如果按照上面这些思路，还需要考虑如何不让积分器产生过多的噪声，（这是非常复杂的环节，比如多斜积分AD），还有积分器时间常数的关系（这个会影响所有信号的幅值），所以就应该只论证积分，不是积分器，只该考虑单独的纯数字计算积分环节。我从一个传感器的1NPLC的静态噪声采样数据做了误差最大的一阶梯形积分，发现积分后的噪声量显著减少，如图1（只截了1ms）。当然了，可能是1NPLC不够长，从统计学考虑时间也就是样本数会影响随机量，从一些文献同样可以得到结论，“The process of integration is a form of low-pass filter, so it is expected that high-frequency noise of a random nature will be attenuated. This has been observed in measured data. If there is coherent noise in the digitized data, then it is possible that this could be enhanced, particularly if the noise，is relatively low frequency. A significant enhancement of lowfrequency noise has not been observed in the data we have processed to date.”意思是理论上确实会增强低频噪声（回复者注：是因为积分会滤波），但是实际上没观测到。file:///C:/Users/35901/Desktop/%E5%9B%BE%E7%89%871.png
楼主
标题置顶
标题高亮
来自 2楼
