Re: ANY(ONE can code

I keep seeing that kids in high school (or even earlier) are being taught to code. But then I have to wonder just what level of coding that is. Do they actually learn programming, or do they just learn to put together a bunch of stuff to create apps?

Bill

There's a difference?

We design hardware that way. We seldom know or care about the internals of components. We design from data sheets or experiments and simulate using ideal components and encrypted behavioral models.

Really good design automation (call it AI) would let us move a bit up the abstraction stack, which would help given that most recent EE grads don't know much about electricity.

Rust is the programming fad-du-jour. Or was a few months ago.

Some people type, some draw. It's a different brain type. Personally, I prefer to visualize a system as a block diagram or a schematic rather than pages of strange words and nested curly brackets.

What's pervasive in my world is voltages and currents, changing together in time.

Since the inductors in your Spice simulations rarely have the data sheet parallel capacitance plugged into the Spice model of your inductor, you clearly don't care as much as you should about the internals of at least some of your components.

The ones that are willing to work for you may not be a representative sample.

Artificial Intelligence does seem to depend on large language models, and since there is a lot of sloppy and careless design around, all of which is likely to be pulled in as grist for the mill of the abstraction process, the artificial designs may not work all that well.

It's notorious that large language models tend to hallucinate from time to time, so we may get a few perpetual motion machines out of the artificially intelligent design process.

I care when it matters. Inductor SRF rarely does. Festooning a sim with a bunch of useless parts slows it down, if it doesn't flat break it. I have one thing I'm running sims on now, a sort-of-gyrated complex-impedance dummy load, that runs long enough for snacks and naps.

I design stuff and people keep buying it. How are your latest designs selling?

If you only design very slow stuff, that may be true.

If the parts are useful in the real world circuit, it pays to leave them in the sim. Something that simulates wonderfully, but oscillates in the real world isn't all that useful.

That usually means that there some parasitic high frequency oscillator in there, constraining the maximun step ti\me to something inconveniently low.

You've set up a vanity electronic design service that sells bespoke electronic design to people too lazy or dim to work out how to use off-the shelf hardware. It's a business model of which I'm deeply envious, but I've got this inconvenient compulsion to be truthful, which puts it out of my reach. My most recent design work was for Haffmans B,V. in Venlo, in 2000-2003, and it's probably still selling into the brewing industry.

If one of their competitors had the wit to set up a four terminal liquid conductivity meter - as I wanted to do, but the boss wasn't prepared to pay for a new (very small) printed circuit board - they might have eaten what market there was.

No. It's just a 20th order nonlinear control system that includes a

No. Your compulsion is to be nasty and insulting because you are insecure and have no imagination.

which

An insulated toroidal inductor would be a great liquid conductivity sensor.

Your idea of being insulted is not getting the flattery you feel you deserve. I do have this inconvenient compulsion to be truthful.

I'm not in the least insecure, and I've got enough imagination to have got my name on three patents. Do try to find more credible insults.

It's a well known solution, though the standard solution is actually two non-progressively wound stacked toroids, as you know if you'd ever got it to work (not that I ever did). They have to be immersed in enough liquid to fill the centre holes in both toroids and provide a return path around both toroids.

I was well aware of the solution at the time (it is cute), but the product - a fake beer bottle to go though a brewery's bottle washing machine - couldn't accommodate such a solution. Haffmans had fixed on a two electrode solution at the time, and I had to invent an oscillator to do conductivity-to-frequency conversion to cover the range from the 300 microSiemems conductivity of tap water to the 300 milliSiemens conductivity of 2% sodium hydroxide solution at 85C.

We needed to add a layer of platinum black to the two electrodes to cover the top end of the conductivity scale, but that got flattened by getting hit by droplets of the cleaning liquids, so I had to imagine and implement a scheme to reinforce the fractal structure of the platinum black.

The four electrode solution would have eliminated the need for the platinum black.

I got the gig because the local electronics consultants hadn't realised how tricky the problems were, in part because my boss at Haffmans hadn't done enough homework when he took on the project. After I'd got the job I spent an afternoon in the Nijmegen University chemistry library doing the homework he should have done. He did have a Ph.D. in chemistry too, but in a rather different sort of chemistry.

The consultants were happy to put my circuit into production, and we got on fine.

Scarcely a circuit which would use use much in the way of wound inductors - they might show up in the power supplies, but at 20GHz everything starts looking like a transmission line where the parallel capacitance is an integral part of the circuit. That was certainly true for the 1 GHz stuff that I did forty years ago.

I am aware that you do "design" fast stuff, though design doesn't seem to be quite the right word for your approach.

Distributed amplifier bias networks tend to use inductors. Some people like those dreadful conical things, but there are better ways.

Lots of people make conicals now that the Piconics patents have timed out.

I accept that we have different approaches to electronic design. My approach is to actually build stuff that works and sells.

So was mine, back when I could find people to hire me. The 1GHz stuff ended up in a product Cambridge Instruments sold. I even got a patent out of it.

"U.K. patent 2139411 "Moving Plate" (also US patent 4614872) on an improved blanking system for charged particle beams, easily adjusted to match a wide range of particle velocities; assigned to Cambridge Instruments in 1983."

The boss wanted our voltage contrast electron microscope to be also capable of EBIC (electron beam-induced current), so the beam-blanking system had to cope with beam voltages from 15kV to 300V - blanking plates long enough (18mm) to bend a 15kV had more than 0.5nsec transit time for a 300V beam. If you could set the electrodes parallel to the beam for the 15KV beam, and rotate them until they were at right angle to beam for beam voltages less than about 800V you could shorten up the region of interaction enough to stay out of trouble.

We did push the envelope a bit further than you seem to.

We did the picosecond master timing system and multi-GHz beam modulators for the world's biggest laser. That was fun. The Livermore people are wonderful to work with. We always seemed to work with a genius female physicist who did the heavy thinking.

Of course your scheme was taken from the Hewlett-Packard Journal, and made a lot more sense back when it was invented.

Now that we've got really low jitter local oscillators - not as stable as atomic clocks, but with quite a bit less jitter - a common clock does make a lot more sense (even if it depends on a lump of sapphire immersed in liquid helium and seems to have been invented in Western Australia for the Australian over the horizon radar system).

One of the more depressing features of the Review of Scientific Instruments is way it reveals that American physicists don't take electronics all that seriously. The one time I dropped your name on a physicists from the NIF he was wasn't in the least impressed.

And why would you care?

HP's vernier-locked digital delay generators and picosecond time interval counters were big and clumsy, not worth copying. I'm not aware that HP did any eo modulators.

Got references? Thought not.

Free ice cream.

It was you who made the claim, here, years ago.

I presume "big and clumsy and not worth copying" means that you tried and failed. I wasn't interested in the electro-optic beam modulators - the last time any of my friends wanted something like that he realised that he could point his laser at a rotating CD and rely on the lines on the disk to get the modulation he needed. He was using a laser to keep track of the flame fronts inside a single cylinder internal combustion engine for Shell. The HP laser interferometer relied on Zeeman splitting to do much the same job. Zygo got a bigger (and more stable) frequency difference with a 10MHz electro-optic modulator, which would have let us move our stage faster in a write-on-the-fly shaped beam electron beam microfabricator that Cambridge Instruments had started work on, but the project got too expensive to complete. It would have cost them 3.5 million UK pounds to finish the job, and they had that - they had to pay that much to buy themselves out of the contracts - but it would have tied up all the engineers they had for 18 months, and that wasn't a practical option.

No. The 5370 time interval counter and their time synthesizer box used a triggered delay-line oscillator and a frequency heterodyne phase locker. That was complex and had huge insertion delays and lots of jitter.

Their (Keysights's) latest time interval counter has lots of jitter too.

This summarizes known (to me) DDG architectures:

Very interesting. How did Wavecrest's stuff do this?

Joe

I'm not sure. I actually have an old/ebay Wavecrest but haven't taken it apart.

I think they triggered a linear ramp and digitized that using a clocked ADC.

That's a great way to make a time interval counter. ADCs have got screaming fast lately.

I meet CE/EE grads who don't know what a state machine is. Much less a software state machine.