TPS92512: Slow pulses to the PWM pin

Hello James,

Can you measure Vout when PWM first goes high along with the current? If the compensation remains unchanged but the output voltage drops during the PWM off-time you can see a current overshoot until Vout gets back to where it is supposed to be because of C13. If this is due to C13 charging the LEDs will not see this current, can you also measure the LED current?

Also note that the minimum on-time of 140 ns is typical.

Best Regards,

Below:  I led measured across a 0.41 ohm resistor note there are 3 parallel paths so total current is 3 times this reading.

Above: yellow V out, Pink: pin 4 Pdim

Below: Yellow I out at off transition,  Pink:4 Pdim

Below: Yellow: I out, Pink pin 8 Vcomp

Below: Yellow: I out, Pink pin 8 Vcomp  Zoomed out

Hello James,

What is Q1 for? What is FB4?

It appears to be minimum oscilloscope picture with COMP. The other I would look closer at COMP during the running and off-time. You can try lowering the switching frequency to see if that changes the overshoot. You can also try a different COMP capacitor value to see if there is change in behavior.

Best Regards,

Q1: There are actually 3 transistors/LED paths that are enabled one at a time to select different colors. During my testing only one path is enabled so I am only showing that to simplify the discussion.

                Note: I shorted out the transistor today and saw no difference.

FB4: this is a ferrite bead max 0.15 ohms at DC.

                Note: I shorted out the Ferrite bead today and saw no difference in the spike.

Below : i put a 1uF cap on Ccomp you can see this stabilizes Vcomp but the current spike is still there.  I also tried going to a smaller cap but it did not change much.

Below: R46 IS CHANGED TO 100K too reduce thew switching frequency.  the resulting spike increases slightly.

One more image below: Pink is the output from pin 10 to the inductor.   Yellow: is the current spike.  the first 3 pulses are too large then it stabilizes.

Hello James,

There are two currents that are involved when turning on. One is the LED current the other is the current to charge the output capacitors. There will be overshoot if Vout is lower then when it was turned off because comp stays constant. This current is going to charge the output capacitor. Comp will also try to compensate for lack of current when it first turns on which can cause overshoot. I'm not sure if you checked but when you changed the comp capacitor did the overshoot change (amplitude and duration)? It should be less overshoot with a larger capacitor but for a longer duration. Also, watch Vin when this happens. If it is drooping when it is switching and you have a input capacitance you can also get overshoot because Comp will not be at the right level for a higher Vin when it initially starts. The Comp hold assumes that Vin and Vout stay the same from turn off to turn on. I would look at trying to increase C13 value so if there is overshoot it goes into the output capacitor. There are other possibilities as well.

Best Regards,

Below you can see light yellow with standard Ccomp 10nF.  Dark yellow is with Ccomp set to 1uF.  the spike goes up a bit but duration is about the same.

Changing C13 to 11uF there is only a 1.86A spike on the LEDs.  However the inrush spike over all in the system increases to 6 amps due to the increased current to charge C13.  I would need to increase my current sense resistor power, it's also a larger EMI spike.

Changing C13 to 2.2uF  brings spike on the LEDs to 3.15A. The inrush spike over all in the system increases to 4.26 amps due to the increased current to charge C13.

Below Current through 1/3rd of the leds across a 0.41 ohm resistor with C13 = 2.2uF

Hello James,

I'm not understanding why it's doing this. Can you look at Vin, Vout, the current spike and COMP (comp will need to be zoomed in to see if it's moving after the PWM goes low, right when the PWM goes high and during the recovery as seen above in the current. The Vin I'd like to see is right at the power stage near pin 2 of the IC with the probe return near the IC as well near pin 9.

Being the response doesn't look a lot different between the 10n and the 1u during the overshoot it seems as though another variable is causing this.

It would be good to see what the duty cycle is at the end of the PWM pulse before it turns back on also including Vin, Vout and current.

Best Regards,

Hello James,

Follow up question to add to my previous post.

I have another question, how long are the wires to the LEDs or far away is the LED connection?

Best Regards,

The LEDs are on the same 4 layer PCB as the converter. The furthest LED is 3.4” from the center of the converter IC.

Critical probe location view.

i used a 4 channel scope to get all the readings at once. the results are not as clear so if you want a zoom of something let me know.

On all grabs,

Channel 1: Vin

Channel 2: Vout

Channel 3: Vcomp

Channel 4: I out

Images are in the order: Pwm on until current V comp stabilizes.

              Zoom: PWM turning off

     Zoom pwm turning on.

Hello James,

This looks like initial turn-on which could be minimum on-time caused. I was more concerned about when you are PWM dimming and Vout is not low. If comp is low and the current peaks as it is doing, unless something else is going on, it is probably minimum on-time. If lowering Vin or the switching frequency helps it points to this as well. You can also look at the switch node when this is happening. It should not be regulating current as comp rises and it looks like it is (at a lower level than the set-point), what does the switch node look like during this period.

Can you look at this during the PWM off/on pulse?

I would try decreasing the switching frequency (possibly with a larger inductor value). It seems the frequency is too high for the duty cycle you are trying to operate at.

Best Regards,

                It could be too high of a frequency but it’s what the TI Simulation tool recommended, at 32v it’s right around the minimum on time but even at 28v where there is more  wiggle room the spike is still there.  If I use a 10uH inductor and set the frequency down to 1 Mhz the spike is right around the same height as the standard current wave which unless i am missing somthign else seems to solve the issue.  I will have to find a way to fit that bigger inductor.

Images below in order

1. Pink Current, yellow Switch node.  ( PWM low to high) Note the extended on time on the first few cycles.

2. Pink Current, yellow Switch node.  ( PWM high to low )

3. Pink Current, yellow Vcomp.  ( PWM low to high)

4. Pink Current, yellow Vcomp.  ( PWM high to low )

5. Current with a 10uH inductor 1mhz Operating frequency.

Hello,

Is this LED current? If it is the converter current it will have a current spike as it is putting current back into the output capacitor. If Vout changes during the off-time there will be a current spike because of the output capacitor.

Changing from 32V to 28V is fairly insignificant for this issue.

Also, the duty cycle starts out higher because Vout is lower than when the PWM first went low. This is how the part works to keep regulation constant during PWM, this current should not end up in the LED, is should just be charging the output capacitor.

Best Regards,