ADC: use floats for ADC averaging

this costs almost nothing and improved accel/gyro calibration

libraries/AP_ADC/AP_ADC.h

@@ -38,7 +38,7 @@ class AP_ADC
 	   The function returns the amount of time (in microseconds)
 	   since the last call to Ch6().
 	*/
-	virtual uint32_t Ch6(const uint8_t *channel_numbers, uint16_t *result) = 0;
+	virtual uint32_t Ch6(const uint8_t *channel_numbers, float *result) = 0;
 
 	// check if Ch6() can return new data
 	virtual bool new_data_available(const uint8_t *channel_numbers) = 0;

libraries/AP_ADC/AP_ADC_ADS7844.cpp

@@ -204,7 +204,7 @@ bool AP_ADC_ADS7844::new_data_available(const uint8_t *channel_numbers)
 // equal. This will only be true if we always consistently access a
 // sensor by either Ch6() or Ch() and never mix them. If you mix them
 // then you will get very strange results
-uint32_t AP_ADC_ADS7844::Ch6(const uint8_t *channel_numbers, uint16_t *result)
+uint32_t AP_ADC_ADS7844::Ch6(const uint8_t *channel_numbers, float *result)
 {
 	uint16_t count[6];
 	uint32_t sum[6];
@@ -230,7 +230,7 @@ uint32_t AP_ADC_ADS7844::Ch6(const uint8_t *channel_numbers, uint16_t *result)
 	// division. That costs us 36 bytes of stack, but I think its
 	// worth it.
 	for (i = 0; i < 6; i++) {
-		result[i] = (sum[i] + (count[i]/2)) / count[i];
+		result[i] = (sum[i] + count[i]) / (float)count[i];
 	}
 
 

libraries/AP_ADC/AP_ADC_ADS7844.h

@@ -28,7 +28,7 @@ class AP_ADC_ADS7844 : public AP_ADC
 	float Ch(unsigned char ch_num);
 
 	// Read 6 sensors at once
-	uint32_t Ch6(const uint8_t *channel_numbers, uint16_t *result);
+	uint32_t Ch6(const uint8_t *channel_numbers, float *result);
 
     // check if Ch6 would block
 	bool new_data_available(const uint8_t *channel_numbers);

libraries/AP_ADC/AP_ADC_HIL.cpp

@@ -54,7 +54,7 @@ float AP_ADC_HIL::Ch(unsigned char ch_num)
 }
 
 // Read 6 channel values
-uint32_t AP_ADC_HIL::Ch6(const uint8_t *channel_numbers, uint16_t *result)
+uint32_t AP_ADC_HIL::Ch6(const uint8_t *channel_numbers, float *result)
 {
 	for (uint8_t i=0; i<6; i++) {
 		result[i] = Ch(channel_numbers[i]);

libraries/AP_ADC/AP_ADC_HIL.h

@@ -36,7 +36,7 @@ class AP_ADC_HIL : public AP_ADC
 
 	///
 	// Read 6 sensors at once
-	uint32_t Ch6(const uint8_t *channel_numbers, uint16_t *result);
+	uint32_t Ch6(const uint8_t *channel_numbers, float *result);
 
 	// see if Ch6 would block
 	bool new_data_available(const uint8_t *channel_numbers);