LEDs Was: Re: Emulation (Was: Emulators of Classic Computers))

From: der Mouse <mouse_at_Rodents.Montreal.QC.CA>
Date: Fri Jan 16 19:18:55 2004

[Multi-reply here. Second-level quotes are me.]
>> [...]: how _small_ do/can LEDs get? (And how much heat do they
>> produce?) I've been pondering something, but in order for it to be
>> workable I need to be able to cover fairly large areas with some
>> kind of display technology at a resolution no worse than about
>> 75dpi. Since "large areas" means dozens of square feet, too large
>> for a CRT, all I've been able to think of are LEDs. But I don't
>> know how practical (or more likely how drastically impractical) that
>> is.

["Rob O'Donnell" <classiccmp.org_at_irrelevant.fsnet.co.uk>]
> This one looks close to your dpi requirement
> http://www.okaya.com/HP3/image/RG160160-C.pdf (if I am translating
> 160 dots in a 63.5mm wide module correctly)

64dpi, yes, that's encouraging.

[Pete Turnbull <pete_at_dunnington.u-net.com>]
> Interesting... Smallest I've seen are surface mount, about 0.1" x
> 0.05". Not small enough for 75dpi, but perhaps you can get smaller
> arrays. However, even LCD displays for laptops are only about 75
> dpi, and I suspect you'd get gaps between the arrays.

Yes...I have a feeling it'd have to be entirely custom. (It probably
would anyway, given what I want to do with it. See below.)

> Power? Well, rule of thumb is around 2V across an LED, at around
> 10mA, depending on the type and brightness you want. [...] about
> 576W per square foot.

Ouch. And that's only 10dpi one way by 20dpi the other. On the other
hand, those LEDs are probably intended to produce enough light for a
relatively large viewing area. My plan calls for a very low total
viewing area attributed to each LED.

> I don't know offhand what the efficiency of an LED is, but a few of
> those would make a reasonable room heater :-)

Efficiency doesn't much matter. If it's sucking 500 watts, it's
dumping 500 watts, and whether that comes out as heat or light makes
relatively little difference - except for what little escapes into
space, it all ends up as heat.

> Of course, you could multiplex them and cut the power to maybe
> 10%-20% of that.

Maybe. See below.

[Brad Parker <brad_at_heeltoe.com>]
> There is a company in the Detroit area (Daktronix) which makes large
> displays using leds. I saw the "panels" being made one day.

> So it's very practical. (and I believe there others who do this also)

> It seemed to turn into an interconnect and testing problem... e.g.
> most of the problem was one of interconnecting the panels and testing
> them.

Well, given what I want to do, interconnect is one of the issues, but I
believe it should be fixable.. I suspect the panels you describe are
probably at very coarse resolution, probably no finer than maybe 4dpi -
most consumers of large panels don't also want fine resolution.

What I really want to build, y'see, is a cellular automaton machine.
Ideally, it would be little modules that abut one another and
communicate neighbour cell state between modules somehow. Then add a
little logic behind each LED, and clock them in parallel.

Power distribution is one issue that would need attention. I believe
clock distribution actually isn't much of an issue, because they don't
need to be truly clocked in parallel; provided you do something like
sample neighbour state on the rising edge and change state on the
falling edge, clock skew across the array doesn't matter as long as
there's little to no clock skew between adjacent cells. Add a few
control signals and a little more logic, and you need a host interface
at only one point (it wouldn't be used in normal operation anyway).
Signals that go to the whole array (like control signals and clock)
could be regenerated by each module, to eliminate the need for central
drivers capable of driving thousands of loads.

This means that no individiual pixel needs to generate much light -
certainly nothing like as much as most single LEDs produce. Taken in
aggregate, a square foot should be capable of (very roughly) as much
light as a square foot of those panels Brad Parker mentions, and should
produce no more than about as much heat, certainly not hectowatts per
square foot.

I don't know whether it's feasible to fabricate LEDs on the same wafer
as logic gates, but if possible that seems like a promising approach to
me. I'm quite sure the technology is up to it; even ten thousand cells
per square inch (100dpi) is nothing, not even if you go overboard with
generality and give each module 512 bits of RAM to hold the automaton
definition (I would be willing to live with limiting myself to a 3x3
neighbourhood). Communication between adjacent modules would need to
be worked out, but that problem I feel sure can be solved. Physical
alignment is just a question of packaging.

That leaves only the money. I'm certain this wouldn't be cheap. :(
Ah well, but that our dreams should exceed our grasp, or what's a
heaven for? (Maybe not quite what Browning wrote, but more or less the
same sentiment.)

/~\ The ASCII der Mouse
\ / Ribbon Campaign
 X Against HTML mouse_at_rodents.montreal.qc.ca
/ \ Email! 7D C8 61 52 5D E7 2D 39 4E F1 31 3E E8 B3 27 4B
Received on Fri Jan 16 2004 - 19:18:55 GMT

This archive was generated by hypermail 2.3.0 : Fri Oct 10 2014 - 23:36:46 BST