热度 1| ||
The basic strategy with non-sinusoidal waveforms is to start with an estimate of how many harmonics might be required, and set the oversample factor to 8. Run the simulation. Reduce the number of harmonics by about 50% and run the simulation again. If the answer does not change, reduce the number of harmonics again. Do the same with oversample. You are looking for the lowest runtime that also produces the correct answer, which will usually occur with a relatively large number for oversample factor, and a relatively small number of hramonics.
Every circuit is a little bit different in terms of the number of harmonics and oversample factor, but the principles are the same.
仿真器经常能看到Oversample Factor的选项,比如HB仿真,这个选项应该如何设呢?大的原则Cadence在Oversample factor的帮助文档里已经给出了,就是上面这段话。
过采样是在时域转为频率时做DFT所需要的,它跟谐波的数量有关。一个时域的波形,要想精确的表示,谐波的数量自然越多越好。谐波数量越多,对采样点就提出了更高的要求。但采样点多,毫无疑问会降低仿真速度。
非线性越多,谐波就多,过采样率就应该更高,结果才会更精确。
疑问:看帮助文档,说2个谐波,4个采样点可以表征,但要想精确表示oversample factor应该为8,32个采样点。那每个波形采样的数量点可以提前确认好吗,不然为什么会有过采样的概念。过采样肯定就是超过了规定的数量? Interesting.
确实,有点像香龙某个定理。