• State Resolved
  • Locked Locked
  • Replies 3 replies
  • Subscribers 62 subscribers
  • Views 147 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

LM3423: LED driver current stability

Alessandro Cominelli
Intellectual 375 points
Part Number: LM3423

Dear Mr. or Mrs,
I am using the component LM3423 to drive 6 LEDs in series with a pulsed current. The goal is to generate a rectangular current pulse of 1.75A with duration 1ms and a repetition frequency of 25Hz.

I followed the datasheet to configure the device in boost configuration and I also simulated the system with TINA simulator to check the result. You find attached the TINA model with all the values of the passive components
LED Driver - Boost Configuration - autosave 22-08-26 11_17.TSC
Here you can see a screen of the schematic:

In simulation everything works fine, but then on the real board I observe very high spikes superposed on the LED current. The average shape is rectangular, as expected, but spikes fully cover it, ranging from 0A up to 3.6A:


To get the image above I placed a resistor of 0.1 Ohm in series to the LEDs and I used a Lecroy oscilloscope with a differential probe with 200MHz bandwidth to get the voltage on the resistor.
The average current shape is well visible and its amplitude is 1.75A as expected. The spikes are very regular and spaced at a frequency of about 1.4MHz. Here in the following you can see a zoom:


The bandwidth of the spikes is very high. You see here the waveform limiting the oscilloscope bandwidth to 20MHz:


The big issue is that the spikes propagate through the power supply electronics and I also have problems on some communication channels on my board when LEDs are on.

I also observed another strange behavior of the device. Starting from the datasheet, I dimensioned the current-limiting resistor Rlim, equal to 120mOhm, so that the limiting current is about 2A, but in this way i discovered that the average current of the pulses is limited to 1A, while spikes actually reach 2A, so I changed the limiting resistor to 0.6A to increase the current limit to 4A. In this way I could get the results I described above.

Is there a way to eliminate or filter spikes? Is it a known issue?

Thank you in advance for your help.
Best regards,
Alessandro Cominelli

  • +1
    TI__Mastermind 18890 points

    Hello Alessandro,

    The spike you are seeing is really scope coupling and not real in the current to the LED.  Please use a current probe on the output and you will see the noise coupling is not there.  One way to check this scope coupling is to ground the scope ground on one ground on the board then the tip to another ground at 100mV per div as in you second scope capture if the glitch noise is still there then it's scope coupling.  

    Thanks Tuan

  • 0 in reply to Tuan Tran74
    Intellectual 375 points

    Dear Tuan,
    Thank you for your answer. I tried observing ground using a simple 10x probe and I see disturbances, as you suggested.
    The same holds with the active differential probe observing the voltage on the resistor in series with the LEDs.
    Does this mean that the LM3423 emits strong electromagnetic disturbances? Is there a way to limit it?
    My problem is that having LEDs on disturbs communication on a I2C bus on the board, due to strong interferences.
    Thank you in advance.
    Best regards,
    Alessandro

  • +1 in reply to Alessandro Cominelli
    TI__Mastermind 18890 points

    Hello Alessandro,

    The most likely issue is your layout that has large current loops when Q1 turns on and off.  Beside fixing the layout you can also try to slow down the switching edges of Q1 by having a gate resistor from the LM3423 GATE to the gate of of Q1.  Start with 10 Ohm and you can increase it to see if it helps with the radiated emission.  This will also decrease the efficiency so it's not the best solution if you slow down the edges too much.

    Thanks Tuan