TMS320F2812 DSP编程之AD采样精度的校准算法

只看该作者 · 2016-7-16 20:27

F2812内部集成了ADC转换模块。该模块是一个12位、具有流水线结构的模数转换器，内置双采样保持器（S/H），可多路选择16通道输入，快速转换时间运行在25 MHz、ADC时钟或12.5 Msps，16个转换结果寄存器可工作于连续自动排序模式或启动/停止模式。在实际使用中，ADC的转换结果误差较大，如果直接将此转换结果用于控制回路，必然会降低控制精度。（最大转换误差可以达到9%左右）
   F2812的ADC转换精度较差的主要原因是存在增益误差和失调误差，要提高转换精度就必须对两种误差进行补偿。

   对于ADC模块采取了如下方法对其进行校正：
   选用ADC的任意两个通道（如A3，A4）作为参考输入通道，并分别提供给它们已知的直流参考电压作为输入（RefHigh和RefLow），通过读取相应的结果寄存器获取转换值，利用两组输入输出值求得ADC模块的校正增益和校正失调，然后利用这两个值对其他通道的转换数据进行补偿，从而提高了ADC模块转换的准确度。
   实现校准的硬件电路在本文中不作描述，在有关资料中可以查到。下面是该算法的C语言实现：

//首先计算两个通道的参考电压转换后的理想结果
//    A4 = RefHigh = 2.5V  ( 2.5*4095/3.0 = 3413 ideal count)
//    A3 = RefLow  = 0.5V  ( 0.5*4095/3.0 = 683 ideal count)
#define    REF_HIGH_IDEAL_COUNT 3413
#define    REF_LOW_IDEAL_COUNT 683
#define  SAMPLES    63
//定义所需的各个变量
Uint16  Avg_RefHighActualCount;
Uint16  Avg_RefLowActualCount; /
Uint16  CalGain;                                              // Calibration Gain
Uint16  CalOffset;                                           // Calibration Offset
Uint16  SampleCount;
Uint16 RefHighActualCount;
Uint16 RefLowActualCount;
//对各个变量进行初始化
void InitCalib()
{
Avg_RefLowActualCount = 0;
Avg_RefLowActualCount  = 0;
Avg_RefHighActualCount = 0;
RefHighActualCount = 0;
RefLowActualCount = 0;
CalGain = 0;
CalOffset = 0;
SampleCount = 0;
}
//获得校准增益和校准失调
// Algorithm: Calibration formula used is:
//
//  ch(n) = ADCRESULTn*CalGain - CalOffset
// n = 0 to 15 channels
//  CalGain = (RefHighIdealCount - RefLowIdealCount)
//                      -----------------------------------------
//                   (Avg_RefHighActualCount  - Avg_RefLowActualCount)
//
//  CalOffset = Avg_RefLowActualCount*CalGain - RefLowIdealCount
//
//  A running weighted average is calculated for the reference inputs:
//
//  Avg_RefHighActualCount = (Avg_RefHighActualCount*SAMPLES
//                         + RefHighActualCount) / (SAMPLES+1)
//
//  Avg_RefLowActualCount  = (Avg_RefLowActualCount*SAMPLES
//                                        + RefLowActualCount) / (SAMPLES+1)
//
void GetCalibParam()
{
RefHighActualCount = AdcRegs.ADCRESULT4 >>4;
RefLowActualCount = AdcRegs.ADCRESULT3 >>4;
if(SampleCount > SAMPLES)
  SampleCount = SAMPLES;
Avg_RefHighActualCount = (Avg_RefHighActualCount * SampleCount
      + RefHighActualCount) / (SampleCount+1);
Avg_RefLowActualCount  = (Avg_RefLowActualCount * SampleCount
                           + RefLowActualCount) / (SampleCount+1);
CalGain = (REF_HIGH_IDEAL_COUNT - REF_LOW_IDEAL_COUNT)
                     / (Avg_RefHighActualCount  - Avg_RefLowActualCount);
CalOffset = Avg_RefLowActualCount*CalGain - RefLowIdealCount;
SampleCount++;
}
//在ADC_ISR中，对其他各个通道的结果进行修正：
interrupt void  adc_isr(void)
{
GetCalibParam();
......
newResult n= AdcRegs.ADCRESULTn*CalGain - CalOffset;
......
}
复制代码

   通过上面的代码，配合硬件电路改动，可以大幅实现提高ADC采样的精度，实现更灵敏、更精确的控制