Fork me on GitHub

Hardware

From:	SMTP%"leichterlrw.com" 21-MAR-1995 03:51:28.33
Subj:	RE: Sticking MV II cards into a MV3400

	I've got a couple of disk controllers (Emulex QD21 and Dilog DQ686) in
	a couple of MVII's that I want to junk.  I've got access to a MV3400
	and was wondering if the Q-bus backplanes are different on these
	systems.  I realize that the mounting equipment is different (MVII's
	in BA23 vs the MV3400 with a BA213), but I can tolerate some
	jury-rigging in this application.

The Q-bus is pretty standard; your boards should work.  The only Q-bus
systems that get a bit odd are on things like the whatever-4000-it-is that
provides a Q-bus interface.  It has some restrictions - but the problems are
in the interface to the rest of the system, not in the Q-bus itself.  Other
than that, unless you have some *really* old Q-18 stuff - well, DEC is pretty
good at designing and specifying buses and the Q-bus was at least their 3rd
generation (after the Unibus and the Omnibus).

							-- Jerry

From:	SMTP%"SMSprovis.com" 21-MAR-1995 04:59:18.20
Subj:	Sticking MV II cards into a MV3400

   One more caution.  While the boards are almost certain to work,
half-size (dual-height) boards must go into the top half of the BA213
backplane (rows A-B).  In the BA123, slots 5 and up have bus continuity
through both halves (rows A-B and C-D), so putting two small boards into
one slot can work.  In a BA23, my old books are not so clear, but the
same may be true for slots 4 and up.  In a BA213, it's one board only
per slot.  Otherwise, it can't miss.  We have mu-VAX IIs (BA123) and a
4200 (BA215 and BA213), and have run several boards in both.  (Moving a
board from a BA[1]23 to a BA21x is easier than going the other way,
mechanically.)

   Getting rid of any 8MB memory boards from your junk mu-VAX IIs?  I
could help.

------------------------------------------------------------------------

   Steven M. Schweda                    (+1) 612-636-8950  (voice)
   Provis Corporation                   (+1) 612-636-8951  (facsimile)
   2685 Long Lake Road                  smsprovis.com  (e-mail)
   Roseville, MN  55113-2537


From:	SMTP%"hoffmanxdelta.enet.dec.com (Stephen Hoffman)" 21-MAR-1995 16:13:02.90
Subj:	Re: Sticking MV II cards into a MV3400


In article <3kl7vg$cpisaba.info.ucla.edu>, ronpisces.fusion.ucla.edu () writes:
:
:I've got a couple of disk controllers (Emulex QD21 and Dilog DQ686) in
:a couple of MVII's that I want to junk.  I've got access to a MV3400 and
:was wondering if the Q-bus backplanes are different on these systems.  I
:realize that the mounting equipment is different (MVII's in BA23 vs the
:MV3400 with a BA213), but I can tolerate some jury-rigging in this application.


  ATTEMPT ANY AND ALL Q-BUS MODIFICATIONS AT YOUR OWN RISK -- CONTACT
  YOUR HARDWARE SERVICE ORGANIZATION FOR FURTHER ASSISTANCE.

  You may (will?) end up with RF emissions and/or RFI problems if the
  enclosure is not properly RF-sealed.

  First, if you are not familiar with working inside the Q-bus enclosure,
  or if you are unfamiliar with the DMA grant continuity, or if you are
  unwilling to make potentially permanent modifications to the card, or
  if you are unwilling to place other cards at risk due to an incorrect
  installation, or if you are unfamiliar with working with static-sensitive
  electronic components, DO NOT ATTEMPT THIS.

  Dual-width cards should install and operate in the upper half of the
  slot in the BA2xx and BA4xx system.

  If it is a quad-width card, what you are considering likely involves the
  electrical modification of a Q22/Q22 card to operate in a Q22/CD slot. 
  (In the BA23 and BA123, these Q22/CD slots are restricted to "memory" and
  "processor" modules, and Quad-width boards can *not* be installed in these
  slots.  The KA640 MicroVAX 3400 processor shipped in a BA2xx enclosure.)

  Quad-width cards may/will cause you problems in the BA2xx or BA4xx series
  Q22/CD enclosures -- you should be aware that a DMA grant continuity etch
  on the lower half of the card is *not* permissible in a CD slot.  (All
  BA2xx and BA4xx series enclosures are Q22/CD in all slots -- most (all?)
  Q-bus cards not intended for use in the BA2xx or BA4xx series will have
  this etch, and it will cause problems.)  If the DMA or Q-bus circuitry
  is connected off the lower etch fingers, you will have real trouble here. 
  If it is "just" the DMA grant continuity circuitry, some surgery on the
  card etch may allow the card to operate in a Q22/CD slot.

  I DO <**NOT**> RECOMMEND MAKING THESE MODIFICATIONS, NOR DO I RECOMMEND
  INSTALLING MODULES IN AN ENCLOSURE THEY WERE NOT INTENDED TO RESIDE IN.
  CONTACT YOUR HARDWARE SERVICE ORGANIZATION FOR FURTHER ASSISTANCE.

  ------------------------------ Opinionative -------------------------------
   Stephen Hoffman, NR EMT-I, WEMT, N1THN        hoffmanxdelta.enet.dec.com
        OpenVMS Engineering, Digital Equipment Corporation, Nashua NH


From:	SMTP%"newsspcuna.spc.edu (Network News)" 21-MAR-1995 21:40:15.23
Subj:	Re: Sticking MV II cards into a MV3400

In article <3kl7vg$cpisaba.info.ucla.edu>, ronpisces.fusion.ucla.edu () writes:
> I've got a couple of disk controllers (Emulex QD21 and Dilog DQ686) in
> a couple of MVII's that I want to junk.  I've got access to a MV3400 and
> was wondering if the Q-bus backplanes are different on these systems.  I
> realize that the mounting equipment is different (MVII's in BA23 vs the
> MV3400 with a BA213), but I can tolerate some jury-rigging in this application.

  Almost all MVII systems (MVII, VSII, VSII/GPX) were delivered in BA23 or
BA123 cabinets. The BA23 is 3 C-D slots followed by Q22 slots, the BA123 is
4 C-D slots followed by Q22 slots. The BA2xx/4xx cabinets used for the later
MicroVAX/VAX systems are only C-D slots. This means that any dual-height
card *must* go in the A-B (usually top) positions in those cabinets. Depend-
ing on the weight of the card and the stiffness of the backplane connectors,
the card may tilt out of its socket. There's a plastic baffle which looks
like a dual-height card that is used to support these cards as well as make
sure the airflow works properly. If you're only moving one dual card it
shouldn't matter.

  You've already observed that the mounting is different (the new systems
are S-box). DEC retrofitted some cards (the KLESI-SA is a regular KLESI
with a S-handle riveted on), redesigned some cards (the CXY08/CXA16/CXB16
are re-layouts of the DHV11/DHQ11 family), and left some cards alone (the
TQK70 is the same board in both cabinets). DEC sells blank S-box fillers
as well as some with cutouts, so you could probably do something that both
looked Ok and maintained the airflow and FCC compliance if you wanted.

  Some boards won't work in the newer CPU's. An obvious example is Micro-
VAX II memory. There are some more subtle issues, though. For example, many
older boards (like the TSV05 controller) and 3rd-party boards didn't work
in KA650/KA655-based systems (3500/3600/3800/3900) due to an error in the
Q-bus interface IC on those processor boards which DEC was very reluctant
to fix. Eventually DEC fixed it - if your system is on DEC hardware support
you should be able to get a new CPU (for the KA650 it's Rev. K1, I think).
If you're not on DEC hardware support, it's probably cheaper to buy a used
CPU on the surplus market (making sure it's up to rev) than to get DEC to
swap it on the time+materials plan. I don't think this affected the KA640
(3300/3400) but I'm not certain.

	Terry Kennedy		  Operations Manager, Academic Computing
	terryspcvxa.spc.edu	  St. Peter's College, Jersey City, NJ USA
        +1 201 915 9381 (voice)   +1 201 435-3662 (FAX)

From:	SMTP%"Info-VAX-RequestMvb.Saic.Com" 22-MAR-1995 16:20:55.54
Subj:	Re: Sticking MV II cards into a MV3400

Stephen Hoffman, NR EMT-I, WEMT, N1THN        hoffmanxdelta.enet.dec.com:
>In article <3kl7vg$cpisaba.info.ucla.edu>, ronpisces.fusion.ucla.edu () writes:
>:
>:I've got a couple of disk controllers (Emulex QD21 and Dilog DQ686) in
>:a couple of MVII's that I want to junk.  I've got access to a MV3400 and
>:was wondering if the Q-bus backplanes are different on these systems.  I
>:realize that the mounting equipment is different (MVII's in BA23 vs the
>:MV3400 with a BA213), but I can tolerate some jury-rigging in this application.
>
>
>  ATTEMPT ANY AND ALL Q-BUS MODIFICATIONS AT YOUR OWN RISK -- CONTACT
>  YOUR HARDWARE SERVICE ORGANIZATION FOR FURTHER ASSISTANCE.
>
>  You may (will?) end up with RF emissions and/or RFI problems if the
>  enclosure is not properly RF-sealed.
>
>  First, if you are not familiar with working inside the Q-bus enclosure,
>  or if you are unfamiliar with the DMA grant continuity, or if you are
>  unwilling to make potentially permanent modifications to the card, or
>  if you are unwilling to place other cards at risk due to an incorrect
>  installation, or if you are unfamiliar with working with static-sensitive
>  electronic components, DO NOT ATTEMPT THIS.
>
>  Dual-width cards should install and operate in the upper half of the
>  slot in the BA2xx and BA4xx system.
>
>  If it is a quad-width card, what you are considering likely involves the
>  electrical modification of a Q22/Q22 card to operate in a Q22/CD slot. 
>  (In the BA23 and BA123, these Q22/CD slots are restricted to "memory" and
>  "processor" modules, and Quad-width boards can *not* be installed in these
>  slots.  The KA640 MicroVAX 3400 processor shipped in a BA2xx enclosure.)
>
>  Quad-width cards may/will cause you problems in the BA2xx or BA4xx series
>  Q22/CD enclosures -- you should be aware that a DMA grant continuity etch
>  on the lower half of the card is *not* permissible in a CD slot.  (All
>  BA2xx and BA4xx series enclosures are Q22/CD in all slots -- most (all?)
>  Q-bus cards not intended for use in the BA2xx or BA4xx series will have
>  this etch, and it will cause problems.)  If the DMA or Q-bus circuitry
>  is connected off the lower etch fingers, you will have real trouble here. 
>  If it is "just" the DMA grant continuity circuitry, some surgery on the
>  card etch may allow the card to operate in a Q22/CD slot.
>
>  I DO <**NOT**> RECOMMEND MAKING THESE MODIFICATIONS, NOR DO I RECOMMEND
>  INSTALLING MODULES IN AN ENCLOSURE THEY WERE NOT INTENDED TO RESIDE IN.
>  CONTACT YOUR HARDWARE SERVICE ORGANIZATION FOR FURTHER ASSISTANCE.

I'm certainly not going to disagree with the sentiments, but FWIW I believe 
it should be possible to use a quad height card in a Q22/CD backplane, 
provided some care is taken.

Because of the way that the CD connections work, I believe that the effect
of the DMA (and BR) grant continuity etches will depend entirely on the
adjacent cards. Putting a double height Q-bus grant card (or double height
module) in the AB slots either side of the quad height card should mean
that these grant etches are not connected to anything. 

I think. I may be wrong.


Mark Iline	systemmeng.ucl.ac.uk
Dept Mech Eng, University College, London. UK


From:	SMTP%"hoffmanxdelta.enet.dec.com" 22-MAR-1995 17:40:04.66
Subj:	Re: Sticking MV II cards into a MV3400


    Hello Mark Iline,

>Because of the way that the CD connections work, I believe that the effect
>of the DMA (and BR) grant continuity etches will depend entirely on the
>adjacent cards. Putting a double height Q-bus grant card (or double height
>module) in the AB slots either side of the quad height card should mean
>that these grant etches are not connected to anything. 

>I think. I may be wrong.

   Ignoring mounting and EMI/RFI, "dual" modules placed in the AB slots
   will work correctly in either BA23/BA123 or BA2xx/BA4xx.  The "quad"
   Q22-bus modules are the problematic ones.

   The CD (lower) section of the BA2xx and BA4xx is not a Q22-bus, it's
   part of the PMI CPU-memory interconnect bus.  Shorting the CD-PMI
   pins in the position that matches the CD-Q22-bus DMA grant continuity
   pins is not something I'd care to try.

   The last time I was swapping Q22-bus boards between the BA23/BA123 and
   the BA2xx/BA4xx series, I had to cut the DMA grant continuity etch on
   the BA23/BA123 "quad" modules to get them to work in the BA2xx/BA4xx.
   (The fellow that designed the particular module I was working with had
   specifically placed a zero-ohm resistor -- a jumper -- across these two
   pins to make cutting this easy.)

   I don't have the KA630/KA640/KA65x module pinouts handy to check what's
   in that pin position in tbe backplane.  If I can dig up a copy, I'll
   check to see what might happen when those two PMI pins are shorted.

   Steve

  ------------------------------ Opinionative -------------------------------
   Stephen Hoffman, NR EMT-I, WEMT, N1THN        hoffmanxdelta.enet.dec.com
        OpenVMS Engineering, Digital Equipment Corporation, Nashua NH

From:	DRKCLU::IVAX         "Mark Iline - Info-VAX account" 23-MAR-1995 12:42:36.99
Subj:	Re: Sticking MV II cards into a MV3400

Hi Steve.


>   Ignoring mounting and EMI/RFI, "dual" modules placed in the AB slots
>   will work correctly in either BA23/BA123 or BA2xx/BA4xx.  The "quad"
>   Q22-bus modules are the problematic ones.
>
>   The CD (lower) section of the BA2xx and BA4xx is not a Q22-bus, it's
>   part of the PMI CPU-memory interconnect bus.  Shorting the CD-PMI
>   pins in the position that matches the CD-Q22-bus DMA grant continuity
>   pins is not something I'd care to try.
>
>   The last time I was swapping Q22-bus boards between the BA23/BA123 and
>   the BA2xx/BA4xx series, I had to cut the DMA grant continuity etch on
>   the BA23/BA123 "quad" modules to get them to work in the BA2xx/BA4xx.
>   (The fellow that designed the particular module I was working with had
>   specifically placed a zero-ohm resistor -- a jumper -- across these two
>   pins to make cutting this easy.)
>
>   I don't have the KA630/KA640/KA65x module pinouts handy to check what's
>   in that pin position in tbe backplane.  If I can dig up a copy, I'll
>   check to see what might happen when those two PMI pins are shorted.
>

I've got a fair understanding of the problem. I'm not 100% sure, but my 
understanding is that the pins on the CD side of the backplane are simply 
connected such that the top contacts connect to the bottom contacts on the 
slot above, and so forth.

I think this originally came about for dual card controllers like the RLV11 
(the RLV211 was a single card, supporting 22 bits) (I think), so that the 
two cards didn't need an 'over the top' connector. The same mechanism 
allows LMI (uVAX II memory) to have its private memory bus (although it 
still needs a ribbon cable), and PMI (11/83 ?), in which the memory cards 
sit above the CPU card, whereas the Q-bus cards are below.

If I'm right about the CD wiring, as long as the CD slots either side of
the non-skunked quad height card are free, you should be ok, since the
grant tracks aren't connected to anything.


Mark

Mark Iline	systemmeng.ucl.ac.uk
Dept Mech Eng, University College, London. UK

From:	SMTP%"leichterlrw.com" 23-MAR-1995 14:08:03.12
To:	Info-VAXMvb.Saic.Com
CC:	
Subj:	Re: 100% CPU usage slows down other processes

	> [I wrote:]
	> Short of giving the real CPU hogs lower priorities, there isn't much
	> you can do on a V5.5-2 system.  V6.0 adds support for a class
	> scheduler, which allows you to assign processes to different
	> classes, each of which can only get some pre-defined fraction of the
	> available CPU time.  For example, you could assign processes that
	> are using a great deal of CPU time to class B, and all others to
	> class A, and the specify that class B is to be allows at most 70% of
	> the CPU.  All the CPU-bound processes together will share the 70%
	> give to class B, while 30% will remain available for everyone
	> else.  Again, however, this is new to V6.0.  (Actually, the
	> mechanism has been there for quite some time, but support and
	> documentation are new to V6.0.  I don't know if the newly-documented
	> $SCHED system call actually works the same way in V5.5.)

	I'm suprised this thread didn't continue with more discussion on the
	above statement. Can you tell me where this class scheduler is
	maintained from? What documentation should I be looking at to utilise
	this V6.0 feature?

The System Services manual, what else?  The basic operational model is that a
privileged process first uses $SCHED (CSH$_SET_QUANT function) to establish
a group of classes and assign class quanta to them.  As processes in different
classes run, VMS will deduct the time they use from their class's quantum.
When a class's quantum reaches 0, no processes in that class will be
scheduled until $SCHED is used to assign more time.

Then the privileged process asks VMS to tell it about all processes
(CSH$_READ_ALL) or only about processes that have not been assigned a
scheduler class (CSH$_READ_NEW), and if it wishes it assigns them
(CSH$_SET_CLASS).  Once it's done, the process can request that an AST be
delivered when new processes enter the system (CSH$_SET_ATTN_AST).  (It must
also arrange to use CSH$_SET_QUANT periodically to refresh the times allocated
to the various processes.  As a safety "dead-man" switch, VMS will disable
class scheduling if too long a time - normally 30 seconds, but this is
settable - goes by without the use of the CSH$_SET_QUANT function.)

I've seen an example of a simple class scheduler, but I'm not sure where.  It
may even be in the VMS V6.0 SYS$EXAMPLES; more likely, it appeared in a
Digital Systems Journal article within the last 6 months or so.  The code
examples from those articles are available on line somewhere, perhaps from
Hunter Goatley's site, but I'm not sure.

By the way, someone pointed out to me that the code for the scheduling facili-
ty was not shipped before V6.0.  (I seem to recall that it was being used
experimentally within DEC several versions earlier, but probably only on
specially-built development versions of VMS.)
							-- Jerry

From:	SMTP%"hoffmanxdelta.enet.dec.com" 23-MAR-1995 14:59:50.98
Subj:	Re: Sticking MV II cards into a MV3400


    Hello Mark Iline,

>>   The CD (lower) section of the BA2xx and BA4xx is not a Q22-bus, it's
>>   part of the PMI CPU-memory interconnect bus.  Shorting the CD-PMI
>>   pins in the position that matches the CD-Q22-bus DMA grant continuity
>>   pins is not something I'd care to try.
>>
>>   The last time I was swapping Q22-bus boards between the BA23/BA123 and
>>   the BA2xx/BA4xx series, I had to cut the DMA grant continuity etch on
>>   the BA23/BA123 "quad" modules to get them to work in the BA2xx/BA4xx.
>>   (The fellow that designed the particular module I was working with had
>>   specifically placed a zero-ohm resistor -- a jumper -- across these two
>>   pins to make cutting this easy.)
>>
>>   I don't have the KA630/KA640/KA65x module pinouts handy to check what's
>>   in that pin position in tbe backplane.  If I can dig up a copy, I'll
>>   check to see what might happen when those two PMI pins are shorted.

>I've got a fair understanding of the problem. I'm not 100% sure, but my 
>understanding is that the pins on the CD side of the backplane are simply 
>connected such that the top contacts connect to the bottom contacts on the 
>slot above, and so forth.

>I think this originally came about for dual card controllers like the RLV11 
>(the RLV211 was a single card, supporting 22 bits) (I think), so that the 
>two cards didn't need an 'over the top' connector. The same mechanism 
>allows LMI (uVAX II memory) to have its private memory bus (although it 
>still needs a ribbon cable), and PMI (11/83 ?), in which the memory cards 
>sit above the CPU card, whereas the Q-bus cards are below.

    Dual cards are, as I've mentioned, not a problem -- assuming they
    are installed in the AB portion of the BA2xx or BA4xx bus, and
    assuming EMI/RFI is handled.

>If I'm right about the CD wiring, as long as the CD slots either side of
>the non-skunked quad height card are free, you should be ok, since the
>grant tracks aren't connected to anything.

    Various members of the MicroVAX family used *both* the over-the-top
    ribbon cable *and* the CD portion of the backplane to communicate
    with the memory modules.  I *know* the KA650 puts out the main memory
    address on the CD pins on the backplane, and reads/writes the data via
    the over-the-top ribbon cable interconnect.  The memory control lines
    are also present on the CD pins.

    If you're right and the pin on the bottom of one card in the CD fingers
    is daisy-chained -- connected to the pin on the top of the next card
    in the bus, etc., then (with a gap between the modules) you're right,
    the configuration may work as long as no adjacent card(s) also use the
    CD.  I would not want to bet on this, and I would not want to leave
    this system for the next fellow to maintain -- it's too likely that
    a simple subsequent module addition or removal will have rather
    unforseen and puzzling consequences.  And if the card accesses the
    Q-signals bus via the CD pins as I could envision some (mis)designed
    modules doing, all bets are (obviously) off.

    In the particular configuration I were working with, one of the
    requirements was stuffing as many modules as we could into the box,
    and cutting the grant was thus necessary -- since the modules were
    backed up against the memory modules.

    Steve

  ------------------------------ Opinionative -------------------------------
   Stephen Hoffman, NR EMT-I, WEMT, N1THN        hoffmanxdelta.enet.dec.com
        OpenVMS Engineering, Digital Equipment Corporation, Nashua NH

From:	DRKCLU::IVAX         "Mark Iline - Info-VAX account" 23-MAR-1995 15:38:42.63
Subj:	Re: Sticking MV II cards into a MV3400

>>I think this originally came about for dual card controllers like the RLV11 
>>(the RLV211 was a single card, supporting 22 bits) (I think), so that the 
>>two cards didn't need an 'over the top' connector. The same mechanism 
>>allows LMI (uVAX II memory) to have its private memory bus (although it 
>>still needs a ribbon cable), and PMI (11/83 ?), in which the memory cards 
>>sit above the CPU card, whereas the Q-bus cards are below.
>
>    Dual cards are, as I've mentioned, not a problem -- assuming they
>    are installed in the AB portion of the BA2xx or BA4xx bus, and
>    assuming EMI/RFI is handled.

Ah, but I wasn't referring to dual *height* cards. The 'dual card' RLV11
controller was composed of 2 quad height cards that communicated over the
CD interconnect. 

>    If you're right and the pin on the bottom of one card in the CD fingers
>    is daisy-chained -- connected to the pin on the top of the next card
>    in the bus, etc., then (with a gap between the modules) you're right,
>    the configuration may work as long as no adjacent card(s) also use the
>    CD.  I would not want to bet on this, and I would not want to leave
>    this system for the next fellow to maintain -- it's too likely that
>    a simple subsequent module addition or removal will have rather
>    unforseen and puzzling consequences.  And if the card accesses the
>    Q-signals bus via the CD pins as I could envision some (mis)designed
>    modules doing, all bets are (obviously) off.

Fair point about maintainability.

I must admit, I'd doubt that any quad height cards that are at all recent
would access Q-bus lines via the CD slots. There's no point, and Q/CD
backplanes were common around the time of the (pre-BA23) 11/23 (in fact
probably with DEC supplied 11/03s, precisely because of the RLV11. As
you've observed, modules were designed to work either in the 'serpentine'
Q/Q backplanes, or the 'straight down' Q/CD backplanes by having only
grants on the CD slots, and jumpers to disconnect these.

>    In the particular configuration I were working with, one of the
>    requirements was stuffing as many modules as we could into the box,
>    and cutting the grant was thus necessary -- since the modules were
>    backed up against the memory modules.

Surely not moving the graphics card on a VSII/RC (5 slot VSII) into a Q/CD 
slot to free up another double height Q-bus slot ;-) ? Naughty people like 
us took the araldite/epoxy out of the other 3 slots.


Mark