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

Posted: 21st Sep 2004
Ooops, sorry forgot to say...

Each of the red tubes has 6 20mA, 2.4v red leds in.
The green/blue tube (which shows up as a white blob) has 4 blue and 2 green 50mA, 3.6v leds in.

Like I say, it's not a very good comparison. One day, I'll sit down and figure out the whole lumens/candelas/radiant flux stuff, but it's not as easy as it should be, cos some give total output, and others give output per degree of viewing angle etc...

monkeys ate my brain

Posted: 21st Sep 2004
You could set a camera to a fixed shutter speed and then try each LED pointed at a wall of something.. ? Might give a reasonable idea (even if it's not very scientific)...

Cake or Death?

Posted: 21st Sep 2004
I have some ~5000MCD LEDs - there are four in each of the Poi you seed in my avatar. Take a look here for the conversion from MCD to LM. For my LEDs, that's a LM of (roughly) 0.4 - compare that to the Luxeon emitters here. So for a corresponding blue luxeon, you're talking a 10x improvement in LM. I'm not sure how much brighter that would appear, but from what I've read, they're significantly brighter than regular Super-Bright LEDs.

One question which springs to my mind is what gives the best LM/W i.e. which gives you the most light for the same amperage. Afterall, these poi will run off 4xAAA batteries - each one is rated for 700mAh so the less current used, the longer they'll last

Cake or Death?

Posted: 26th Sep 2004
Right, I just forked out more money than I'd have liked for 6 luxeons, 2 red, 2 green and 2 royal blue. I have all the bits and peices now to build some uber poi - just need to plug it all together (once the luxeons arrive). Wish me luck!

Cake or Death?

Posted: 13th Dec 2004
Blimey, how did I miss this whole discussion, you guys could have done with my help last year building some 1 Watt Luxeon LED's into your sequencers

Wizard of LED poi !

Posted: 18th Dec 2004
FireGeek: are you using emitters or stars ?


some links you might find interesting/useful:

2xAA mag light VS (probably) 1w Luxeon
https://home.mchsi.com/~lambda2/newpill6.jpg

also
https://home.mchsi.com/~lambda2/kit2.htm
https://home.mchsi.com/~lambda2/kitinst.htm

so little time. hope to get back to this soon.
vim

Posted: 20th Dec 2004
Plugged them in and nearly blinded myself - these things are DAMN bright! I made the mistake of getting stars rather than emitters. I'll get some emitters when I get round to ordering some more - should minise the spread between the rgb elements.

Those links are for white luxeons aren't they? Or were you suggesting the power source was of interest?

The Luxeons I have at the moment draw about 300mA each, so a little under a watt if all three are on at the same time.

Cake or Death?

Posted: 26th Dec 2004
Wahoo! Just got a PWM signal out of my little PIC @ ~2.5Khz. I have a button which alters the duty cycle and hence the brightness over 8 steps. Time to crank up the PWM signal - 2.5KHz will leaves gaps of about 10mm when spun

Cake or Death?

Posted: 15th Jan 2005
yep, for the white - as an example of the kind of light a single luxeon will produce. (for anyone interested)

I _was_ very interested in the DC-DC converter, but have since changed my mind (again) and decided to use AAA cells (primarily because one 123 lithium doesn't, IMO, weigh enough and two cells is difficult because the 12F675 pic isn't happy at anything over 5v).

I'm still playing with the idea of using a lithium polymer cell (as in mobile phones), but they have to be monitored for a drop below 3v and need a special charger etc etc - not KIS. The nice thing about them is that they are rechargeable and give 3.7v.

It's great to see you're making progress; I hope to be making a little of my own as soon as a new programmer arrives and I work out how to use this In Circuit Serial Programming thingumi.
I'll leave the big guns with you and your luxeons though: I'm going to see if I can get anything usable out of the 20mAh supplied by the pic.

Posted: 15th Jan 2005
hey Firegeek or Vim,
whats the PIC and programmer your using? is the pic small enough to fit into a glowstick?

i wanna try to make my leds strobe in a glowstick...

thanks

Posted: 23rd Jan 2005
Hi Dan,



I was using the PG2C and it worked perfectly for me. I think it cost USD $12 on ebay.

I no longer have a serial port so I'm now using one of these, and paid AUD$50 for the privilege. Works a charm.



You shouldn't have much trouble fitting a 5mm LED and pic into a large glowstick, particularly if you solder the LED directly to the pins on the pic; coin cells would seem to be the most suitable power source.



I'm using the 12F675 , though if you're just starting you may find more example programs and help available for the PIC16Fxxx chips.

Posted: 23rd Jan 2005
Lol Vim,

Guess what programmer I was using? That's right, a PG2C.. and guess what I'm using now? Kit 128 from the same site! hehe, great (read cheap) minds think alike eh?

I've almost got ICSP working.. I can program my 16f767 with code, but for some reason it chucks up an error when it gets to doing the configuration bits I've got a DIP-packaged version on it's way to try without funky ICSP stuff. It could well be my dodgy soldering trying to get such small pins onto wires I can handle!

I've borrowed my dad's etching kit, so pretty soon, I should have a prototype board I can try things on - can't wait!

Cake or Death?

Posted: 23rd Jan 2005


If this helps, I'm using this config:
; Internal Oscillator ON (_INTRC_OSC_NOCLKOUT)
; 4MHz, onboard Osc. NOCLKOUT don't send any clock output on pin 3.
; Power Up Timer ON (_PWRTE_ON)
; Brown Out Reset ON (_BODEN_OFF)
; Watch Dog Timer off (_WDT_OFF)
; Master Clear pin off (_MCLRE_OFF)
; Code Protection off (_CP_OFF)
; EEPROM Data Protection off (_CPD_OFF)

__CONFIG _INTRC_OSC_NOCLKOUT & _PWRTE_ON & _BODEN_ON & _WDT_OFF & _MCLRE_OFF & _CP_OFF & _CPD_OFF


With a bit of guess work I managed to get the 12F675 working with ICSP, but to actually get it working 'In Circuit' will be a pain because all the pins have to be isolated from the application circuit while it's being programmed. I suspect it might be easier if there where more than 4 pins to play with, then they wouldn't have to share duty. I ended up breadboarding two circuits, one application and one ICSP so I can swap the pic between the two.
I'm a little annoyed that the Kit only supplies power from USB when programming...

A few days ASM practice and I'll see if my Ferric Chloride is still viable, and that the copper hasn't corroded off my PCB stock ;-)

Back in business!

Posted: 25th Jan 2005
Got software PWM working a charm; 128 steps is more than enough for the amount of light produced by 5mm LEDs.
Tonight I'll see about adding some lookup tables so fades are more linear.

Programming in asm has it's down sides, but on the other hand you get SOOOO much more space.



FireGeek: any news?
I've been meaning to ask: what are you using to switch power to the LEDs? I've been thinking the irlml2502 would do the job.

Posted: 26th Jan 2005
Vim, what frequency do you have that PWM going at? I gave up with software in the end! Do you use caps to smooth it or something?

No new news as yet. I tried a PIC16F767, but wasn't able to program it properly. I'm not sure if it's the new programmer, my dodgy soldering, ICSP, or something else yet. I have some more chips - DIP packaged so hopefully I'll be able to try those. If they work, I'm back on track

I use a little mosfet per Luxeon.. can't remember the part number off hand - I'll have a look when I get back in this evening.

I remember someone saying that humans see light as a log function, so you might try something like that to adjust the fading. Reading values out of a table is another thing on my to-do - I thought I probably do the same as you, using a lookup to smooth the appearance of the fade.

Have you given any thought as to how you'll store light sequences? Like red for 3 secs, blue for 3 seconds, fade to green etc. etc. I've given it some thought but not really come up with a solution I feel happy about.

..and yes, ASM is the way forward

Cake or Death?

Posted: 26th Jan 2005
It's amazing how many glow-toys are out there with fixed colours. By that I mean there seem to be a lot out there that do one or maybe two colours, but none that will reproduce the full range as ours (hopefully) will. Definitely a gap in the market there.

Cake or Death?

Posted: 26th Jan 2005
Alright,

If you guys figure out how you want to program your chips, I'll be well up for getting my [Old link] software to output the right format. The ideas that spring to mind are:

- a lookup table of values for a whole pattern, with a speed setting (not so good for fades)
- opcodes for solid/strobe/fade etc, which take values so you can build functions out of little modules

Keep up the good work!

monkeys ate my brain

Posted: 26th Jan 2005
FireGeek:

I used the Kit182 pdf file and the docs for the 628A to work out where all the ICSP connections should be on the 12F675. It was a lot easier than wading through the MicroChip ICSP documentation.

This was the important bit:

Written by:

---

Here is a photo of how we setup Kit 182A to program a

628A chip. VPP to pin 4, GND to pin 5, VCC to pin 14,

CLK to pin 12, DAT to pin 13. Pin 10 to GND with a

10K resistor. www.kitsrus.com/jpg/k182_6.jpg

---

What had me for a while was that VCC == VDD.





I'm haven't worked out the frequency yet. From memory I think the current (ugly) pwm loop takes the equivalent of about 30 instructions: about 30uS with the internal 4mhz clock (I think). Which would mean that at at it's dimmest the LED would be lit only once every 4000uS or so.

I could be very wrong regarding the time per instruction. I'll get back to you.



Caps between the pic and the LED, or the pic and the fet, will smooth the output if it's not fast enough, but I'm hoping they wont be required.



oh, and I was hoping that I could assume that the light produced vs voltage was exponential...

(a while later) damn. They're all over the place. The luxeon III is nothing like exponential, and the 5mm LED I'm using now almost is. Then there's junction temperature, etc ,etc. . I think I'll just do it by eye for the moment



I've not given much thought to sequence storage or input for a while. I still like the idea of holding the toy up to the monitor, and I think it's doable.



It will be nice if we manage to get this working. Restrictions on fire seem to be increasing, so I feel we need to push glow-toys into the 'affordable quality' range asap. I'm hoping that, even if what we are doing here doesn't make it to market directly, at least those in the market already will move things along a bit.



mo-seph:

I like the opcode idea: it's code heavy but data light (get it, get it? ), which is fine by me at the moment because I think there will be plenty of room for code.

ohohohoh! (and duh)... I played around with a little poi applet of my own a while back, but drawing the fading trail took a long time (actually it was changing the alpha value of the colour that seemed to take cpu time). It just occurred to me that instead of blanking the offscreen image (if you're doing it that way) every frame, just paint over it with a transparent background colour. The amount of alpha you use will determine how long the trails are!





...sometimes I think my brain has it's on/off switch wired wrong...




EDITED_BY: vim (1106789248)

Posted: 27th Jan 2005

Written by: vim

---

It just occurred to me that instead of blanking the offscreen image (if you're doing it that way) every frame, just paint over it with a transparent background colour. The amount of alpha you use will determine how long the trails are!

---

Great minds....

monkeys ate my brain

Posted: 27th Jan 2005
I hope your code's faster than your estimate vim - 4ms is waaay too slow once you start spinning these things around. I reckon you'll need some caps in there to smooth things out unfortunately. I originally tried writing some PWM code running on an 18f1320 @ 40Mhz. It was probably badly coded, but whatever I tried I never got it fast enough. Good luck!

I thought I'd try starting with an exponential / log and see if it improves the look of the fade.. then improve it by eye. Wait and see what works I guess!

I too wrote a basic poi program based on that very same idea

I know it's a way off (learnt to walk before you run right? ), but there's a feature I've been toying (get it, get it? :P) with. Aerotech's allow you to make your own programs through a series of taps. I'm not sure if they're stored permanently, although it'd make sense if they were? Anyways, my idea was that you'd have an infra-red tranceiver on each head. You then point it at a base unit on your PC which allows you to program new sequences (built using mo-seph's light-sequencer, or something like it). Now for the really cool bit.. if you have a few people with these toys, or ones that accept the protocol we'll develop, you could get the poi to teach each other. Imagine it - you're out spinning and you see this great sequence someone's come up with. You ask how they did it, and they just point their poi at yours - voila, you have the same cool sequence!

As well as that, you could keep all the sequences people come up with on a website somewhere - I've registered www.hyperlights.co.uk with this in mind So people can share the kewl stuff they come up with.

I don't think it's possible with the PIC I'm using to write to program memory, but I'm sure we could work out how to get a little eeprom working with the units.

Cake or Death?

Posted: 27th Jan 2005

Written by:

---

and they just point their poi at yours - voila,

---

Great minds...

Posted: 27th Jan 2005
I was thinking of making a simple serial protocol, which you could do with light flashes, and then you could have a java applet which flashed your computer screen to transmit the patterns - then you don't need any hardware etc. and you can do it in any internet cafe you come across. (Although explaining to the owner why you were rubbing a 5 foot staff on their computer screens might be interesting).

Then it would also be easy to get the toys to transmit it.

The units I have so far are:
strobe( time, speed )
fade( time, startLevel, endLevel )
hold( time, level )
stepSequencer( stepTime, number of steps, step settings )

The step sequencer is useful for doing "knight rider" style effects: if you have a bunch of different lights all running a step programme, it's easy to coordinate things between them. You can probably even get them to write letters in super cheesy style...

If you've got suggestions for others, I'll try and knock them up, and if you figure out an opcode spec, I'll make some output.

monkeys ate my brain

Posted: 27th Jan 2005
LOL @ the staff comment. It'd be pretty cool though - take your PDA out to spinning sessions and write new sequences out there

Would an IR transeiver be able to pick stuff up from the screen d'you think?

Are those effects per colour or what?

Cake or Death?

Posted: 27th Jan 2005
They're per "block". You can arrange any number of blocks in any positions on the toy. Each block is only a single colour, so to do full colour stuff, you have to overlap rgb blocks and make a sequence for each one.

No idea how much IR a screen gives out; I was just going to use visual spectrum light. The main problem is screen refresh rate - 60Hz -> 30 bit/s max bandwith -> 15 b/s useable -> about 2 bytes a sec -> about 2 secs per sequence unit (opcode + 3 parameters) -> 10s - 1 min depending on pattern complexity. Which isn't ridiculous, but it's a bit slow.

Would almost be quicker and easier to use a small microphone and the speaker. Spectrum games on tape, anyone?

"I've got four billion flashing sequences on my ipod..."

monkeys ate my brain

Posted: 30th Jan 2005
Sounds like it shouldn't be too difficult to make some sort of output our PICs could use?

How would you sense the light from the screen? I would've thought an LDR would be too slow?

Microphone could be pretty cool - with some clever DSP stuff you could also use it to sync to music!

Cake or Death?

Posted: 30th Jan 2005
do you guys have any experience with using a 555 timer?

Posted: 30th Jan 2005
I spent an entire day trying to build a battery enclosure for three AAA batteries. I forgot to KIS. It sucked. Two of those CR2032 and a nice factory made, PCB mounted battery holder are starting to look very attractive.


dan:
not a lot myself, but if you're trying to strobe an LED: this Google Search ought to provide you with more than enough information.

Posted: 31st Jan 2005
...ah, but they're not rechargeable! ..er, are they?

I used some ~10mm (ID) aluminium tube for my last set of LED Poi. It held 4 AAA batteries, and formed the chassis for the poi. I had the LEDs positioned around it, shining up the tube. Worked pretty well.

Have you thought about using one of the standard battery holders you can get from maplin, farnell or RS?

I'm planning to construct comething using a mould and some resin for my new luxeon beasties

Cake or Death?

Posted: 1st Feb 2005
well... no, but...
much trouble this is.

soooo. just a random thought: < ehehe, deleted> ahh, yes: why hasn't anyone stuck a couple of luxeon III stars in a staff? I'm thinking that the amount of light produced would have to blow away anything else around at the moment in terms of shear visibility... no? and there is stacks of room for batteries, heat sinks, etc.


back to the battery holder: Something like this (ebay link, probably won't be around long) would be very nice, but I don't seem to be able to find one (3xAAA battery holder in triangular prism or squat cylinder shape. the link is/was to an LED torch).
I used some aluminum tube to make an LED torch a while back. The AAA batteries fit perfectly end to end. The problem is that I've been trying to make a discreet package that contains all the electronics, and is separate from the diffuser. I thought it would be nice to be able to remove the diffuser and use your poi/staff as a torch (via an override button so you don't have to navigate your way back to the tent using 'fit-inducing-strobe-with-sparkles_#3.seq').

I think it's doable, but I need a cnc mill, a lathe, and possibly an injection molding machine. Though I have had some success just using hot glue and duct tape..

in other news:
my calculations where correct, and yes, it was a little under 4ms. Totally Horrible. Today I re-wrote a lot of code,added the tables (though I'm not sure if they are worth it with the little tri-colour LED I'm using), basic sequencing, and reduced the number of light levels.
I took it down to 16 steps between off and full brightness, just to see how little I could get away with, loaded it into my mostly-hot-glue-and-duct-tape-with-a-few-bits-of-electronics-embedded prototype, and give it a spin:

oooohh! aaaaah! (well, I was impressed)

I've set the colours to fade on and off at different speeds, resulting in a fairly continuous morph (did we end up calling it that?) from colour to colour, but with odd patches of black and white.
16 steps gives a nice steady transition with this LED, and doesn't strobe at all even when the LED is spun without a diffuser. With a decent diffuser I think 32 steps would be fine, but unnecessary.
For lights with larger range the circuit is going to have to contain FETs anyhow, so adding a few capacitors into the circuit isn't going to be a big deal, but for this I was trying to get away with as few components as possible. I think two (tri-colour LED and PIC chip) is about as minimal as is possible, and I'm very happy with the results so far.

I think the resin idea is a nice one, and very flexible in design terms (though not so flexible in physical terms!)

Posted: 1st Feb 2005
oh. Just in case I didn't mention this before.



I looked into using PET (Polyethylene Terephthalate) bottle preforms as poi/staff heads a while back. It turns out (can't find the articles again) that PET will often perform better vs polycarbonate in drop tests. Someone else also 'tested' a PET bottle preform with a hammer and was unable to do more than scratch it. YMMV: I have not managed to get my hands on any so can't test this.



There are a huge range of shapes, sizes and colours 'available' (unless you happen to live where I do, it appears ), and all are smooth and uniform.




EDITED_BY: vim (1107262482)