On Wed, 2006-03-22 at 10:09 -0500, Jack Carroll wrote:
> On Wed, Mar 22, 2006 at 01:17:57PM +0100, Hans Kristian Rosbach wrote:
> > 
> > Also, would not these pins be less vulnerable to noise than the
> > ones at the other far end, thus maybe routing would not need to be
> > so strict? After all, the more bits you get, each one will have less
> > and less actual impact on the image. So my logic is that the first
> > bit toggles a color value of 0/128, so it is absolutely critical to
> > get this one right. But the 9'th bit toggles 0/0.5 so some deviance
> > probably would not be noticed unless it's more wrong than right.
> 
>       In an analog situation, that would often be true.  For instance, in
> the design of a simple DAC ladder network, more tolerance error and ringing
> is acceptable on the lower bits than on the higher ones.
>       This is a digital situation, though.  When the clock edge strobes
> the data word into the DAC registers, each bit has either settled at the
> register input or it hasn't, so after the clock the bit will be either
> completely correct in the register or completely wrong.  So every bit's
> track on the board must meet the appropriate design rules for impedance,
> length, and stray reactive loading.  Also, the very high clock rate makes
> the transmission line delay through the data and clock tracks a significant
> item in the total timing budget.  It's necessary to account for the skew
> between the clock edge and each bit, when calculating timing margins.  It
> would almost certainly be necessary to make every bit line the same length,
> or close to it, then adjust the delay in the clock line to fit.  The signals
> coming the shortest distances would require some extra bends so that the
> transmission line delay is the same as the signals coming from more distant
> pins.

If I had only thought a little more I would have remembered that line
length is crucial, thanks for reminding me.

Regarding the last bit being correct or not in the digital domain, that
is actually what I meant. Do we really care if it gets set to the wrong
value if it's for example less than 10% of the time? Not having it at
all would possibly be wrong 50% of the time. Would it not be better?

-HK

_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to