Hello everybody,
I'm designing a LED driver for a lighting system that is about 1 meter away from the driver (due to mechanical constraints). This lighting system is made of 6 strings of 4 LED. Each string runs at a maximal current of 1A, and a maximal forward voltage of 13V (typical 12V). The purpose of it is to provide lighting for a vision system, that has an acquisition rate of 5kHz, but the lighting system has to be on 50% of the time only, at this frequency. A 0-5V logic signal comes as input of the board to enable / disable it. Each string current has to be adjustable using a dedicated potentiometer. Our system's main power supply is 24V, but supplying lower voltages is not a problem for us.
First attempt was to use the LM3409. Functionnaly it seemed fine, I used the IADJ pin for analog dimming (through a potentiometer) and the EN pin for strobing. The issue is that as soon as I have several strings working, I can (depending on the individual settings of the potentiometers) get an unstable lighting. From one line to the other of the same picture, I get up to 20% of difference in luminosity. This is not acceptable for sure. This was also possible to reproduce with the EN always to 1 (no strobing)?
My design was clean, according to my own experience. As it is close to the LM3409 eval board, I ordered 6 of them to try to reproduce it with a board designed by someone else (TI). The result was exactly the same. The outcome of this, is that each driver would emit electromagnetic perturbations which would disrupt the 5 others, most likely through the IADJ circuit.
I decided to abandon the LM3409 approach, and even the switched mode approach, and switch to a linear regulator. To avoid dissipating too much power, I will have to supply it with a lower voltage - maybe 14 or 15 volts - but as already said, this is not an issue.
Problem is: I couldn't find in TI's catalog a linear constant current source allowing 1A, providing analog and "digital" dimming. I tried using concurrent's LT3080 which can be used as a current sources, and playing with transistors I could get something working, but not at 5kHz. 1 kHz would be barely reachable, and this is a no go for me.
What I found in TI's catalog is TLC5926. If I parallel all of its outputs, I get enough current for my application, current can be controlled through R-EXT and it has /OE. Only issue is that it is OFF by default and would require an SPI like communication to turn it off. I would like to avoid using a microcontroller just for this. That's not really cost effective, makes the whole system unnecessarily complex and adds extra steps in production.
Application report SLVA346 provides a solution to this by using a NE555 for "simulating" a SPI command. Is it really usable in real life? What would be the risk of switching the TLC5926 to another mode than the normal one, knowing that we can never assume that the incoming enable (strobing) signal is in 0 or 5V state?
I know that's a broad topic, somehow complex. Thanks in advance for all the help and tips you can provide to me.
