Searching \ for '[EE] Driving LED matrix displays?' in subject line. ()
Make payments with PayPal - it's fast, free and secure! Help us get a faster server
FAQ page: techref.massmind.org/techref/displays.htm?key=led
Search entire site for: 'Driving LED matrix displays?'.

Exact match. Not showing close matches.
PICList Thread
'[EE] Driving LED matrix displays?'
2006\02\23@022923 by Peter Todd

picon face
Anyone have any experience with this? I just started a project that will
be a cube with 6 led matrix displays, one on each side, using 5x7 led
modules. Each side will end up as a 35x35 matrix that will then display
a 3d rendered wireframe image of a cube. The part that makes it all
worth doing, is I'm incorporating a high-resolution 3-axis tilt sensor
into the center of all this, so while you can rotate the physical cube,
the virtual, rendered cube, will always remain stationary.

I'm using LEDs rather than off the shelf LCD displays, because finding
1-1 aspect ratio LCD's is rather hard, and the ones that I did find had
rectangular exteriors due to connectors and what not. Hard to make into
a nice, almost seemless cube.

The modules I got, the cheapest I could find, have 100ma peak currents
and 11ma average current ratings. So... I was going to base my design
around simple ULN200x and UDN200x source and sink drivers with 74hc574's
to drive them. The drivers can do 500ma source or sink so that should be
fine.

5V is well above the LED's forward voltages, but of course they will
have an associated duty cycle so the average current is correct. My plan
is to make use of the watchdog timer, and have the first thing the pic
does is disable all outputs and wait for x cycles, so even in a reset
loop I won't fry my display. Disabling all outputs will work by pulling
the 74hc574 output enable lines low. Sounds reasonable to you guys? What
do commercial designs do? I found a bunch of current drivers in my
research, but they all seemed aimed at smaller LED displays, or even
just one or two LEDs, and were fairly expensive.

One challenge I see is power. Peak power, for all leds on, is about 16
amps per side! Or about 100amps for the whole board! That's just nuts...
Is there such a thing as a simple, self-contained switching regulator
that can handle a decent amount of current? I'd like something as simple
as a 3-pin regulator, no external components or anything. I'm trying to
have this thing completely built by March 11th, which I know is utterly
insane, but trying is fun!

Of course if all else fails if I just make the assumption that I'm
displaying wireframes, which means very few leds are on, and I reduce
the brightness, current requierments become *much* more reasonable, well
within the ability of a 7805 or two, but I'd like to avoid such
assumptions if it's reasonable.

The powercord going to this sucker will be fairly beefy BTW, it's gotta
allow the thing to hang off the ceiling... and be somewhat kid-proof.

--
spam_OUTpeteTakeThisOuTspampetertodd.ca http://www.petertodd.ca

2006\02\23@032323 by Jose Da Silva

flavicon
face
On February 22, 2006 11:45 pm, Peter Todd wrote:
> Anyone have any experience with this? I just started a project that
> will be a cube with 6 led matrix displays, one on each side, using
> 5x7 led modules. Each side will end up as a 35x35 matrix that will
> then display a 3d rendered wireframe image of a cube.

That's only 7350 LEDs.
Maybe you should add another LED to indicate you have power to your
circuit in case someone needs reassurance the circuit is turned on ;-)

{Quote hidden}

You can try reducing current by stacking the voltage planes. 6 boards...
that would be 100amps/6 = 17amps.
Divide by 4 and you are needing 4amps x 24 layers, or 120vdc to power
your tower of 5v x 24layers.

One set of boards driving your LEDs etc runs at 0-5v,
Another set runs on the 5-10v layer,
Another set runs on the 10-15v layer,
Another set runs on the 15-20v layer.....

Another thing is, if you are running identical pictures on each side,
then you can also reduce the circuitry by 6 by putting 6 LEDs in series
controlled per driver. You'll need a different driver most likely since
you'll be needing a higher voltage.

2006\02\23@033315 by William Chops Westfield

face picon face

On Feb 22, 2006, at 11:45 PM, Peter Todd wrote:

> I just started a project that will be a cube with 6 led matrix
> displays, one on each side, using 5x7 led modules. Each side
> will end up as a 35x35 matrix

Hmm.  That's equivalent to an LED-matrix sign 1 character high by
210 characters wide, which is... very large for that sort of thing.

> One challenge I see is power.

No Duh.  7350 LEDs.  Ouch.

> Peak power, for all leds on, is about 16amps per side! Or about
> 100amps for the whole board! That's just nuts...Is there such a
> thing as a simple, self-contained switching regulator that can
> handle a decent amount of current? I'd like something as simple
> as a 3-pin regulator, no external components or anything.
>
Sure.  There are LOTS of these little DC-DC converters aimed
at telco and comm equipment, that convert ~48V DC to your choice
of output voltage at frankly incredible efficiencies in amazingly
tiny packages.  200W per cubic inch or so.  For instance, less
than 2.5x1.5x0.5 inches, 5V at 40A:
<http://www.artesyn.com/powergroup/productinfo/
quarter_brick_48v_single_ultra_info.htm>

We sprinkle them like candy through the larger cisco routers,
for instance.  Don't ask how much they cost, though.  And don't
ask me where you get the 48V supplies to provide at the input,
either...  (On the other hand, these can frequently be found in
discarded last-generation comm gear.  I could probably come up
with quite a few in a hour's worth of dumpster-diving (although
the current crop probably output 1.8 to 3.3V...)

Since this isn't the first time you've run into power problems,
I think you should do some research on distributed power.  Just
because you need 5V at 100A doesn't mean you need a SINGLE 5V
regulator that can deliver 100A; a half-dozen 12A supplies MIGHT
be a lot easier...

Of course, you won't be doing "all LEDs on", since you said
"wireframe", and the peak current isn't appropriate since that's
for a momentary pulse in a multiplexed display (max 6*35 LEDs
on at a time at maybe 50mA; only about 10A required!  Less than
that since the wireframe won't light up full rows on all sides
at once.  Probably.)  I hope you realize that this thing is
likely to be to hot to handle.  That 10A at 5V works out to about
50 watts, which will heat up something this size (less than 10inches
on a side?) quite happily to an impressive temperature.

And it's already been done (in 2d, anyway) with laptops and their
internal 'drop sensors.'
  http://osxbook.com/book/bonus/chapter10/ams2hid/

BillW

2006\02\23@114012 by Robert Rolf

picon face

Peter Todd wrote:
> Anyone have any experience with this? I just started a project that will
> be a cube with 6 led matrix displays, one on each side, using 5x7 led
> modules. Each side will end up as a 35x35 matrix that will then display
> a 3d rendered wireframe image of a cube. The part that makes it all
> worth doing, is I'm incorporating a high-resolution 3-axis tilt sensor
> into the center of all this, so while you can rotate the physical cube,
> the virtual, rendered cube, will always remain stationary.

So when kids shake the cube, what happens? A typical tilt sensor is going to
go crazy, as will the software. You need a combination of tilt and
rotation sensors to keep the 'virtual cube' stationary in space.
I would suggest testing your sensor filtering and display
algorithms first or you may be quite disappointed with the final result.
And if shaking the unit makes the display go berserk, you can bet that
the kids will shake it vigorously.

Given the effort you are putting into this, it might be worth looking
at using tri color LEDs so that you have something even more colorful
when you're done.


Robert

2006\02\23@163957 by Peter Todd

picon face
On Thu, Feb 23, 2006 at 12:29:30AM -0800, Jose Da Silva wrote:
> On February 22, 2006 11:45 pm, Peter Todd wrote:
> > Anyone have any experience with this? I just started a project that
> > will be a cube with 6 led matrix displays, one on each side, using
> > 5x7 led modules. Each side will end up as a 35x35 matrix that will
> > then display a 3d rendered wireframe image of a cube.
>
> That's only 7350 LEDs.
> Maybe you should add another LED to indicate you have power to your
> circuit in case someone needs reassurance the circuit is turned on ;-)

That's a good idea, but you know, sometimes LED's fail, maybe I should
bolt a neon tube to it, for redundency?

{Quote hidden}

Hmm... Sounds like a bit of a routing nightmare to me. I'd be running
things in series, so I'd have to "route" the voltage from one board to
another. If it's just running the boards in series, they'll use vastly
different amounts of current as the pictures change, probably not going
to work.

Thinking outside of the box is good though. For that matter, why must it
be a box? :)

> Another thing is, if you are running identical pictures on each side,
> then you can also reduce the circuitry by 6 by putting 6 LEDs in series
> controlled per driver. You'll need a different driver most likely since
> you'll be needing a higher voltage.

Ahh, see, they aren't identical pictures, they are 3d-renderings of an
identical object, but from different angles.

Nice idea though, very sneaky for the right application...

--
.....peteKILLspamspam@spam@petertodd.ca http://www.petertodd.ca

2006\02\23@170234 by Robert Rolf

picon face
>>Another thing is, if you are running identical pictures on each side,
>>then you can also reduce the circuitry by 6 by putting 6 LEDs in series
>>controlled per driver. You'll need a different driver most likely since
>>you'll be needing a higher voltage.
>
>
> Ahh, see, they aren't identical pictures, they are 3d-renderings of an
> identical object, but from different angles.

So why do you then have to keep the virtual image stationary as
the box is moved? And isn't the image going to get rather distorted
as the box is rotated and the edge LEDs get used instead of ones more
central to the surface planes?

R

2006\02\23@171217 by Peter Todd

picon face
On Thu, Feb 23, 2006 at 12:33:14AM -0800, William Chops Westfield wrote:
>
> On Feb 22, 2006, at 11:45 PM, Peter Todd wrote:
>
> > I just started a project that will be a cube with 6 led matrix
> > displays, one on each side, using 5x7 led modules. Each side
> > will end up as a 35x35 matrix
>
> Hmm.  That's equivalent to an LED-matrix sign 1 character high by
> 210 characters wide, which is... very large for that sort of thing.

Definetely, though size wise, it'll be only 4inches on each side, the
dot pitch for the LED's is 0.1"

> > One challenge I see is power.
>
> No Duh.  7350 LEDs.  Ouch.

It'll be the most LEDs I'll have had in any project!

{Quote hidden}

Thanks! Those look excellent, exactly the sort of thing I've needed for
this and a number of other projects. That said I'll do some looking to
find some even smaller ones, those particular ones probably won't
physically fit in my application, I genuinly need something the size of
a 7805 or two.

> We sprinkle them like candy through the larger cisco routers,
> for instance.  Don't ask how much they cost, though.  And don't
> ask me where you get the 48V supplies to provide at the input,
> either...  (On the other hand, these can frequently be found in
> discarded last-generation comm gear.  I could probably come up
> with quite a few in a hour's worth of dumpster-diving (although
> the current crop probably output 1.8 to 3.3V...)

Digikey has nice ready-made 48V power supplies, seen 'em in the
catelogs, lots of different types too.

> Since this isn't the first time you've run into power problems,
> I think you should do some research on distributed power.  Just
> because you need 5V at 100A doesn't mean you need a SINGLE 5V
> regulator that can deliver 100A; a half-dozen 12A supplies MIGHT
> be a lot easier...

Probably true. I already want the power to be contained to each side, as
I want to build all 6 sides using identical circuits.

> Of course, you won't be doing "all LEDs on", since you said
> "wireframe", and the peak current isn't appropriate since that's
> for a momentary pulse in a multiplexed display (max 6*35 LEDs
> on at a time at maybe 50mA; only about 10A required!  Less than
> that since the wireframe won't light up full rows on all sides
> at once.  Probably.)  I hope you realize that this thing is
> likely to be to hot to handle.  That 10A at 5V works out to about
> 50 watts, which will heat up something this size (less than 10inches
> on a side?) quite happily to an impressive temperature.

Not to mention getting power *too* it becomes quiet difficult, I was
hoping to be able to use nice coiled up audio cables, they are very
resilient and bouncy.

Good point on the heat, didn't think of that one at all. LEDs are what,
10% or so efficient? Even assuming that much is turned into light
energy, that's hot alright. And it's *4inches* on a side, even worse.

Working through the numbers a little better... Peak current for the LEDs
is 100ma. At that current the datasheet says my max duty cycle is 7% So
I could apply the peak current to each row at 1/35th duty cycle and
still be well within spec.

35 collums * 0.1A = 3.5A per side * 6 sides = 21A

More reasonable...

Something I realised right now reading the datasheet... That max peak
current, of 100ma, is reached at a forward voltage of 2.75V So am I
corrent in saying that I can't drive those LEDs from a 5V source, even
if I used really low duty cycles, because the peak current would be well
over that 100ma rating?

> And it's already been done (in 2d, anyway) with laptops and their
> internal 'drop sensors.'
>    http://osxbook.com/book/bonus/chapter10/ams2hid/

Er, funny story about that... I was telling this idea to my electronics
teacher, who in the field of electro-mechanical art is very well known,
and he said he thought it was brillient, I should totally do it. He also
told me he had done almost the exact same thing, 10 years ago, with one
rotating monitor held in a gyroscope-style mechanical movement with a
286 running the graphics and a joystick to move it. What's worse, is the
guy who curated it, happens to be good friends with the same guy who I
sell most of my work too...

Fortunately his version never worked all that well, and didn't get much
press. He also didn't think anyones done that idea as a proper handheld
cube, so I'm still gonna make it.

--
petespamKILLspampetertodd.ca http://www.petertodd.ca

2006\02\23@172915 by Peter Todd

picon face
On Thu, Feb 23, 2006 at 09:39:26AM -0700, Robert Rolf wrote:
>
> Peter Todd wrote:
> > Anyone have any experience with this? I just started a project that will
> > be a cube with 6 led matrix displays, one on each side, using 5x7 led
> > modules. Each side will end up as a 35x35 matrix that will then display
> > a 3d rendered wireframe image of a cube. The part that makes it all
> > worth doing, is I'm incorporating a high-resolution 3-axis tilt sensor
> > into the center of all this, so while you can rotate the physical cube,
> > the virtual, rendered cube, will always remain stationary.
>
> So when kids shake the cube, what happens? A typical tilt sensor is going to
> go crazy, as will the software. You need a combination of tilt and
> rotation sensors to keep the 'virtual cube' stationary in space.
> I would suggest testing your sensor filtering and display
> algorithms first or you may be quite disappointed with the final result.
> And if shaking the unit makes the display go berserk, you can bet that
> the kids will shake it vigorously.

Well to be exact I'm planning to use a 3-axis acellerometer,
http://www.sparkfun.com/commerce/product_info.php?products_id=252, as my
tilt sensor. So I know that it's only referencing to gravity, therefore
already it's ignoring any movement parallel to gravity, such as spin.

My guess is that any vigourous movement will cause a lot of weird stuff
to display from the non-gravity accellerations present. But the display
should get more stable as movement stops, because I'll always be
directly referencing gravity to decide what to display.

It's not a perfect situation, but I think for a first try it'll make for
a decent effect. I'm designing the system to be modular, so at the very
center will be a seperate circuit board that just does the tilt
processing and then tells the display boards exactly what I want
displayed. So when the first version doesn't work will enough, I'll swap
the tilt board out. I'm also planning to try to leave some physical
space in case I need to have an acellerometer some displacement from the
center of the unit, so as to better measure rotation from the
accellerations.

Come to think if it, I'll bet you kids would love seeing the display go
berserk after some vigorous shaking... haha, anyway, my intended
audience is really crusty art collectors who will probably not even look
at it and instead nod their heads at how brillient a concept it is... If
they ever buy it, it'll be kept in some box somewhere, never seen again.

> Given the effort you are putting into this, it might be worth looking
> at using tri color LEDs so that you have something even more colorful
> when you're done.

Quiet frankly, nah, I don't mean to ignore your suggestion, but I'm
lucky enough as it is to get my LEDs as cheap as I have, $0.60 each, and
that requiered a decent amount of research. Tri color will definetely
blow my budget unfortunately.

--
.....peteKILLspamspam.....petertodd.ca http://www.petertodd.ca

2006\02\23@173909 by Peter Todd

picon face
On Thu, Feb 23, 2006 at 03:02:32PM -0700, Robert Rolf wrote:
> > Ahh, see, they aren't identical pictures, they are 3d-renderings of an
> > identical object, but from different angles.
>
> So why do you then have to keep the virtual image stationary as
> the box is moved? And isn't the image going to get rather distorted
> as the box is rotated and the edge LEDs get used instead of ones more
> central to the surface planes?

You are totally right and the image will definetely get distorted.
You're looking at a projection of a cube, at an angle, rather than a
cube. It's like the difference between a hologram and a standard photo.

As for why... Well here is my dry-art talk explanation. My concept with
this is that I want to present a situation where it isn't clear what is
more real, the physical object of the cube, or the virtual creation that
is projected onto it. Both will clearly be consistant with physical
reality in some way, the real cube moves when you move it, the virtual
cube will fairly consistantly not move when you move it.

Work comes to worst I put the thing in a multi-axis gyro-type mechanical
setup and put rotary encoders on each axis. Then making the virtual cube
stable is a simple matter of reading the encoders and doing the math.
I'll decide if I think I want to go in that direction once I have a
physical tilt-based model made.

--
EraseMEpetespam_OUTspamTakeThisOuTpetertodd.ca http://www.petertodd.ca

2006\02\23@174359 by William Chops Westfield

face picon face

On Feb 23, 2006, at 2:28 PM, Peter Todd wrote:

> Good point on the heat, didn't think of that one at all. LEDs
> are what, 10% or so efficient?

Hmm.  You'd think someone would be putting ultra-bright LED chips
in dot-matrix packages, give than they're used in high-visibility
signs and whatnot.  Those you ought to be able to get by driving
at 5mA average current or less...  It might be worth checking if
such displays are available elsewhere...

I've also seen dot-matrix displays in up to 16x16 LEDs, which would
make a 32 cube a bit easier...

BillW

2006\02\23@181652 by Jinx

face picon face
> for instance.  Don't ask how much they cost, though.  And don't
> ask me where you get the 48V supplies to provide at the input,
> either...  (On the other hand, these can frequently be found in
> discarded last-generation comm gear.  I could probably come up
> with quite a few in a hour's worth of dumpster-diving (although
> the current crop probably output 1.8 to 3.3V...)

Amazing coincidence Bill - I was fortunate enough last night to be
invited to rummage through Nokia's warehouse and take whatever
I wanted from the tonnes of old gear going to the tip/recycler

Bagged myself several Swichtec 48V/50A supplies and Finlandia
48 -> 5 & 12 convertors, as well as a bootload of fancy high-tech
switchgear. LEDs, switches and assorted pricey hardware and
exotic PCBs all over the place. Stepper-driven gold-plated tuners,
cable, extruded cases, relay banks etc etc

Happy as a pig in .... ;-))

Through a family contact I hear of other throw-outs in all kinds
of businesses, but there's simply too much to take advantage of
it all


2006\02\23@184133 by Marcel duchamp

picon face
Jinx wrote:
 > Amazing coincidence Bill - I was fortunate enough last night to be
> invited to rummage through Nokia's warehouse and take whatever
> I wanted from the tonnes of old gear going to the tip/recycler
>
> Bagged myself several Swichtec 48V/50A supplies and Finlandia
> 48 -> 5 & 12 convertors, as well as a bootload of fancy high-tech
> switchgear. LEDs, switches and assorted pricey hardware and
> exotic PCBs all over the place. Stepper-driven gold-plated tuners,
> cable, extruded cases, relay banks etc etc
>
> Happy as a pig in .... ;-))
>
> Through a family contact I hear of other throw-outs in all kinds
> of businesses, but there's simply too much to take advantage of
> it all

Hmmm... do any of these involve alarm systems accidentally turned off?


2006\02\23@192358 by Jinx

face picon face
> Hmmm... do any of these involve alarm systems accidentally
> turned off ?

Is "some" "any" ?

Nah, all legit. BTW, Aucklanders might notice their cellphone
coverage is a little patchy at the moment. Sorry 'bout that ;-)

Seriously, if you're prepared to look, there's all manner of old
equipment destined for the dump that can be ratted

2006\02\23@221117 by Jose Da Silva

flavicon
face
On February 23, 2006 01:55 pm, Peter Todd wrote:
{Quote hidden}

Yup, you would have to treat each board in terms of constant current
because some boards would have all the LEDs lit, while some boards
would have to bleed current through despite having no LEDs lit.
Takes thought, but certainly can be done.

> Thinking outside of the box is good though. For that matter, why must
> it be a box? :)

No, that's upper management speak.
For embedded, the joke is "think inside the box".    ;-P

2006\02\24@065450 by Howard Winter

face
flavicon
picon face
Jinx,

On Fri, 24 Feb 2006 12:16:07 +1300, Jinx wrote:

> I was fortunate enough last night to be invited to rummage through Nokia's warehouse and take whatever I
wanted from the tonnes of old gear going to the tip/recycler

<<JEALOUS!!!>>

I did once get invited to dumpster-dive at a University's science building, but most of it was 40+ year-old
lab equipment that really was useless these days, and the majority had no markings to even give a clue what it
was.  From the tonne-or-so of stuff only a few kgs were worth saving.  Got some nice 10-turn vernier dials,
though - one day I'll find a use for them!  :-)

Cheers,




Howard Winter
St.Albans, England


2006\02\24@065555 by Howard Winter

face
flavicon
picon face
Marcel,

On Thu, 23 Feb 2006 15:41:29 -0800, Marcel duchamp wrote:
...
> > Through a family contact I hear of other throw-outs in all kinds
> > of businesses, but there's simply too much to take advantage of
> > it all
>
> Hmmm... do any of these involve alarm systems accidentally turned off?

I don't think he meant *that* kind of "family"!  :-)

Cheers,


Howard Winter
St.Albans, England


2006\02\24@072246 by Jinx

face picon face
> > Hmmm... do any of these involve alarm systems accidentally turned off?
>
> I don't think he meant *that* kind of "family"!  :-)

Oh, goombahs. Hey, I don't know nothin' 'bout no goombahs, capice ?

2006\02\24@083509 by Alan B. Pearce

face picon face
>Well to be exact I'm planning to use a 3-axis acellerometer,
> www.sparkfun.com/commerce/product_info.php?products_id=252,
>as my tilt sensor. So I know that it's only referencing
>to gravity, therefore already it's ignoring any movement
>parallel to gravity, such as spin.

Except it will not ignore gravity. Gravity will appear as a constant offset
in the Z direction for which axis of the accelerometer is measuring that
direction.

>My guess is that any vigourous movement will cause a lot
>of weird stuff to display from the non-gravity accellerations
>present. But the display should get more stable as movement
>stops, because I'll always be directly referencing gravity
>to decide what to display.

I would be tempted to blank the whole display if you sense any "high
frequency" movement on any axis, as a 'negative reward' for this behaviour.

2006\02\24@162816 by Jinx

face picon face
>> take whatever I wanted from the tonnes of old gear going
>> to the tip/recycler

> <<JEALOUS!!!>>

Well, Howard, I've.....um.....I've always coveted thine sporran

> Got some nice 10-turn vernier dials, though - one day I'll find
> a use for them!  :-)

Sad to say, as we all know, something's not worth anything if
you don't use it, no matter how much it used to be worth

2006\02\24@185856 by andrew kelley

picon face
<about noise and vibrations and accelerometers>
At Umass Dartmouth, we participated in the DARPA Grand Challenge;
although we never made it to any of the competitions, the INS system
worked.  Now the Inertial Nav System worked using three accelerometers
and gyros.  The group that did INS used a Kalman filter which required
a bit of time to settle; but once it was settled(converges faster w/
more variance), they threw it in a car with a laptop and gave it its
starting coordinate.  They drove it around ring road (about 1 to 2
miles, few minute drive @ 25mph).  They took the data file and
overlaid it on a google map.  The error was IIRC about 100 ft towards
the end of the trip.  This was done on a PC though..  The kid that
designed most of the software and got it workin ended up working at
Sippican..

--
andrew

2006\02\26@010722 by Peter Todd

picon face
On Fri, Feb 24, 2006 at 06:58:56PM -0500, andrew kelley wrote:
> <about noise and vibrations and accelerometers>
> At Umass Dartmouth, we participated in the DARPA Grand Challenge;
> although we never made it to any of the competitions, the INS system
> worked.  Now the Inertial Nav System worked using three accelerometers
> and gyros.  The group that did INS used a Kalman filter which required
> a bit of time to settle; but once it was settled(converges faster w/
> more variance), they threw it in a car with a laptop and gave it its
> starting coordinate.  They drove it around ring road (about 1 to 2
> miles, few minute drive @ 25mph).  They took the data file and
> overlaid it on a google map.  The error was IIRC about 100 ft towards
> the end of the trip.  This was done on a PC though..  The kid that
> designed most of the software and got it workin ended up working at
> Sippican..

Thanks for the info, sounds very encouraging!

What sort of acellerometers/gyros were they? Simple MEM devices or more
advanced?

I'd really like to do a proper 6-axis tilt/rotation setup and learn how
Kalman filters work eventually. But I think the 3 axis tilt setup's a
good start at least, especially with a two week schedule... Oh well,
that's what summers are for.

--
petespamspam_OUTpetertodd.ca http://www.petertodd.ca

2006\02\26@011659 by Peter Todd

picon face
On Fri, Feb 24, 2006 at 01:35:06PM -0000, Alan B. Pearce wrote:
> >Well to be exact I'm planning to use a 3-axis acellerometer,
> > www.sparkfun.com/commerce/product_info.php?products_id=252,
> >as my tilt sensor. So I know that it's only referencing
> >to gravity, therefore already it's ignoring any movement
> >parallel to gravity, such as spin.
>
> Except it will not ignore gravity. Gravity will appear as a constant offset
> in the Z direction for which axis of the accelerometer is measuring that
> direction.

Well what I'm saying is that by soley measuring tilts relative to
gravity, there is no way to detect if the device is being spun.

If my x and y are reading 0, and z is reading +1G, I know I am oriented
upright. But the device could very well be spinning with x and y
centered on the axis of rotation and therefore experiencing no
centrifugal force.

I haven't put much detailed thought into exactly how the math will work
yet, but my thoughts are that I'll measure tilt by looking at the
relative porportions between x, y and z. So any movement will "look"
like gravity to the math, even if the sum total of the vector is higher
than 1G. So someone moving the cube side to side will cause the image to
sway back and forth as the force exerted on the cube moves from side to
side.

> >My guess is that any vigourous movement will cause a lot
> >of weird stuff to display from the non-gravity accellerations
> >present. But the display should get more stable as movement
> >stops, because I'll always be directly referencing gravity
> >to decide what to display.
>
> I would be tempted to blank the whole display if you sense any "high
> frequency" movement on any axis, as a 'negative reward' for this behaviour.

Nah, I was just at a whole art exhibition put on by some quiet famous
artists exploiting various "glitches" in computer games.

Let people have their fun by messing up the inputs, it's only temporary
anyway.

--
@spam@peteKILLspamspampetertodd.ca http://www.petertodd.ca

2006\02\26@012641 by Peter Todd

picon face
On Thu, Feb 23, 2006 at 02:43:57PM -0800, William Chops Westfield wrote:
>
> On Feb 23, 2006, at 2:28 PM, Peter Todd wrote:
>
> > Good point on the heat, didn't think of that one at all. LEDs
> > are what, 10% or so efficient?
>
> Hmm.  You'd think someone would be putting ultra-bright LED chips
> in dot-matrix packages, give than they're used in high-visibility
> signs and whatnot.  Those you ought to be able to get by driving
> at 5mA average current or less...  It might be worth checking if
> such displays are available elsewhere...

Well, these ones I'm using because they are the absolute cheapest I
could find anywhere. Nothing else I found was even close to the 0.59each
price I got.

> I've also seen dot-matrix displays in up to 16x16 LEDs, which would
> make a 32 cube a bit easier...

Definetely! Definetely! I could probably put the driver circuitry on the
same board in that case. Right now I'm doing a two board design, routing
for the leds and LED drivers behind on a slightly smaller board.

Can't say I found any 16x16 LEDs though in my searching, 5x7's are
probably much more common due to their use to display text. 8x8's are
common, but only in corse dot pitches.


I found an interesting project similar to mine BTW:

http://www.ok.sfc.keio.ac.jp/2005/web/en/project/zagon/

They use 16x16 LEDs... Looks like they couldn't fit everything in the
one cube though, they've got a seperate driver board. Their cube is
smaller than what I have planned though, 2.5" aparently.

--
KILLspampeteKILLspamspampetertodd.ca http://www.petertodd.ca

2006\02\26@044457 by William Chops Westfield

face picon face
On Feb 25, 2006, at 10:42 PM, Peter Todd wrote:

>  found an interesting project similar to mine...

In a slightly different direction is Mark Lotter's "Cubatron":
  http://www.nw.com/nw/projects/cubatron/
(uses LOTS of PICs, BTW.)

Huh.  I used to work with Mark, sorta, a long time ago.  Interesting
that he's been doing 'art.'

BillW

2006\02\26@063050 by Peter Todd

picon face
On Sun, Feb 26, 2006 at 01:44:53AM -0800, William Chops Westfield wrote:
> On Feb 25, 2006, at 10:42 PM, Peter Todd wrote:
>
> >  found an interesting project similar to mine...
>
> In a slightly different direction is Mark Lotter's "Cubatron":
>    http://www.nw.com/nw/projects/cubatron/
> (uses LOTS of PICs, BTW.)

I'll say, lots of money too, I'll bet he spent at least $3 a board...

I recently saw another neat one in the same idea, don't have the url
handy, but it was a 10x10x10 cube made of dead bug connected white leds.
Standard matrix driven from the looks of it, no tech details but the
videos were nice and it had a wonderfull spartan appearence.

3d matrix cube would find it on google I think.

> Huh.  I used to work with Mark, sorta, a long time ago.  Interesting
> that he's been doing 'art.'

Hey, that's what happens when you start going to burning man!

The light strings he has look neat, hopefully they'll sell...

--
RemoveMEpeteTakeThisOuTspampetertodd.ca http://www.petertodd.ca

2006\02\26@064044 by Peter Todd

picon face
Still working on nailing down the design for my LED cube...


I've realised that I need a part that can source 3A of current at about
3.3V with a 0.5% duty cycle. I've found IC's that can *sink* current
like that, but none that can *source* it. Now I have found SOT-23
packaged FETs that can do that just fine, so worst-case I just solder a
pile of them down, but I'm curious if there is something akin to a 7x or
8x "FET array", like the UDN2981 parts, on one through hole IC. Pretty
sure I've throughly searched digikey and google, but hey, sometimes it's
just the right search terms...

Any idea if such a beast exists?

--
spamBeGonepetespamBeGonespampetertodd.ca http://www.petertodd.ca

2006\02\26@224015 by andrew kelley

picon face
> What sort of acellerometers/gyros were they? Simple MEM devices or more
> advanced?

IIRC they are/were AXDLxxxx somethings..  So yes, standard MEMS..

> I'd really like to do a proper 6-axis tilt/rotation setup and learn how
> Kalman filters work eventually. But I think the 3 axis tilt setup's a
> good start at least, especially with a two week schedule... Oh well,
> that's what summers are for.

They've written whole books on them.. there's some pretty good
tutorials on them.. Mainly useful if you have multiple sources for
information about the same thing (and filtering noise)

andrew
> --
> TakeThisOuTpeteEraseMEspamspam_OUTpetertodd.ca http://www.petertodd.ca

More... (looser matching)
- Last day of these posts
- In 2006 , 2007 only
- Today
- New search...