Tarbell is making me insane

[Allison]

>
>Subject: Re: Tarbell is making me insane
>   From: Grant Stockly <grant at stockly.com>
>   Date: Wed, 07 Nov 2007 12:07:22 -0900
>     To: "General Discussion: On-Topic Posts Only" <cctech at classiccmp.org>,
> Pete <petechapman at frontiernet.net>
>
>
>>Welcome to S100.
>>
>>I don't have info on tarbel 1011 seris FDC. The schematic on hand I
>>could comment more.  However 8" DD with 2mhz 8080 is very difficult
>>to do as single density is already pushing the 8080 through some very
>>tight software loops.
>
>I think the crystal on the tarbell has been changed from 4MHz to 
>2MHz.  I do know that one of the failure modes is for the Tarbell to 
>"crash" within a wait state.

Changing the crystal would half the data rates making it right for
5.25" floppy.

One of the modes all the contrllers that use hang mode (wait), is
sector never found. Wrong clock rate will insure that.


>I read the manual and when it started talking about not using 
>"Active" wait state generation I was given a hint.
>
>The tarbell as I received it was using XRDY for wait states.  My 
>front panel uses XRDY.
>
>When I was developing the Altair kit I found that the 1k memory card 
>would NOT deposit or examine because the PRDY was being driven by the 
>front panel.  The 1k memory card would have to basically over power 
>the front panel buffer.  This would cause some nice transients too.
>
>So this morning I changed the tarbell from XRDY to PRDY and it 
>worked.  It has worked fine every time I have tried it since, so 
>maybe I found the 'BUG'.

It's always Prdy.    The 1K card uses it becuse the rams installed were 
Sllooowwww and wait states were needed.  I'd not use that card ever.
When using A FDC that asserts wait for syncronization you really
don't want memory that insert wait states as the CPU is alreay slow 
enough and slow memory is a bad thing.

>So, WHY do all these old cards use 8T97 type drivers to force the 
>wait line?  I guess its only the MITS display board and 1k memory 
>board that drive the wait line regardless of if they are addressed or 
>not.  Whats wrong with these guys using open collector parts with a pullup???

Both work.  The logic however uses the 8T97 and tristates it when 
not needed (board not addressed) which is a "legal" way to do it
though it seems like more effort.

Of course using high is active and mixing than with low is active was 
a dumb thing too.  Even Tristate asserts a better low than a high.

>I'll report on how robust it is later on.  I've had several false 
>starts with this setup...
>
>>Also sounds like oneshot problems.  Check cpu timing.  Even small timing
>>errors tend to magnify bus noise issues and incompability problems.
>
>I checked that.

Good, keep checking it as those oneshots are not reliable in my book.


Allison



>Grant