DIY LED driver for reef lighting

[terahz]

Oops, that sux. Sorry.

However, 141F is fine, given that the thermal shutdown is at 320F.

I've had 7 XPGs running at 18mA (lowest DAC setting) for about 1hr now. Vin is 25V, Vfw is 18.37V. Once I put my pinky on the IC, I don't feel any heat for the fist 4 seconds, then 3 seconds later it is too hot to hold. I used my computer's fan controller to measure about 120F, but I don't know how accurate that is given the tiny chip and the clunky probes the controller has.

If I turn Vin up to 26V, the chip becomes considerable hotter. Going to Vin 24, takes about 15 seconds of touch to become really hot, but I wouldn't call it too hot to hold.

At 22Vin the chip is barely over ambient temperature.

So, If switching frequency can be correlated to temperature, it looks like this:

400kHz - slightly above ambient.
450kHz - gets to "too hot to touch" after about 7-10 seconds (110F is probably a good measure of where 'too hot to touch' starts)
500kHz - too hot to touch in about 3 seconds.

That would support what you guys are observing as well.

Still, I wouldn't be worried about it at 150F. It might be worth tweaking Roff and Coff to get a lower frequency for some people.

If you are worried about the heat, add a second Coff(C5) on top of the current one (in parallel) and I'm sure that will make a big difference.

Don't forget to report back guys.

 

[shirley386]

Still, I wouldn't be worried about it at 150F. It might be worth tweaking Roff and Coff to get a lower frequency for some people.

If you are worried about the heat, add a second Coff(C5) on top of the current one (in parallel) and I'm sure that will make a big difference.

Don't forget to report back guys.

Would that address the issue of it not dimming down?

I'd like it to dim down to "moonlight" or "I can't be sure it's on"

Thanks
Shirley

 

[terahz]

Shirley,

I don't know how to answer that as I'm not sure what you expect in terms of dimming down. I still don't know how low in terms of current your drivers get. Assuming the circuit is running correctly, you are as low as it gets. Changing the frequency will not make a difference either way though. Maybe you can tell us the voltage between L+ and L-? That might give us a idea how low your current is at the lowest setting.

I don't think you will be able to dim 14 XT LEDs to the level of 14 3mm LEDs if that is what you're after. Even at 12mA, the 7 LEDs I have are still enough to light up my "lab" (brightness adjusted to about what I see with my eyes, yes I know it depends on the monitor but it is as good as I can do for visual representation):

If you need dimmer than that, you either have to PWM with something that supports 11bits+ of PWM to get in the microamps. (The analog dimmer certainly cannot get that resolution in the lowest points).

Or just get 5-10 cheap 3mm LEDs from anywhere and use them as moonlight. You can probably drive them directly from your micro.

Hope that helps.

 

[AHutch]

A Mastec power supply would certainly make things easier...

So I swapped out the 7.32k Roff for a 16.5k Roff and it changed a few things. For one, Iled dropped from 108mA to 38mA at the lowest DAC setting. The LEDs are considerably dimmer now on the lowest setting. Also, the chip is no longer too hot to touch at this setting. I don't have an accurate way to measure the temperature but lets just say it's no longer a concern.

I grabbed pics of a few traces on the lowest and highest settings.

LED - 7 XT-E
Vin - 27.55V
Roff - 16.5 kOhm
Coff - 470pF
L1 - 33uH
Rsense - .22Ohm
Cin - 2.2uF
Cout - 2.2 uF


Lowest setting:
DAC (2) 9mV
V R6R7 126mV
Vout 18.83V
Iled 38mA

PGate

RoffCoff

Fsw appears to be about 750-800kHz at the lowest setting.

Highest setting:
DAC (4095) 5V
V R6R7 1.29V
Vout 23.19V
Iled 960mA

PGate

Fsw appears to be about 125-150kHz at the highest setting.

The Arduino's onboard voltage regulator stopped working before i could capture the Roff Coff trace on the highest setting. I will say that there is a lot of noise on the highest setting. The scope picked up voltage spikes at the same frequency on the 5V source line connected from the arduino to the driver.

 

[terahz]

A Mastec power supply would certainly make things easier...

The second channel on mine died 3 days after I got it... still haven't fixed it.

So I swapped out the 7.32k Roff for a 16.5k Roff and it changed a few things. For one, Iled dropped from 108mA to 38mA at the lowest DAC setting. The LEDs are considerably dimmer now on the lowest setting. Also, the chip is no longer too hot to touch at this setting. I don't have an accurate way to measure the temperature but lets just say it's no longer a concern.

Yay! Good news for Shirley as well. I guess I was wrong about the frequency not affecting the current.

Lowest setting:
DAC (2) 9mV
V R6R7 126mV
Vout 18.83V
Iled 38mA

That is really odd, how on earth is a voltage divider boosting the voltage of the dac?!

On a side note, I just measured the current for the first 300 steps of the dac and the strangest thing happened. The dimmest setting for me is actually at about 200 (245mV dac output, 65mV at the IC). Not sure why, but that gave me 11mA.

Can you try it on your build ? Are the LEDs getting brighter as you move from 5 to, say, 50?

Thanks for the photos! Can someone explain what they mean?

 

[shirley386]

So I swapped out the 7.32k Roff for a 16.5k Roff and it changed a few things. For one, Iled dropped from 108mA to 38mA at the lowest DAC setting. The LEDs are considerably dimmer now on the lowest setting. Also, the chip is no longer too hot to touch at this setting. I don't have an accurate way to measure the temperature but lets just say it's no longer a concern.

Now that sound promising! Can you see any difference in intensity at the highest setting?

 

[BeanAnimal]

Hrmmm.... did you break into my house and take a picture of my bench?

I just moved my stuff over to this corner to give her bacl the utility room... but in there I had almost the exact same shelf layout as yours with the equipment in the same places... kind of creepy.

 

[shirley386]

Well, I just looked and Roff is R5 and I've already got a 16.5k there. I've got some extra 30k that I used on R7. Would that be worth a try or do you think it would just smoke the whole board! :headwally:

 

[terahz]

Hrmmm.... did you break into my house and take a picture of my bench?
<snip>
I just moved my stuff over to this corner to give her bacl the utility room... but in there I had almost the exact same shelf layout as yours with the equipment in the same places... kind of creepy.

Haha, that's crazy. I promise to put a scope under that bench meter eventually

Well, I just looked and Roff is R5 and I've already got a 16.5k there. I've got some extra 30k that I used on R7. Would that be worth a try or do you think it would just smoke the whole board! :headwally:

Shirley,
I really hope you don't smoke the board . Yeah, I already knew you had 16.5K for Roff, that's why I suggested adding a second Coff (C5) on top of the current one. The frequency is affected both by Roff(R5) and Coff(C5). Since your resistor is already 16500, you can double up on the capacitor. If you just solder a second on top of the first one, that will give you the same result.

 

[shirley386]

Shirley,
I really hope you don't smoke the board . Yeah, I already knew you had 16.5K for Roff, that's why I suggested adding a second Coff (C5) on top of the current one. The frequency is affected both by Roff(R5) and Coff(C5). Since your resistor is already 16500, you can double up on the capacitor. If you just solder a second on top of the first one, that will give you the same result.[/QUOTE]

I'll give it a shot. Nothing more to lose at this point.

 

[shirley386]

Houston - I think we have a problem

Houston - I think we have a problem

terahz - I stacked another cap as you suggested. I don't think the chip gets quite as hot as it did but it's still to hot to hold my finger on. The flickering at low setting has stopped though. It it still pretty bright as the lowest setting but probably not as bright as it was. Also, at max power it does not seem to be as bright as it was. I think that just reduced the brightness across the board.

In a effort to make sure that I had not ordered (or was sent) the wrong components, I got my big magnifier with the little "bifocal" spot on it, opened the blinds so the light would come in and managed to read the numbers on the resistors. I then found this site to decode them:

http://www.hobby-hour.com/electronics/smdcalc.php

(Hope it's alright to post that here)

Here's what I found:

For R6 the BOM calls for 91K ohms but the Mouser number listed is for 91 ohms.
For R7 the BOM calls for 30K ohms but the Mouser number listed is for 30 ohms.

I ordered them according to the Mouser number in the BOM spreadsheet. What do you think would be the result if I swapped these two resistors for the values that you have listed?

 

[terahz]

Shirley, good to hear the flickering has stopped. As for brightness, I think that's just what it will be. These LEDs are fairly bright (that's kind of the point), so it is hard to really dim them down with analog voltage.

Maybe some high precision PWM chip (like the PCA9685PW, as BeanAnimal suggested) will do better job at it. The data sheet suggests that it can handle up to ~13bits of accuracy with PWM, so it should be able to dim the LEDs down to 0.25mA vs the current ~25mA we get.

Regarding the LEDs, R6 and R7 are a simple voltage divider. In general it shouldn't matter much what their value is, it is the ratio that is important. However, I seem to remember the high resistance causing some problems at some point for the dac... hmm. I'm also running with 91 and 30, so I think you're good there.

I guess if you want to use what you have, you can PWM the EN pin (remove U2 and put PWM on the top right pin (the one who's trace goes between C4 and C5/R5 and ends in the middle of U1). But as I said before, I'd just get a handful of cheap LEDs and create your a moon light that way.

 

[nyquist]

So, I setup a driver like DWZM's, and here is my schematic:

I am just trying to get it to work without an actual PWM source for now, so I have a switch hooked up to the enable pin


I have tried the above using between 7 through 13 LEDs as the load. What happens is this:
(V1 = voltage across 1 ohm resistor; Vled = voltage across all the LEDs when on)

7 LEDs: V1 = 106mV, Vled = 23V (just under 3.3V per LED)
8 LEDs: V1 = 85.6mV, Vled = 25.6V
9 LEDs: V1 = 73.4mV, Vled = 28V
10 LEDs: V1 = 66.5mV, Vled = 30.5V
11 LEDs: V1 = 60.5mV, Vled = 33V
12 LEDs: V1 = 55mV, Vled = 35.6V
13 LEDs: V1 = 51mV, Vled =37.9V (just under 2.9V per LED)


So as I increase the number of LEDs, the current goes down and accordingly the voltage drop per LED goes down. Am I clearly missing something?

Why isn't my constant current driver supplying constant current? (I also setup two independent circuits the same way, both behave the same, so I am quite sure I didn't just set something up wrong, unless my schematic is wrong) Anyone have any insight for me? It would be greatly appreciated

 

[terahz]

Shirley, Bean and others, I've ordered a 12bit PWM chip to give it a try with the lm3409s. I'll let you guys know how that works. It should be able to give linear 4000+ steps at up to 1MHz.

 

[kcress]

12b = nut case..
Why not 16bits!
No! no! 24bits!!

 

[shirley386]

Shirley, Bean and others, I've ordered a 12bit PWM chip to give it a try with the lm3409s. I'll let you guys know how that works. It should be able to give linear 4000+ steps at up to 1MHz.

I did the PWM mod that you mentioned. This looks much more promising. It really dims now. I'll build a few more and test them. Also less components to solder!

It still started to flicker at the highest setting. I could hold the eraser end of a pencil on the chip and it stopped. When I took the pencil off it had begun to melt the rubber eraser. I was able to turn the power supply down a little to make it stop flickering without noticing any difference in intensity. I think the flicker must be a part of the thermal shutdown.

 

[terahz]

12b = nut case..
Why not 16bits!
No! no! 24bits!!

Heh, this time I wasn't looking for 12b, just trying to get higher 8 bit and that was the first one that I saw so I bought it . But 16mil steps sounds like fun too

I did the PWM mod that you mentioned. This looks much more promising. It really dims now. I'll build a few more and test them. Also less components to solder!

It still started to flicker at the highest setting. I could hold the eraser end of a pencil on the chip and it stopped. When I took the pencil off it had begun to melt the rubber eraser. I was able to turn the power supply down a little to make it stop flickering without noticing any difference in intensity. I think the flicker must be a part of the thermal shutdown.

You might be right there. Either way it is good to lower the PSU to as close as possible to the LEDs Vfw + ~0.5V. It is more efficient that way.

 

[shirley386]

terahz - I think I remember a few pages back you had used an ethernet cable to power your LED string(s). How is that working out? I was thinking about that too and looked up the specs on Cat5 cable and I think the current was about 500 mA. Other people have mentioned that it might not be a good idea as Cat5 is intended for data, not power. I'd like to get your experience.

 

[terahz]

No problem running cheap ethernet on mine for a couple of years now.

 

[89Delta]

Have read through this thread and am very impressed with everyone's work on the various drivers. I was thinking of trying my own build using the AL8805 LED Driver. Has anyone had any experience with these at all?