Connect with us

PCBs correct first time?

Discussion in 'Electronic Design' started by Leon Heller, Nov 25, 2003.

Scroll to continue with content
  1. Ian Buckner

    Ian Buckner Guest

    I did a board once where the documentation for a through hole
    custom chip (255 pin grid array, one pin missing as a "key") did not
    specify which way up it was. Yup, did it the wrong way.
    Fortunately, I was able to cut one pin off and load it from the
    wrong side of the board - took me some time to convince
    myself it would work.

    Then there was the time the layout person created a footprint
    for a 256 pin QFP from the vendor's data, which said:
    Pin spacing: 0.020" (0.5mm). Guess which was the _real_ dimension.
    That one meant that the pads were one pin off along the length
    of each side. Boy, was I popular with the tech that had to bend
    each lead just a little bit to fit ;-)

    Regards
    Ian
     
  2. maxfoo

    maxfoo Guest

    most chip companies will give or sell an evaluation board, hook em all
    up and test em as a prototype of your final layout.
     
  3. No, I let them decide what to do. So 32mil holes maybe get drilled
    as 35. I trust they have a good standard rack, and good common sense.

    Once made a mistake, thought I had box-headers with machined pins
    for 32mil holes, but it were square pins that really need 40mil.

    I was able to use a 'press&fit' method for the protoboard, requiring
    quite a bit of force (used a vise), but payed a bit more attention
    to the production run ;) Re-drilling is not an option, drills the
    plating out.
     
  4. John Larkin

    John Larkin Guest

    If all the eval board does is prove that the datasheet is correct and
    that the part works, why bother? I haven't used an eval board ever,
    and that covers a couple of hundred million dollars worth of product
    design.

    How would you incorporate into a breadboard an eval board for a
    650-pin FPGA running at 120 MHz? Complex, high-speed designs can't
    sensibly be breadboarded. They need the same correct-the-first-time
    discipline that ICs need, for much the same reasons. One saving grace
    here is the FPGA, which can be reworked an unlimited number of times
    without pads falling off the board.

    First-cut FPGA designs seem to always have lots of bugs, as does
    software, because the designers *know* they don't have to get it right
    the first pass, and that they can fix bugs in relative privacy. But
    hard ASIC designs tend to be right the first time, or at least have
    incredibly low per-function error rates, because the consequences of
    design error are so expensive and so publicly visible.

    John
     
  5. EEng

    EEng Guest

    I don't. In my readme1st.txt file that accompanies all design files
    to the board house, I specify that all dimensions are FINISHED
    dimensions. It is their responsibility to adjust accordingly and most
    good board houses know this. For a .035 hole, I know they're going to
    add to the drill size so that when thru plated, it comes out .035, and
    they know I'm going to demand they re-do the job on their dime if they
    don't give me what I ordered.

    As designers, we have the responsibility of ensuring our designs do
    what they're supposed to do, the way they're supposed to do it, just
    as board fabricators know that is their responsibility to give us what
    we ordered. I won't do biz with any board house that requires I make
    the adjustments for them. Am I supposed to ask each board house
    whether they hot oil level or hot air pressurize? Its not my job
    except to specify that these are finished dimensions. Let's keep the
    responsibilities where they belong, and frankly if I don't get what I
    ordered, I'm not paying for it, and you know they understand that.
     
  6. EEng

    EEng Guest

    That only proves that THEIR board is correct, not yours.
     
  7. John Larkin

    John Larkin Guest

    I find it fun to glue up the cardboard PCB dummies and then make the
    whole product enclosure as a cardboard-and-duct-tape mockup. Install
    fans, get some incense sticks from the Chinese gift shop around the
    corner, and this is great for air flow fiddling. You can install real
    heatsinks with bolt-on power resistors and measure thetas. Air flow is
    far more peverse than electromagnetics, and the results are often (at
    least to me) wildly counter-intuitive.

    John
     
  8. John Larkin

    John Larkin Guest


    Here's some of my standard stuff; the OEM things are even hairier.
    Every board you see here was shipped to paying customers while still
    rev 'A'.

    http://www.highlandtechnology.com/

    We did a very complex picosecond-precision controller box for an
    excimer laser (used for a serious fraction of the world's IC
    lithography), from first customer contact to delivered, working first
    article in 60 days. We dropped everything else and put all five
    engineers on it (*and* its full-rack test set!); we were very careful
    to get it right the first time because there was no time to spin it
    again. Every single circuit design was new.

    Looking back on the project, it was amazingly efficient, precisely
    because we *had* to get it right the first time. Firing up the first
    unit was wonderful: things just worked, bam-bam, like dominoes falling
    over.

    None of this is miraculous; anybody can usually get things right the
    first time it they really want to. If the people who put up buildings
    were as sloppy as most EE's and programmers, we'd all be camped out in
    rubble.

    John
     
  9. I keep meaning to try your suggestion, but my circuits have been
    mostly boringly low power and the temp. differences (while sometimes
    important) are easier to "see" with a ribbon surface probe. Things get
    strange when you go below a tenth of a degree.

    As well as the incense, maybe adding some fresh oranges to the
    cardboard dummy to make it into a shrine of sorts would improve the
    chances?

    Best regards,
    Spehro Pefhany
     
  10. Fred Bloggs

    Fred Bloggs Guest

    Can you give us a link to this ribbon probe?
     
  11. I agree, and that's exactly what I do for production boards, but for
    prototypes, doing that can double the cost, because they charge an
    extra . Fiddling the hole sizes involves some risk, and it means
    another difference between the prototype and the production boards,
    but risk/cost it seems worth it. AP circuits is one of the most
    restrictive with only 8 drill sizes standard and $9 US extra for each
    size that is not in the standard set. I might not care if some of the
    parts are bit looser than optimal to save $72, especially if it will
    still pass the annular ring DRC check.
    It's a cost-benefit thing. If I want an EXACT prototype (routed
    outline, V-grooves, panel size the same as production, hole sizes and
    finish identical, solder mask, solder paste? and silk screen
    resolution identical and so on) then it's going to be SO expensive
    that it may not make sense $-wise unless the run is going to be rather
    large. There are different prototype houses that do that sort of work
    (including carbon printing, dielectric printing, milling, various
    plating options and so on). I've seen cases where the prototype cost
    almost as much as 1,000 boards, which seems a bit inefficient if
    that's all you want.

    Best regards,
    Spehro Pefhany
     
  12. Mike Page

    Mike Page Guest

    Can I ask what field you work in ?

    I would put the cost of a board spin far higher than $100, which would
    only cover the plotting charge. You have to pay for actual PCBs on top.
    Then you have to cost the time involved in generating and releasing /
    archiving the schem, BOM, and artwork. You have to order the parts, pay
    for them, and finally assemble them. You or someone you pay to do it for
    you. How can all that be done for $100 ? Granted you will get some
    economies of scale by testing several variations within a build.
     
  13. Greg Neff

    Greg Neff Guest

    On Thu, 27 Nov 2003 14:42:37 GMT, Spehro Pefhany

    (snip)
    We look at the drill table, and if there are some drills that are
    close (within a couple of mils) then we will edit the pad stacks in
    the PCB layout accordingly. We then re-run all the DRCs, and
    regenerate the gerber files. This keeps the number of drill sizes
    down, and can avoid having to load an extra set of drills in the
    machine.


    ================================

    Greg Neff
    VP Engineering
    *Microsym* Computers Inc.
     
  14. Greg Neff

    Greg Neff Guest

    On Thu, 27 Nov 2003 10:48:32 -0800, John Larkin

    (snip)
    (snip)

    We did a custom protocol converter in 10 days, including PCB,
    enclosure, software and documentation. This was from first customer
    contact, through prototype delivery, to delivery of 15 production
    units. The customer paid dearly for this, and this project was very
    hard on our staff and suppliers. We are used to tight schedules, but
    this was absurd. This project was probably an order of magnitude less
    complex than the one you describe, but as you said there was no time
    for a second spin.

    We find that most errors are caused by rushing, IOW not following the
    procedures and checks. Rushing results in mistakes. We call it the
    RTF (Rush To Fuckup) ratio. We don't have an equation for this yet,
    maybe someone can offer something up.

    A PCB can be built with no errors, if all of the right tools are
    available for simulation, and if all of the procedures and checks are
    in place and properly followed. It's a function of resources, skill,
    and discipline.


    ================================

    Greg Neff
    VP Engineering
    *Microsym* Computers Inc.
     
  15. Okay, but that changes the production boards, rather than just the
    prototypes. It's a good way to check for sizes that are close, but
    different, especially on things like mounting holes.

    Best regards,
    Spehro Pefhany
     
  16. Greg Neff

    Greg Neff Guest

    On Thu, 27 Nov 2003 23:08:13 GMT, Spehro Pefhany

    (snip)
    Right, I missed the bit in your post about the 'proto house'. We use
    the same PCB manufacturer for both prototypes and production. What
    drills are in the standard rack? What's the penalty for using
    something not in the rack?


    ================================

    Greg Neff
    VP Engineering
    *Microsym* Computers Inc.
     
  17. Pick AP circuits as the whipping boy, as they're about the most
    restrictive (only 8 standard sizes) with their minimalist proto 1
    service:

    #70 .028"
    #65 .035"
    #58 .042"
    #55 .052"
    #53 .060"
    #44 .086"
    1/8 in. .125"
    #24 .152"

    Beyond that it's an extra C$13.50 per size.

    Best regards,
    Spehro Pefhany
     
  18. John Larkin

    John Larkin Guest


    I think what's being debated here is whether it's worth the 'trouble'
    of trying to get it right the first time. My opinion is that the
    habits acquired in the quest for correct-the-first-pass have huge
    longterm benefits, and, with experience, the actual effort is minimal,
    certainly less than the time wasted in doing it a second or third
    time.

    The very first powered flight of the space shuttle took off vertically
    and got a full crew to orbit and back safely; that was no prototype.
    It was later, when the culture got sloppy, that they began to crash.

    John
     
  19. YD

    YD Guest

    When in doubt print out top and bottom layers, tape them together and
    check against the parts.

    - YD.
     
  20. budgie

    budgie Guest

    I gather from your post that you are in the A-E category, not F.

    I don't claim to be perfect, but I do put in that effort to get it
    right first time. Although a quantum leap below JL's stuff, it still
    wastes my time and resources, as well as my clients' time, if I stuff
    up and have to do a go-round. As John put it, "None of this is
    miraculous; anybody can usually get things right the first time it
    they really want to. " Generally I choose to.

    In software, as someone else pointed out, the cost/embarrassment/time
    implications of software ****-ups are in a different universe to
    hardware. Most often, software foul-up recycle time is measured in
    minutes.
     
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

-