steppers are rated 20A max, consequently, switching 10-20A into an inductive load creates quite a bit of noise. > > The problem I am seeing is not uncommon for chopped steppers, ie that the chopper settles on a subharmonic of the switching frequency (20kHz) because of crosstalk. > > I have tried moving stuff around on the prototype board and looking at various grounding strategies. I have separated AGND and DGND in the schematic and then joined them, but that makes KICAD treat them together as a net. What should I do? Put in a 0 ohm resistor where I join them to keep filled zones separate?
In the absence of somebody that really knows what they are talking about, I will try and help. I recently had problems like this with a switching power supply. I built a 1 inch diameter, circular wire loop across the two leads of my oscilloscope. Then I hovered the loop over the board and used it as a noise detector. I found distinct sources of noise and these would sharply fall off if you moved the loop away from them. Then the game becomes to put distance between those sources and anything that cannot tolerate noise. You can use ground islands which have isthmuses to connect them together. But I would never forget that it usually takes a current loop to generate noise, and the size of that loop determines the amplitude of the noise. The problem with chopping up your ground plane can be that you extend the size of the loop. You always want to get the "return path" (i.e. ground, or ground plane), directly under the "source path" so that the loop is only the thickness of your board layer, and the orientation of that loop is generating noise parallel to the board, not perpendicular to it. Think of the loop as being partially on one layer, and partially on a layer DIRECTLY underneath it. Hope this helps. A few more prototypes may lead you to answer. Dick
