> > [1] Your contrast won't be as good as it would be
> >     with static drive.  Even a 2:1 mux severely
> >     limits your contrast
>
>      Up till a month ago I would have agreed with
>      this whole heartedly. I have, however, been blown
>      away by the custom duplex displays we are now using.
>
>      I find no perceptible difference in the contrast
>      between these new displays and our older static
>      drive displays (and they were _very_ good static
>      displays).

It's possible to adjust things so that you can get good contrast with
multiplexed displays--don't get me wrong.  On the other hand, static
displays have the definite advantage that you can just hook them up for
maximum drive and they'll pretty consistently look great.

>      Yes, you have to pay attention to display fluid,
>      temperature range, and drive voltage.  But there's
>      no longer, in my mind, any contrast-driven reason
>      to discriminate against duplex driven displays.

If you can manage to consistently drive the display optimally, this is
true.  But the display very definitely goes from being something you can
just drive hard independent of circumstances to being something that you
must drive 'just right'.

BTW, it would seem a simple improvement to an LCD display would be to
aligh the two polarizers such that the clear parts of the display were
just a tiny bit dark, but such that a small amount of drive would make the
display get more clear (and a bit more would make it get darker).  It
would seem this should allow the unlit segments to match the background
perfectly, improving contrast.  Anyone know if this has been done?

> > If you use a bias voltage of 1/(1+sqr(2)), however,
> > you can get a contrast ratio of up to sqrt(3+2*sqrt(2)):1
> > [about sqrt(5.8:1)].
> >
> > The PIC 16C924, however cannot bias the display with
> > such a voltage.
>
>      I don't have the datasheet here, but as I recall
>      the '924 uses an external resistor ladder to generate
>      the drive voltages.
>
>      Won't the correct selection of resistor values give
>      the drive bias you want?

Unfortunately not the way the PIC is designed.  Generally, when driving a
multiplexed display you want to ensure that all segments on an inactive
row are driven the same.  The normal driving pattern, assuming bias Q,
is... [for each parameter, there are two values; the pins should go
between the two values 'in phase']

Active   row: 0 and 1
Passive  row: Q and 1-Q
Dark  column: 2Q and 1-2Q
Blank column: 0 and 1

The segments thus receive the following voltages:
Dark  column - Active  row: 2Q and -2Q
Blank column - Active  row:  0 and   0
Dark  column - Passive row:  Q and  -Q
Blank column - Passive row: -Q and   Q

Note that the magnitude of voltage on a segment on an inactive row is
indepent of whether its column is on or off; it's Q regardless.

The difficulty with the PIC is that the magnitude of "blank column,
passive row" is equal to the difference between the two center voltages,
while the magnitude of "dark column, passive row" is the difference
between an outer voltage and its nearest center voltage (note that if the
two outside gaps aren't equal, this will create a DC bias that may destroy
the LCD).  For the PIC to use the optimum drive on a duplex LCD, it would
need to have seperate inputs for the passive row and column voltages.

> > [3] It may be easier to wire things if you use two
> >     displays, each of which is internally 2:1 multiplexed.
>
>      Big yes to this.  The simplification in board
>      layouts that came with the move to duplex displays
>      alone made the exercise worthwhile.

Personally, I have something of a bent for 3:1 multiplexing, in part
because I came up with a technique that gets decent contrast without any
biasing nonsense (ideal RMS on:off, using 1/3 bias, would be sqrt(11/3);
my method produces an RMS on:off of sqrt(3) using clever timings alone.
Unfortunately, while the method would also 'work' with a 2:1 display and
the programming would be simpler than with 3:1, the contrast ratio would
actually be no better.

FYI, the technique would also be possible for displays muxed higher:

2:1 or 3:1 --- sqrt(3/1)
4:1 or 5:1 --- sqrt(11/5)
6:1 or 7:1 --- sqrt(21/11)

Note, however, that the 7:1 mux require a 128-step display frame; I
suspect the approach would be practical for--at most--a 5:1 display mux.