Using two RL02 drives on OS/8? (cont'd) - test-drb@ccmp.vtda.org

25 Jan 2014

...
  On Sat, 21 Dec
2013 22:36:42 -0600, you wrote:

 The last time I had a chance to tinker with my
malfunctioning PDP-8/A  
with two RL02's was way back in Feb. 2013... here is the most relevant  
tail of the old thread:

  The connections from the drive select outputs on
the RL8A checked
 perfectly to the line receiver inputs on the RL logic board. While I
 appreciate Rick's taking the time to make me a selectable version of
 the RL02 oscillating seek.... he included a bug at no charge too ;)

 Specifically, the constant at 0230 (0100) selects the appropriate
 drive and resets it all right (0101 in AC). But - later on down the
 program, the SEEK (03) command is issued with another RLCB... however,
 the AC is forced to 0003 by the microprogrammed 7325.

 (As is probably apparent by now, the AC must have 0103 in it to select
 Drive 1).

 I edited the program to use a TAD 0231 and put the constant at 0231.
 So instead of the CLA CLL CML IAC RAL (load AC with 03 the hard way,
 an old-time DEC programmer's way to save one word when memory was a
 precious commodity) :) the AC now loads with 0103 and lo and behold,
 Drive 1 seeks merrily away! Restore the two constants to Drive 0 and
 THAT drive seeks. OK.

 However, all was not lost... during the couple of hours of chasing my
 tail wondering where the LSB of the drive select was going, I found
 that I had inserted the header into the RL logic board crookedly and
 bent two pins, one in the wrong hole and one shoved aside! :(

 Not only that, the line driver chip which provides the drive select
 signals had come from the factory with NO solder at all on its Vcc pin
 16!
   =:^O
 Fixed that too.

 I think the OS/8 packs have been wiped out by now... various FAULT
 lights coming on, especially on the 2nd drive in the chain... going to
 build another system pack with VTserver which takes a good half-hour
 or more at 19200 baud. Hope the servo tracks are ok, otherwise I'll
 need to buy a couple of good packs! OK. Last week I finally fired up the 8/A again.
The problems had  
actually gotten worse during the prolonged rest... I couldn't get  
either drive to work this time (fault lights on both, as soon as the  
Ready lights come on, even before any accesses). After much hair- 
pulling and 'scope troubleshooting, I finally found (inside drive 1)  
that I had improperly crimped one of the IDC connectors to the ribbon  
cable, inside the bottom drive in the rack, where it was making a  
variable resistance short between the two sector pulse drive lines!

So, having fixed that, I verified that Rick's oscillating seek program  
works on either drive without faulting. Enter the constants for drive  
0, drive 0 seeks no matter whether the first or second drive in the  
chain as long as the unit 0 plug is inserted. Same for drive 1. Both  
are not being selected simultaneously.

I also entered a short program to get the two status words, and that  
reads the status of either drive correctly (idle, spin up, locked on  
track 0, spin down, write protect button pushed).

Now, the original problem is still there, that OS/8 won't boot and the  
RL02 faults (with either the original OS/8 pack that may have been  
wiped, AND the remade one). Doesn't matter which drive is used as  
Drive 0, or whether the new or the original OS/8 pack... Hit the BOOT  
switch and Drive 0 audibly moves its heads, the fault light flashes  
briefly, then it goes back to being ready (and of course OS/8 is not  
running).

The interesting part is if both drives are loaded, then the fault  
light on *both* drives flashes, then they both go ready again!
I haven't recently tried disconnecting the second drive cable and just  
running with one drive, but that wasn't working before, either.

I am not going to be able to sort this out (unless I use my ancient  
Tek 7D01 logic analyzer triggered by the fault signal) without a  
diagnostic pack...

Does anyone have the diagnostic program set for PDP-8 and can put them  
on an RL02 disc pack? I could either mail a disc to you, or I could  
pay for one if a spare is available.
thanks for any help!
-Charles 
OK. Got a serial link to my PC going, and successfully downloaded various RL02 diagnostics
from the "diagpack2.rk05" image. The seek test works, as does pack verify.
However, either drive 0 or 1 flags an error on the read/write test promptly.:

WD1 0235
WD2 0000 (the two status words from the RL02, and they are normal);
ER 2001 (if I'm interpreting this right, this is the Error Register which is showing
only an OPI - Operation Incomplete error);
CB 0115 (<----------- what is this? ----------)
CA 0000 (cylinder address)
SA 0100 (sector address)

As I posted on another thread, having instructions for these AJxxxx routines would really
help, although I can figure much of it out from the RL02 manual, especially if I just use
the defaults when prompted.

What I think this is telling me is that the problem is somewhere on the controller card,
as neither drive will write but both can read without errors.

There is a "Diskless RL8A test" (AJRLAC) but I can't figure out how to use
it. Anyone got instructions for this one? thanks :) 
Read over the RL8A Technical Manual, specifically the write section.

CB = Command Register B, 0115 is Drive 1, memory field 1, Write
command.
CA = Command Register A (which also happens in this context to be
Cylinder Address :) Head 0, Direction Forward, Cylinder 0).

So - the proper write command is being issued on the first
cylinder/sector address of the platter, but OPI (operation incomplete
in 200 ms) is the only flag returned. That tells me the fault is
likely somewhere in the write logic as it appears that nothing is
being written at all, not just incorrect data or at the wrong time
which would set other error flags. Will report more when I get another
chance to work on it! 
I think I found the problem (aside from the defective cable previously
described), and "Y'all ain't gonna believe this"...

I toggled in a couple of simple programs to write each register
(Command A, Command B, Word Count, Sector Address) and read them back.
On a few occasions I noticed the SA was not reading the same as
written. And... the returned bit pattern changed when I moved the RL8A
card!

As it turns out, my quad extender card was not lined up perfectly in
the chassis (which is hex-height, and so is the RL8A, but it only uses
four of its six finger sets). I put the RL8A directly in the slot and
re-ran the diskless controller test and it worked flawlessly. Also the
read/write tests now work without all those errors too.  I can't
believe I spent hours and hours chasing this :P

However, my OS/8 pack still won't boot. Interestingly, the same boot
behavior is seen with the scratch packs that I used on the read/write
tests (flip the boot switch, the heads move, the fault light flashes,
drive goes back to being ready). So it looks like the OS has been
wiped out and it's time to make another pack on the 11/23+ with
vtserver. We'll see how that goes soon...