I am trying to power a Phlatlight csm360 with 6A.

I have used the design suggested in the National AN-2041 application note for the LM3434 driver chip and built a board that should deliver the 6A. For the LED, this equates to an Vled of about 15V.

At the moment the current driver is setup for 5A, using a 12m Ohm sense resistor. According to the application note, the LM3434 tries to keep the Vsense = 60mV, thus this should give the 5A.

We are using a 100W 16V power supply. The LM3434 states that it is capable of Vout = 0.9(Vin) = 14.4V. At 5A, the LED requires about 13.8V.

I have now done some tests on the board I build. When shorting the outputs, I measure exactly 60mV over the sense resistor. This shows that the design indeed regulates the current.

But, when connecting any resistive load <3 Ohm (which will keep Vout < 15V), or when connecting the LED, I expect to see the same, ie. that the Vout is changed to achieve V = I x R with I kept at 5A). Still, the Vout just goes up until it reaches the supply voltage (16V). I then measure 70mV to 100mV over the sense resistor, equating to 6A to 8A into the load/LED.

Clearly something is wrong, as the LM3434 should keep this Vsense at 60mV.

Does anyone have any suggestion as to why my board is not regulating to the set current? The design is a copy of the circuit in AN-2041 (page 2), but using the 12m Ohm sense resistor.

Any ideas?

Thank you guys!

 

...though I'm still unclear on the purpose of driving an 8 Amp (84W @6A) LED, I would like a photo of it in operation though.

The project is to replace a 400 Watt high-bay factory light with the Phlatlight csm360 LED. The heatsinking has been sorted out, and with a bench supply, even though this LED can handle 100deg C, it reaches about 70 deg C maximum.

A perfect current source has infinite voltage.

The current source hitting the supply line is called "clipping" in audio amplifiers.

You need a higher voltage power supply...

Yes, the output voltage of the driver does clip at the supply voltage, but that is then already at much higher than was needed, and the output current is then much higher than was set with the sense resistor.

If we up the voltage supply, the output current just goes up some more? According to the LM3434 application note, the Vout can equal 90% of Vin, which should seem to be enough?

 

Update: We have been discussing this design and issues with National Semiconductor, who makes the LM3434. They have come back and suggested that we have minimum 5V headroom between then supply voltage and the Vled output of the current driver.

We got a bigger power supply (24V/9A) and tried to power one of the LED's with it. Still the current just goes up until 8, 9A and the supply just clips.

We have also ordered and received the evaluation board (LM3434SQ-20AEV) that this design is mostly based on. This is discussed the National Semiconductor AN-2041 application note. According to this application note, the eval board can take in 30V and output up to 18A. We set it to 4A using the onboard pot and an ammeter as load. Then we connected the LED. Only 2A could be measured. We turned up the pot and at 3.6A the eval board (I think it is the pot again) went up in smoke.

The fact that we used the eval board now rules out all doubts about mine's layout/thermal issues.

Everything the LM3434 application engineers tells us says this device should work for 6A, 14.3V, but I am wondering if its not time to go back to the drawing board, select a new device.

Does anyone have any suggestion for a base for a current driver that can manage this? I'm even considering a linear source again using no switcher. (But then the heat sinks just goes up like crazy).

Any ideas would be great!

 

Are you using a non-inductive current sense resistor? A typical wirewound resistor has a certain amount of inductance, which will throw off your measurements.

Interconnecting wiring also has inductance.

 

Are you using a non-inductive current sense resistor? A typical wirewound resistor has a certain amount of inductance, which will throw off your measurements.

Interconnecting wiring also has inductance.

I am using a 10 mΩ "solid metal alloy" sense resistor. From datasheet, 0.5nH - 5nH inductance. http://www.vishay.com/docs/30100/wsl.pdf

This is the same resistor as National Semiconductor uses in their evaluation board for the LM3434.

I will look at the wires connecting the sense resistor to the chip, though. They are routed as a diff. pair, and are surrounded by ground plane, but I will report back.