Connect with us

a couple good lines

Discussion in 'Electronic Design' started by John Larkin, Feb 9, 2006.

Scroll to continue with content
  1. Jim Thompson

    Jim Thompson Guest

    “Hire some philosophers — scientists who appreciate simplicity,"
    Vreeland said. "Engineers love complexity too much and as a result
    complexity is running amok.”

    Isn't that bassackwards?? Philosophers (scientists ??) are the ones
    who give you hare-brained theories based on inadequate data.

    Had a call this morning from Raytheon (old Hughes Tucson) wanting
    information on my 1967 design MC4024 ;-) That's what non-failure and
    simplicity are all about.

    ...Jim Thompson
     
  2. Phil Hobbs

    Phil Hobbs Guest

    And your data for that statement is....? (Win could have a field day
    with that one, but I wanted to sound a nonpartisan note first.)
    Right, use only 53 transistors, and give them all catchy names. Your
    die size is now 1% of the area of one bond pad, but oh well. ;)

    Cheers,

    Phil Hobbs
     
  3. Pooh Bear

    Pooh Bear Guest

    This engineer appreciates simplicity, especially the elegant variety. It's a mindset thing IME.

    Graham
     
  4. Genome

    Genome Guest

    I like......

    "People don't kill chips, complexity kills chips," Chayut said.

    Chayut called for a way to start using all the transistors available to the
    designer, even considering implementing redundant architectural designs that
    would double up on the paths and in case of failures take over operation of
    the chip.



    Chayut was that Indian bloke who called up yesturday and tried to sell me a
    mobile phone. After I told him I'd give him two kettles for his daughter he
    pointed out that he was a Hindu but I trounced him by pointing out that his
    university wasn't as good as the Muslim one, I have inside information from
    a Muslim person. I then went on to ask him how I didn't know he was part of
    a 419 scam so why should I give him my bank details so he tried it on a bit
    and then I totally wasted him by telling him that his wife was probably a
    fat **** who didn't know how to cook and I wouldn't marry his daughter
    because she would make a similar instant transformation afterwards and then
    he broke into tears.



    DNA
     
  5. Joel Kolstad

    Joel Kolstad Guest

    There's an example from a former Tektronix engineer in the chapter in Jim
    William's analog design book on vertical amplifiers where he briefly discusses
    how they took a 300 transistor circuit (replicated 4 times over in an IC) and
    cut it down to about 30 (x4) while simultaneously improving specs that
    mattered and only reducing others (that didn't matter as much) a little.

    I imagine this kind of effort is still expended on analog ICs... On digital
    ICs, it seems that the synthesis tools are usually good enough that
    significant transistor reductions seem to come about more from
    architecture/algorithm choices than clever use of gates, 'flops, etc.

    Hey Jim, how many transistors do you see in the average analog IC you do these
    days?
     
  6. John Larkin

    John Larkin Guest




    Everyone knows that debugging is twice as hard as writing a program in
    the first place. So if you're as clever as you can be when you write
    it, how will you ever debug it?

    - Brian Kernighan
     
  7. Genome

    Genome Guest

    Oh no you don't....

    'It's simple'

    'But it's broken'

    'So, I copied someone elses and made it simpler'

    'But it's broken'

    'Yes, so it was their fault... I made it simpler though.. See, now it
    doesn't cost so much'

    'But it's broken'

    'What, like it's my fault you can't find the right salesperson?'
    .........

    DNA
     
  8. Jim Thompson

    Jim Thompson Guest

    50 years of observation.
    He couldn't have a field day with a corn cob ;-)
    Excuse me? I only count 46, and my original data book gave NO names.
    What pray tell are you looking at? There were other devices called
    4024... make sure you're looking at a TTL-compatible VCM.
    Don't you have that backwards? A bond pad would typically be 4x4
    (mils). IRC the MC4024 is about 50x50 (mils).

    ...Jim Thompson
     
  9. Jim Thompson

    Jim Thompson Guest

    That's the digital mindset.
    Sno-o-o-ort ;-)

    ...Jim Thompson
     
  10. Jim Thompson

    Jim Thompson Guest

    Most of the stuff I'm doing today is SOC (system-on-chip), and the
    device counts can run in the thousands.

    But your typical jelly-bean part, for example an OpAmp, is probably in
    the 15-20 device range (per OpAmp).

    My first OpAmp design, the MC1530/31 was 16 transistors (1963).

    ...Jim Thompson
     
  11. Guest

    I've heard plenty of hare-brained theories from engineers, based on
    totally inadequate data - medico's are worse, of course.

    Scientists are usually pretty good about relationg their theories to
    observable facts - after all, that is what they are trained to do.
    Where they are reliably hopeless is in design for production - a good
    scientific instrument only has to work long enough to collect enough
    data to lets it;s owner write one paper, and it doesn't matter if it
    has to be recalibrated by a graduate student before each reading, or
    left to cool off for four hours between readings.
    I used the MC4024 in 1972 - didn't like it much. The MC4046 was much
    nicer. Nowadays the MC4024 has all the notalgic charm of the uA709 -
    I've finally stopped having the nightmares where I'd have to use one or
    other of them in a real design ...
     
  12. mw

    mw Guest

    I don't understand this. First Chayut says complexity is bad. Then he
    argues for adding complexity.

    Why add redundant schemes to solve problems that may never happen? Then
    you have to test and troubleshoot even more complexity.

    mw
     
  13. Joerg

    Joerg Guest

    Hello Jim,
    Which company's? The low end stuff with some elementary circuits on the
    fringe of a uC is rather boring, nothing to write home about IMHO. The
    real stuff is probably the Marvell line and others. They often don't
    publish much but one has to go the old fashioned rep route.

    In the beginning SoC looked like buying a kit to bake a cake instead of
    doing it from scratch. But I assume that's not true anymore these days.

    Regards, Joerg
     
  14. Joel Kolstad

    Joel Kolstad Guest

    His definition of "bad complexity" is "any circuit that didn't show up in an
    undergraduate introductory course on design." His definition of "good
    complexity" is "brute force replication of functionality... if we try to do
    the same thing enough times, sooner or later surely it'll work?"
     
  15. Guest

    There is a least one bad line

    "People don't kill chips, complexity kills chips," Chayut said.

    I've seen plenty of bad engineers screw up relatively simple circuits,
    while good engineers can produce complex circuits that work like a
    dream, straight out of the box, if you give them enough simulation and
    prototyping time.

    I don't think complexity - is of itself - the real problem. With a
    complex circuit you do have to spend time breaking it up into more or
    less separate modules, and sorting out tidy interfaces between the
    modules, but once you've done that you can farm out most of the work to
    merely competent engineers.

    Life gets difficult when you work for pointy headed managers who can't
    cope with the complexity of managing non-interchangeable engineers, and
    try to insist on parcelling out the work on a first-come first-served
    basis.

    The sort of support software that people like Chayut are peddling
    doesn't make any difference there - software is only as good as the
    people driving it.

    On one occasion at Cambridge Instruments I had to take over a complex
    board that wasn't working right, despite being simulated to the hilt at
    the design stage - and when I looked at the original simulations, the
    fault was clearly visible, if not as bad as it turned out to be on the
    prototype board. Fixing it was easy - I relaxed the specification
    involved (since I'd written it, nobody got too fussed). It we'd caught
    ithe problem before the board had been laid out, we could have done
    better.
     
  16. Phil Hobbs

    Phil Hobbs Guest

    You were complaining about modern digital designs using too many
    devices. In a modern process, a 50-transistor chip would be too small
    to see, which is what I was getting at. (And a 15% error for the
    transistor count of a device whose schematic I've never seen isn't too
    bad, I don't think. I've never even laid eyes on a 4024, though I've
    used a few 4044s.)

    It would be great if it were possible to breathe on every small section
    of a big processor design and optimize it, but it isn't. This is not an
    excuse for sloppy design, or for schools turning out EEs who don't know
    which end of a soldering iron to hold, but people just don't have the
    capacity to manage that much complexity.

    Cheers,

    Phil Hobbs
     
Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day

-