2
\$\begingroup\$

I am designing a two-channel PWM LED driver. It will switch 24V LED strips at 12 amps. Each strip has two white LED chips. Each channel complements the duty-cycle of the other one (i.e. while one channel's duty-cycle is 20%, the other is 80%). Thus I am able to adjust the CCT (correlated color temperature).

I designed and coded the circuit. I don't have enough LED strips that provide 12 A, so I am using two rheostats for load.

The driver circuit schematic is shown below:

schematic

simulate this circuit – Schematic created using CircuitLab

I am monitoring the both Drain to Source PWM signals on an oscilloscope. When I change the duty-cycles of the channels, the one with the wider duty-cycle creates voltage spikes over the narrower one's signal and it changes voltage level of the other's signal on own switching time. (Oscilloscope screen photos are below.)

0% Yellow (CH1) - 100% Blue (CH2)

0% Yellow (CH1) - 100% Blue (CH2)


10% Yellow - %90 Blue

10% Yellow - %90 Blue


30% Yellow - 70% Blue

30% Yellow - 70% Blue


50% Yellow - 50% Blue

50% Yellow - 50% Blue


70% Yellow - 30% Blue

70% Yellow - 30% Blue


90% Yellow - 10% Blue

90% Yellow - 10% Blue


100% Yellow - 0% Blue

100% Yellow - 0% Blue


How can I eliminate these disruptive effects of both switching on each other? I tried to design a snubber according to NXP AN11160. But it didn't eliminate this effect.

My power stage schematic is: regulator stage

Two regulator used in power stage. LM317 is down 20-30VDC input to ~10VDC to drive MOSFET's gates. L78L33 for feeding the MCU and other ICs..

My pcb layout is: PCB_Layout

My probe setup is: Probes NEW setup: shortened ground lead

Power Supply wave form while switching : Blue PS signal


I am able to suppress voltage spikes by adding a resistor between the gate of power mosfets and drain of driver mosfets. My drain to source signals of M1/M2 is :

Signals of CH1 & CH2

My another problem is audible noise at heavy load. C1 and C2 in the circuit are 4u7 ceramic caps and while the PWM duty-cycles are different from 100%, the caps make audible noise. The switching frequency is 400Hz, that is below the audible borders.. I tried to use aluminum electrolytic caps instead of ceramic ones. However, electrolytic capacitors heated and exploded after a while.

How can I solve audible noise?

\$\endgroup\$
  • \$\begingroup\$ Show layout. Show oscilloscope probe setup. \$\endgroup\$ – winny Jan 29 at 17:22
  • \$\begingroup\$ @winny , I added the pcb layout and oscilloscope probes setup above. \$\endgroup\$ – OzkulA Jan 30 at 5:21
  • \$\begingroup\$ The small spike is caused by your ground leads on your probes being too long. Take a look here: electronics.stackexchange.com/questions/136123/…. But your main issue is load regulation. Please measure Vsup with one channel on your scope. \$\endgroup\$ – winny Jan 30 at 7:15
  • \$\begingroup\$ I shortened ground lead of probe but spikes are still seems. I also added photo of the power supply signal. The PS was adjusted 23.5VDC input at 10.9A load. The wave form of the PS voltage seems vary between 21 to 25VDC while the two mosfets are switching... \$\endgroup\$ – OzkulA Jan 30 at 9:00
  • \$\begingroup\$ Excellent probing! That drop is significant! You need better regulation/lower impedance. How much of that is resistive? \$\endgroup\$ – winny Jan 30 at 9:12
1
\$\begingroup\$
  1. Voltage seems to jump up at the time when both channels are off -- so your power supply has poor load regulation or you have too much wires resistance that connect power supply with your circuit.

  2. Voltage spikes are generally not dangerous here and caused by fast switching off of M1 and M2 transistors. To mitigate, you can add 100..1k resistor between M1/M2 gate and driver circuit (that is, drain of M4/M6).

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ @Ivd, voltage spikes are critical for me. Because, I use current sense IC (INA302) at the input for current sensing. When INA302 senses over 21A flowing across the sense resistor (0R001), the driver starts the interrupt sequence to kill the PWM. My rheostats adjusted 2Ohm and about 15A current flows the driver while input is ~30VDC. Although there are 5-6A to current limit level, the SC protection actives because of the voltage spikes. so, it is necessary to suppress the spikes as possible. I will try adding resistor between power mosfets' gate and drain of driver mosfets. \$\endgroup\$ – OzkulA Jan 30 at 6:30
  • 1
    \$\begingroup\$ I connected a 270R resistor between M1/M2 gate and M4/M6 drain and it suppressed the voltage spikes at the turn off period of mosfets.. Thank you for your advice. \$\endgroup\$ – OzkulA Jan 30 at 12:44
0
\$\begingroup\$

If I understand correctly, you are wondering about the voltage step (circled in red) and not really about the spike (red arrow).

enter image description here

Frequency is low (~400Hz) and this step is a significant fraction of a millisecond long, so we're not dealing with high speed stuff. So I'm thinking about resistance, not inductance. Or a misbehaving power supply.

First thing to check would be simple ohmic voltage drop over pcb copper. Notice the big red copper pour has quite a lot of tracks slicing into it, which makes it much thinner than it looks, and current has to run around a rather convoluted path...

enter image description here

You can investigate by probing the voltage with your scope between actual ground on the left side of the pcb, and the source terminal of the FETs. Got some suspicious voltage drop? Then you can solder copper wire like this, to see if it works better:

enter image description here

Also check voltage ripple on the power supply, but since you're using high current it matters where you probe. Wire resistance will add voltage drop. If you measure much lower voltage ripple at the output of the supply than on your board, then something has higher resistance than anticipated and you'll have to find it. If ripple is very high on the power supply output terminals, then... the power supply has too high output impedance.

My another problem is audible noise at heavy load. C1 and C2 in the circuit are 4u7 ceramic caps and while the PWM duty-cycles are different from 100%, the caps make audible noise.

Why did you put these caps on the output? They won't perform any smoothing at all and they short the AC part of your PWM signal, which will draw high ripple current (this is why your electrolytics fried). It would be best to remove them...

Also the phase of your signals is wrong:

Each channel complements the duty-cycle of the other one (i.e. while one channel's duty-cycle is 20%, the other is 80%)

This isn't what is happening on the scope screen... It looks like both LED colors are on at the same time.

If you want both outputs to be complimentary, you should change the phase of the signals (or invert one output) to make sure only one is active at any given time....

With the signals as they are currently, and both LED colors ON at the same time, this will draw double the current, ie 24 Amps. If you used a 15A power supply thinking the instantaneous current would never exceed 12A because both LEDs aren't ON at the same time, then this may be the cause of your high ripple voltage...

| improve this answer | |
\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.