| ??? 11/15/05 15:32 Read: times |
#103750 - Periodic interference can make them fail Responding to: ???'s previous message |
Frieder said:
You might find a median filter (read f.e. 5 values, sort them, then take the middle one) to give good and robust results. Median filters and similar methodes can be very annyoing, if periodic interference is superimposed to signal! When I was at CERN we had big troubles in reading out our MX7-chip mounted at the LEP (large electron positron storage ring), because we used ready made data aquisition software, which sorted a portion of about 30 samples and removed the five most deviating ones afterwards. The result then was some strange noise, we couldn't find an explaination for. What has happened? Due to a hum loop, hum was introduced to signal, which could have been nearly totally averaged out, when, yes, when all the samples would have been taken into consideration with the averaging process. But, removing even only one of the samples will result in an imbalance, and the averaging can fail! The same with median filters: If periodic interference is superimposed to signal, then removing a number of samples and focusing on certain other ones which seem to be representative can result in big errors. Kai |
| Topic | Author | Date |
| Weight Scale Avarging method | 01/01/70 00:00 | |
| Filter providing two time constants | 01/01/70 00:00 | |
| median filter, simulate algo on PC | 01/01/70 00:00 | |
| Periodic interference can make them fail | 01/01/70 00:00 | |
| Acknowlegded. | 01/01/70 00:00 | |
| Agreed | 01/01/70 00:00 | |
| 50Hz sampling | 01/01/70 00:00 | |
| Anti-aliasing filter will help | 01/01/70 00:00 | |
| what this is all about | 01/01/70 00:00 | |
| Moving average window. | 01/01/70 00:00 | |
| Average on condition | 01/01/70 00:00 | |
| is that not SOP (standard operating proc | 01/01/70 00:00 | |
| Hallelujah! | 01/01/70 00:00 | |
| meaningless | 01/01/70 00:00 | |
| What are you using that one for in a wei | 01/01/70 00:00 | |
hello from chhatraliamd for weight scale | 01/01/70 00:00 |