DIY led driver

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
I need a 12v 5a led driver, and I'm thinkingof making one with the LM2678 switching regulator. How would I do constant current Just set the regulator to 12.1v and then use a resistor?
 

Mike S

Newly Enlightened
Joined
Apr 29, 2011
Messages
132
Check out this PDF. There are several variations of this circuit floating around.

http://www.ottomat.hu/Kapcsrajzok/aramgenerator3.pdf

I've tried to built it several times and failed to get the full 3A output. If you have a good understanding of OP AMPS, then it shouldn't be a problem.

Another option is to use the switching regulator to drop voltage to a few parallel low dropout LED drivers.
 

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
I'm thinking I might use 14 7135s in parallel. I'm using 5xcree xml easy white leds, I could power it off of a 4s lifepo4 pack which would give around 12.8-12v for much of its life.
 

jpou

Newly Enlightened
Joined
Mar 31, 2012
Messages
19
I'm thinking I might use 14 7135s in parallel. I'm using 5xcree xml easy white leds, I could power it off of a 4s lifepo4 pack which would give around 12.8-12v for much of its life.

And have 14 points of failure? Why not get one or two normal drivers?
 

Th232

Flashlight Enthusiast
Joined
Dec 25, 2008
Messages
1,064
Location
Sydney, Australia
I'd rethink that as well. With a fully charged pack you're dissipating something like 12 W of power through the 7135s. Very wasteful and more than a bit toasty. Why not just buy a TaskLED H6CC instead?
 

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
really? Nominal 3.6, but that drops quickly to 3.2 and the forward voltage is 11.7 the TaskLED H6CC is pretty expensive.
 

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
at 3.6v its 82 percent efficient with 7135 drivers, but at 3.2v its 88 percent efficient and once its down to 3v its 97.5 percent efficient. Its distr ibuting 11 watts at 3.6 volts per battery, but that is about the same as the H6CC

Most normal drivers don't power cree xml easy white leds if i understand currently.
also the h6cc needs a full volt of voltage drop.
 
Last edited:

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
I've been thinking about controlling it with an attiny chip. I think I'll have a thermistor next to the leds to check the temperature and three modes. It will switch through them based on presses. The first mode will apply voltage to all 14 7135 chips vdd, the second mode will apply power to 7 amc 7135 chips vdd, the third mode will apply power to 3 of them.

I just had a novel idea. From my understanding pwm does not increase efficiency of the leds, it just decreases power draw. I could however connect a pot to a larger attiny chip and have 14 "steps", I would ivied the pot into 14 steps, the top having all 14 of the 7135 chips being powered, and the bottom only having 1. This would wallow the efficiency to increase as the power dropped. How does this sound?
 

uk_caver

Flashlight Enthusiast
Joined
Feb 9, 2007
Messages
1,408
Location
Central UK
If you have 5xXML what supply voltage and circuit arrangement were you thinking of, how were you thinking of using a step-down converter, and what LED current were you hoping to get?

In your second post you talk of a 4s lithium pack but I can't quite see how that fits with 5xXML and a buck converter.
 

jpou

Newly Enlightened
Joined
Mar 31, 2012
Messages
19
I've been thinking about controlling it with an attiny chip. I think I'll have a thermistor next to the leds to check the temperature and three modes. It will switch through them based on presses. The first mode will apply voltage to all 14 7135 chips vdd, the second mode will apply power to 7 amc 7135 chips vdd, the third mode will apply power to 3 of them.

I just had a novel idea. From my understanding pwm does not increase efficiency of the leds, it just decreases power draw. I could however connect a pot to a larger attiny chip and have 14 "steps", I would ivied the pot into 14 steps, the top having all 14 of the 7135 chips being powered, and the bottom only having 1. This would wallow the efficiency to increase as the power dropped. How does this sound?

And so you will need 14 FETs (or 28 if using N-P FET pair) to switch power to 14 7135s... Talk about increasing board space...

I could however connect a pot to a larger attiny chip and have 14 "steps"
smallest attiny with ADC is about 3mm x 3mm size. Remember that you have 14 7135s and 14 (or 28) FETs. Overall pot idea is good but WHY you don't go proper driver route instead of 7135s...
 

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
I thought I could just switch the power to the vdd.

I have very low voltage overhead. Could you explain to me a proper driver route? It would appear I would need a buck/boost driver to use a proper route. ALso at 1a current the easywhite leds vf is about 12.5. Couldn't I just switch power to the vdd of the 7135 chips? have the attiny apply power or not and tie the grounds?

Please help me understand. I have not done a lot of electronics. For testing i'm using a NFET and a transistor as a constant current source, however direct4l off the batteries, the leds do not turn on. I'm using a boost converter for testing, but Its too big for a light and then I have the inefficiency of boosting and then regulating and burning off the extra voltage as heat.
 
Last edited:

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
@caver I have 5 cree xml easywhite 12v leds. I would wire them in parallel. I'm hoping to run them at 1 amp each, for 5 amps total, all in parallel. I will most likely run then slightly below that incase the vf is different on them so if one is a current hog it doesn't go over spec. at 1 amp the vf is about 12.5 volts. The other option is to wire them all in series and power them with a boost converter. I don['t want to use a prebuilt driver like the hyper boost.
 
Last edited:

uk_caver

Flashlight Enthusiast
Joined
Feb 9, 2007
Messages
1,408
Location
Central UK
And so you will need 14 FETs (or 28 if using N-P FET pair) to switch power to 14 7135s...
Not really - the MCU can drive 7135s directly, if they're connected conventionally on the low side of the LEDs and any turned on together can have their Vin pins connected together, driven by a single MCU output.
The current required is minimal so an MCU should be able to handle a decent number (as suggested by the various Chinese drivers with an MCU driving numerous 7135s).

In any case, to turn on any number from 1 through 14 of 14 output-paralleled 7135s would only take 4 outputs, driving 1,2,4,7 chips respectively, and turning them on in a slightly modified binary pattern (take desired number, output directly if 7 or less, otherwise add 1 and output).
 

uk_caver

Flashlight Enthusiast
Joined
Feb 9, 2007
Messages
1,408
Location
Central UK
@caver I have 5 cree xml easywhite 12v leds. I would wire them in parallel. I'm hoping to run them at 1 amp each, for 5 amps total, all in parallel. I will most likely run then slightly below that incase the vf is different on them so if one is a current hog it doesn't go over spec. at 1 amp the vf is about 12.5 volts.
Ah - I'd been thinking or regular XM-Ls - my mistake.

With the NFET circuit, how much voltage overhead should it be taking to turn on?
 

uk_caver

Flashlight Enthusiast
Joined
Feb 9, 2007
Messages
1,408
Location
Central UK
As for using 7135s, it does seem like the potential voltage overhead on fresh cells could be a bit large.

What capacity do you have for cooling the LEDs/driver?
 

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
I haven't tested the minimum required to turn on, but it is more than 1 volt. I'm right now running it sat about 18 volts, just for testing. I don't have a definite plan for cooling yet, however I plan on making a housing on a lathe and cutting fins going from the front to the back on my high school's mill or my hand router with a jig to fit on my wood lathe which has indexing. I'm thinking it will automatically step down output as temperature increases above a certain threshold. I've though about it and if I used 7135 chips I would have 14 levels, 1 through 14 chips active. As temperature increases It would lower the levels until he temperature levels out. I thought of pwming between active chip levels, but I don't think it is neccesary.
 

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
the discharge curve of a lifepo4 is about:
typical-lifepo4-discharge-curve.jpg

i ordered 10 a123 batteries from ebay, but I dobt they are really a123. i'll do a discharge test soon.
But much of the life is between 3 and 3.25 volts depending on my batteries discharge curve.
Little of it is at the full 3.6 volts fresh off the charger.

I'll run some thermal tests in solid works once I get it setup on my compeer and reinstall windows.
 
Last edited:

uk_caver

Flashlight Enthusiast
Joined
Feb 9, 2007
Messages
1,408
Location
Central UK
I guess with a total voltage that low (I really have been somewhat distracted today), 7135s probably would be OK, and would certainly give smooth-ish control, though turning on 1 through 14 of them might give a limited number of effective steps (the difference between 'N on' and 'N+1 on' or 'N+2 on' might be pretty hard to see for larger values of N).

If you had a spare output, you could turn on the 7135s in clusters of 1,1,2,4,and 6 and with PWM control for one of the '1' outputs, you could get continuous control over the whole range while having most operating levels being PWM free, and have the ability to dim right down to zero.

If you're controlling via a pot, it might be worth trying to make the overall 'feel' linear even if that involves some post processing of the pot result, or some pre-mangling of the pot input voltage (it's surprising what a resistor or two can do to nonlinearise a linear pot).
 

Hmmm

Newly Enlightened
Joined
Apr 26, 2009
Messages
147
I was considering pwm control between levels and then switching to more 7135s at a lower pwm for higher levels, but i'm not sure If I will do that. Smooth control would look nice. I can control all of the "linearness" when programming, I think.

That second idea sounds very interesting. I think I understand what you are talking about. I would pwm only one of the 7135 chips and the rest would be either on or off. That sounds very good and would help with efficiency.

Am I correct that with this low of a difference between supply and output voltage a linear regulator like a 7135 is a good choice?
 
Top