Forums > Technical Discussion > LED skunkworks: Wish list/brainstorm.

Posted: 1st Feb 2005
oooh.. lots of interesting stuff to be thought about here - nice work! Sounds like you're making some serious progress - any chance of some pics (bad pun I know!) / vids?

What speed (hz) is your PWM loop up to now?

I have a similar sounding prototype formed from plastic cups, elastic bands, and yes, you guessed it, gaffa tape

I reckon one of my Luxeon 'heads' in each end of a staff would be expensive, but would look truly unbeleivable!

I had a look at those bet bottles - good idea ...quite phallic though! hehe

*ahem*

being able to use these as torches would be handy, although a little risky without some sort of diffuser - these Luxeon LEDs really are VERY bright!

Some more PICs arrived this morning - I should be able to work out whether it's the programmer I'm using, or my shoddy SMD soldering that's holding me back.

Cake or Death?

Posted: 1st Feb 2005
no pics or vids I'm afraid. I've been trying to buy a camera off ebay for the last three weeks or so, even though I probably can't afford it.

umm, eew, math. The PWM loop runs at 30khz.
I think.

well, there's room for a motor, I guess
Alien Se...
yes, *ahem*


bright, and HOT, from all accounts.

Good luck with the new PICs, I hope they work!

Posted: 3rd Feb 2005
30khz.. in software?!! *bows down*

I shall be installing some sort of heatsink-type-affair in there to deal with the heat.

Still no luck with the PICs for the time being.

I got some hamster water bottles the other day when seem to be pretty robust and comfortably hold 4xAA batteries. Feels like a pretty good weight to me too - not as heavy as fully fueled fire poi, but heavier than tennis balls

Cake or Death?

Posted: 9th Feb 2005
Just found out that using a CRT as a serial link isn't new idea (they never are.... )

check this out:
https://datalink.fries.net/index.html
especially:
https://datalink.fries.net/transfer.html

But that means that it works, which is encouraging. Although needing a CRT not an LCD is a bit limiting.

I'm off for a week of microprocessor shenannigans with a mate. Hopefully I'll have something nice to post for you guys here soon...

monkeys ate my brain

Posted: 9th Feb 2005
interesting stuff.. I wonder if the same applies to todays modern (high refresh rate) TFTs.. ?

Cake or Death?

Posted: 24th Feb 2005
Gahhhhhhhhhhhh technical specs melting my brain! Okay, trying to wade through this and throw around ideas...

strobe = standard... could also strobing through different color sequences instead of just endlessly blinking or alternating-- done fast enough it could be like a dithered fade

Morphing/fading/ribbons = EXTREMELY good

Bouncing = good = zigzag or even crisscrossy patterns

and then combinations of all those!

controls in the handle = probably too darn hard to use to be worth the bother

synchronization = quite good especially if it could be done for a group... imagine programming patterns that go from performer to performer!

as many colors as possible and the ability to choose/program color schemes = absolute must (and there's no such thing as RGY. R and G make Y, and just how do you get blue with that palette?) CMYB is for print only and near useless when working with light, just stick with RGB and keep everyone's life simple.. and hexidecimal is a darn sight more convenient than 0-255 values... I was also under the impression that LEDs make almost absolutely pure hues)


We need poi that don't just light up at the ends, but along their entire length, from the grip on down. The trouble with that is that unless the different sections were isolated and needing their own batteries each, swivels would have to be incorporated, and the whole thing would probably have to be at least somewhat flexible. I guess rigid could also work, though it would make wraps impossible. Another spiffy thing to do to would be to hook up a flag along the length.

What would also be TRULY FANTASTIC is if the poi would -draw- stuff as they travel through the air. You know, like those LED clocks that make numbers from a little wand that wiggles back and forth. Picture this: Doing a butterfly with an actual, stationary outline of butterfly wings hovering in the air in front of you. Or just spinning one poi in the air and having it make blooming flowers or a kaleidascope. ..Or if you could get the timing/motion sensing stuff right somehow, doing a flower move with the poi changing colors as they travel out from your body or come in towards the center. Maybe some kind of proximity sensor arrangement where you wear something on your belt or put it on the floor and the poi change colors as they get near/far away from it?


One of those message wands now can scan in and show pictures you draw on paper-- and it looks pretty small, too.
Scanning Picture Wand

"Ours is not to question The Head; it is enough to revel in the ubiquitous inanity of The Head, the unwanted proximity of The Head, the unrelenting HellPresence of The Head, indeed the very UNYIELDING IRRELEVANCE of The Head!" --Revelation X

Posted: 24th Feb 2005
Hi all,

Right back from playing with micros. Been using Atmel AVRs, which totally rock. Lovely in circuit programming, do lots of the hard stuff for you. Looking at a 2312, which is *just* large enough to fit a software USB stack and some PWM stuff in memory, although I might need some more memory for a full on light sequencing setup. (also been playing with the ATMegas, which are lovely but the package size is way too big)

PWM:

I don't think you can use caps to smooth PWM stuff; you could if LEDs were nice linear things, but they're not at all linear, so you have to switch them fully on for a small portion of time, rather than partway on all the time. You could smooth them optically using phospors, though

The loop should be quite small, though; it's a compare for each output channel (possibly with a bit set), a compare for your counter (possibly with a setting back to zero) and a counter increment. On my atmels it's quick enough to not flicked when jiggled violently, and that's writing it in C...

monkeys ate my brain

Posted: 24th Feb 2005
Oh yeah and forgot to mention--

What about padding?

Doesn't anybody want to make LED poi that are nonlethal, or does bashing yourself repeatedly with hard objects like an idiot make people feel more manly or something?

"Ours is not to question The Head; it is enough to revel in the ubiquitous inanity of The Head, the unwanted proximity of The Head, the unrelenting HellPresence of The Head, indeed the very UNYIELDING IRRELEVANCE of The Head!" --Revelation X

Posted: 24th Feb 2005
I'm building some which will be covered in foam

Cake or Death?

Posted: 25th Feb 2005

Written by: Gnarly Cranium

---

What about padding?

---

Silicone sealant all the way

monkeys ate my brain

Posted: 17th Mar 2005
How're everyone else's new poi coming along? My development will be stalled until I can figure out how to get my new PICs (16F767's) working

Cake or Death?

Posted: 19th Mar 2005
Glad to see people are still working things here!

I've resumed development of my LED poi.. The original designs were using AVRs and solid state accellerometers to change the LED color based on the speed of rotation. We had basic functionality, but were having a real hard time with the PWM to make it smooth.

I've switched to some PICs, 16F628 I think, and have finally got the C compilers set up, which should give me some more flexibility (my assembly is crap).

I've also been reading over code for scrolling LED signs, and mechanically scanned toys (like the 'message wands'), and I think I'm wrapping my head around the idea enough to attempt a polar coordinate version, where the 'frame' isn't made up of a row of columns, but rather a polar arrangement of columns, but it could be scanned in the same way. The only problem is refresh rate, but if I get some more practice with threading, I might be able to use the accelerometer to figure out the refresh rate (since acelleration would be proportional to speed, and better yet, acelleration at the top of the swing will be 2G less than the acelleration at the bottom of the swing (thanks to gravity) which would even let us keep images vertical).

Anyway, one step at a time, right? I look forward to seeing more about what everyone pulls together, and I'll be posting source code as it becomes functional.

And I don't even want to open the interface pandora box right now! Firmware first, interface second!

peace,
sam

Posted: 20th Mar 2005
The major problem with making PWM LED Poi is the fact that you need to flash them so damn fast! That's why I resorted to hardware

The wands etc. that you mentioned work by repeatedly sending a message to the same point in space. I suspect that because the speed at which poi spin, you're unlikely to be able to reproduce the effect as nicely. Still, it'll be interesting to see what you come up with

I've got some tiny accelerometers to play with too. I've not yet tried to hook them up though - I figured I'd get the LED's flashing first. Where did you get yours? I think the use of accelerometer's is going to be instrumental to getting the kind of patterns we're after!

Cake or Death?

Posted: 20th Mar 2005
I'm using the same accellerometers, I think the ADXL150 to be specific. Its a single axis 50g unit with good sensitivity. I got 3 of them free through Analog's samples program. Just make yourself a company name and request samples! I get most of my specialized ICs that way, and then if I need more than 2 or 3, I buy after that.



I've got an 8 LED 'pseudo array' like the 'waving wands' working nicely. The idea is that you have an image, say on an 8x8 grid, and you have your vertical array of LEDs display the first column, wait a set period of time, and then display the next column, then wait, then the next column. When it hits the end, just cycle back to the first and keep going.



The code for smiley face looks something like this:



Code:

---

x = 16; // Time to complete entire image in ms

while(1) {

PORTB = (0b00111100); DelayMs(x/8);

PORTB = (0b01000010); DelayMs(x/8);

PORTB = (0b10010101); DelayMs(x/8);

PORTB = (0b10100001); DelayMs(x/8);

PORTB = (0b10100001); DelayMs(x/8);

PORTB = (0b10010101); DelayMs(x/8);

PORTB = (0b01000010); DelayMs(x/8);

PORTB = (0b00111100); DelayMs(x/8);

}

---

If you look close, you can see the smiley face in the actual bits in the code. This code works, and when I wave the LED bar around you can see the smiley face. I'm not sure what the ideal delay time is, but I've been playing around with it and getting some OK results.



The image produced looks like ticker tape with smileys printed on it, each smiley is bumped up against the next. You could put a blank line after the last row if you wanted to 'separate' the images.



I'll try to take a picture of it tonight.



For poi, people may need to either spin really fast to get a good image, or the people watching may need some chemical assistance to improve their persistence of vision.



Otherwise you might just see a weird trailing after image, which could be cool all by itself.



Next step is getting some good calculations for poi speeds, and considering something other than 'one bit' color...



Peace,

sam
EDITED_BY: phidauex (1111347337)

Posted: 20th Mar 2005
Oh, and getting more than 8 LEDs in the vertical array is going to be important! I'm thinking if I use 4 blocks of 8 leds common cathode, use a transistor on one of four I/O pins to activate the cathode for that row then one of 8 I/O pins on the other port to activate the individual LEDs, then I can put those four blocks in a row and get a single column of 32 LEDs.. Which should give us some good possibility for images, while still leaving some I/O open for the acellerometer and a button or two for changing modes.

I'm sure that I can multiplex that fast enough to get a good image with one bit color, but I'm uncertain if I can use PWM on that and keep it refreshing fast enough. Only time will tell.

peace,
sam

Posted: 21st Mar 2005
Sounds pretty cool to me. I got all the chips for the prototypes I'm working on the same way as yourself. IIRC I've got ADXL202 and 210's because they give PWM output. I need all three CCMP ports for the PWM output so rather than using analog inputs I'll just count the digital output.

Do you have any idea how many G's a poi-head pulls? Obviously it's depndant on the length of chain, speed of spinning, but some ball-park figures would be good. I did some calculations a while back which suggested my accelerometer's would be too sensitive - a figure of around 15g and maximum load springs to mind!

Cake or Death?

Posted: 24th Mar 2005
Finally! I got me new programmer working and even managed to program the chip using ICSP (In-Circuit-Programming)



Hopefully I'll get the 3 PWM channels working over the easter weekend. Once that's in place, I can start loading funky patterns from tables



I get what you're saying about the 8 LED block, but I'm a bit confused about the 32 LED bit - any chance of a diagram?
EDITED_BY: FireGeeK (1111678610)

Cake or Death?

Posted: 26th Mar 2005
The acceleration of an object spinning at a fixed angular velocity is equal to velocity squared dividied by radius, or a = (v^2)/r.

The circumference of a standard set of 18" poi is about 113" (2pi*r), or 2.87 meters.

At a 'regular' spinning speed, I make about 10 revolutions in 4.8 seconds (measured by stopwatch). Thats .48 seconds per revolution, or 2 revolutions per second, or 120 RPM.

At a 'fast' spinning speed, I double that and move up to about 4 revolutions per second, or 240 RPM.

If I'm making 4 revolutions per second at my fast speed, that means the speed of the poi head is 11.48 meters per second.

Plugging that into the above acceleration equation we get 293 m/s^2, or about 30Gs. Serious acceleration!!

So my spinning probably ranges from between 10Gs at slow speeds to up to 30Gs at 'full bore'. You can run the numbers for your own spinning style and length of poi if you want. Just remember to convert your units, and remember that the speed of the poi and the length of the poi are the only thing that matter, mass is irrelevant for these calculations.

I do like the idea of PWM output from the accelerometer, but do they offer that in a higher G unit?

----

The 32 LED array concept is similar to the 8 LED array concept, but with another step added in the middle...

Imagine we have an 8x4 grid of LEDs. 8 high, and 4 wide. Each row is connected to one output pin (so you are using an entire port to do all the rows). When that pin goes high, all 4 LEDs in the row turn on.

Now, instead of running each of those LEDs to ground like you would normally, connect the anodes of each column, and connect that to a transistor, which then leads to ground. The base of that transistor is then connected to another output pin on your controller. So now you have 4 blocks of 8, each of which has all their anodes switched on and off by 4 more outputs.

To display an image on this screen, you slice your image into 4 columns. You activate the first row by setting your first transistor output pin high, 1000. Then you set your 8 output pins controlling the rows to the first column of pixels. Since the transistor is only turned on for ONE of the columns, only the first column illuminates. Then you turn off those 8 pixels, and switch off the transistor for that column, and switch on the transistor for the next column, 0100. Activate the rows for the next pixel. Continue working your way down the columns, and when you get to the end, go back to the first. Since you aren't putting a delay between these actions, the 'scanning' down the rows is not perceptable, and you see all 32 LEDs operating in a 25% duty cycle, and you see a complete image, just like the scanning on a TV.

You can practice that, and it is fairly easy.

The next step is to turn our 8X4 in to a 32X1 array. Without changing the circuit schematic, arrange your columns in a big row. Now, instead of displaying a square image, you display a pattern on that row by activating each block of 8 in turn. Since you switch back and forth between those 4 blocks of 8 so fast, you see the entire thing appear to light up at the same time.

Displaying your first column on your virtual 32X32 display is the same as displaying the first frame of your 8x4 display. Set a loop to display that row for a set period of time, maybe 5 ms. Then switch to the new 32 pixel row, and display that, one block of 8 at a time, for another 5ms. Scan your display like that.

It gets tricky because effectively you are having to handle refreshing an entire image AND refreshing individual 8 led blocks, at different rates. Add animation, and you are refreshing entire images at yet a different rate.

I'm still a little shakey on the actual code, but you are going to want to use your timer interrupts to help keep track of when to switch rows, and when to switch frames. I'll post some C code once I get it ironed out.

peace,
sam

Posted: 26th Mar 2005
Unfortunately, Analog don't do a higher G PWM outputting accelerometer

Certainly sounds interesting.. I reckon you're going to have a hard time making the code fast enough for this to work well though. Good luck!

Cake or Death?

Posted: 19th Apr 2005
Hi guys! hope you all have more time than I do at the moment. How are things?

I just stumbled across the 3W RGB emitter/star by Prolight. They are another Luxeon clone, but this one has 3x1W emiters in it!

Cant find a link to the manufacturer; these links may not work for long:
https://www.ledsales.com.au/graphics/rgb_prolight_led.jpg
https://www.ledsales.com.au/coming_soon.htm (halfway down)

https://www.led-bulbs.com/eShop/10Browseprorgb.asp?Category=Prolightrgb

Posted: 21st Apr 2005
Wow! I get the impression they're not out yet, but I shall certainly be getting myself some of those. The ~1cm seperation of the luxeons in [Old link] is a bit of an issue as you can see the seperation. Also, the different colours have their maximum output at different places so the colours don't mix as well as they might. With what you've linked to there, that shouldn't be an issue. Also, it should save a bit of space which could make for even tighter designs!

Nice one!

Cake or Death?

Posted: 30th Jun 2005
I might try to build a mechanical accelerometer, as it would provide a changing pulse frequency in time with the spin rate, instead of PWM that would need to be run through code to provide a clock tick (don't need to know how strong the force is, just the rate of change). I just need to work out how to use interupts, so the pulses can modify the length of a pattern cycle.

You aren't thinking or really existing unless you're willing to risk even your own sanity in the judgment of your existence.

Green peppers, lime pickle and whole-grain mustard = best sandwich filling.

Posted: 15th Mar 2007
If you're looking for durability, rechargeability and programmability, its as simple as getting 2 mobile phones. They're pretty cheap now, and they have places to attach swivels and string. The screens are bright, they last for a couple of hours on a charge, and writing screen-color changing apps, even playing mp3s on them is dead easy.

Want to synchronize? Simply get Bluetooth-capable phones and write an app for that.

Downside: probably hurt quite a bit when struck ;-)