X5R vs X7R MLCC

My irony meter pegged again.

The models aren't crap for simplicity, they're crap because nobody cares about making them good. The OPA140 model won't converge unless the supplies are exactly symmetrical, for instance. Lots of other models don't show realistic supply current, or don't limit their outputs at the supply rail, or don't have noise, or do linear interpolation from tables so that they're nondifferentiable in many places.

There's nothing preventing them from making good behavioural models of op amps. We aren't talking about simulating a CPU at the transistor level.

Cheers

Phil Hobbs

Don't underestimate the virtues of women, or cars, or women in cars.

We usually use the Universal Opamp model, tweaked to be similar to the part we plan to use. It runs fast and is well behaved. If something unusual matters, like recovery off a rail or PSRR, we breadboard the amp for that. Behavioral models generally can't be trusted.

I guess that a full-transistor model of an opamp would be accurate but would run too slow.

I think that Analog Devices rushed a lot of models into LT Spice, or maybe just had a lot of mediocre models around already.

I recently simulated a current-output MDAC into a THS6022 current-mode power opamp. Current modes should play happy together, right? The sim was beautiful; the real thing oscillated hard enough to drag down the supplies. I got lucky: the fix was to remove all the compensation parts.

I wish Digikey sold luck.

I have the complete AD library. They're still crap, but better crap than lots. The ADA4899 model has a 1/f corner of about a megahertz, iirc. The real part is around 10 Hz.

Except for the giant output capacitance of your average MDAC.

Well, they sort-of do--you can get transistors with the pinouts upside down. ;) TI's version of the MAX803 comes in three different SOT23 pinouts as well.

Cheers

Phil Hobbs

There you go, whining about a mere 1e5.

I took that into account, and shoehorned the TI model into LT Spice. I was so proud of that beautiful simulation!

I was also counting on the phase lag of the compensated dac-amp being cancelled by a lead elsewhere in the circuit.

I once determined that every possible SOT23 and SOT143 transistor pinout has been used by someone.

Whining, pissing, moaning, _and_ complaining in fact. ;)

CFB models are often way off in the stability properties. The THS3091 model claims that it will work well with a 1-ohm input resistor. It won't.

Plus every possible numbering scheme. We ignore the datasheet and use the old Motorola scheme. (You folks do too, iirc.)

Cheers

Phil Hobbs

I owe you a beer or two!

I was already in doubt today that I could make a current source from an OPA134 and a FZT851. It converged, but delivered simply impossible results running on +10/-5V :-(

Yesterday, an AD8065 model delivered -85 mV running from GND and +10V.

The ADA4898 model (the new one) does not like to converge. It seems that happens even easier when there are more than one of them in a circuit. I have no idea what must go wrong for that to happen.

The old ADA4898 model had ridiculous noise. But V-noise is it's only reason for existence. The ADA4898 is not yet in the LTspice distribution. Probably they have found more problems.

cheers, Gerhard

Based on some years of interaction, I expect that we could have quite a good time together one of these days, beer or no beer. ;)

Cheers

Phil Hobbs

You can have a bigger party on a boat.

Nonsense. Our "computing power" certainly isn't limited. The chip guys did far better simulations with the hardware they had available forty years ago than you can do with the board level models today. It's all about the models.

And bill slowman is a spherical prick.

I'm definitely unkind to John Larkin, but he does deserve it. That may make me a prick, but "spherical" seems to be irrational.

Spherical means you are a prick from any viewpoint.

But not a prick from any of them. Make up what passes for your mind.

A prick is asymmetrical.

It's that old saying about get ideas and attitude into kids at a young age and you have the man. Neither of my lads showed much interest at all in electronics, or computing, where I ended up, but you don't own kids, only have them on loan and they will make their own minds up. One is currently working in admin. while the other is doing a business degree, which may in fact be more use to him.

May be some genetics, but imho, much more about example and inspiration from the family and other possible role models.

Never had the luxury of hole cutters back in the day, so we scribed a circle, drilled small holes around the circumference with a hand brace, knock out the center and file round to size. When you have done that for half a dozen or so B7G, B9A and the odd octal valve base, you help build practical skills that last a lifetime.

Starting with valve tech, right up to date with arm cpus has made it an interesting journey and still love the work...

Chris

That is something I contest. Learning a language takes a young kid three years or more. As an adult, it took me less than a year to learn French. No learning is ever complete though and the sort of errors I make are different from those of a French native.

Jeroen Belleman