FW: RK05 question

From: Paul Koning <pkoning_at_equallogic.com>
Date: Wed Jul 14 08:16:55 2004

>>>>> "Gooijen" == Gooijen H <Gooijen> writes:

 Gooijen> Hmm. So in your mind experiment you keep the 12 sector
 Gooijen> pulses and the track pulse, and deliver the track pulse with
 Gooijen> 16 sector pulses to the hardware of the drive logic. Using
 Gooijen> the 12 sector pulses will probably produce a more stable 16
 Gooijen> sector pulse signal as the PLL is kept in lock in shorter
 Gooijen> intervals.

 Gooijen> But what would the PLL circuit do when it gets 12 sector
 Gooijen> pulses (that is fine) *_and_* at some odd moment the extra
 Gooijen> track pulse ...?

If the loop bandwidth is right, it will ignore the track pulse as
noise. That assumes that the sector pulses are spaced equally
everywhere including at the track start.

 Gooijen> I was thinking of *only* the track pulse and get the PLL
 Gooijen> locked on that pulse only, assuming that the stability of
 Gooijen> the PLL is sufficient for one revolution. From the N divider
 Gooijen> is would be easy to get the sector pulses. So, I cover all
 Gooijen> the sector slits on the hub of the disk.

I was thinking that the "right" design would work with unmodified
packs, by electronically ignoring the marks it doesn't need.

 Gooijen> But as you said (I did not know that part and did not check
 Gooijen> on an RK05 pack I have at home), the track pulse is not
 Gooijen> halfway two sectors. Some separate binary comparator could
 Gooijen> check the N divider outputs and then you simply set a binary
 Gooijen> number to the comparator when it should give the track
 Gooijen> pulse.

The key question is what the positions of the slots are. If the
sector slots are equally spaced, they can drive a sector pulse PLL
(and the track pulse can simply be taken from the track slot). If
there's a spacing irregularity at the track start, then you'd need to
create a high frequency clock from the track pulse, and divide it down
to produce the sector pulses at the right spacing. That wouldn't be
too hard either; for example, if you create a clock at 360x the track
pulse rate, you get 1 degree resolution on your synthesized sector
pulses...

          paul
Received on Wed Jul 14 2004 - 08:16:55 BST