Knight Rider + PIC?

[cpemma]

I've been thinking of making a longer back & forth Knight Rider, and come up with a few ideas and a query. 

• 16 LEDS - 4584 schmitt inverter for clock and latch, 4029 4-bit up/down counter, 4514B 4-16 line decoder. 1 pot (speed control).
• 20 LEDS - dual opamp triangle wave generator, 2 x LM3914 bargraph drivers. 3 pots (speed + min/max limits)
• More LEDs - above, more LM3914. One snag with the LM3914 is limited and sink-only current for the display, needing discrete PNP switchers. With a source output I could use 8-channel NPN chips to boost current.
• Use a PIC?

This last one's got me wondering if I should start playing with these devices. Example here suggested I could get parts count down and maybe add fade easily, but it's only 8 LEDs.

I see PICs with up to 31 I/O lines, then there's possibly multiplexing; so I'm asking the PIC experts like JustDIY and others, what sort of display can you run with a PIC and not too much else? Which PIC do I need? Must be back & forth, 20 or more LEDs.

[justDIY]

if you want the "true" knight rider effect, where you have a dim tail of leds following the bright bouncer, I think the lm3914 based circuit is going to be best for this.

you can accomplish the same in a PIC, using a lot of programming (as the author in your link points out using PWM)... this also requires a discrete pin for each LED, since you'll need to be turning it on and off

if you want a 'simple' led chaser, where the dim tail is left up to persistance of vision rather than a cooling down filament, a simple small PIC (like a picaxe), 8 pin atmel, or whatever, is all you need.

here I have 24 leds (rgb 8x3) controlled by an 18 pin pic, but actually I was only using 4 control lines from the pic itself (check out the videos near the middle of the post):

http://projects.dimension-x.net/archives/51

The number of leds connected is only limited by the speed of the microcontroller you're using, and I think that limit is some huge number, even a slow uC can push bits very fast.  The patterns seen in those videos appear to be "mirrored", this is from the mathmatical formula I was using to generate them, and purely intentional.   Making the led "bounce" from one side to the other would be easier