Commodore TED (C16, C116, Plus/4) ve C128 tasarımında parmağı bulunan Bil Herd, Commodore'daki kalite tercihleri hakkında bir şeyler yazmıştı dün (16 Şubat 2019). Noktası virgülüyle dedikleri şöyle:
"So let me add to what I have said about quality and cost. As I have stated, I would not have spent an additional dollar not needed in order to make a system have supposed increased longevity beyond the design goals. A dollar not needed is by definition waste, and to put it in perspective the build cost for the C116 was between $9-$10 dollars not counting the box and skid.
That doesn’t mean that I didn’t build in quality by appropriately spending money, a part that is cheap and negatively affects yield, costs more than a part that does its job for a few cents more. Also when I did a cost analysis for my boss I included yield costs, I.E. the true cost to build something. I also included the cost to rework vs the cost to strip and throw in the trash. At these quantities you had to be dedicated to the truth, or the truth would hunt you down and beat the crap out of ya.
There are a lot of free or cheap ways to improve quality, being aware of isothermal lines and component spacing all affect component life, as does specifying a realistic working voltage. If I knew a cap could be exposed to 25V I would specify a working voltage of 50V (my rule of thumb is double it and then round either up or down to the nearest standard voltage) That’s not to say that purchasing and production wouldn’t later cost reduce the voltage for price or availability and I supported these activities as that was their job. To be sure we didn’t always get it right, and it wasn’t all that unusual to switch vendors only to see a blip on the QA numbers as our production quantities would quickly determine the real life attributes of a new vendor or part. When dealing with high quantities the truth comes out, usually quickly.
So would I spend an extra dime to improve quantity? Yes if my evaluation was that it was needed, and one of my dimes was probably equivalent to near a dollar for people not getting our quantity discounts. I also was young and tireless in my pursuit of a better/cheaper way, and on one project I really did find near a dollar savings by going to a custom inductor/transformer, all I had to do was accurately specify it but that was something I already knew how to do as a custom transformer company (I can’t remember the company name but the engineer’s name was Pat Pascale) had taught me back when I did instrumentation before working at CBM. I was kind of a natural high after saving a buck, that was a big thing for me, even though it was a somewhat unnatural way to cop a buzz. >
Speaking of instrumentation, I knew how to design equipment to very high standards, and that did help me to design cheap consumer goods, as long as I continuously challenged my own assumptions and history and was open to not-so-pretty or pure methods.
Speaking of which, my life changed when we got the C116 back in its final form from CBM Tokyo. I had wondered how they were going to deal with the heat of the 5V regulator. This was the days before Low Drop Out (LDO) regulators and we needed at least 3V drop meaning an 8V supply. The reality is by the time you calculate all tolerances and the AC range of 95-132 VAC you would have at least 4V. If the regulator supplied 1Amps and had a 4V drop that meant it had a dissipation of 4W.
What the engineers did as they put the final C116 together is they assumed that the system would always pull a minimum of 200ma (or so) no matter what (NO datasheet had minimums in it, nobody really tested for it). They consequently offloaded that to the 20 Ohm 5W power resistor you see on the layout of the C116. At .2A and 4V drop that means .8Watts (think heat) have been offloaded from the regulator. When I saw this technique based on a reasonable but new assumption of minimum current draw I giggled maniacally, and never looked back as I had in one instant learned to set aside the pride of over-designing instrumentation in lieu of Commodore cost sensitivity. They had also bent a piece of metal to the available space, typical CBM trade-off of high quantity where material costs were low and the onetime NRE cost of specifying a custom bent part was negligible.
Heh, If the part still ran too hot I was prepared to step in and specify a hard black anodized layer on the heat sink as shiny smooth surface is sub-optimal for heat transference, but it wasn’t needed. (I knew this from over-designing instrumentation)"
-- Bil Herd, 2019.02.16
Commodore
