At 08:20 PM 3/20/01 -0500, Brian Hechinger wrote:
> > Problem isn't building a SCSI controller, but the fact that you want it to
> > look like an MSCP controller from the computer perspective.
>
>how does this complicate things?  i mean, are we talking major complications
>or minor ones?  if i could get all the specs for SCSI, QBus and MSCP would an
>electronics genius (i view him as such) have a chance at creating such a 
>beast?

Yes it is quite "doable" but requires some serious motivation. There are 
cheaper alternatives:
         1) Goto to government, college, and liquidation auctions and bid
            on everything that has DEC gear on it. Then rip it apart and
            look for SCSI controllers. If you bid right you'll find one after
            only $200 - $300 in bids, if you resell the left overs on Ebay
            you can probably break even (and get one for "free" that way)
            (you will also get literally a ton of vax stuff)
         2) Watch the "crannies" on Ebay, sometimes people sell systems that
            have SCSI controllers in them but list them as the complete system.
            Look in the pictures, ask the seller questions.
            I bought three this way in one transaction!
         3) Buy an old PC chassis, put a 10MB ethernet card in it and a couple
            of IDE drives and use NFS mounts. (this is the cheapest/fastest
            way to get SCSI)
         4) Buy some parallel I/O cards (these are often pretty cheap) and
            write a software SCSI driver.
         5) Design a Qbus card that talks MSCP and SCSI.

Now the latter will "cost" in the neighborhood of $3,000 (trust me, I've 
priced it) Fortunately you can do the whole thing these days with one FPGA 
and some
buffer RAM so the components are fairly simple. DEC has a patent on MSCP 
that has yet to expire, so you will need to either license the patent 
(probably darn near impossible) or sell only "plans" so that people can 
make one themselves (which is allowed by patent law)

If you get to the point where you have the board and it works, then you can 
probably sell 100 - 200 for $75 each, which after your costs will make you 
about $15 each. If you do sell 200 then you'll again make back your investment.

> > Having it talk raw SCSI isn't a good option. No boot programs exist, and
> > you'd also have problems booting it even if you did write you own boot
> > proms.
>
>hmm, well if it were simple it would have been done already. :)

KZQSA's are "cheap" for that reason. Direct SCSI on the Q-bus, no driver 
support.

>so:
>drive <- SCSI -> MSCP <- QBus -> VAX
>does that look like what we would need?

Yup. MSCP is documented in the UDA50 programmers manual and in the patent 
(I don't have the number handy but you should be able to find it fairly 
easily from the archives.) The Qbus is very well documented but its quirks 
are less so. I suggest the Xilinx Spartan II FPGA for the design because it 
is dense, relatively cheap, and infinitely reprogrammable.

--Chuck