Blind simulation without understanding

Full confession: That is my MO since years. I have done bench testing on maybe 30% of projects last year that involved the design of new or revised circuitry (not EMC works and such). The other 70% were born totally in cyberspace, meaning on this here computation machine. And they work, they are in production, they make my clients money.

I bet Jim's percentage of total cyberspace designs is 100% like it is with almost any IC designer, or close to that. With the exception of honey-do projects, of course :-)

Naturally, this does not mean one can approach the computer without a solid plan of actions or without a solid circuit design know-how.

P.S.: When young, engineers must experiment and solder, not matter what.

Total cyberspace in that "testing" is done only on the simulator... no more breadboarding. BUT most initial work-up is by pencil and paper and Algebra.

Pretty much the same way... except I, personally, am the "process house" ;-)

...Jim Thompson

I still do about as much breadboarding as simulation, but a lot of designs get neither.

Lots of parts, especially fast ones, don't have decent Spice models. By the time I evolve a model and verify it against a real part, I may as well just breadboard the real circuit.

Recall my recent experience with the AD8014 bootstrap ramp generator? The simulation, using ADI's model, was useless.

Besides, messing with Dremels and soldering irons and test equipment is fun.

John

It's my experience, both in real life and simulation, that slightly too much feedback, combined with unfortunate oscillator response (too many poles in the amplitude response), causes it to overdrive itself, forcing bias down, cutting it off, causing it to recover, restarting the cycle.

I once made an amplitude regulated VFO with a peak detector and error amplifier (to give better regulation than a light bulb, at lower power levels as well, though the implementation as-is contains a Vbe tempco).

The oscillator, in the linear (non-saturated) range, is roughly an integrator; ideally, it's an exponential integrator, because as gain is varied (by bias, in this case), the oscillation should grow or decay exponentially with time, not linearly. The toe of that exponent can be approximated as linear, so it therefore acts as an integrator to a reasonable degree.

Now, the detector and error amplifier have a nice long time constant, to keep ripple out of the loop and ensure low distortion. This absolutely trashes the phase margin, because it looks like an integrator at some upper frequency as well.

I simply put a pole-zero compensator in the loop and it went from unconditional squegging (even under variation in coupling capacitance, detector capacitance, etc.) to rock steady action. It doesn't even have a startup transient, it just goes... bloop and stays there, with 10ms of applying power.

Tim

there.

mathematical

Those last examples are at the level of a competent technician, not a wizard.

Bits of math, rules of thumb, and reflexive sanity checking are part of it, but real wizardhood requires a disciplined fecundity of imagination. A wizard can come up with a completely new circuit topology, or five, or ten, as needed, and has the grip to make one or more of them work really well for the given requirement. Creative goal-post moving is also characteristic of wizards. Of course many of these approaches may have been used before by others, but that's beside the point--the imagination, intuition, grip, and technical taste are the essence of wizardhood. Academic ideas of originality are useless except in getting tenure, and possibly getting or avoiding patents--a wizard does what works, regardless of whose idea it is. Someone, no matter of what professional stature, who is too attached to his own ideas is not a wizard. (SED is a good place to apply that rule--we have genuine wizards, budding wizards, former wizards, wannabe wizards, and wizards-in-their-own-minds.)

Doing your own math, building your own circuits, and critiquing the circuits of others analytically develops intuition and technical taste far faster than simulations. (Note that to be useful the critique has to be _analytical_, based on engineering goals such as cost, space, performance, power consumption, and reliability. Reflexive disdain for certain parts and design techniques, and those who use them, doesn't qualify. SED-style pissing contests emphatically don't qualify.)

Cheers

Phil Hobbs

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

hobbs at electrooptical dot net

Yeah, I had to buy Mathcad for a client's project last year and thought about maybe using that for the math stuff at the beginning of a design. But ... ain't happening, pencil and green engineer's pad are faster.

Sometimes vendors are and then it gets ugly. Like my favorite job this afternoon: Did the regular cleaning of the kitchen sink waste pipes and sure enough now that all the gunk is out there's a tiny leak. Because this stuff is all cheap plastic these days. Hurumph.

Yes, I remember. However, last year I had a few designs where super fast pulse stuff happened all on SPICE. One client actually told me that they do not believe in simulators and I guess were a bit surprised when my cyberspace circuit produced roughly the same plots on their scope that it did on my computer. Despite the fact that I used magnetics in an unorthodox way.

Oh yeah, it sure is. Now if there was a xx GHz USB-sampler that would make it even more fun. I've got a USB spectrum analyzer since a month and that thing really kicks butt. I have to put da big machine on the check list now so I won't forget to turn it on once in a while so it's memory battery won't dry up.

and

whatever, I

Years ago I bought Mathcad, but found it quite cumbersome to use. If I can't solve an equation on paper, I use Excel or, quite often now, simply use PSpice... the behavioral modeling capability is so good, I can set up and solve a transcendental equation in minutes.

Yup :-( I just got an announcement that it's also wet under our kitchen sink :-( ...Jim Thompson

you're

and

whatever, I

I use Mathcad 6.0 to do photon budgets for clients, and I'd like a slightly newer version. Either of you gents interested in selling?

Cheers

Phil Hobbs

-- Dr Philip C D Hobbs Principal Consultant ElectroOptical Innovations LLC Optics, Electro-optics, Photonics, Analog Electronics

hobbs at electrooptical dot net

and

whatever, I

Could be worse. I was just outside and in the distance I heard a major kablouie, made the windows rattle. Then you could hear stuff clanging and shattering back to earth. But ... no sirens, nada, nobody seemed to fuss about it. And I have no idea where it came from. No, not from my lab :-)

ts

ng

f you're

ebra and

atever, I

on

We came back from Lyon to find that a 15mm copper pipe had pin-holed through. We called a plumber who not only cut out the offending section - and replaced it with new cooper pipe, hooked up with compression fittings - but also noticed the coupling to the base of the hot water tank was leaking.

When he pulled it apart, the hard rubber washer came out in two segments that had obviously separated a while ago. A new washer stopped the leak.

The stuff was probably installed fifty years ago. Cheap plastic may start leaking earlier, but nothing lasts forever.

-- Bill Sloman, Nijmegen

and

whatever, I

There are some...

John

uire

rs,

parts

oring

re

=A0If you're

algebra and

whatever, I

ng on

is

no

er

ly newer

Try E-Bay - I bought a cheap out of date version there a few years ago, for my employers, whose computer manager feared Linux and wouldn't let it near any of their computers.

I'd go for Scilab running under Linux if I need the functionality.

-- Bill Sloman, Nijmegen

Amazing! They don't give a risetime spec in the manual but at 8GHz bandwidth it should be good enough.

$10k, yikes ...

Well over $5k, seems to be geared towards the TDR market.

That's careful selection of problems and careful shepherding of students. Real life can be messier.

Scarcely. That was years ago. I'd like to put a simulation together that would "squeg" - it would probably take all of five minutes if I could get hold of a set of VBIC model parameters - but it's not exactly on my "to do" list.

Back when I was a graduate student I had to strip a few hundred turns off the inductor to stop my first-ever Baxandall class-D oscillator from squegging. I hadn't had access to any kind of coil-winding machinery at the time, which made it particularly irritating.

The non-squegging with MOSFETs was reported here a few years ago, not by me.

-- Bill Sloman, Nijmegen

e a

You get the same problem in a second order phase lock loop. Floyd M. Gardener's "Phaselock Techniques" spells out the problem (rather more elegantly than you have done) and the solution.

The Baxandall Class-D oscillator doesn't have any feedback, and Peter Baxandall called what he was seeing "squegging" because he didn't have a clue what was going on.

-- Bill Sloman, Nijmegen

you're

and

whatever, I

Someone's propane tank ?:-) ...Jim Thompson

You can get a nice used 11801 and a 20 GHz sampling head for around $1500. But it's not very portable.

John