Bad Circuits in the Wild

"Jim Thompson" wrote in message news: snipped-for-privacy@4ax.com...

:-) Yeah, anyone who knows you knows you have no reason to make others look bad. As I say, some people have a hard time believing anyone would be critical of their designs for reasons other than as a means to make them look bad.

Yep, agreed!

snip...

driven

had

poof! --

Cut....

That's a digital shunt regulator. What's wrong with that? Why is it any different in concept than an analog shunt regulator or any more prone to failure?

More and more design is disappearing in to the realm of code development. Look at DSP for example. More and more things are being done in Code and less and less in hardware.

There's no way in hell you could do a modern device like a cell phone, or an I-Pod by soldering transistors and resistors together.

Would they have had a leg to stand on? In other words, could you have dug up prior art if they had sued you over it?

Sure, pretty much eveyrone used to do that.

Not providing the schematic doesn't add much hinderence to someone out to pirate your design, but it does seriously preclude your customers from doing their own troubleshooting. Not worth it, IMO. And today, turning a physical PCB back into a layout is potentially time-consuming but trivial; bypassing programmable logic and software copy protection -- while always doable -- can take up so much time and be done by few enough individuals that it's no longer profitable to do so. (And unlike software, where people will spend hundreds or thousands of hours to crack your protection just to demonstrate their "skills," a similar thing doesn't seem common with hardware!)

---Joel

Lots of other reasons. I've run into some really strange one. One set of design specs required that a specific exotic hybrid chip be used for the final product. Never mind that it was too slow, too expensive, too big, wrong Vcc, sucked too much power, and made little sense. The VP of engineering was on the board of the chip vendor and he was trying to save his investment.

Another company had a surplus of some odd parts that were purchased for a product that never sold well. My instructions were to incorporate these parts into the next product to save a few pennies. I dutifully did what I was told, only to be blamed for the inevitable purchasing crisis, where both the availability and price of re-ordering this odd part were problematic. I suspect that the engineering time, board revisions, and paperwork required to incorporate a more standard part wiped out all the savings of using up the odd parts inventory.

Another odd design was the result of what happens when production drives the design. One company had a series of problems resulting in final products that did not meet specifications. Most of those products eventually required component values selected during board test. Needless to say, that was rather expensive. Rather than do a rigorous tolerance analysis and incoming component inspection, engineering was ordered to include enough adjustments to insure that the product met specifications, no matter how crappy the parts that purchasing was ordering. The result was a rather odd design, that had more trimpots and adjustable capacitors, than it had semiconductors.

I've lost count of how many sample parts were disowned by their manufacturer. The level of paranoia at one company was sufficient to require that all specified parts have at least 2 independent vendors and sources. For key semiconductors, the more, the better.

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

It's no more a shunt regulator than a resistive voltage divider is: It had no feedback and had just been manually "calibrated" to work, whereas anything called a "regulator" needs to have feedback to adjust its behavior past on what the load does.

Still, there's nothing "wrong" with it, but it doesn't really meet the "robustness" criteria that many designers strive for when cost isn't a huge priority. Here I define "robust" designs as those where specific consideration was given to doing no damage (or at little as possible) to the device when reasonably foreseeable component failures or application failures occur (i.e., microcontroller dies, someone shorts the output, etc.).

Yes, but the choice to move control logic from "discrete" logic to programmable logic shouldn't compromise the robustness of the device. E.g., for something like a DSP motor controller, if you put a .45 slug through the DSP (and nothing else), I expect that the motors should either stop or perhaps keep doing what they were doing prior to the DSP's vaporization; they shouldn't start wildly oscillating or otherwise go nuts.

Here's a good example of a not-very-robust design: I have a power supply that loses regulation and rails the outputs (!) when you key up a 150mW transmitter within 1m or it (and it's a dual-output 36V @ 7A supply, so it has plenty of muscle behind it to destroy anything connected when it goes bonkers). Now, it was dirt cheap -- easily 1/3 or less of what a comparable Agilent or Tektronix unit would cost -- so while I cringe to say it, for the price its performance is "adequate." Although I still intend someday to go and attempt to fix the EMC problem!

---Joel

No. The patent was issued prior to the founding of my employer. We never did pay any royalties because the product in question didn't sell enough units to get the attention of the much larger companies IP department. The product in question was also scheduled for retirement shortly, so our attorneys just stalled until the problem went away. We were not told to not send any more great ideas to the magazines, but everyone there understood that it was a bad idea.

I'm looking at the pile of TV and audio boxes sitting next to my computer. No schematics on any of it unless I purchase the overpriced service manual. Some require that I be an "authorized" service center or something in order to get my hands on the schematics and to order components from the manufacturer. The computer is worse, as there are few printed schematics available for any of the boards inside. It probably doesn't matter as computers tend to follow the chip vendors "reference" design, which is easily obtainable by purchasing the overpriced developers kit.

Sure it is. One product I designed was on a tight release schedule. We let everything slide except the product. That meant the service manual was delayed about 10 months. For the first 10 months, we had few problems with defective units or field failures. Hardly anything came back trashed. Then the manuals, with the schematics were released. The dealers immediately attempted to repair, modify, tweak, tune, adjust, etc the radios. Butchered radios were returned in levels sufficient to have the company reconsider its dealer service program. If we hadn't released the schematics and repair manuals, life would have been so much nicer.

PCB reverse engineering is usually a waste of time with more than 2 layer boards. However, there are companies that will desolder a multilayer board, and use a bed of nails probe board to reverse engineer the internal traces. As you mentioned, it's time consuming, and fairly simple with the proper setups.

Well, I've never reverse engineered a digital design. Analog and RF designs are fairly common. I never tried to extract the exact schematic from the board. If I needed that, espionage is a much simpler method. What I wanted was a function diagram (block diagram) of the board, along with the applicable voltages, waveforms, gain distribution, and bandwidths. That's quite different from what someone just looking for a way to repair the board is interested in obtaining.

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Hi Jeff,

I was thinking test equipment and high-end commercial stuff; you're certainly correct that for consumer goods the schematics -- if available at all -- had a price tag attached to them and were sometimes considered "secrets." (Heck, a couple years ago I paid some Russias some pittance for a copy of a Sony stereo service manual!)

I meant the schematic/service manual should be available "in general" -- a delay due to getting the product out the door as a first priority is entirely understandable.

Personally I'd be tempted to terminate the service contracts with any service centers who routinely butchered radios -- their skills just aren't up to the standards I'd want them to have. I'd be worried that their lack of skills would reflect badly on my company!

---Joel

no

huge

the

failures

development.

E.g.,

the

perhaps

that

transmitter

of

Now, it

Tektronix

performance

the

I disagree, you can put all kinds of fail safes in the code plus multiple redundancies, even multiple processors, voting routines and other tricks to improve reliability. Secondly you can bias the output and use watch dog routines so that a processor failure or other component failure or false starts will shut down an operation.

If you put a .45 slug in it, it dies, so what? If you put a .45 slug in you, you die too! Wildly oscillating motors would be the least of your problems.

BTW fix your power supply, there's something wrong with it.

You could make the same argument for the "core" of a traditional (analog) motor controller. I think you'd be opening yourself up to a lot of liability if you didn't take the (traditional) approach of using (relatively) physically separate circuits -- even if they're implemented digitally in the form of, e.g., little "supervisory" CPUs -- to independently watch for fault conditions. I wouldn't personally market a motor controller that didn't have such features! (Something like the device Joerg worked on where the CPU just died if you stopped the clock... that's arguably "reasonable," -- in that no one gets hurt just because a few tiny transistors fry. It's the cases where the potential failure of a system can be "reasonably" expected to result in physical injury to other people or devices where you need to start adding that robustness to avoid liability problems... You're aware that every major IC manufacturer out there requires written approval from the president or similar before they'll sell you parts to use in life-support equipment, nuke reactors, etc., right?)

Human physiology is not always as robust as I'd like. :-)

I'm convinced it's working 100% "as designed." It's just that the design isn't robust!

They also outlive their components by many years, and usually have no illuminating discussion of the solution - merely a schematic, occasionally with a sketchy (and sometimes incorrect) explanation of its workings, often not. Actually quite similar to poorly written cooking recipes, which don't usually outlive their ingredients, but which often do have fictional requirements or poorly understood steps that are faithfully copied down by cooks that don't understand what's really going on, and thus what's important, what's not very important, and why.

Sure helps to dampen the enthusiasm of kids in school when they find a neat-sounding circuit they want to try, and it's A: not possible, since the components it depends on haven't been made for 20 years or B: it doesn't work due to some perpetuated error, which they don't have adequate knowledge (yet, if ever) to sort out.

--
Cats, coffee, chocolate...vices to live by

Reminds me of the time I was asked to build this

a 40khz tv light source repeater

I told my client,'Sure..but I'll cost extra to figure out if it'll work. That tuned circuit part looks quirky to me.' (I had little knowledge of photodiodes at the time.) Client says 'No..no ..just build it' I said..'There's no part number for the photodiode.' Client says 'There can't be that much variation in photodiodes.. Just try any generic part.'

So I did.. The circuit didn't work.

The photodiode + tuned circuit didn't produce enough signal to make the schmitt change states.

I ended up designing my own circuit that did work. Used a rail to rail comparator + mosfet. I didn't see any need for a tuned circuit. The optical source was to be right on top of the photodiode.

D from BC British Columbia Canada

The software equialent is to make something that works.. for yourself in your specific setup. "Hey it works.. must be alright!". Only for others to have it bomb on them because of incorrect assumptions and shortcuts.

Oh btw.. there is this company that ships products that isn't finished. But customers keep buying it like drugs. I think it's based somewhere in redmond :)

As for circuits, I rather see them as a suggestion than anything else.

As for #2 see:

for a deliberate movement in the schools to foster "self esteem" because of misguided studies that apparently showed that as a way to build real scholastic achievement.

A year after Pygmalion in the Classroom appeared, Nathaniel Branden published The Psychology of Self-Esteem, introducing another magic bullet and setting off an entire educational movement. Fostering self-esteem as Branden actually described it?an internalized sense of self-responsibility and self-sufficiency?could have been positive. But the movement focused instead on having a favorable opinion of oneself, independently of objective justification for that favorable opinion. From the 1970s through the 1990s, low self-esteem took on the aura of a meta-explanation. California went so far as to establish a task force on self-esteem, which predictably concluded in its 1989 report that ?many, if not most, of the major problems plaguing society have roots in the low self-esteem of many of the people who make up society.? And since low self-esteem was the problem, high self-esteem was the solution. The educational romanticists bought into it unreservedly. Children were to be praised, because praise fosters self-esteem. If criticism were unavoidable, the criticism should be cocooned in layers of praise, because criticism undermines self-esteem. Classroom competitions should be avoided, because they damage the self-esteem of the losers.

Once again, an appealing story turned out to be false. The landmark change in the scholarly consensus occurred in 2003 when a comprehensive review of

15,000 studies on the relationship of self-esteem to the development of children, headed by a scholar who formerly had been sympathetic to the self-esteem movement, concluded that there is no empirical evidence that improving self-esteem raises grades, test scores, or, for that matter, has any positive effect whatsoever. Once again, you wouldn?t know it by visiting classrooms. If anything, the assumptions of the self-esteem movement are more firmly embedded in educational practice now than they have ever been.

Robert H.

Yes, this is when teachers in grade school replaced the word loser with nonwinner. "Good job Billy your a nonwinner!" Mike

Joel,

I began adding the notice that all my published circuits had been personally built and tested after reviewers of two technical paper I wrote claimed that circuits in the papers would not work if actually built. What they didn't know is that I had built them and tested them extensively. One was an infrared travel aid for the blind and the other was a light flasher circuit. The aid for the blind had been tested with dozens of blind people and had won an IR-100 Award.

Simulation is fine, but it doesn't necessarily detect real problems that can come from circuit board leaks, inductance problems, noise, etc.

The circuits in my Radio Shack lab kits were built and tested multiple times, mainly to check for errors in the instructions I wrote. I built every circuit twice from the instructions alone and then Radio Shack engineers built the circuits.

Forrest M. Mims III

snipped-for-privacy@4ax.com...

e

ve

e know which

with

it

run

gn

snipped-for-privacy@4ax.com...

ote

have

ple know which

g

d with

cuit

g run

sign

Back in the day, there used to be this huge book of Op-Amp "recipies", many of which I suspect were unbuilt and untested. Likely some of that experience rubbed off in your direction... ?

-mpm

My mind is 7 3/8" (same as my hat size). I don't know if that small, medium, or large.

I hate big schematics, cut into small pieces. Trying to follow the signal flow, through literally hundreds of mislabeled connections, splattered over multiple pages containing oversized component symbols and unfathomable notes, is (to me) frightening. I much prefer to have a giant schematic, on preferably one or two pages, than dozens of small schematics. In their radio products, Motorola duz it right, with huge foldouts (sometimes 4ft long), containing as much of the schematic as practical.

Drivel: Once (and only once) upon a time, I decided to do something different. Instead of scribbling the schematic in the traditional left to right signal flow, I did it in the opposite direction. I find it difficult to read right to left schematics, so I figured it would do something between inspire comments to incite a riot. I was fully prepared with an arsenal of excuses (i.e. Hebrew goes from right to left). Absolutely nothing happened and nobody said a word. Drafting merrily laid out the circuit board, tech pubs merrily printed it as it was drawn, and none of the dealers ever mentioned it. Eventually, somone in drafting asked if I would do it again, as it was easier to layout the TO-5 transistors without having to run a trace to the base between the emitter and collector. Sigh.

--
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Yeah, me too... although I prefer schematics that were drafted for C-sized paper... or D-sized if absolutely necessary: C-sized schematics still look entirely readable on 11"x17" printouts, where D-sized is starting to get painful there... although D-sized on 13"x19" is still (just) tolerable. Of course, to some extent how readable things are depends on how large the symbols were drawn in the first place -- ORCAD's provided libraries tend to make everything rather larger than necessary, IMO.

I don't mind multiple pages where a conscious effort has been made to minimize connections between the page. For a radio it's easy to stick, e.g., the LNAs on one page, mixers/demodulators on another, power supplies on a third, etc. But if it can all comfortable fit on one or two pages, so much the better.

("right to left" schematics...)

Things like labeling net names clearly, bundling up signals into busses in a sane manner, and drawing reasonably clear symbols are generally a higher priority than which way signals "flow." Not to mention that probably 80+% of people just don't care at all -- I've had techs tell me that purposely placing two mechanically identical connectors far apart rather than snuggled up right next to each other is "stupid" because it caused them more layout work... and this after we'd previously blown up PCBs due to having identical connectors immediately adjacent to one another and people mis-plugging cables! It also took me awhile to convince some techs that shrouded or otherwise keyed connectors were a good idea since -- even though, again, we'd been blowing up boards due to cables being plugged in backwards -- they didn't want to have to kept more stock items around when, e.g., a 0.1"-spaced double row header can just be cut to any size necessary from a single long strip. (Interestingly, each the very cost-competitize PC industry does this these days -- things like IDE or floppy drive connectors are shrouded on all PCBs I've seen for a number of years now, even though way back in the late '80s and early '90s no one bothered... probably took some number of fried boards to convince them as well...)

---Joel

Say, I was going to snip this, but then I remembered a Simpsons quote.

Nelson is imagining himself in a laboratory, weighing the Three Stooges' brains and finding the heaviest. Nelson: "Hmm, Moe -- 3.2 kilograms!"

*Slaps across all three brains*

Oh, and let's not forget the favorite of tech manuals... sheets bigger than your desk, there's one small board per each page, and each signal goes EVERYWHERE before reaching its destination. It's like Jackson Pollack in black and white with a square. When there's a bunch of lines routed in generally the same direction, they are inevitably bunched together in a horrible jumble of pinstripes that makes following a signal a challenge for an experienced Where's Waldo reader! Maybe companies intentionally publish these horrible drawings in the hopes that repair techs won't be able to figure them out, thereby earning them more business. I can't bring myself to imagine they used the same drawings internally...

Tim

-- Deep Fryer: A very philosophical monk. Website @