How does this block diagram convert input current to per unit value?

This is from

Field-Oriented Control of PMSM Using Hall Sensor

openExample("mcb/FocHallExample")

From Current controller subsystem. I couldn't wrap around how this method converts the input current to per unit value when I don't see a division of a base value, like how we analyze transformer problems. Can someone explain to me how this successfully converts the current to per unit value? Or point me to a resource that I can study, because I can't find anywhere online that explains this.

Answer 1

[Walter Roberson]

Input values are multiplied by 2^6. Note that Qu<<6 has the same effect as Vu*2^6 for integer-type values, except that the undefinedness of overflow conditions are expressed slightly different ways.

After that, the sfix32_En17 takes the bits and applies an internal division by 2^17 .

175 * 2^6 --> 11200

11200 / 2^16 --> 0.08544921875

So an input of 175 generates an output of 0.08545

Answer 2

[MathWorks Motor Control Blockset Team]

This implementation assumes a 12-bit ADC, where the full-scale ADC count is 4095. The measured current is represented in per-unit (pu) form with respect to the neutral-to-peak value of the sinusoidal current. Therefore, the effective ADC count corresponding to 1 pu is 2048 (approximately half of the full-scale range).

To convert the offset-compensated ADC measurement to a per-unit value, the ADC count must be divided by 2048, which is equivalent to a right shift by 11 bits.

The implementation uses a fixed-point data type with 17 fractional bits. When a signal is cast using Data Type Conversion in Stored Integer mode, the underlying integer value is preserved and only the binary-point location is reinterpreted. In other words, the bit pattern remains unchanged while the decimal point is positioned according to the specified fixed-point scaling.

Since the per-unit conversion requires an effective division by 2¹¹, and the output fixed-point format requires 17 fractional bits, the value is first shifted left by 6 bits (17 − 11 = 6). As a result, when the signal is reinterpreted as a fixed-point number with 17 fractional bits, the output directly represents the per-unit current. This approach avoids an explicit division operation and achieves the scaling efficiently using only bit-shift operations.