DIY LED driver for reef lighting

[O2Surplus]

Kress

Thanks for the heads up on the Idea for the additional heat sink. I've got some solid copper heat sinks originally designed for use with vertical mount 78 series through hole regulators. If this new board that I've created runs too hot- I'll find a way to re-task them to help with the thermal issues.
Thanks again for your help.

 

[O2Surplus]

double post

 

[der_wille_zur_macht]

Of course it always depends on the dissipation. If that board has been run fully loaded for long enough to reach thermal equilibrium and the part was not too hot it's fine as is.

This version only exists in EAGLE to date, so I can't talk about real performance yet. But I expect it to be fine. With careful component selection, I have run terahz's board up to 3A without a problem, and (absolutely no offense to terahz) I would expect my board to have better thermal performance than his.

 

[O2Surplus]

Well Guys- I've got some good news and some bad news. Good news is- My driver works! Bad news- It works too well. I was aiming for 2000 ma @ 28 volts, but I got 2700ma instead. I'm going to have to play with some of the component values to bring the current down to the desired level. I'm hoping that I can make the adjustment with a different value sense resistor, but there doesn't seem to be very many different values available in the 2512 package. Anybody have any ideas?

 

[Krakonosh]

Well Guys- I've got some good news and some bad news. Good news is- My driver works! Bad news- It works too well. I was aiming for 2000 ma @ 28 volts, but I got 2700ma instead. I'm going to have to play with some of the component values to bring the current down to the desired level. I'm hoping that I can make the adjustment with a different value sense resistor, but there doesn't seem to be very many different values available in the 2512 package. Anybody have any ideas?

What size are you looking for ?
Mouser has about 1000 in 2512 (6332 metric) 1W current sensors.

 

[der_wille_zur_macht]

Well Guys- I've got some good news and some bad news. Good news is- My driver works! Bad news- It works too well. I was aiming for 2000 ma @ 28 volts, but I got 2700ma instead. I'm going to have to play with some of the component values to bring the current down to the desired level. I'm hoping that I can make the adjustment with a different value sense resistor, but there doesn't seem to be very many different values available in the 2512 package. Anybody have any ideas?

Can you post your component values and all your variable targets (Vin, Vout, frequency, etc.)?

 

[O2Surplus]

DWZM- Here's the target values that I was aiming for.

Nominal input voltage: VIN 28.V
Nominal output voltage (# of LEDs x forward voltage): VO 25.4V
Switching frequency (at nominal VIN, VO): fSW 200kHz
Average LED current: ILED 2,000mA
Expected efficiency: η 98.00%

Here's what's on the PCB

Chosen Roff 15,000Ω
Chosen L1 10.μH
Chosen Rsense .1Ω
Chosen CIN 10.μF
Chosen Cout 1.μF (forgot to include this one)
Estimated Rd 1.5Ω
Cin1 = 10uf Cin2 = 10uf
Ruv2 =49.9kohm
Ruv1 = 6.98Kohm

Any Ideas where I screwed up? I'm actually very happy the driver didn't blow up in my face! I'm looking forward to redesigning the board, if I can get a handle on what component values I'll need up front.

 

[der_wille_zur_macht]

What load were you driving when you did the test run and got 2.7A instead of 2A? Did you confirm that Vin and Vout were at your target values?

Can you post part numbers for the inductor and transistor? And your Coff value?

Besides Rsense the component values that most affect Iled are Roff/Coff and L1. The datasheet calls out that Coff should have short connections to the IC's timing and ground pins. You got the timing pin right but I can't see that the other end of the cap has a short, direct route to the chip's ground pin. Wonder if there's some interference.

You wouldn't happen to have a scope to look at the output, would you?

It looks like you have more input capacitance than you need but I can't see how that would cause this.

 

[O2Surplus]

DWZM-

Sorry I forgot to include the value for C-off. I used the default value ( 470 pF) and used the same mosfet that you'd posted earlier about, P/N FDD4685, with a 10 mH inductor P/N SLF12565T-100M4R8-H . I was driving a large BridgeLux BXRA-C4500 led array when I took the current readings. The led arrays nominal Vf is listed as 25.4 volts. I don't own a scope right now, but I may just have to borrow one to further investigate the issue if my component values don't point to the problem.

 

[Baghyyy]

As before, for analog dimming, all components are required. For PWM dimming R4, R6, R7, R8, R9, C6 and U2 should be omitted.

I can soldering all the components and then use the driver with "PWM" or with "ANALOG" changing only the jumpers?

 

[terahz]

kcress, good point about the cooling. I hadn't actually done a load test yet so didn't know how much the board would heat up. That's what I did last night. I loaded up 16 cree XPGs at 1A with 50V psu and after about 30 mins the hottest part on the board (the mosfet) was about 67C (152.6F). So to that, I've made the following change:

That, plus a little bit better mosfet should make things a little cooler.

absolutely no offense to terahz

None taken. What did you end up using for Q1? That seems to be the heater on the board . I'll be ordering some more stuff and was considering the IRFR6215TRLPBF.

BTW I've also added future electronics part numbers for most parts in the BOM. Prices there seem to be 30-40% lower! Unfortunately they don't have everything we need to make a full order from there.

I can soldering all the components and then use the driver with "PWM" or with "ANALOG" changing only the jumpers?

You can, but for PWM dimming you will still need to provide 5V to the leftmost pin (the one that is furthest from D1) on the settings header to set the maximum current. It might be easier to just remove C4 and R4. That will leave IADJ on the lm3409 open and will set the current at max.

 

[Baghyyy]

You can, but for PWM dimming you will still need to provide 5V to the leftmost pin (the one that is furthest from D1) on the settings header to set the maximum current. It might be easier to just remove C4 and R4. That will leave IADJ on the lm3409 open and will set the current at max.

but if we put another jumper here, moving a little down more jumpers and the resistors?

 

[kcress]

kcress, good point about the cooling. I hadn't actually done a load test yet so didn't know how much the board would heat up. That's what I did last night. I loaded up 16 cree XPGs at 1A with 50V psu and after about 30 mins the hottest part on the board (the mosfet) was about 67C (152.6F). So to that, I've made the following change:

That, plus a little bit better mosfet should make things a little cooler.

I'll be VERY interested in seeing what that change brings. Thanks.

 

[Baghyyy]

terahz
with this modification could be made to use with either PWM or with ANALOG?

 

[terahz]

I'll be VERY interested in seeing what that change brings. Thanks.

Will certainly report when I get them.

terahz
with this modification could be made to use with either PWM or with ANALOG?

Yes, this will certainly work.

 

[NanoNana]

so it looks like the new design is gonna run a bit cooler but are there any performance improvements? We are still going to be limited to 1A right? at least it will run cooler, Im going to be using some heat sinks from the mosfets on old PC power supplies to keep mine cool on the v0.4 versions unless there are significant performance upgrades with the v0.5

Let us know what you come up with and thanks again for all the hard work!

 

[terahz]

I wouldn't say it is limited at 1A. My mosfet is probably not very well suited for the higher currents. DWZM said he was able to run at much higher currents with a different mosfet without it getting too hot, so at this point it is about tuning the component selection, rather than making design changes. By the end of the week I should have a few different mosfets to test.

I'll probably send the pcb order tomorrow for v0.5, so my guess is it will take at least a couple of weeks before I can do a comparison.

BTW, I have a bunch of PCBs at v0.2(good for PWM dimming) and v0.4. If anyone wants some, let me know.

 

[Baghyyy]

I'll probably send the pcb order tomorrow for v0.5, so my guess is it will take at least a couple of weeks before I can do a comparison.

Why not try to make the order with my configuration of the PCB with jumper PWM and analog?
Here the file .brd:
LM3409.v05.brd

 

[der_wille_zur_macht]

I wouldn't say it is limited at 1A. My mosfet is probably not very well suited for the higher currents. DWZM said he was able to run at much higher currents with a different mosfet without it getting too hot, so at this point it is about tuning the component selection, rather than making design changes.

Yup, indeed. The MOSFET in your original BOM got hot enough at my currents to unsolder itself. Amazingly the circuit still worked but the part was literally floating in a pool of liquid solder. Definitely not good long term. The MOSTFET I have on there now runs totally fine and is cool enough that you can touch it without burning your finger. BUT it has a lower voltage limit - so there's definitely a tradeoff involved - as I said earlier, I really think it needs to be emphasized that unless you are operating at EXACTLY the same parameters as someone else, you should validate component choices for the power-handling components (L1, Q1, D1) to make sure they can handle your power levels.

 

[der_wille_zur_macht]

None taken. What did you end up using for Q1? That seems to be the heater on the board . I'll be ordering some more stuff and was considering the IRFR6215TRLPBF.

I used this:

https://www.mouser.com/Search/ProductDetail.aspx?R=FDD4685virtualkey51210000virtualkey512-FDD4685

I chose with this process:

1) Find parts with appropriate voltage limit
2) Narrow to parts with appropriate Qg (datasheet calls out instructions for this)
3) Choose a part based on max current, Rds, and max power dissipation

The power it has to dissipate is basically a function of Rds at any given power level - the difference between your original part (~.6ohms) and my part (~.03ohms) makes it pretty clear why mine doesn't get as hot.

The new part you listed, IRFR6215TRLPBF, has higher Qg than the datasheet wants (which might not matter if your frequency is low enough) but it has a fairly high Rds. It looks like it might not be quite optimal at higher currents but it might be OK at your lower target current.