0
\$\begingroup\$

I want to dim a 5 meters long LED strip reaching all across a room with a PWM signal from an Atmega328p. First the schematic and the specs:

Schematic

Specs:

fsw = 1000;
Vi = 24; % also same as VDS
VGS = 12; % driving gate-soruce voltage
Vp = 4.7; % plateau voltage
Vth = 2.6; % threshold voltage
Io_max = 5;
tr = 30*10^-9; % desired rise time
tf = 30*10^-9; % desired fall time
Qtotal = 17*10^-9; %total gate charge at operating point (VDS, VGS)
Crss = 21*10^-12% aka VGD
Ron = 0.0029; % On resistance at operating point
R_JA = 50; %Junction to Ambient resistance
Vf = 1.2; % Forward voltage drop of the diode on gate at Ig_turnOff
Coss = 326*10^-12 % Output capacitance
Ls = 20*10^-9; % Estiamted stray inductance

%Turn ON:
Ig_turnOn = Qtotal/tr % Required Gate soruce Current
RG_turnOn = (VGS-Vp)/Ig_turnOn

%Turn OFF:
Ig_turnOff = Qtotal/tf % Required Gate sink Current
RG_turnOff = (Vp-Vf)/Ig_turnOn

I've chosen PWM dimming over linear dimming because first, it is simpler and doesn't require an inductor or buck IC such as HV9910C. Secondly, dimming is linear and covers entire brightness range when PWM is used. Thirdly, changing the strip length won't require me to change current sense resistors.

But I got two concerns:

  1. There will be fast transients seen at the LED strip because it is directly connected to the drain. I don't want the LED strip to become an antenna and cause EMI all across the room. In linear dimming circuits such as the one below has a filter built in:

linear dimming circuit example

The PCB will be four layers (signal/PWR - GND - GND - signal/PWR), the MOSFET and driver are really close to the terminal. I think this is the best solution for EMI on the board but this still doesn't help with the long wires of the strip. Do I really need to lengthen the rise and fall times? Or maybe I shouldn't worry.

  1. What is the difference in terms of efficiency between this PWM dimming vs Linear dimming. The PWM method has fewer components but larger current is being switched and also the LED's are lit to their full brightness in ever cycle. I know that LEDs are very inefficient when used around the maximum current rating. For linear dimming, LEDs are always under low current where they are more efficient in terms of Percieved_Brightness/Power

IAUC80N04S6N036ATMA1

UCC27516

\$\endgroup\$
5
  • \$\begingroup\$ I've just tested PWM and linear (aka analog) dimming, I've applied the same average voltage (8.3V) to two identical 12V rated led modules from a regular buck converter and Arduino PWM controlled IRLZ44N. PWM version was brighter. I swapped the modules and the results were the same. That's interesting. This shows linear dimming has no single advantage over PWM dimming. \$\endgroup\$ Commented Apr 25 at 0:49
  • \$\begingroup\$ Since diodes are not linear devices, whichever waveform with the highest peak voltage will deliver the most power and so be brightest. For an obvious example, consider a 3V LED driven at a constant 2.5v vs. 5v PWM at 50% duty cycle. Both have the same average voltage but only one generates any light at all. This doesn't really say anything about analog vs. PWM dimming though, just how LEDs work. \$\endgroup\$ Commented Apr 25 at 1:19
  • \$\begingroup\$ Low frequency PWM can also create a banding/strobe effect on cameras. Try and see the strip through your mobile camera at different PWM duty levels. You might notice it. \$\endgroup\$ Commented Apr 25 at 3:35
  • \$\begingroup\$ To do analog dimming, you can't just adjust the voltage, you need to regulate the current. \$\endgroup\$
    – winny
    Commented Apr 25 at 6:53
  • \$\begingroup\$ It was naive of me to to assume LEDs as resistive load. I've redone the test. The power of each LED module is as follows: PWM: 8.16V * 0.0579A = 0.47W & Analog: 9.23V * 0.051A = 0.47W and by a visual inspection I can say they are equally bright. I wish I had a lux meter. Also I've experienced slight strobe effect on my phone camera but I was expecting that and it's not an issue for me, I can go for 2kHz just in case. \$\endgroup\$ Commented Apr 25 at 14:59

1 Answer 1

0
\$\begingroup\$

I think this is the best solution for EMI on the board but this still doesn't help with the long wires of the strip.

The bold part is wrong.

Running high \$di/dt\$ through long wires will cause EMI because it'll radiate (A 5-metre wire can work like a monopole antenna at around ~30 MHz). With PWM you can't get away from high \$di/dt\$ because the full current flows in every switching period regardless of its duty cycle unless you slow down the rising and falling but this will increase the losses of the switching element, the MOSFET in your case.

But with a switching regulator, you keep the high \$di/dt\$ in a smaller loop (L1 inductor, Co capacitor, D1 diode). The \$di/dt\$ in a switching regulator will be much less thanks to its ramping shape instead of pulsating shape. The output is DC (yes, it has ripple but is low in amplitude) so the long wires will be less of a problem compared to PWM. Supply input wires will carry switching currents but will be less of a problem if they are kept short enough.


What is the difference in terms of efficiency between this PWM dimming vs Linear dimming.

I don't see much of a difference. Of course, a DC-DC regulator will have higher losses (IC's consumption, inductor's DC resistance, catch diode's switching losses etc).

A switching regulator (LED driver) will bring flexibility over a wider input voltage range. Plus, your PWM switcher has no overcurrent detection but the switching regulator does.

\$\endgroup\$
5
  • \$\begingroup\$ I'll add 47uH inductor such that one end will see LED- node and other end will see anode&drain node. I thought of E = 1/2*L*I^2 and E=VQ and t=Q/I (L=10^-4, V = 24) so that even for very low currents such as 1A, LED- wire will take 1us to go from 0V to 24V and vice versa. Inductor will give back the stored energy instead of wasting it as heat like the MOSFET. The PWM frequency is very slow compared to these numbers so it will not impact dimming. I don't need overcurrent protection because the main supply has it. What do you think? How can I know what is the proper inductor size? \$\endgroup\$ Commented Apr 25 at 16:24
  • \$\begingroup\$ @CaveScientist I'll add 47uH inductor such that one end will see LED- node and other end will see anode&drain node. Keep the Schottky diode D1 in your circuit and add a filter capacitor across the LED load, you'll have an inverse buck which is no different than the HV9910C circuit. Inductor will give back the stored energy instead of wasting it as heat like the MOSFET. D1 is needed for this. The PWM frequency is very slow compared to these numbers so it will not impact dimming ... How can I know what is the proper inductor size? The lower the frequency, the larger inductor you need. \$\endgroup\$ Commented Apr 26 at 6:51
  • \$\begingroup\$ ... I can't make a calculation at the moment but 47 uH for 1 kHz seems really low to me. Consider \$V = L di/dt\$ instead of energy storage. \$di\$ is the inductor's current ripple which will determine the inductor size and loss. \$dt\$ is the on-time of the switch. \$V\$ is the input - output voltage difference. \$\endgroup\$ Commented Apr 26 at 6:55
  • \$\begingroup\$ Inductance shouldn't depend on freq because it's not for buck operation, it's just for slow transition. I need to determine a fixed duration. I've changed pwm freq to 3.9khz. But also, the duty cycle is divided into 255 steps. That means minimum duty cycle is 1/3900/255 = 1us. The inductor shouldn't smooth the PWM signal thus the transition duration must be max 100ns. This is the worst case (max current) it must be much less normally. For 3A, L = 24*100*10^-9/3 = 0.8uH. My formula gives 1.6 uH so they are close. Maybe the self inductance of wires is sufficient and I should ditch the inductor. \$\endgroup\$ Commented Apr 26 at 15:28
  • \$\begingroup\$ Even if I use 1kHz the inductor shall be max 10uH. I may put small inductor I'm still considering but I hope my calculations are correct. If I increase the inductance it will operate in CCM mode and I will have an open-loop buck converter. It has a different dimming characteristic but would still work. \$\endgroup\$ Commented Apr 26 at 15:32

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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