Date   : Thu, 20 Jul 2006 15:51:54
From   : Peter Coghlan <Peter.Coghlan@...>
Subject: Re: What types of

>
>DFS 1.20 has some issues with drives stepping faster than it
>expects. DFS 1.21 fixes this.
>

Hi Jonathan,

I sent you a note about this some months ago but I never heard
anything back. Somebody else said you were probably busy with
elections at the time.

I have researched carefully the "Drive error 10" Drive not ready
error which results when using some newer 3.5in drives on bog
standard 8271 equipped BBC micros.

The issue is not to do with stepping times or any other timings
that can be set in the DFS or the 8271. The DFS does not know
or care that a drive might step too fast. It only cares if they
step too slowly.

The problem is that some newer drives suppress the index pulses
while they are stepping the heads. I believe this is a "feature"
to improve seek times with newer hardware. However, the BBC does
not have the circuitry needed to take advantage of this.

The bog standard BBC disk controller ciruit has a some flip flops
and stuff driven from the index pulses that come from the drive.
When the index pulses are lost, even very briefly, this piece of
circuitry drops the READY signal to the 8271 and the 8271 then
duly flags a drive not ready error.

Your patch for DNFS 1.2 works around the problem by retrying
indefinately when an drive not ready error is encountered. It does
not attempt to fix anything to do with stepping times, timeouts or
anything to do with any timing. I have not yet but hope to get
around to finding what implications if any this patch has when a
real "drive not ready" condition arises.

I have come up with a hardware work around for one of my drives. I
used a couple of extra transistors to couple the index sensor signal
directly to the index pin on the drive connector, bypassing the
electronics in the drive that suppress the signal. This worked well
for a while but is giving problems again now. I suspect I made a dodgy
solder joint as fitting the components in the space available was
very fiddly. Hopefully I will get a chance to look at it again soon.

I also looked into the idea of implementing a hardware work around
in the BBC disk controller circuitry but I could not come up with
anything straightforward that would do the job there.


                     Regards,
                       Peter Coghlan.