Loading...
Please wait, while we are loading the content...
Similar Documents
Extracting a more accurate position from a quadrature signal using the SRI
| Content Provider | Semantic Scholar |
|---|---|
| Author | Bruijnen, Dennis |
| Copyright Year | 2002 |
| Abstract | T i e Siii has been made to get more information oiit of a quadrature signal produced by an encoder. If you only use the quadrature signal for controlling an actuator the accuracy is limited. It will contain the so-called quantization noise. (It looks a little bit like noise with an maximum amplitude of half an encoder step.) Most controllers require knowledge of the position and the velocity to obtain a good tracking. When using an encoder the problem is to get an accurate velocity signal not to mention an acceleration signal. When differentiating the noisy quadrature signal the noise will be emphasized. Wnen enlaj-ging the gain, the closedloop system will become faster unstable. Using a low-pass filter to get a better position and speed signal is not always a good solution because of the produced time delay. Of course you could change the setup like buying another more accurate encoder but if you could use the same setup and upgrade the encoder with some cheap add-on it would be better. The SRI is a prototype of such add-on. It collects the times when a quadrature event takes place. At such event the position is very accurately known namely halfway the encoder step. Every sampletime the SRI will fit a line through the last couple of collected time stamps and extrapolate it over the calculation time of the fit. With this you have an accurate position determination, much better than when using the quadrature signal itself. You can also derive a better velocity signal and even an acceleration signal. A sample frequency of 100 Hz has already been achieved with a first order polynomial fit through four points with a permitted calculation time of 0,001s. When putting a small fluctuation on a constant input for the motor the position and velocity go nicely through the quadrature signal. A sample frequency of 1000 Hz is easily reachable but that hasn't been done yet. The only thing to do is calculati~g fzster than one szmpletime and that is possible because the algorithm takes about 2 10-~s of time. And another fact is: the faster you calculate the fit the more accurate will be the prediction of the position because less extrapolation is needed. |
| File Format | PDF HTM / HTML |
| Alternate Webpage(s) | https://pure.tue.nl/ws/portalfiles/portal/4283821/640660.pdf |
| Alternate Webpage(s) | https://pure.tue.nl/ws/files/4283821/640660.pdf |
| Language | English |
| Access Restriction | Open |
| Content Type | Text |
| Resource Type | Article |
