I once asked my boss why we were putting up with the time wasters higher management was telling us to do, rather than getting the real work done. His answer was that it was better to be fired in six months for not getting the job done than today for not doing what the big boss wanted. In six months there is a good chance someone else will be in even more (schedule) trouble.
Nice.
The last one made me smile - that's the "just get the job done" solution there.
-- John Devereux
Those little LTM8023 bricks are very handy. Price is OK, and they are reliable and very clean. They're good for 2 amps each, and parallel easily. The U2 site was just in case we needed more FPGA core current, which we don't.
We did make our own Cuk for -12, and bought the big ugly 24-to-12 isolator block.
Like the current shunts?
John
from the
for
worked until I
unmarked, but it
about the
illumination for the
antenna (not
the thing
a wire
a broken
wiring had
headers?
together,
cables?
than
great
10% done.devices
Those that I've looked at didn't look trivial, but they didn't look innovative either - more filling a gap in the market than creating a new market.
When you have a tolerably precise idea of what you can offer for a practicable price.
My specifications included a lot of discussion of what we needed to do and quite a few suggestions of how we might do it. If the guy who did the detailed design of the digital electronics had paid much attention to the suggestions, the maximum sampling rate might have been twice as fast, but he - correctly - decided that the project would never make it into production, and cut a few corners. It wasn't exactly a self- fulfilling prophecy, but it didn't help. Clever engineers can be a mixed blessing.
The risks that you are taking don't appear to be all that nerve- wracking.
Because a complicated board can take that long.
I reworked the electron beam tester timing circuits for pulse generator to drive a an electron-spin resonance machine - the coarse timing was to have been based on a 500MHz clock driving ECLinPS synchronous counters, and the fine timing on on the MC100E196 (which hadn't been available when we were building the electron-beam tester, and - with it's 20psec quantisation - it wouldn't - in any event - have satisifed my pointy-headed boss's demand for 10psec time resolution which drove us much further into new technology than was sensible, to the point where the project proposal that I'd put in based on GigaBit Logics GaAs parts had been written tongue-in-cheek to demonstrate that the 10psec specification was a "bridge too far").
Since the time delays generated by the MC100E196 are painfully temperature dependent, the board was to include hardware for converting these time delays into mark-to-space ratios, and eventually low-pass filtered analog voltages, which could be digitised to monitor the actual delays - all 128 of them - in a millisecond.
By then I could no longer buy ECL static RAM, so the programmed pulse widths were to be stored in CMOS static RAM and clocked out - four at time - to be translated to ECL levels and serialised.
It was a complicated board, and took several months to design. Latching the data to program the MC100E196 to dealy the pulse edges happened after the latching of the data to program the synchronous counters to generate the pulse edges, so it had to be doubled buffered. I recall that Jim Thompson ran into the same problem with one of his designs at the same time and asked for help here - which I was able to provide, but Google's advanced groups search can't find it, so it may just be a pleasing delusion.
-- Bill Sloman, Nijmegen
from the
on for
worked until I
unmarked, but it
about the
illumination for the
antenna (not
the thing
a wire
a broken
wiring had
headers?
together,
cables?
than
great
10% done.devices
this
and
things
stackups
Which I do before I design it, not after.
Why would I do that? I design things with a high probability that they will do what I intend. And the great majority do, without breadboards, on the first spin of the deliverable PCB. Engineering consists of building things that have a low probability of failing. Bridges, airplanes, skyscrapers, laser controllers.
a...
You can do 10 ps delay generation, or time-interval measurement, with
10KH ecl and an FPGA. Or even all FPGA, but the TCs will need compensation.That sounds like a bummer, when you spend so long working on a board that the parts become obsolete before you get it to work.
Sounds like a slow, expensive failure. We prefer to not do that.
John
The secret of success in management is to be into the next job, preferably with another company, before your mistakes get attributed to you ;-)
--
"For a successful technology, reality must take precedence
over public relations, for nature cannot be fooled."
(Richard Feynman)
Whenever a boss told me to do something I didn't want to do, I said "no" or just didn't do it. So far, it's only got me fired once, and that turned out for the best.
John
I remember one guy lake that. He played on the big boss' softball team, too. From Jr. Grunt to management in nothing flat.
It wasn't that what they were asking was wrong, just a waste of time. As we got later, the waste got bigger. Fortunately, my boss at the time was one of the best I've worked for. He felt his position in life was to hire good people and then run interference for them. ;-)
I
it
the
had
Desigining and building one of a kind test equipment.
?-)
As I said, you don't push the envelope. You are basically doing minor variations of the same thing over and over again, which does make the design process rather predictable, and correspomdingly less than interesting.
sSure. It's mostly turning the handle, but that kind of engineering isn't anything to boast about. Cambridge Instruments was the last place where I worked where we did stuff that pushed the state of the art, and did thngs that our rivals didn't seem to be able to copy.
When I was working at Nijmegen University, I introduced them to surface mount components, emitter-coupled logic and mixed signal connectors, some ten years after I'd done the same thing at Cambridge Instruments, who had been remarkably slow to get into surface mount components - I only did it because they were essential for sub- nanosecond rise-time circuits.
You can now. FPGA's have gotten a lot faster in recent years. 10KH ecl is a bit slow for that kind of work, and when I searched for examples of circuits that you could still buy, Google came up empty. Back in
1995 I published comment in Rev.Sci.Instrum. being rude about a paper that has made a fuss about 10KH being four times fast than TTL, when ECLinPS was four times faster than MC10KH - not their only or most serious short-coming by any means. It's difficult to see why anybody would use 10KH today unless they were seriously fixated on DIP packages.I'd bought ECL static RAM for the electron beam tester in the UK in
1989-91 and we'd had no trouble getting it to work. The system as a whole took a while to get running but the ECL memory wasn't one of the problem areas. In 1997, in the Netherlands, I found that the only manufacturer who still seemed to be making ECL static RAM - Cypress - wasn't prepared to even start talking about selling it to Nijmegen University, so I didn't waste any time designing it into the new machine.The design never got far enough forward to have had the possibility of working - after I'd put in roughly a year putting together a detailed design involving several boards, and then a second, rather cheaper detailed design, the customer within the university found out that he was closer to retirement than he'd realised when he wasn't allowed to take on any more graduate students who would have been the people actually using the machine. The design had been finished by then and I'd just started wresting with Orcad's peculiarly horrible new printed circuit package, which - at the time - wouldn't let you do manual routing at all, so I was a bit peeved. The programmer who'd been writing software for the system wasn't best pleased either, and went off and had a baby, which seemed to cheer her up, along with her recently acquired husband
It didn't get far enough to be a failure. I'm pretty sure that it would have worked brilliantly if we'd managed to turn it into hardware. The local practical problems had been sorted out when we up- graded the previous - all-TTL - delay generator with the minimum amount of ECLinPS and a few ECL-to-TTL and TTL-to-ECL converters, so I knew that we could put surface mount components on four layer boards with the gear we had at the university, and the conceptual design was essentially a simplification of the electron-beam tester test timing system, which had been working as intended in the prototype tester (along with the rest of the prototype hardware) for about a year when that project was cancelled.
-- Bill Sloman, Nijmegen
I wouldn't recommend the highway option, unless you're prepared for the consequences. There's little profit in it for anybody concerned. Just quietly get yourself out of there.
In my first electronics position, the only way to get a message upstairs was to resign with notice..........but I was (luckily) offered my job back after the issues identified were paid some overdue attention (ie crap product).
The same company suspended my ass three times, reinstated it twice and fired it three times .... all in one memorable week, a few years later.
Nothing anyone else could do about it; the department manager was apparently having personal problems. Refusing to sign a piece of paper that had nothing to do with the job. A smarter man would've taken my earlier advice.
RL
tly
.Stuff that you claim is new and pushing the envelope, mainly because you don't know all that much.
In the same way that you know the scientific evidence for anthropogenic global warming isn't to be relied on, and want us to ignore the risk that the global climate could get to be 2C warmer by the end of the century.
You are ignorant and over-confident, which doesn't damage your business as much as it might because you don't know about the bleeding edge stuff that is being written up in the academic literature, and think that it is cutting edge when it percolates down to off-the-shelf products that you can buy from Digikey. Not buying stuff that you can't buy from Digikey (or Farnell) is actually a pretty good strategy, but it does preclude you from claiming to push the envelope.
sf as
n't
Actually, the logic prompted Cambridge Instruments to can the project was not that it wouldn't sell, but that they couldn't rely on selling enough machines within the following eighteen months to give them the cash flow that they needed to cover the costs of getting the machine into production.
The situation was complicated by the fact that the pointy-headed boss who had insisted on the 10psec quantisation in the first place had been demoted from technical director to manager of the electron-beam tester project about half-way through the life of the project, much to his disgust, and he'd proceeded to betray the team by spending his last six months in the job setting up a new company (which still existed a few years ago) based on two of the companies better senior engineers (neither from our team), while failing to sell the electron beam tester.
When he actually resigned, the marketing department - who had hated his guts for years, not without reason, since he was a dishonest (if very intelligent) bastard - had had to go through his list of sales prospects and figure out who was actually likely to buy a machine.
They figured out that they could sell twelve in eighteen months, which wasn't quite enough. There were probably more prospects than had been listed - the pointy-headed boss hadn't been trying all that hard - but the marketing department were in a particularly sceptical mood at the time/
o y rThere was a management failure by the man who ran the science faculty workshop, who had let the project chew up a couple of person-years of work for the benenfit of a potential customer who turned out not to be in a position to use it, but the technology didn't get to a point where it could be said to have failed. I had essentially no contact with the guy who had commissioned the project - he had a long established relationship with the guy who had developed the amazing - if excessively handcrafted - TTL-based pulse generator that he was using at the time, and since I got on fine with our guy I was happy to let him do the talking. Obviously, I shouldn't have been, but I liked the people involved, and went out of my way not to upset them more than I had to. Implicitly telling my friend that he should have been using ECL ten years earlier was not something that I enjoyed doing, but I didn't have any choice about that.
-- Bill Sloman, Nijmegen
e
That would be 1981 when I was working for ITT-Creed in Brighton, on the communicating word-processor that was going to replace the TWX/ Telex machine (and eventually did - for a few years - in Germany and Sweden). I was spending most of my time reviewing software and communications protocols. We did point out to marketing the hardware would actually make a handy office computer - it was pretty much exactly equivalent to the IBM PC, but with an 8086 rather than a 8088
- but marketing didn't think that they could sell more than 150 a year in the UK into that market.
-- Bill Sloman, Nijmegen
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.