Math and electrical desgin

It's best to be fluid and confused for a while before getting rigorous. Bottom-level thinking about a new problem can create insights that can affect top-level architectures and requirements, and specs. How can you create requirements if you don't know what's possible?

Or invent new ones.

Top-level requirements documents are very difficult to change in some organizations. They require signoff by too many managers. Important or just sensible changes can go into a swirl of management approvals and disappear in the process.

I recently discovered a semiconductor processing system that has four expensive energy metrology subsystems that never worked but have been installed in all systems since 2003. My gear collects useless data from them and sends it to a computer that ignores it. It would be impossible to move enough management to fix this. Figure roughly a million dollars a year wasted for 17 years now.

Bottom-level things, at the board level, are easier to fix. Just do it. Management can't read schematics.

--

John Larkin         Highland Technology, Inc 

The cork popped merrily, and Lord Peter rose to his feet.  
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

George Herold wrote in news: snipped-for-privacy@googlegroups.com:

Barefoot water skiing.

Running off a whole rack at pool.

Any multivariate system quickly involves linear algebra; the FFT is a matrix, and multiply-accumulate operations are pretty important (inner product) in signal processing.

Thinking about differential equations is intuitive and easily explained to people who can visualize things like mass and velocity. After that, let computers do the actual work. Anything interesting is nonlinear anyhow.

I taught one course in nonlinear computer simulation that consisted of the students all coding the filling of a toilet tank. They all did it well and enjoyed it.

--

John Larkin         Highland Technology, Inc 

The cork popped merrily, and Lord Peter rose to his feet.  
"Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"

Yes to all the above.

The response to such "waterfall" processes is are the various "agile" processes. The agile processes do have some real advantages w.r.t. waterfall, but the zealous acolytes are too dumb to realise that they have different disadvantages.

A key issue is the absolute reliance on test-driven design and a bottom-up approach.

TDD is great for ensuring your last low-level tweak didn't change things - but it is of little use in getting the right design. Too many people think "it works because it passes the tests", without giving any thought to the comprehensiveness of the tests.

As old timers know, "you can't test quality into a product".

When creating a new plant, Intel duplicates absolutely everything, down to the ridiculous last detail (e.g. exact type of obsolete PC). They know how fragile semiconductor plants are, and want to remove all possible variables.

Such low-level fixes can have "unintended consequences" at the system level.

You've apparently never done it.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Phil Hobbs wrote in news: snipped-for-privacy@electrooptical.net:

Going over my design ideas with supportive co-engineers is one thing. Having one or more unable to see what you've come up with 'providing' stupid input is entirely another, and therefore is not a favorite thing to do. I'll design it, they'll do their pieces. No need for a white board discussion, just a white board project declaration, and any issues they have can be brought up while they are doing their assigned part.

You've apparently moved toward the Larkin attitude of presumimg things which you know no facts regarding as it relates to me. That is a character and integrity hit you do not want. But you live as you like.

Which is not what I was referring to, at all, and sounds boring, I agree.

Collaborative design at a whiteboard may go something like this example of a phase-sensitive laser microscope. Somebody says, "How about using a heterodyne interferometer with an acousto-optic cell and a Wollaston prism to recombine the beams? That way we can make the whole thing common-path so we can use normal galvo scanning." (A real example from long ago--an English guy improving on an idea of mine.)

Then everybody chimes in with questions such as the following.

"You'll need at least a milliwatt of laser power to get to the shot noise in that bandwidth. That's about 200 kilowatts per square centimetre--will your sample stand that?" (Leading to a discussion about

3-D heat conduction in solids.)

"The topological phase shift in the galvo scanner will screw up the beam recombination in the return path, won't it?" (This is a real issue that constrains the scan range.)

This isn't rock-throwing at all--everybody wants the thing to work, and ideally they're all friends. After half an hour or so, everybody has a much better idea of whether the scheme can be made to work, and the final version probably doesn't look much like the initial proposal.

It's better if there are two or three schemes being discussed, along with lots of ideas for improvements. You also need a few harebrained notions tossed in. (*)

It only works if nobody minds having their pet ideas demolished, because that's definitely going to happen. Personally I'm very glad when it happens to me, because it may save a lot of wasted work. I have lots of ideas, so I'm not at all attached to the bad ones.

Not intending to be insulting at all, but as your first paragraph suggests, you've probably never done what I'm attempting to describe. (Or perhaps I just didn't describe it well enough initially.)

Cheers

Phil Hobbs

(*) The use of helicopter propwash to knock ice off high-tension wires allegedly came out of a discussion like that.

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

I have experienced that. It works when a team of engineers is assigned to a product/product line and they trust each other and the company is not in a layoff cycle. In other words, it is rare. My persecution (and experienc e at the moment) is that companies are doing the matrix form of running pro jects.

In the matrix org people are replaceable and even the project managers are replaceable. Good team interaction is not budgeted into the schedule and i t does not occur. My company at the moment is doing a project with input f rom 5 sites. It is going better than I expect, but it is not nearly as sat isfying as working in a project structure. Also, it is sort of working bec ause the majority of engineers are very experienced and are not falling int o the matrix structures pitfalls. Heaven help a company when they assign s ignificant responsibility to younger engineers in a matrix org.

That's the Intel "copy exact." I guess it's good for them, in building fabs all around the world, but locks in dumb stuff, theoretically down to the brand of soap in the bathrooms.

Like improved reliability. All it takes is some people with intelligence and guts to look over a change and decide that it's an improvement. I guess there aren't enough such people around. The original designers of things are often gone a few years later.

I'd love to ask the guys who designed that metrology thing why they integrated a fast edge twice, and buried it in noise, before handing it to us to time stamp. Nobody even knows who designed it.

We create a design notes file for all our products, and we don't lose it. If you might consider changing a bottom-level design, it helps to know the original intent.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

We did a brainstorm session once, with DR, about reflections off a liquid metal sphere. That was wonderful. You and DR did, in about 20 minutes, astounding math. My function was just to make a lot of random noise that could be filtered.

Old cars, like MGs, and old Honda motorcycles, had carburetors that obviously couldn't work. There was way too much static friction for a little bit of differential air pressure to move the needle valve. But there was a lot of engine vibration too.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

Just so. It smacks of "here there be magick".

Or not.

A slightly different case with different causes, but the 737-max debacle indicates how a low-level tweak by inexperienced people can cause unintended consequences in a domain outside their experience.

As someone who likes to believe in cause and effect, the cockup theory of history is usually the case and the conspiracy theory of history is less often the case.

Now get taken over a few times, have the faceless MBAs do some "rationalisation", and see the end result :(

Conspiracy is the harder thing to do.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

Pretty good quality noise, that.

I have a really big bag of tricks for doing those sorts of calculations, mostly consisting of amusing facts, such as that in the ray model, a shiny sphere scatters parallel light exactly equally into 4 pi steradians, and that the scattering cross section is twice its projected area. (This is easy to show for a plane obstruction--the hole it punches into the plane wave scatters just like the obstruction, except

180 degrees out of phase. Add them back together and you get the original beam.)

That's how I do photon budgets for instruments all the time, but it's way more fun doing it live.

How's DR doing? Have they started paying him what he's worth?

There's a recent physics coinage for that: "stochastic resonance". Like that famous remark of Voltaire's about the Holy Roman Empire, of course stochastic resonance is neither stochastic nor resonance.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

You are correct, though it is rather muddled in scientific literature:

Ok, intuition. That's super-important in engineering.

That's where the right feel for Maxwell's stuff needs to come in. If someone can't develop an understanding of electrical, magnetic and electromagnetic fields it may be better to pick another career. Maybe sales, makes more money anyhow :-)

--
Regards, Joerg 

http://www.analogconsultants.com/

No, that's impossible. It defies all common sense.

(This is easy to show for a plane obstruction--the hole it

He is sort of a systems architect and futurist, looking ahead at where semiconductor fab may go. I think his pay is OK now. It would be cool to repeat the fun we had. There is maybe a 5% chance that we will.

Any time you need a little chaos, I'm available.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

Great article. Mo is a speech pathologist and bird lover, and I'll send it to her.

--

John Larkin         Highland Technology, Inc 

Science teaches us to doubt. 

  Claude Bernard

That's why it's an amusing fact, and therefore memorable. The law of reflection tells you that tipping a mirror by an angle theta steers the beam by 2*theta, so going from the centre to the edge, the reflected angle rotates through 180 degrees. So you do illuminate all 4 pi steradians. The growth of the circumference with radius is exactly balanced out by the obliquity of the surface and the solid angle vs. deflection, so the illumination is uniform.

Here's hoping.

Sure, we should figure out something to do together again.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

The moral of this story is to not trust Larkin's "common sense".

--

  Rick C. 

  --- Get 1,000 miles of free Supercharging 
  --- Tesla referral code - https://ts.la/richard11209