Cleaning up

source/Core/BSP/Pinecilv2/configuration.h

@@ -165,9 +165,9 @@
 #define NEEDS_VBUS_PROBE          0 // No vbus probe, its not connected in pcb
 #define CANT_DIRECT_READ_SETTINGS   // We cant memcpy settings due to flash cache
 #define TIP_CONTROL_PID             // We use PID rather than integrator
-#define TIP_PID_KP                3000
-#define TIP_PID_KI                5
-#define TIP_PID_KD                300
+#define TIP_PID_KP                35
+#define TIP_PID_KI                8
+#define TIP_PID_KD                200
 
 #endif /* Pinecilv2 */
 

source/Core/Threads/PIDThread.cpp

@@ -98,19 +98,23 @@ template <class T, T Kp, T Ki, T Kd, T integral_limit_scale> struct PID {
     const T target_delta = set_point - new_reading;
 
     // Proportional term
-    const T kp_result = (Kp * target_delta) / 100;
+    const T kp_result = Kp * target_delta;
 
-    // Integral term
-    integration_running_sum += ((target_delta * interval_ms) / 3000);
+    // Integral term as we use mixed sampling rates, we cant assume a constant sample interval
+    // Thus we multiply this out by the interval time to ~= dv/dt
+    // Then the shift by 1000 is ms -> Seconds
+
+    integration_running_sum += (target_delta * interval_ms * Ki) / 1000;
 
     // We constrain integration_running_sum to limit windup
+    // This is not overly required for most use cases but can prevent large overshoot in constrained implementations
     if (integration_running_sum > integral_limit_scale * max_output) {
       integration_running_sum = integral_limit_scale * max_output;
     } else if (integration_running_sum < -integral_limit_scale * max_output) {
       integration_running_sum = -integral_limit_scale * max_output;
     }
     // Calculate the integral term, we use a shift 100 to get precision in integral as we often need small amounts
-    T ki_result = (Ki * integration_running_sum) / 100;
+    T ki_result = integration_running_sum / 100;
 
     // Derivative term
     T derivative = (target_delta - previous_error_term);