DIY LED driver for reef lighting

[89Delta]

KRavEN,

I've downloaded both zip versions of your driver design and from the location on github, but I get error trying to open the eagle files. Says it's invalid data in file...any ideas??

Thanks

[shirley386]

I got all the parts to build several of Terahz's LM3409 I2C drivers and got one put together. I connected it to a Hydra with a sketch that searches for I2C devices. If found it so I felt good about that. I soldered up a string of 12 LEDS for a test run. I loaded a sketch that I thought would dim them to about 50% and connected the Hydra to the driver, power supply to the driver, and LED string to the driver. I powered everything on and WOW did they light up! I was surprised that they would be that bright at 50%. To further test it I changed the value of the dimmer and reloaded the sketch. No change, still very bright. No matter what value I enter they do the same thing. Then I unplugged everything and took the Hydra with the I2C connection, ground, and 5V off. The only things connected are the power supply, driver, and LEDs. Still does the same thing. I checked as best I could with a high power magnifier but I can't see a short anywhere.

Kinda mixed emotions; yes it works but it doesn't work right!

Any suggestions? I think tomorrow I'll just start over building another driver.

[89Delta]

Great thread everyone...with all of you contributing something I thought i'd throw it out and try to see if the LT3496 is a good candidate. If I read everything online correctly I can run 10x 3W LEDs @750mA across all three outputs the chip has using a 30V PS in conjunction with the LM2841 I believe. I know its meant for use in RGB applications but I thought WTH...

[shirley386]

Now I'm wondering if I've got the LM3409 on backwards. If I'm looking at the board with the heat sink away from me, the little dimple is on the top left side, on the side next to L1. I'm now noticing a small white dot on the silkscreen on the opposite corner. Should the dot on the chip line up with the dot on silkscreen?

Sorry to be a pain!

[89Delta]

Yes Shirley,

All chips have a tell-tale sign of where the correct alignment is for the chip or any other for that matter...lol. Call it a "cheat sheet".

Originally Posted by shirley386

Now I'm wondering if I've got the LM3409 on backwards. If I'm looking at the board with the heat sink away from me, the little dimple is on the top left side, on the side next to L1. I'm now noticing a small white dot on the silkscreen on the opposite corner. Should the dot on the chip line up with the dot on silkscreen?

Sorry to be a pain!

[noplay180]

How many drivers would i need to run 225 3 watt drivers are they dimable

[89Delta]

You would need 38 of the CAT4101 Drivers if you go that route or 13 triple drivers ....unsure of the LM3409HV though.

[terahz]

Originally Posted by shirley386

terahz - I've got the parts ordered for your LM3409 driver with I2C dimming. Could you post a sketch snippet of the code you're using for dimming?

The basic code for telling the driver to dim looks like this:

PHP Code:

    byte Program = 64; // Program command
    byte Device = 96; // Driver Address
    int value = 500; // (0-4095)

    byte b1 = byte((value / 16));
    byte b2 = byte(value % 16);
    
    Wire.beginTransmission(Device);
    Wire.write(Program);
    Wire.write(b1);
    Wire.write(b2 << 4);
    Wire.endTransmission(); 


Originally Posted by shirley386

Then I unplugged everything and took the Hydra with the I2C connection, ground, and 5V off. The only things connected are the power supply, driver, and LEDs. Still does the same thing. I checked as best I could with a high power magnifier but I can't see a short anywhere.

Kinda mixed emotions; yes it works but it doesn't work right!

Any suggestions? I think tomorrow I'll just start over building another driver.

Check the voltage between C4 and R4 (pin 2 of the lm3409). It is likely that it is shorted to something high and overriding the DAC voltage.

Originally Posted by shirley386

Now I'm wondering if I've got the LM3409 on backwards. If I'm looking at the board with the heat sink away from me, the little dimple is on the top left side, on the side next to L1. I'm now noticing a small white dot on the silkscreen on the opposite corner. Should the dot on the chip line up with the dot on silkscreen?

It is unlikely that you got it backwards and still have working LEDs.

[terahz]

Originally Posted by noplay180

How many drivers would i need to run 225 3 watt drivers are they dimable

Depends on the current you plan to drive them and if you're ok running strings in parallel.

If you drive the LEDs at around 700ma, you can easily put 3 parallel strings of 14-15 LEDs each per [email protected] (48V in). That will mean you need 5 drivers. Yes, they can dim the LEDs too.

for the CAT4101 drivers, that is about 32 ICs or about 11 tripple channel boards (if my math is right).

[ZenMastr1968]

Ok, could use a little help. I didn't find the lower amp LM3409 lists for THz's design, so I'm going to try to run through the spreadsheet myself. Can some one give me an indication of a good target for frequency in step 2? Up above that section it says Fsw should be between 100 - 1,00kHz. Is that the same frequency and is that 100 Hz or 100kHz to 1,000kHz (and isn't that 1mHz)???

[ZenMastr1968]

and as a follow up - what is Cout on THz's design? I don't see that referenced in the data sheet, but do see C2 on the output side going to ground - is that it??

[shirley386]

OK, I've built two working drivers. I built a "test string" of 6 XT-E CW and 6 XT-E RB. While it will certainly dim, it does not dim as low as I expected. With the minimum value in Terehz's sketch (16), it only dims to what I would call 50 - 60 %. I had hoped that it would dim down to almost nothing.

Any ideas?

[terahz]

Originally Posted by ZenMastr1968

Ok, could use a little help. I didn't find the lower amp LM3409 lists for THz's design, so I'm going to try to run through the spreadsheet myself. Can some one give me an indication of a good target for frequency in step 2? Up above that section it says Fsw should be between 100 - 1,00kHz. Is that the same frequency and is that 100 Hz or 100kHz to 1,000kHz (and isn't that 1mHz)???

If you don't want to get into details about the frequency, use 400kHz. It is what is used in the application notes.

Originally Posted by ZenMastr1968

and as a follow up - what is Cout on THz's design? I don't see that referenced in the data sheet, but do see C2 on the output side going to ground - is that it??

Yes, C2 is the output cap.

Originally Posted by shirley386

OK, I've built two working drivers. I built a "test string" of 6 XT-E CW and 6 XT-E RB. While it will certainly dim, it does not dim as low as I expected. With the minimum value in Terehz's sketch (16), it only dims to what I would call 50 - 60 %. I had hoped that it would dim down to almost nothing.

Any ideas?

Ok, few things here. Can you measure the voltage at the point between R4 and C4 for the highest and lowest values you can use (I assume when you use 15, the LEDs shut off?).

Do the same at the point between U3 and R6.

[shirley386]

Originally Posted by terahz

Ok, few things here. Can you measure the voltage at the point between R4 and C4 for the highest and lowest values you can use (I assume when you use 15, the LEDs shut off?).

Do the same at the point between U3 and R6.

Lowest value that will allow lights to come on is 2
R4 - C4: .01
U3 - R6: .00

Highest value: 4095
R4 - C4: .08
U3 - R6: .31

These are the reading with the multimeter set on 20V. If I've done my math correctly then:

Lowest value that will allow lights to come on is 2
R4 - C4: .2 V
U3 - R6: .00 V

Highest value: 4095
R4 - C4: 1.6 V
U3 - R6: 6.2 V

[terahz]

Ok, I'm not sure how you compute the second values but I'll take your word for it. The U3-R6 value being 6.2V is strange because the max for the DAC is 5.5V... Are you sure you're feeding 5.0V to the board? I can't see how you'd get more than that at U3-R6.

1.6V at the highest value also seems a bit high, it should be more like 1.3V but that could be explained by lower tolerance resistors. If you want to get a perfect range, you'd have to adjust the values of R6/R7 divider to get a perfect 1.24V at the max range. However I have a feeling you're feeding a bit more than 5V to the board, which would translate in higher output voltage as well.

Either way, it looks like the circuit is operating pretty much as it should. You're supposed to get between 1.3V-0.1V at R4C4. That is the dimming range of LM3409.

Can you measuring the current though the LEDs at the two ends of the dimming range? The moment you hit 1.3V on R4C4, you should be at the maximum current you've built your board for. At the lowest dimming range, you should have only a few mA going through the LEDs.

If that is the case and it still feels like the LEDs are too bright at a few mA current, it is likely that your eyes are playing tricks on you and in fact they are dimmed way down.

[shirley386]

I'm computing it from the 20V setting on the multimeter: 20 * .31 = 6.2. I realize now that that is not correct as I checked the voltage from the Hydra and I didn't have to do the calcs.

I feeding it from a Hydra with the USB BUB powering it. I checked the voltage @ 5.02 V going from the Hydra to the driver.

I'm wondering if there is a short somewhere that is allowing some current from the 48 V power supply to "get involved". The LM3409 gets too hot to hold your finger on. That's about the only place I see a possibility for a short. I wonder if I fed too much solder into the hole on the back of the board and the "puddle" that I think is under the chip is shorting to a pin(s)? That thing is a witch with a capital b to solder!

[terahz]

What is the config of your board? How much current, how many LEDs etc?

If the LM3409 is to hot to touch, something is definitely wrong. Have you seen my video showing how I solder it?

If in fact you don't have to multiply the measured voltage, than 0.01V-0.08V is no good. Are the LEDs really bright?

You can try to remove the lm3409 by placing your board flat on a pan and on the stove. Just keep the heat at medium and watch the board, you will see it melt and you should be able to lift it with tweezers. Then try to resolver it.

Let us know how it goes.

[shirley386]

Originally Posted by terahz

What is the config of your board? How much current, how many LEDs etc?

If the LM3409 is to hot to touch, something is definitely wrong. Have you seen my video showing how I solder it?

If in fact you don't have to multiply the measured voltage, than 0.01V-0.08V is no good. Are the LEDs really bright?

You can try to remove the lm3409 by placing your board flat on a pan and on the stove. Just keep the heat at medium and watch the board, you will see it melt and you should be able to lift it with tweezers. Then try to resolver it.

Let us know how it goes.

I'm using your BOM for the components. The string of LEDs consist of 6 XT-E CW & 6 XT-E RB.

I watched your video and you make it look easy. The problem I was having is holding it exactly in place to prevent the pins shifting to a neighboring pad. I finally clamped it with tweezers with one leg of the tweezers on top of the board holding the chip down and the other one on the bottom with an alligator clamp holding the tweezers down. Then I could gently nudge it until it looked like the legs were lined up correctly and started soldering and cleaning up with solder braid.

I'll try to cook one of the boards and see if I can get the chip off and redo it.

[terahz]

Ok, so you have built them for 1A and 48Vin. 12 XTs at 1A should require a bit less than 38V. You might want to add another 2 LEDs in there to reduce the load on the IC (since it has to regulate more than 10V down at 1A).

Again, measure the current that is going through the LED strings. Just break the LED circuit somewhere and with your multimeter in current mode, complete the circuit. If that is more than 1A (which might happen if something is shorted and the LEDs start drawing their max current) then you definitely want to redo the board. If it is at 1A, then at least you know the main part of the board is correct. Then you'll want to look at U2/U2 R6-R11 and C6 for shorts/bridges.

Also, I've had a board from china come shorted, so it is possible that the board is bad too.

Good luck.

[AHutch]

Hi folks. i could use some help with the lm3409 driver board. I built two boards based on Terahz's v0.7 design and used the components in the attached spreadsheet.

lm3409_component_worksheet.xls

The configuration consists of 9 XT-E LEDs in series, a single I2C driver board v0.7 and a 28v(27.65v actual) power supply.

Both boards seem to function fine at full on though the total LED current is lower than expected. I probably need a higher input voltage however i am happy with the lumens output at full on.

At the highest setting (4095), R4/C4 voltage is 1.242v and LED current is 300ma.
At the lowest setting (2), R4/C4 voltage is .130v and LED current is 100ma.

I see two problems:

First, based on previous comments it sounds like the LED current should be much lower than 100ma at the lowest setting.

Second, on the lowest setting the lm3409 gets very hot to the touch. At the highest setting, I feel no heat from any component on the board.

I tweaked my design last minute and the Roff and Cin values aren't ideal based on the configuration. Could this cause problems with heat when the lm3409 needs to switch at a higher frequency?

I used a 28v supply that was "free" and a 32v supply would probably allow me to reach a higher current. Could the lower voltage be the cause?

I don't see any visible shorts on the board and I find it strange that both boards are exhibiting the same symptoms.

Any help would be greatly appreciated... Thanks

[terahz]

I have no idea how the board operates if Vin is lower than the LED voltage.

I'd just remove one of the 9 LEDs and try again. Otherwise try with the 32V PSU.

Other than that, the r4c5 voltage ranges are pretty much spot on, so my bet is on the lower Vin. Given that this is a switching regulator, it is probably going crazy trying to drive the LEDs.

EDIT: the Datasheet suggests that the regulator stops switching in the case you have, and the output voltage starts dropping with the input voltage. That is probably why you see only 300mA at max.

[AHutch]

Thanks Terahz. I took out one of the LEDs. On the highest setting, the LED current increased to 980ma without excessive heat from the lm3409c. On the lowest setting, 108ma with heat.

So... let me ask what I should have asked in my first post. Should the lm3409 even get warm in a conservative configuration like this one?

[terahz]

I've only tested the board with a big load (48Vin, 14 LEDs at 1.5A), and the IC got just a little warm. The mosfet was doing just fine with the chunky heatsink. I never really let it on the dimmest setting for a very long time, nor did I think it would be hotter than on higher current.

I definitely wouldn't expect it to be 'too hot to touch'.

I would also think that the minimum current would be lower... I can't exactly remember how low I got when I measured things. What is the ripple current of the inductor that you have?

[BeanAnimal]

Originally Posted by der_wille_zur_macht

First though, here is my version of the LM3409 design:

http://code.google.com/p/hpled/sourc...trunk%2Flm3409

I went and built some drivers and a spreadsheet to calculate the required values (no idea this thread existed). Checking google for existing calculators for LM3409 values, I (oddly) found my way back to my own home forum...

In any case, it looks like I wasted 2 hours building a spreadsheet that has already been built....

I did however find a slight error in the spreadsheet that DWZ has posted.

The Forumula in cell B29 should be changed to read
=B7+B23/2

instead of
=B30+B23/2

The ILmax value in step 3 should be derived from the user chosen Iled value. The Iled value in step 4 is the Iled current that results from the Rsense choice.

have fun...

(and don't mind me, the LED hater... I was never here)

Cell B24 references

[BeanAnimal]

Originally Posted by terahz

If the LM3409 is to hot to touch, something is definitely wrong.

The datasheet indicates that PFET with a Qg > 30 can cause excesive heating and that moving the Cin from VIN-VCC to VCC-CSN can help...

Just a thought.