations aren't great simulations of real power rails.

It's more a response based on experience. Supervising junior engineers leaves its scars.

The fundamental idea was sound - it just took a lot of engineering effort to make it work. The shadow mask was a good idea, but it did need a great deal of engineering effort to come up with a version which could be make it work. Committing that kind of effort to a version of the idea that can't be made to work can be a very expensive a error of judgement.

My own example of a nightmare bad idea wasn't one of mine - when I was working on phased array ultrasound at EMI Central Research around 1977 when one of the brain scanner guru's decided that the brain scanner scheme would work for ultra-sound, despite the fact that ultrasound doesn't follow a predictable path inside the human body. I was rude enough about it that I got let off working on the project, and close enough to the team that did get stuck with job to see just how much of a disaster it was.

Probably not. Once marketing have been sold on a idea, it's hard to unwind the fantasies.

Cap tempco won't matter if it's in an amplitude servo loop.

Some ceramic blends have nearly linear C-V curves high up their voltage limit. A small AC voltage on top a big DC bias won't have much distortion. Then use two caps in series, to cancel the residual c-v curviture. Distortion in the A stage will be tiny, and is then filtered 12 dB/octave by the integrators.

Basically nobody mentions C-V effects on data sheets.

In real life, at audio frequencies, opamp distortion will usually dominate.

The best way to get a sub-PPM distortion sine wave is digitally, with a good 20-bit DAC like AD5791.

Capacitor datasheets don’t contain anything useful except package size, reel specifications and the part numbers vs. nominal values.

The interesting stuff is in the characteristics sheet, or more recently on the manufacturers characteristics websites, such as Murata’s SimSurfing and whatever Samsung‘s thing is called today.

There are some makers that don’t seem to have any of that kind of stuff, so sensible folk don’t buy their caps, at least not in values and sizes where there’s likely to be a CV problem.

Alpos have made giant ceramics less necessary, so most of the CV troubles I need to design around involve moderate values in very small packages.

Cheers

Phil Hobbs

Murata has good technical info here, lots of characterized components

On 8/3/25 7:46 AM, john larkin wrote: <...>

Samsung has a graph on this datasheet for a 100nF 10V 0402:

On 8/3/25 7:46 AM, john larkin wrote: <...>

I forgot the page number in my previous post.

Section 2-7 has a graph of capacitance variation with bias in this Samsung datasheet for a 100nF 10V 0402 part.

Aluminum polymers basically make big ceramic bypass caps unnecessary.

I think most people use way too many bypass caps anyhow. A typical IC data sheet says to use a string of different value caps as close as possible to every power pin. On a multilayer board with power pours, that's silly.

When it might matter, I test caps for leakage vs voltage, destruct voltage, C-V curves, ESR, ESL, reverse voltage behavior, stuff like that. It's interesting.

Yeah, some datasheets have 0.1 uF data. For the same outline, the layer thickness s changes pretty radically with value and voltage rating.

The drop in epsilon depends on the E field, and of course E=V/s. Thus the

Cheers

Phil Hobbs

So you are hostile to yourself?

A bit like Chromatron being a bad idea until Trinitron fixed it. Everything necessary to build Trinitron tubes was already in place because of the unworkable Chromatron. All they had to do was replace the gun with one with three cathodes and get rid of the deflection wires. And a good idea was born out of many years pursuing an idea which was never going to work for a 26 inch home TV set.

I think you may as well say I don't like bananas.

Exactly when have you seen a real circuit designed by myself?

Everyone knows that.

Yeah well I'm sure the junior engineers would have got over it quickly. They may also have had opinions which they didn't tell you about but I guess if they're junior then you wouldn't care.

I'm not sure I'd call it an error of judgement if there is no way at the time to determine whether it's an error or not.

My experience with rude people is that they do tend to get held back. I'm sure you made your point that this idea for a brain scanner would not work but why be rude about it?

I was takling to my kids about this on Friday. We conjecture that any dielectric has a limit on volumetric energy storage. Ceramics gracefully relinquish C, and film caps just fail.

Most aluminums sort of zener and limit voltage, until they overheat. I've seen an aluminum polymer that failed hard without warning at about 1.5x rated voltage; we don't use that type.

But some people still bypass voltage sources in Spice!

Silly ideas are often a path to good ideas. Play with them instead of clubbing them to death as infants.

It's kind of a habit because that's how you would usually want it in a real circuit. It's likely that some of my simulations have them because that's what I'd want in a real circuit. A better simulation might be to have some resistance and/or inductance in series with the simulated voltage source. But that can sometimes increase the simulation time or make it come to a shuddering halt with singular matrix or other issue.