Billy.Pettit at wdc.com
Thu Jun 8 19:38:23 CDT 2006
Tony Duell answered:
> However, the operation of a scanner is not too complex to reverse
Rather you than me. The only scanner I looked inside had several
> I guess that we still can't get you to accept one, even for free, without
> open hardware source.
Surely you all know me by now. I am not going to depend on something that I
I'm still glassy eyed about this statement. So I thought I would tell you a
little about some work I did in the past. In OKCity, I was responsible for
all the customer support documentation on the MPI floppies, cartridge and
fixed disk drives. In the early 80's I had to do a study and recommend ways
to reduce costs - publications had become a huge cost empire within
Engineering. And whole product families were being delayed while the
elaborate manuals were written and prepared. Customers were complaining
about the enormous cost of maintenance.
So I sample polled the largest customers for our drives. It was a couple of
hundred out of 2500+. At that time, the only ones who replied that they
needed schematics were a couple of repair services. We asked on theory of
operation, app notes, etc. Same thing. None of the customers paying the
bills used any of the elaborate documentation except for the interface
The conclusion reached was that schematics were not required and they were
dropped. If any customer complained, I was to supply him engineering
drawings as schematics. No customer ever complained. (Note the word
customer - the ones who spent the money. A few third party types weren't
happy, but they were a real thorn in our side then.)
Theory of Ops was dropped. So were wire lists. Parts lists stayed - always
like to sell parts at slight markups. (Sarcasm implied.)
At that time two other trends were starting. First, most products no longer
needed card cages. ICs allowed all the logic of a drive to be on one PCB.
And we certainly did not want end users mucking up those boards by trying to
repair them. (Also, many of the circuits were tuned in PCB fab; swapping
out an IC wouldn't always fix the problem. It had to be tuned to be
Second, the first ASICs were used; I believe the first ones for us were on
the 8" products. Physical size and board space were suddenly critical. The
shrink to the 5.25" Wren series was really tough for engineers who were used
to designing individual boards bigger than the entire Wren.
So - the MPI official policy was to no longer provide documentation. The
customer did not demand it, wasn't willing to pay for it. With ASICs and
single board drives, we didn't want them to repair the units. The customers
were pleased with the change - they were able to reduce their maintenance
staffs, their spare parts costs and their cost of ownership. I'm cutting
out a long section out here about the reliability of replaceable components
vs limiting connections, adjustments, power etc. It is the same argument
used on anything electronic.
The point to all this is that a series of conscious deliberate decisions
were made to prevent what Tony wants to do: repair to a component level.
Repair cost money, was unreliable and required massive support. The world
would only pay for cheaper product. The ASPs for disk drives moved from
$3000+ to sub $50 in 15 years while the capacity went from 16 MB for Wren I
to a terabyte today. A repairable drive was a luxury that nobody would pay
for. And as time has proved, was unnecessary.
Since that time, more than 20 years, Tony Duell is the only person I've met
who insists on repairing his personal work system to the component level.
Everyone else, wants it to be cheap and reliable. And that includes the
most fanatical military buyer.
What's next? My last three employers all had long range programs to get the
entire drive electronics and interface down to one piece of silicon. And
they want to eliminate the connectors as much as possible. Hard drives will
be soldered to PCBs before the end of the decade. Optical drives are a
little harder because of the form factor. They will achieve one piece of
silicon, but it will probably be mounted on the OPU and have a 4 wire serial
As everything moves to SOC (System On A Chip) designers become firmware
engineers. And already, there are plans to make firmware easier and more
modular. The goal here is to have the system design itself from a few high
I understand Tony's feelings - I've encountered similar viewpoints many
times in other fields. But it makes me sad, because he can't share a lot of
the fun with us. We can't send him scanned manuals and he can't send us
some of the work he has done. And what about photos from our conventions,
videos of the speakers, software that is on CDs not floppies? The exchange
of knowledge that bonds the rest of us is lost with Tony. And he has lots
of knowledge that we would enjoy seeing.
More information about the cctalk