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.

LM3409HV-Q1: LED Driver Minimum EN Switching Frequency

Part Number: LM3409HV-Q1
Other Parts Discussed in Thread: LM3409HV, LM3409

We have an LED Light Bar design using the LM3409HV driver powered by +48Vdc to run a string of 12 Cree XPL LEDs. The PWM switching frequency of the driver is set at 400KHz (Roff 24K3, Coff 470pf). We use the EN input to strobe the LEDs to synchronize with an external camera, sort of like a flash but running up to 1KHz with a 100us ON duration (10% Duty Cycle). Everything works as expected until we drop the EN strobe frequency to around 60-80Hz (still 100us ON). Below that frequency, there is a large delay before the LEDs turn on (70-80us compared to 2us normally). I can see switching noise on the LED power through the inductor that corresponds to the PWM switching frequency, but no increase in LED voltage until 70-80us later. I ran a test on an LM3409HV Evaluation Board we have and found that a similar issue happens with EN input around 80-90Hz.

Is this the expected operation?

What is the minimum EN switching frequency that can be used?

What changes do I need to make to ensure the LEDs turn on fast when EN is driven by a single pulse?

I've attached the driver portion of our schematic. Only two LEDs are shown, but it is a string of 12.

  • Hello,

    Possibly what is happening is it is going into low power shutdown.  You can look at VCC to see if it is falling below the UVLO threshold.

    "8.4.1 Low-Power Shutdown
    The LM3409/09HV can be placed into a low-power shutdown (typically 110 μA) by grounding the EN terminal
    (any voltage below 0.5 V) until VCC drops below the VCC UVLO threshold (typically 3.73 V). During normal
    operation this terminal should be tied to a voltage above 1.74 V and below absolute maximum input voltage

    Another possibility is the output voltage is falling low enough that the Coff pin cannot charge.  This can be seen by looking at the output voltage and/or the Coff pin.  If this is the case a separate Roff resistor can be added, see Figure 27 of the datasheet.

    For low PWM frequencies or longer off-times you can use the UVLO for PWM dimming.

    Best Regards,

  • Thanks for your suggestions Irwin.

    I was checking through the e2e support answers again and came across this thread: LM3409: LM3409 PWM Dimming Question in low frequency. It looks to be the same issue I’m asking about, I was a little confused by it at first because of the mention of 30Hz Fsw, vs the 200KHz fsw. I’m guessing the first one refers to the external PWM input to EN and the second is the chips internal PWM for current control? Anyway, I tried implementing the fix you suggested by adding a series diode/resistor (1N4148/56K) between VCC and COFF, and I can now drive the EN pin down to <1Hz without affecting the turn-on time of the regulator.

    In your response to my question you suggest a similar fix based on figure 27 of the datasheet. Using formula 16 in that note suggests I need (24K3 x 48Vdc)/(3A x 0.075Rds) = 5.1M resistor. I went with the 56K originally based on your response to the previous question. Should I be using something closer to 5.1M instead of 56K? What implications are there with either resistor value (56K vs 5.1M)? I’m inclined to stay with the 56K resistor since I tried it and it seems to work, I just want to make sure I’m not going to cause other issues with the design.

    You also mention below that UVLO can be used for PWM dimming for low frequency applications. I can see how that would work in the datasheet, too bad there’s not a note or example with that suggestion? Is there any advantage of using UVLO for external PWM control over the series diode/resistor from VCC to COFF? Does one method have a faster turn-on than the other?

  • Hello Doug,

    Yes, the 30 Hz would be the PWM frequency and the 200 KHz would be the switching frequency.

    The previous post is quite a bit different then what your setup is including their Roff resistor being much lower, Vin being much lower.  VCC is also referenced to Vin on this part since it has a high side P-FET driver.  Adding 56 Kohm from Vin to Coff with your 24.3 Kohm Roff resistor will cause the switching frequency to go up (off-time gets shorter), it will also increase your average LED current.  By using Vin the added Roff will always be in the circuit.  The diodes are there for when Shunt FET dimming is implemented, you are not doing this so they won't do anything (the diodes are not needed in your case).  What you have will work however you will want to reset your switching frequency back to what you want.  If your output voltage varies much you will lose some regulation accuracy since the off-timer is what determines peak to peak current ripple.  I would raise the added Roff resistor so it has less influence on the off-timer as long as the start delay isn't too long. 

    Don't use that formula for what you need.  I would look at this as a voltage divider.  If your output is low, Roff is the lower resistor of a divider from input to ground.  If Vin is 48V, the divider has to get above 1.24V, I would go minimum of 2V maybe higher depending on how slow it comes back on (from testing).  That would be 558 Kohm for the second Roff.  If the delay is too long I would lower it further since this resistor is what is sourcing Coff when Vout is low.  You may want to increase the 24.3 Kohm Roff resistor to get the switching frequency back to where it was (as well as average current).

    If you want the added resistor out of the circuit for regulation reasons I can give you a method to do that though it will add more parts.

    The two things I mentioned are two different issues.  One is when VCC goes into UVLO.  The other is if the output goes low enough it cannot charge Coff from Vout appropriately.  If the added Roff fixed your issue the PWM probably isn't causing your issues.  Between EN and UVLO PWM dimming the response in your application would look the same time wise.  The difference is the UVLO method doesn't go into low power shutdown when it is low so you can keep it off longer, or use lower PWM frequencies without going into low power shutdown.  Looking at VCC with respect to VIN and looking at Vout and Coff will tell you which is causing your issue.  Perhaps both may be causing issues.

    Best Regards,

  • Hi Irwin,

    I measured Vin-Vcc, during normal operation this is 6.0V. When I drop the input PWM frequency till the turn-on delay occurs, it changes to roughly 5.8V. According to the datasheet, UVLO won't kick in until Vin-Vcc < 3.73V, correct? So this isn't the issue?

    I also had a look at Vout and Coff but can't really see anything obvious. Both show changes when the PWM frequency is dropped but that was to be expected. I'm not too sure what I should be looking for?

    My internal switching frequency has changed from 430KHz without the 56K resistor to 670KHz with so I will need to tweak Roff1 somewhat. I also have Roff2 connected from Coff to Vcc as per the other post, not Vin. I haven't played around with the resistor value yet but will do that shortly.

  • Hello Doug,

    Yes, if VCC isn't hitting UVLO threshold that is not what is going on.  On your design you can place your second Roff resistor to Vin, VCC is only 6V below that.

    Best Regards,

  • There must be something else going on still. I tried resistor values from 560K down to 56K between Vin and Coff, and none of them corrected the startup delay. Connecting the 560K resistor from Vcc to Coff corrects it. My internal switching frequency drops from 670KHz to 650KHz with the 560K resistor so I’ll still need to tweak Roff1 to drop it back to approx. 400KHz.

    I also notice that it only takes a difference of 1 or 2 Hz to change between normal operation and the delayed turn-on, for instance changing from 62Hz to 60Hz. When its just on the edge like this, touching the scope probe to Vcc is just enough to cause the circuit to operate normally again. If I lower the PWM frequency below that however, touching the probe to Vcc makes no difference.

  • Hello Doug,

    I would go back and look at VCC with a scope again when you are running into this issue.  The added Roff resistor will cause current to be supplied to VCC and with 56 Kohm that is probably too much current being added.  By adding current the VCC will hold up longer.

    Best Regards,

  • I actually have a 560K resistor now from Vcc to Coff, not the 56K. Vcc is at approx 40V and Coff is at roughly 2V, so current is being drawn from Vcc in that case, not added? With the 560K resistor I was getting turn-on delays again at roughly 30Hz PWM input. I lowered the resistor to 220K and it started working at 1Hz again.

    Based on the scope probe loading I mentioned in my last post, I tried the 220K from Vcc to ground and found it corrects the turn-on delay as well. I have a 10uf cap on Vcc to Vin, the datasheet suggests at least a 1uf bypass on Vcc. Could 10uf be too large?

  • Hello Doug,

    VCC is negative with respect to Vin.  This is a P-fet driver so the gate is driven low with respect to Vin.  Because the added Roff resistor is pulling VCC down it forces Vin to charge VCC at the same current as the resistor is sinking to ground, plus what the IC needs to run from the internal linear regulator.

    Are you looking at VCC with an oscilloscope?  You can also look at the gate drive and see if the amplitude of the drive drops.

    I think 10 uF should be fine however you could try lowering it and see if it changes your behavior.  The other thing to try is PWMing the UVLO signal.  This can be done with a diode tied into the UVLO divider so you still have UVLO.  Anode to the center of the divider, pulling the cathode low to bring UVLO under the threshold of 1.243V.

    Best Regards,

  • Hello Doug,

    I haven’t heard back from you, I’m assuming you were able to resolve your issue.
    If not, just post a reply below (or create a new thread if the thread has locked due to time-out)

    Best Regards,

  • Hi Irwin, I've been busy testing the solutions. So far:

    1. 220K from Vcc to Coff. This was the first solution I tried. It works but increases the internal switching frequency (fsw) so I would have to change Roff to compensate.
    2. 220K from Vcc to Gnd. This also works and doesn't impact fsw. I have concerns about the implementation of this method. It doesn't seem to follow the presumed issue so I'm not 100% sure why it works or if it will have other negative impacts.
    3. Change Enable to drive UVLO instead. I removed R1, R2 and J2 from my schematic above and added a jumper from J2 (U4 side) to UVLO. This also works and doesn't impact fsw. I understand how this works and will use this method as a solution. 

    You can mark this as resolved. Thanks for your help.


  • Hello Doug,

    Using the UVLO is the better solution in this case.  The other, pull current from VCC at that level may have issues when going into low power mode so the route you are going is better.

    Best Regards,