| ??? 05/11/06 17:01 Modified: 05/11/06 17:07 Read: times |
#116000 - There's no golden rule! Responding to: ???'s previous message |
Richard said:
What priniciples, KAI, would you recommend? How should one set about to avoid contaminating the GND that one needs for reference with the currents that we need to send back to the supply? I have tried to give an answer here: http://www.8052.com/forum/read.phtml?id=115740 I recall the relevant lines again: Kai said:
Ok, I say, keep the awful heavy currents isolated from the chassis, by routing the according ground return currents directly to the minus pole of battery. This you can do with the grounds of generator, starter and eventual ventilation motors.
But then, connect the chassis to the minus pole of battery and use the chassis as reference plane for all the other electronic circuitry, assuming, that the associated ground return currents running through the chassis are only small. I think here of decentral board computers, navigation systems, sensors, distance warners, small lamps, door locks, etc. etc., means all that stuff, that has to do with communication and sensoring but not with controlling of huge currents. In this case I suggest to use shielded cables with the shield connected to the chassis at both ends and routing these cables in close distance to the chassis plane in order to minimize loop areas. If additional high current applications are to handle in this car, then I would locate the according power control units near the battery and route all the associated ground return currents (assumed to be huge again!) via separate cables, means isolated from the chassis, to the minus pole of battery. If there are sensors mounted near the high current applications, think of something located in the rear of the car, then these can still be referenced with the cables shields to the chassis plane. Here the ground plane performance can help to minimize the inductivity of these sensors cables, which would make them otherwise antennas eventually picking up short wavelength radiation. It's just an illusion to get a "clean" GND. Don't even expect that it's clean. The question is only, what can I do to get a "better" GND and how to deal with the ground noise? I know an EMC company doing nothing else the whole day than to find out, how the ground plane of a PCB must be splitted to give optimum performance in high current applications! There's no optimum solution which can be achieved by following a simple design rule! Putting everything on the same ground plane will result in least inductance but increased voltage drops if high currents are flowing along. Routing everything ground star point like, on the other hand, results in lowest voltage drops on ground, but at the same time in most nasty antennas, hugh inductivities, complex impedances and ground loops. The best idea is to start with having everything connected to the same reference plane and then deciding what must be routed separately. In the car application of Mahmood, he can draw all the circuitry, where ground lines must be routed separately, near the battery. Then he can minimize the inductances between the modules sitting there and which should not be connected to the ground plane. The rest circuitry can then profit from the low inductance of ground plane, which is almost noise-free then. Ok, think about what happens if Mahmood decides not to connect the chassis to the battery. Then he must connect every modul, from the lowest current sensor application to the hihgest current stuff star point like to the ground terminal of battery. But when he must connect the moduls to each other, then every additional connection will form a ground loop. Or, he might want not to connect ground from module A to module B and to open the shield at one side of the cables? Finally he ends with a set-up, where hugh complex impedances of the whole ground routing will make his filtering efforts to become unpredictable or even to crash, where, if once an optimum routing and shielding scheme having found, everything turns bad again if he only wants to add one single additional modul. He can be lucky with this methode, but how much time will he need to find the optimum wiring scheme and how much copper he will need to bury into the car... And as if this would not be enough, he has to deal with a modulated 400MHz application, which cries for the use of a reference plane and the use of shielded cables with the shield connected to GND at both sides!!! How shall he connect the modules without ending with complex impedances of the whole ground routing? How will he force the 400MHz application not to turn the whole car into a paradise for electromagnetic waves of all kind??? Kai |
