At 11:55 AM 2/4/98 -0500, Allison J Parent wrote:
>A brief review of the problem:
>
>GIVEN:
> Q-bus VAX systems are common.
> MFM drives (RD5x) are getting old and scarce
> RQDX series controllers are usually in the box
> SCSI and other controllers are generally had to find and not cheap
>
>Problem:
>
> Hard disk storage with the following desirable characteristics:
>  
>	low cost
>	widely available supply
>	inexpensive or common interface
>	Bootable by standard vax rom boots(desirable)
>
>Solutions?
>
> SCSI and IDE drives are currentlly available and both can be had in 
> the .5...8Gb sizes at resonable costs(new) and wide availbility.
>
> Interface for both are now sub problems to be solved.
>
>   -IDE no available qbus interface
>   -SCSI existing boards expensive and most have boot or disk
>    size limitations.
>
>Which leads me to this:
>	
><Just in case the usual SCSI vs. IDE flamewar is in danger of starting her
>
>No intention as I could care less which but in the interest of a low cost 
>board(and disk system) I suggested it.  
>
><I just want to point out that the Q-bus is far slower than either of them
><the transfer speed difference, if indeed it exists, doesn't matter.  Anyw
><one argument I can see against IDE is the 18" max cable length (or whatev
><it is, short anyway), but in a BA23 or BA123 that could be easily met (ca
>
>It is a valid limit.  But most Q-bus boxes can accomodate several 3.5"
>IDE drives in one drive bay(even ba23) as they are far smaller than RD54
>and it would take 3 or more IDE(or 3.5"scsi) to equal the power and heat 
>load.  Outrigger boxes are the not needed.
>
>As an example of the above is I could put a TK50, RX33 with any 3.5"  
>above it in the bays of a BA23.  There is room, power and cooling  
>adaquate to support it.
>
><last time I had a multilayer board made we paid $70/each qty=20 for a boa
><that was a bunch smaller than a dual-height Q-bus board.  That's just for
>
>That is expensive for a board and some shopping can be much lower.  
>However, a down and dirty PIO (and possibly DMA) IDE could be done as 
>two sided (two layer) dirt cheap.  Many of the older q-bus cards are 
>two sided (DLV-11s, DRV-11B/P, RX11...).  In fact a simple PIO IDE
>would easily be wire wrappable or could be done using the DRV-11P
>logic as a foundation.  
>
>For those not familiar the DRV11P was a quad wide wirewrap card with 
>base Q-bus logic and DMA already laidin for custom designs.  The 
>circuits for that are in many of the Qbus PDP-11 books from the 
>early to mid 80s.  While those parts and boards are unavailable
>the logic is copyable and could be put in PALs/GALs or done with 
>common TTL MSI.
>
>Again I'm not advocating a flame war.  SCSI needs micoprocessor support 
>to get the protocal right, to keep the vax from really slogging and that 
>is a significant cost adder that does have it's advanatages.  Also SCSI 
>is a more expensive connector than IDE.  IDE does not require this and 
>IDE to Q-bus is a dirt simple (READ CHEAP) PIO interface. Sure it's not 
>bootable but what's needed to boot?  An RX50/33 off a RQDX1/2/3 or an 
>RX01/2 all of which are common and cheap and adaquate to load a booter 
>that knows this proposed card.  
>
>Now SCSI, if you can keep the intelligence to a cheap micro like the 
>8751 series would be cheaper to do than a 808x based design.  Again 
>performance vs cost is the battle.  With all that has been said a $200
>SCSI card is harder to achieve.
>
>In either case the less you put on the card and the less custom logic 
>the better cost wise. 
>
>The smc37c65 chip is out of production for a while now the SMC37c655 
>sort of replaces it.  Being focused at PC applications that chip also 
>has serial ports, parallel printer port, and IDE interface (and some 
>100 pins). 
>
>To be fair to SCSI there may be smarter chips out there to make the 
>task easier.  That would have to be researched (and their cost and 
>availability too!).
>

>
>Also shipping(customs, duties) across the ocean is a factor.
>
>Allison
>
It looks like IDE will be the way to go,  Allison has done a good proposal
here.  Now we need to collect up the data sheets and make them availible so
that a hardware design could be started.  I would suggest that we stay away
from surface mount components if we plan on making bare boards availible to
users to assemble themselves also while it may not be as steamlined it might
be better to stick with off the shelf thru hole components rather than
programible ones like FPGA or some such.  One thing is for sure the simpler
we make this project the more likely we are to get a board up and running!

Good Idea Allison!






****************************************************************************
************
Rick Copeland
rickgc@calweb.com                           Information Systems Manager
(916) 568 - 6744 x36 voice
Intermag, Inc.
(916) 568 - 6775 fax
****************************************************************************
************