BNC pricing

I think it is much more that aerospace is a more stable industry. While innovation is important, it is not so easy for a startup to come in and steal the lunch of the big players. Product life cycles are decades best developed by mature companies. So their employees tend to be less competitive but rather in the company for the long term and therefore cooperative.

The semiconductor business is still at a state where any day a new idea can revolutionize the industry and challenge the any player, any time. Product life cycles are months so there is lots of pressure to produce. So like the companies themselves, the employees tend to be more competitive and not looking for a long term relationship with their employer and not so cooperative with their colleagues or vendors.

Ship it with sacrificial M-M barrels!

All grist for the mill, thanks.

Cheers

Phil Hobbs

AFAIK for semi fab, that does nor work too well, even for something simple like transistors. Company X makes (say) a 2N2222 (which covers a multitude of sins) and US Army wants company Y to be second source. So,the same equipment used by X is bought and installed at Y and half of the production engineers are sent from X to Y for training and production tweaking. SIX MONTHS later (and same support), production from Y sort-of "matches" data sheet but items are not useful for circuits that need and use X parts. Do not remember the 2N type, but this is a true observation.

At least, for digital logic, one seems to get what i call digital results: works / does not work. Naturally, that could happen even if the test parameters are quite different between X and Y. ..and THAT has happened, a number of times.

"Copy Exact" is VERY hard to implement in practice. Doping levels,diffusion times, etc and etc can be EXACTLY the same in the new fab, and the results can still be s*it. And this also has happened. Reason as to why? Precisely because (after careful analysis) it was not "exactly" the same. Above is back to "simple-minded" transistor fab. The quartz tubes used for diffusion were NEW in the NEW fab; they were CLEAN and so gave a different-in-practice diffusion schedule.

One can almost guarantee that there will be something to "byte" the so-called "exact".

There is a guy that stands by to help the SNAFU..his name is MURPHY.

e" the

Semi technology and process control has come a very long way in those 50+ y ears. Intel's Copy Exact actually seems to work for them. It's sort of a gl orified version of a preflight checklist.

Does lead to some absurdities for their suppliers, but the equipment busine ss has always been brutal.

Cheers

Phil Hobbs

Soldering parts to a circuit board is a very different "process" from ion implanting and plasma etching. The semi QC guys don't seem to understand this. They want us to, for example, do statistical analysis of a power supply rail powered by a 7805, looking for mathematical trends that would suggest a drift towards a limit. Think hundreds of false positives on any given product. We test the supply, but we figure that if it's within the regulator specs, it's OK.

Tell me about it--I spent nearly a decade designing process control instrum ents, for contamination of various sorts and lithography critical dimension control.

Pareto charts, six-sigma, 'continuous improvement', all that sort of stuff . On the other hand, if you've ever been to one of those 7:30 AM production meetings in a fab, you know what kind of pressure those folks operate unde r. Having 500 boxes of wafers backing up in your sector while you frantical ly run monitor wafers to find some problem is _not_ my idea of fun.

Cheers

Phil Hobbs

Once you collect a bunch of parameter measurements (like my 5-volt rail voltages measured over a few hundred units) then there are holy wars about various bizarre algorithms that extrapolate disaster.

Good grief, if a 7805 regulator is out of spec, we'll replace it. It never happens, but you can frighten yourself if you analyze enough data points. If you analyze enough trends, and enough potential causal relationships, you WILL find scary trends and launch any number of Tiger Teams.

If you go to wikipedia, and start with statistical process control, you can get into a link maze of really strange stuff. A big company has to have a QC department, and it grows to be its own cult.

I really liked Copy Exact, which in the past generated a substantial revenue stream for me. The problem was that the computahs needed to also be exactly the same. After a few repairs on some truly ancient PC's, I suggested that they purchase a few old spare machines and monitors, from me of course. Finding an identical machine was not easy and seriously time consuming, but I managed to find a few old 386 machines that were identical to the one's in use. Although my markup was astronomical, I agreed to an expenses only annual service call to verify that they were still functional. The problem was that the machine was sitting next to an open tub of some mildly corrosive chemical. Fumes from the tub would slowly convert the PC case and circuit boards into rust. It took about a year to kill a PC. The spares were needed to keep the line running while I furiously searched for yet another identical replacement.

The problems could have easily been averted by simply moving the PC to a better location, or by substituting a more modern equivalent PC. However, Copy Exact (or whatever it was called about 20 years ago) prevented any such changes. It was discussed, recognized, seriously considered, but in the end, it was deemed better to have me continue my restorations than to risk loosing the process because of some obscure change. The farce continued until about 1998, when all the accumulated computers were declared to be a Y2K hazard. They were summarily replaced with nice new hardware and software without drama. However, all was not lost for me, again thanks to Copy Exact. They put the new computer exactly in the same location as the previous 386 machine, where it promptly corroded away, this time in about 6 months. I received a frantic midnight phone call asking if I had any of the old 386 machines that could be put in service immediately, while they investigated the problem with the new machines. My horde of 3 backup machines was delivered and installed in about 2 hrs.

Companies with "mission critical" computing functions often have an aversion to computer upgrades due to the cost and risks of downtime. They will try to keep the old hardware and software running for as long as possible. I used to help maintain a call logging computer at a PSAP (public safety answering point) that was running SCO Xenix, 10 years after it had been officially discontinued. Many machines, such as NC machining centers, are running equally antiquated hardware and software.

Yeah, I really liked Copy Exact.

How did you (or they) arrange that the copy had exactly the same rust as the original?

Competing with legal and HR. ;)

Cheers

Phil Hobbs

Not so much. It might have been a modest revenue stream, but wasn't to be. All their piled-on QC and Copy Exact nonsense was getting to be a nuisance. The last demand was a requirement that we submit a spectral response curve, well into the UV, on every unit shipped.

That happens in this business: you live off the good customers and learn off the bad ones. It was educational.

They did not buy the design; we had an NDA and we own the IP. It's not worth lawyering up to sue them. We have better uses for our time, like designing new stuff.

What's weird is that even apparently failed projects usually turn out to be valuable in the long run. That's one reason to just move on. It's not as if we had one Great Idea and lost it; we have scads of Great Ideas.

We have a very good corporate attorney, but we wouldn't bother him over something like this.

What's the expression? "Living well is the best revenge."

And software engineering, but the king is IT.

Cute. I/we/they didn't. It was generally recognized that rust accumulation was not desirable and did not directly cause a failure. The corrosion damage was primarily to the copper traces on the PCB's, copper wires, and component leads.

However, rust did precipitate a minor sub-farce. I could easily have had the cases sand blasted and painted to look like new. However, someone decided that there was a risk that the residual volatiles in the new paint might cause problems with the process. Therefore, I couldn't rebuild a rusted computer to look like new and had to use whatever I could find at various scrap dealers and used computah dealers. Some of my replacements were seriously ugly.

The most difficult problem was finding antique and small capacity 3.5" PATA hard disk drives that actually worked. I don't recall the exact sizes. At one point, I had a 256MB CF (compact flash) card installed inside a 3.5" PATA drive as a "functional equivalent" substitute[1]. I probably could have crammed the CF card inside the 3.5" drive by removing all but one platter and letting it spin, but didn't want to risk getting caught violating the Copy Exact dictum. So, I went to a disk drive data recovery service, and had them rebuild a pile of used drives with all new heads and platters, clone the firmware tracks, rewrite the servo tracks, do a primary format, and build a new bad track table. Don't ask what that cost. Average life proved to be about 3 years, which was good enough.

Ok but I wonder why it would not be acceptable to use "any" computer capable of running the required OS and software because it would not be an exact copy, while it would be acceptable to use the "same" computer but in different physical condition.

One could easily argue that the performance of two similar but not the same computers that both have zero defects is more similar than between to "exactly the same" computers, one of which is defective.

POTENTIAL customer. Screw them. If you can deliver the same product at the same price they got nothing to say about how you run your business. A deal like this smells like government.

I say let them stay potential.

Of course we all have our price, make $35 million a year with five people I might consider it, but nobody would offer that.

That one obligingly stole my design and went away.

A current customer, also in the semiconductor biz, now wants us to fill out a spreadsheet to document what it costs to make the products that they buy. They'll allow us 10% overhead on labor, 5% on parts, and a 5% final profit. We haven't had time to fill out that form.

Easy solution. Buy the assembled board from Lowland Technology at your normal list price.

That ridiculous BS has been tried in other industries and as you can well imagine it didn't work out so well for the folks that had the time.

--sp