Q-bus I/O project

John Wilson wilson at dbit.com
Sat Oct 10 23:53:16 CDT 2015


On Sat, Oct 10, 2015 at 11:38:45AM -0400, Paul Koning wrote:
>Did you check the PHY latency for GigE?  I know it's insanely large for
>10G-BaseT, but I don't remember if GigE is still reasonable.

I was worried about that, since other PHY data sheets list high-ish numbers
(hundreds of nsec) as if that's good news.  But I can't find any mention
of latency at all on the Micrel data sheet.  (I chose this part just for
coming in a TQFP package that I can solder.)  So I guess I'll just try it.
Hopefully if the thing works at all, at least it'll support programmed I/O
and interrupts.  DMA would be nice but it'll have to have very fast
turnaround, which will probably need hardware help on the PC end (and
anything for PCIe is a much fancier project -- PEX 8311 or a big FPGA).

On Sat, Oct 10, 2015 at 02:10:25PM -0400, David Bridgham wrote:
>Bus drivers (along with voltage level conversion) have been a problem. 
>You seem to be using 74LVC parts.  Are they close enough?  Though I also
>see a bunch of FETs down there.

I'm hoping the 74LVC244s will work for receiving (I've used them for
TTL levels before), even if technically they don't follow the bus specs.
I have a lifetime stock of DS3862Ns (which do) but I'd rather go RoHS.
Yes the drivers are FETs, with gate resistors to slow down the rise/fall
times a bit.

>I've used OSHPark in the past for small boards but for 45 sq.in. boards
>they'd seem to be awfully expensive, even if using their medium run
>pricing.

Yeah it was $218.75 for three of these.  Which is a lot, but then again no
it isn't, given the size.  And they do such beautiful work...  Also I feel
bad about all the puny adapter boards they've done for me, where they
probably lost money.

>The fingers look gold plated.  It could be just the ENIG finish but I also
>see little tails running off the bottom as if you're setup for hard gold
>plating.  Does OSHPark have that capability?

Not that I know of.  This is just ENIG and won't last long.  What look like
tracks leading off to a plating bar are just me trying to be cute by
imitating DEC's finger shapes.

>I was hoping for the Pick 'n' Paste machine to come along and save me from
>having to do all that by hand but that project seems to have faded away.

Oh seriously!  I've been hoping one of the homebrew SMT assembly robot
projects would "take" too, at a reasonable cost.  Making prototypes is
bad enough ... but what if I get it working and then I'd want to make
dozens of these things, at 4+ hours each?

>I'm curious why you didn't use resistor networks instead of discrete
>components.

In this case I wanted to have FETs to disconnect both legs of each
terminator for when the board is being a peripheral (so a bussed R-network
wouldn't work), and anyway discrete resistors are cheaper and easier to
find in any value and wattage.  More sqinches but it's going to be a
dual-height whether it's full of parts or not so that's no problem.

Has anyone tried active termination on DEC busses?  I'm not sure how
to make the terminator power supply "weigh" enough to quickly resist
pulls in either direction, especially if it's a switcher.

>I also love that your design tool put omegas on the silk screen for the
>resistor values.

It didn't want to!  Each one of those is two superimposed strings, one with
spaces where the omega goes, and one in the "Symbols" font with the omega
(or mu).  OCD is its own reward.  Well I really do like being able to stuff
a board this size w/o looking at the schematic or parts list even once.

John Wilson
D Bit


More information about the cctalk mailing list