木桩

注册日期:2014-11-04
访问总量:813722次

menu网络日志正文menu

回答玉质先生


发表时间:+-

玉质先生:

抱歉,因为旅行,回复晚了。

这个题目我不熟悉,下面的回答也许是答非所问,但我尽力了。

 

Question 1:  Non-integer sampling ratios: The sampling period is rarely an integer divisor of the signal period. In practice, an ideal integer N of samples per cycle does not exist. 

Answer:   Assume we have N arbitrary sampling points, x1, x2 ……..xN.   There is a unique polynomial p(x) of degree N-1 which interpolates these N points. The integral is then approximated by 

image.png


Question 2: Zero-crossing misalignment: Sampled data points generally do not align precisely with the signal's zero-crossing points.

Answer: I do not understand your question well.  If you mean x1 and xN above do not agree with a and b, which are zero crossing points of the function, then you can use the zeros of p(x) to determine a and b and 

then use  

integral-p2.png  to  approximate the integral.


Question 3: Dynamic signal periods: The actual signal period T fluctuates over time and must be measured continuously in real time. 

Answer:  Again, I do not understand your question well.  If you are talking about what I indicated in question #2 above, then you can choose x1, x2 ……..xN  to completely cover a cycle, that is, the point before x1 is of opposite sign as x1 and the point after xN is of opposite sign as xN , then compute the integral as indicated above.  Two such cycles of opposite signs would then be a signal period.


Together, issues 1, 2, and 3 severely degrade integration accuracy.

4. The current solutions, such as FFT windowing, are highly CPU-intensive and impractical for real-time AC power monitoring.

Questions:

? Is there an existing time-domain solution available today?

Answer:  Are my answers above a time-domain solution ?

? If it not exist, would you be interested in reviewing and improving a proposed real time solution?

Answer:  I am not familiar with this topic enough.


浏览(925)
thumb_up(2)
评论(10)
  • 当前共有10条评论
  • 老周二 回复 木桩:

    过奖了,廉颇老矣,能饭否?你才是万维最聪明的女士,现在万维大不如以前了,短视频充斥网络,没有多少人愿意阅读文字了。

    屏蔽 举报回复
  • 玉质 回复 木桩:

    木桩老师, 您作为一名纯数学数值分析教授, 已尽力了。谢谢!

    屏蔽 举报回复
  • 木桩 回复 蒋大仁勇:

    蒋先生好!我也怀念我们曾经一起做数学的日子。你好聪明哦,那些很难的题目,你都能解答

    屏蔽 举报回复
  • 木桩 回复 汝谐毕:

    欢迎毕先生驾到,向先生请安 !

    屏蔽 举报回复
  • 木桩 回复 玉质:

    玉质先生,你的问题我琢磨了好半天,但我还是没看明白,所以我是在答非所问,让你失望了,抱歉。 也许我平时写些有关数学的文章,博友认为我是数学家了。 请不要误会,我不是教授,更不是数学家,我仅仅是一个数学发烧狂。

    屏蔽 举报回复
  • 蒋大仁勇:

    木桩妹,你和,很久没有露面了,很怀念我们和古国人弟一起做初等数学的时光。

    屏蔽 举报回复
  • 玉质 回复 玉质:

    Just asked AI - AI answered -

    Time-domain innovation is always welcome, especially if your method can solve these integration errors without requiring the mathematical overhead of high-order interpolation at the boundaries.

    屏蔽 举报回复
  • 汝谐毕:

    我踮起足尖说:您确实来自金星。

    屏蔽 举报回复
  • 玉质 回复 玉质:

    "N-point interpolate is too expensive to use in the Microcontroller (微机) " - It's too time consuming for a Microcontroller to implemet it.

    The original questions are related to how to modify Trapezoidal, Midpoint, and Simpson’s rules for the real world DSP applications. There is a gap between pure math numberical analysis and engineering numberical analysis math. I tried to fill the gap.

    Thanks

    屏蔽 举报回复
  • 玉质:

    木桩老师, 您作为一名纯数学数值分析教授, 已尽力了。谢谢!

    There is a gap between pure math nuberical analysis and engineering numberical analysis math.

    For an example, N-point interpolate is too expensive to use in the Microcontroller (微机) for the large N. 但您已尽力了。谢谢!

    屏蔽 举报回复