Timeline for Determining the mean and standard deviation in real time
Current License: CC BY-SA 3.0
13 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Jan 21, 2016 at 13:53 | comment | added | graup | I am having trouble understanding the calculation of a. If you do it like this, a will always be negative, which doesn't seem to make a lot of sense. Also, what exactly is T and tau? | |
| Mar 1, 2012 at 17:54 | comment | added | Hilmar | My bad. You are off course correct. I didn't transcribe the filter coefficients correctly | |
| Mar 1, 2012 at 13:27 | comment | added | nibot |
It does have 0 dB gain at DC. Substitute z=1 (DC) into H(z) = a/(1-(1-a)/z) and you get 1.
|
|
| Mar 1, 2012 at 13:22 | comment | added | Hilmar | I think there may be a mistake in here. The single-pole filters don't have 0 dB gain at DC and since you are applying one filter in the linear domain and one in the squared domain the gain error is different for E<x> and E<x^2>. I'll elaborate more in my answer | |
| Feb 29, 2012 at 20:51 | comment | added | nibot |
Woops! Of course I meant 0 < a < 1. If your system has sampling tmie T and you'd like an averaging time constant tau, then choose a = 1 - exp (2*pi*T/tau).
|
|
| Feb 29, 2012 at 20:50 | history | edited | nibot | CC BY-SA 3.0 |
added 70 characters in body
|
| Feb 29, 2012 at 20:40 | history | edited | nibot | CC BY-SA 3.0 |
added 373 characters in body
|
| Feb 29, 2012 at 20:26 | history | edited | nibot | CC BY-SA 3.0 |
Add some analysis.
|
| Feb 29, 2012 at 17:32 | history | edited | nibot | CC BY-SA 3.0 |
Added link to github gist
|
| Feb 29, 2012 at 17:27 | history | edited | nibot | CC BY-SA 3.0 |
Corrected the range requirement for $a$.
|
| Jan 20, 2012 at 4:44 | comment | added | schnarf | This is similar to Jason R's approach. This method will be less accurate but a little faster and lower on memory. This approach ends up using an exponential window. | |
| Jan 20, 2012 at 2:02 | comment | added | Dilip Sarwate |
Could you explain how $a$ determines the length of the running average? And what value of $a$ should be used? The specification 0 > a > 1 is impossible to meet.
|
|
| Dec 6, 2011 at 9:41 | history | answered | nibot | CC BY-SA 3.0 |