LM3429 current limiting and UVLO hysteresis

I might know what is going on here. Is there any way you can take scope shots of the nDIM and COMP pins when you are having issues at the lower input voltages? That could help confirm or disprove a couple of things. Thanks.

Channel 1 is nDIM, Channel 2 is Comp. These were measured with the input voltage set at 6V (below expected UVLO threshold).

In taking the measurements I did notice that if I didn't have the probe connected to the nDIM pin before it went lower than expected, connecting it would cause it to trigger the UVLO

.

Hello,

I calculated a few things from your schematic.  UVLO turnon should be around 6.57 volts, turn off calculated would be around 4.4 volts.  R32, R33 and R41 are used to calculate this along with the 1.24 volt reference and 20 uA hysteresis current.

Ignoring any losses and time delays:

Switching frequency is around 500 KHz

Output current setpoint 0.667 amps

Duty cycle around 83% at turn on threshold meaning only 17% duty is going to the LEDs at this input voltage (0.667amps/0.17=3.91 average current during inductor off time).  Buck boost operates similar to a flyback converter, turn on to charge inductor, turn off to send energy to LEDs, this operates with continuous current in the inductor so the peak to peak charging is a small percentage.

Current ripple of the inductor at turn on threshold is about 1.09 amps (10 uH and 500 KHz).  Peak inductor current is about 4.46 amps (3.91+1.09/2), again this includes no losses or time delays so the peak will be higher in actual operation.  Since R30 will limit the peak current below this the LM3429 won't be regulating, it will be in current limit.

Page 34 of the datasheet shows a design with similar output power and limits the input voltage down to 10 volts.  The current sense resistor on that design is 0.041 ohms so your design should have a lower value since it is trying to run at 6.5 volts input.  The peak current setting should be much higher than 4.2 amps.  Make sure the components including the inductor, L2, can handle the peak current this needs (don't want L2 to saturate).

Something to try would be to change R34 to a higher value to lower the overall LED current to see if it operates correct (perhaps 1/2 or 1/3 the LED current setting).  If there is still strange operation I would return this to the basic circuit without any of the adjusting circuits to get it working first.

At 6.5 volts input the input current will be higher than 3.5 amps so the component choices and layout need to be able to handle this.  It also means there are high di/dts on the board so layout for noise prevention is important as well.

As for measuring signals the oscilloscope probe loop must be small otherwise it will pick up noise from the converter switching.  If the probe has a ground lead measure the ground connection with the probe tip (measure the scope ground connection with the probe), it should be flat, if there is noise the probe loop is picking it up.  I mention this because the voltage on the comp cap cannot change as fast as the oscilloscope picture sent.

Very helpful feedback, thank you. Wasn't aware of/had forgotten about many of those points.

I'm curious about a few things, both so I can modify my existing design, and also further my understanding of these parts;

• How did you calculate the hysteresis on the UVLO pin? I knew there was some built in but it was unclear to me how to calculate this with my own values.
• How would a reduction of switching frequency affect the duty cycle/performance overall?
• I think I have designed well enough for noise prevention, based on the datasheet guidelines. All the LM3429 grounds are tied to the DAP, high current loops are minimized, etc. I'd rather not have to redesign the whole PCB again if I could avoid it, is there anything else I could do to reduce noise if this was the issue?

I'll give the probe ground a measure first thing in the morning and test these suggestions. Thanks again.

Hello,

UVLO calculation. 

Turn on threshold determined by R32, R33 and 1.24 volt reference, 1.24/10K = 0.124 mA, 0.124 mA*43K = 5.332V.  5.332V+1.24V = 6.572V. 

Turn off threshold uses same R32 and R33 divider R41 and 20 uA current source from nDIM.  Using a current source makes hysteresis easily adjustable.  20 uA*12.4K = 0.248V.  New divider voltage is 1.24V - 0.248V = 0.992V.  Current in R33 is 0.992V/10K = 0.0992 mA.  Current in top resistor is 0.0992 mA - 20 uA = 0.0792 mA (the current sourced from nDIM goes through R33 but not R32).  Voltage drop across R32 is 0.0792 mA * 43K = 3.406V.  UVLO turnoff is 3.406V + 0.992V = 4.40V via the nDim pin, it will shut down before that due to the LM3421.

If it needs to turn on by 6.5 volts then turnon UVLO threshold should be 6.5 volts.  If it needs to operate down to 6.5 volts input the turnoff UVLO threshold should be 6.5 volts.  At these lower voltages the input current is fairly high.  If these thresholds are set too close together the wire and input voltage drop can cause the input and driver to oscillate (turn on, current causes input losses so UVLO turnoff theshold triggers so it turns off).  Also, the lower the input voltage the higher the peak current limit needs to be set.  The current is already pretty high at 6.5 volts input compared to what the LED current is.  Just power (Vin * Iin) in versus power out (Vout * Iout), and you have to add efficiency in there too.  Pin = Pout/efficiency.

Reduction of switching frequency will not change the duty cycle or necessary currents during MOSFET on and off time.  The peak current can be affected due to the ton and toff and inductor value (example, leaving the inductor the same and cutting the frequency in half will double the current ripple making the peak higher, Iave + 1/2Ipkpk, calculated from V = L*(di/dt).  This changes if the converter goes discontinuous.

Generally lowering the switching frequency of a switching power supply will make it easier to get higher efficiency at the cost of larger magnetic components.  EMI can be a bit easier to deal with also.

I changed the resistor values to 16.2k and 3.6k, which should give me a cut off of 6.5v and a turn on of 6.82V (assuming my calculations are right). I'm seeing this as much lower, however. Seems like it might be due to R41. I replaced R41 with a short and I notice the hystersis is closer to 6.8/7V, but I don't know why that might be.

How does R41 affect the hystersis?

Hello,

R41 does affect the hysteresis. You can see that in the equation from the previous post. The hysteresis is a 20 uA current source from the IC. This will cause a voltage drop across R41. Turn-on threshold there is no current source, turn-off threshold there is 20 uA current source. All three resistors are in the equations, R31 and R32 for turn-on, R31, R32, R41 and 20 uA for turn-off, see the previous post, it shows the math.

Hello,

If the LM3429 is regulating the peak current will not hit the peak current set-point but your last comment makes me think it is not in regulation but running on peak current limit until the voltage is high enough.  If the peak current is not high enough to allow regulation the compensation will rail positive.  At low input voltage check the Comp pin (pin 2) to see if it is high.

If the current limit looks like it's tripping early there are a few possibilities.  Look at the current sense pin with an oscilloscope.  The oscilloscope ground loop needs to be very short to do this otherwise the signal will not be clear enough to tell what's going on.  See if it is hitting the 0.245 volts (per the datasheet).  It can be as low as 0.215 volts but I wouldn't expect to see it that low.

What does the layout look like for the current sense resistor.  0.060 ohms is low so if there is additional trace resistance that will factor in to the current limit set-point.  The other thing to watch for is noise tripping the current sense early.  Also, if the leading edge blanking isn't sufficiently blanking out the turn-on of the MOSFET it could false trip.  This would be caused by using a large MOSFET and/or a slow turn-on time.  Using an oscilloscope will show what it is.

For calculating the UVLO you can look at the previous comment.  Make sure to add in the hysteresis current (20 uA that comes out of nDIM after it is enabled) when calculating turn-off threshold.  That is how the hysteresis is done on this part.