[dorkbotpdx-blabber] Kicked in the Charlieplex part II.

Donald Delmar Davis don at defazio.digithink.com
Tue Jan 9 18:51:43 EST 2007


I am going to experiment with charlieplexing using the mega8 and the  
tiny26 while I wait for my parts to come.

I am going to experiment with the 4094 cmos shift register which is  
supposed to be able to drive the leds w/o the need for current  
limiting resistors,

See: http://www.kmitl.ac.th/~kswichit%20/xtimer14094/xtimer1.html  
(the only thing he doest really do is to meantion the specific family  
which is all over the place)

I will be sinking them using the venerable hc595 which pretty much  
does spi which most atmels do in hardware.  If this doesnt work I can  
redeploy them to drive the relays I picked up. The 595 has morphed  
into these power based parts for similar apps its kind of weird.

Thanks for the inputs. On to the outputs.


I love this group.

DdDD



On Jan 9, 2007, at 8:09 AM, Jared Boone wrote:

> Thomas Lockney wrote:
>> On 1/8/07, Jason Plumb <jason at noisybox.net> wrote:
>> Jared Boone wrote:
>> > Anyway, I'm probably boring everybody stiff with my jibber- 
>> jabber, so if
>> > anybody wants to talk more about this stuff, e-mail me directly.
>>
>> Let's worry about fluff when we hit 3-digit subscription levels,  
>> eh?  :)
>>
>> Yeah, we've still got a bit before that's a big concern (though  
>> we're making progress in that direction at least ;~)
> Cool. I was just worried I was sidetracking the conversation and  
> totally distracting from the problem Don wanted to solve. I have a  
> tendency to do that...
>
> Another dose of jibber-jabber:
>
> So if he wants to get this thing done on the cheap (and not spend  
> too much time), how best to do that? It already sounds like he  
> wants to use CMOS drivers (or a microcontroller), which certainly  
> keeps his voltage options open. So there's a few options I can  
> think of on how to actually do the multiplexing:
>
> Direct-drive from a microcontroller (uC): this limits the total  
> amount of current to what the microcontroller and its package can  
> handle. And, of course, there's also an upper limit to the number  
> of pins you can control (both package and clock speed limitations).  
> Practically, if you have twenty pins or so, you've probably  
> exceeded the limit of Charlieplexing -- the duty cycle possible per  
> LED gets to be very low, which requires high pulse current.
>
> uC feeding shift registers: The trick here is being able to control  
> all three states of each pin (high, low, tri-state). I don't recall  
> any shift registers supporting shifting of tri-state enable bits.  
> So you'd have one flip-flop for each high/low state and another for  
> each enable state. You could have a multi-serial bus, shifting bits  
> into 16 8-bit shift registers at once, controlling 64 high/low bits  
> and 64 enable bits worth of tri-state buffers -- think of it as  
> four Charlieplexed 16 x 15 grids. Of course, with that many signals  
> available, you could just feed 16 8-bit shift registers that are  
> just high/low. You'd have 128 signals, which could be organized as  
> four by 16 x 16, which actually gets you more LEDs than  
> Charlieplexing, because you don't have to manage the tri-state bit.  
> Plus, I think it would be infinitely easier to code.
>
> uC using shift registers and a 4-bit decoder: screw Charlieplexing,  
> use a 4:16 decoder for LED strobes, and then use the shift  
> registers to turn on the LEDs. The decoder would require some  
> external transistor drive to keep it from frying. So from the  
> microcontroller, take four pins for the decoder. Then 16 bits for  
> 16-bit shift registers, and a bit or two for clocking the shift  
> register and such. For 21 pins, you'd get 16 x 16 x 16. 4,096  
> LEDs?! For hardware, you'd have two 8-bit shift registers per 16x16  
> LED grid, and a drive transistor. So that'd be 32 8-bit SRs, 16  
> drive transistors, and one uC.
>
> It seems like the last solution would be awfully cheap -- about $15  
> or so for the shift registers, $1 for the 4:16 decoder (or two more  
> shift registers), $5 for the uC, and maybe $5 for drive  
> transistors. The expensive part now is mounting all those LEDs and  
> wiring them together.
>
> Am I missing something, or is this a pretty decent solution?
>
>     - Jared
> _______________________________________________
> dorkbotpdx-blabber mailing list
> dorkbotpdx-blabber at dorkbot.org
> http://music.columbia.edu/mailman/listinfo/dorkbotpdx-blabber

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://music.columbia.edu/pipermail/dorkbotpdx-blabber/attachments/20070109/07626c84/attachment.html


More information about the dorkbotpdx-blabber mailing list