CCS: Sampling vac for designing a UCD3138 controlled PFC

Hi,

The first code:

iv.vin_sum = iv.vin_raw + iv.vin_sum - (iv.vin_sum >> 2);
iv.vin_filtered = iv.vin_sum >> 2;

can be re-written as follows:

iv.vin_sum = iv.vin_raw + 0.75*iv.vin_sum (shift right by two bits of vin_sum divides it by 4, and we are subtracting this from the original vin_sum, so only 75% of the original vin_sum will be added to vin_raw to give us the new vin_sum. This will result in a first order IIR filter of vin_sum, with k=0.25)

iv.vin_filtered = iv.vin_sum/4 to scale it correctly

This will remove any noise in the raw value, and iv.vin_filtered will not be the same as iv.vin_raw

Hope this makes sense.

Regards,

Swami

Hi,

Sorry, I missed your follow up question.

The implemented filter is a first order infinite impulse response.

Typically this is represented as: yn = k*xn + (1-k) * yn-1

The example code takes advantage of the fact that the raw value is scaled to 1/4 of the sum value.

So k=0.25 is very easy to accomplish with a right shift by 2 bits.

As an example if k=0.125 i.e. 1/8 then the code will look as follows:

iv.vin_sum = (iv.vin_raw>>1) + iv.vin_sum - (iv.vin_sum >> 3);
iv.vin_filtered = iv.vin_sum >> 2;

For other values of k that can not be obtained by bit shifting, you will have to make sure that the coefficient can be obtained with just integer math, so it may not be easy to get exact floating point coefficients, and it may take more bandwidth of the processor.

Regards,

Swami