starcalc
/
flm01
1
0
Fork 0
Code Issues Pull Requests Releases Wiki Activity

avr: Improve the accuracy of pwr measurements and prevent an underflow condition in a boundary case by rounding down the multiplication constant. Thanks Mario!

This commit is contained in:
Bart Van Der Meerssche 2010-09-08 17:53:57 +02:00
parent 8d1ccbc8fc
commit fdc2e35e0f
1 changed files with 12 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
mote/v1/avr/main.c
View File

@ -287,8 +287,18 @@ void send(uint8_t msg_type, const struct sensor *measurement, const struct state
pulse_count = aux->pulse_count_final; pulse_count = aux->pulse_count_final;
sei(); sei();
MacU16X16to32(value, (uint16_t)(labs(rest)/65536), 242); // Since the AVR has no dedicated floating-point hardware, we need
value /= 1024; // to resort to fixed-point calculations for converting nWh/s to W.
// 1W = 10^6/3.6 nWh/s
// value[watt] = 3.6/10^6 * rest[nWh/s]
// value[watt] = 3.6/10^6 * 65536 * (rest[nWh/s] / 65536)
// value[watt] = 3.6/10^6 * 65536 * 262144 / 262144 * (rest[nWh/s] / 65536)
// value[watt] = 61847.53 / 262144 * (rest[nWh/s] / 65536)
// We round the constant down to 61847 to prevent 'underflow' in the
// consecutive else statement.
// The error introduced in the fixed-point rounding equals 8.6*10^-6.
MacU16X16to32(value, (uint16_t)(labs(rest)/65536), 61847);
value /= 262144;
if (rest >= 0) if (rest >= 0)
value += pulse_count*3600; value += pulse_count*3600;