/diy/ - Do It Yourself

>>1269940
>>1270165
One more advantage of this circuit is that R doesn't need to be very large. Using the internal 1.1V reference, each count represents ~4mV, so 50 counts represents ~200mV. To get a reading of "full scale" 50 at 20mA from the representative branch, R can be 10 ohms.
>sensor module
What kind of sensor module? Maybe you can use the tiny to do the sensing for you.

>>1270186
I'd choose a timebase of 320kHz to start, just to keep the inductors reasonably smol. This can be gotten with a 64MHz PCLK setting, a 1:1 prescaler, and a period of 200. Just set the OC1B pin up for PWM and load OCR1B with the desired output level. A simple proportional control loop with some damping and a saturation detector (undervoltage cutoff) should suffice.
>How would you calculate L, Ipk, ton and toff?
The usual equations, cribbed here from a TI app note, assuming worst-case operation at 3.0V Vin with an pessimistic efficiency of 70% for the two strings of three-LEDs each with Vf=3.4V at 20mA:
duty cycle = 77%
L ripple current = 57mA
I(Lpk) = 40mA + (57mA / 2) = 70mA
L = 120uH
Isw(max) = 230mA
which is readily available in 5mmx4mm SMD packages rated at 0.4A.
BUT
when I run the same numbers for a 120mA load, the required inductor is only 39uH, also readily available in 5mmx4mm SMD packages rated at 0.8A. Weird. Can any power anons explain where I must have surely fucked up?