[Simh] RD10 is probably not the Burroughs IA2

[Timothe Litt]

On 27-Feb-16 14:19, Bob Supnik wrote:
> Thanks, Tim. Burroughs made a lot of fixed head disks at the time. I
> can't identify the model, but the IA2 (see page 7-4 of the B6700
> Hardware Handbook, on bitsavers) seems like a possibility. It has 7552
> sectors per surface vs 8000, but Burroughs sectors were longer than
> DEC sectors (180 x 8b = 1440b vs 32 x 36b = 1152b), so perhaps DEC
> format had more sectors per track.
>
Maybe.  But I don't think it's the IA2.  Probably the previous
generation Burroughs disk.

The RC10 manual (DEC-10-I5AA-D) states that the disk makes one
revolution in approximately 34ms. 
(17 ms access time).  That corresponds to 1765 RPM, an unlikely number
since motors tend
to be a multiple of the line frequency (50/60 Hz).  1800 RPM is 33.33
ms/rev.    That's "approximately
34 ms", and my guess at the correct speed.

The IA2 spins at 1500 RPM, or 40 ms/rev.  Although that works at 50 Hz
(1250 RPM, 48 ms), it's
possible that the spec quoted in the handbook is for 50Hz.  In that
case, the motor run at 60Hz would
be 1800 RPM, and the numbers work.  I didn't see power requirements in a
quick look thru the handbook,
though it does say "Printed in the USA."  50 Hz specs would be a
stretch.  So maybe.  But doubtful.

I doubt that DEC would have misstated the rotational speed and access
time by 18%.  So either
Burroughs specified the IA2 at 50Hz, or it's a different drive.  Next, 
let's look at the geometry.

Back of the envelope:  The IA2 has 151 segments/track vs. 80
segments/track for the RD10.  That's almost
double the number of sectors/track.  Looking at the bits: 1440/1152 -
Burroughs has just 25% more
(data) bits.

The sector overhead (2 chars of sector address + 1 LPCC + ?? analog
trailer/sector gap) is small.   I think
more than 80 sectors/track would have fit.  Switch to a bigger envelope: 

   Burroughs: 1458b/sector * 151 sectors/track = 220,158 data
bits/track.  Say another 100 bits/sector for
                        analog stuff (e.g. write gap).  Call it 235,258
raw bits/track

   DEC:             1170b/sector * 80 sectors/track = 93,600 data
bits/track.  101,600 raw bits/track.
                        That leaves 133,658 bits left over - more than
were used!  We could have fit another ~100
                        sectors/track at that density.  Even if we
wanted to stick with 2 digit BCD sector address,
                         surely we'd have gone to 90 sectors/track.

There is a Burroughs paper from 1963 where they indicate that they used
8 bits of overhead per WORD.
(48 data bits + 6 check bits + 2 "spacer" bits = 56 bits/word.)  There's
no indication that the RC10 did
that, though 6 bits of LPC/sector does seem skimpy.  But even if it
did,  the data bits/sector would rise
to 1,682, or 134,560/track.    So 142,560 raw bits/track.  We'd still
get an extra 55 sectors/track.
(And this doesn't account for the extra bits that Burroughs would have
had, as their handbook is
 in net capacity.)  We can't explain away the density difference that way.

That paper also indicates that they used a dedicated clock track -
whether they still did that 5 years
later is unknown.  The RP02/RP03 used a dedicated surface. 

Digging further, the RC10 manual tells us that the drive-controller
interface is parallel; any clock
recovery is handled in the drive.  And there's reference to an 'address
track', so probably the
sector overhead is just the LPC.  That doesn't change the analysis. 
There's an unexplained factor
of 2 in density between the RD10 and the IA2.

So unless my arithmetic is wrong, or there was a serious "marketing
consideration"  - I don't think it's the
IA2....  Leaving that much unused capacity on the table doesn't make any
sense.

The RD10 (probably) spins faster with (definitely) lower bit density
than the IA2.  Given circuit speeds,
that might have been a reasonable trade-off.  But since bit density has
always driven disk design, I'd
guess that the RD10 was the previous generation Burroughs disk - at 1/2
the density.  They probably
slowed down the IA2 so the controller could handle the increased bit
rate from the heads.

I haven't found a better candidate.

> While the 18b- and 36b-groups used the same disk buyout, they went to
> different vendors for drums. The Type 24 and RM09 came from Vermont
> Research; the RM10B from Bryant.
>
> /Bob


-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4994 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://mailman.trailing-edge.com/pipermail/simh/attachments/20160228/9131d307/attachment.bin>