| ??? 11/05/04 14:55 Read: times |
#80458 - RE: How good is this scheme? Responding to: ???'s previous message |
Prahalad said:
You have mentioned nothing about the Vref dervation source. As Vref must not be greater than Vcc, it could be wise to gain Vref from Vcc. Why not adding a little LC filter, with ground terminal of this C connected to analog ground? This would heaviliy dampen digital noise on Vref and guarantee, that Vref isn't greater than Vcc. How much would doing this help and I see no GND loop as well since transformer windings are galvanically isolated. Yes, but only for DC currents! There IS a connection between both windings at higher frequencies. And this will result in a ground loop for digital currents! Interwinding capacitance between both windings can be several hundreds of pF, have a look yourself by the help of a C-meter. Assume an interwinding capacitance of 300pF. Then this gives an impedance of about 1 / (2 x pi x 10MHz x 300pF) = 53 Ohm at 10MHz. At 100MHz this impedance is only about 5 Ohm! You mean, If I am connecting the two gnd planes Agnd and Dgnd under the ADC then at the same meeting point I should connect all other Gnd in my circuit [ apart from Agnd and Dgnd ]. The chasis Gnd i.e. The Third Phystical Earth wire should be conncted at the meeting point. Yes. Do you mean laying inter plane tracks should be avoided at all and if it is really required to run a track from one plane to other plane then at the plane crossing place a resistor of some value say 10 Ohms must be put in the track. Yes, but 10 Ohm is too small. Something in the range of 1kOhm would be better. Did you mean a small resistor when you said impedence or some other thing? This can also be a soft ferrite. More concretely spoken: What traverses the barrier gap must be filtered, so that noise from either side is stopped at the barrier. But simply connecting capacitors at each side from traversing track to individual ground wouldn't do the job, as then both planes look short circuited for high frequencies at this point. So, it's necessary to introduce a series component, which limits currents running over involved ground planes. This impedance should be as high as possible, and this for frequencies up to 1GHz, of course. In many cases something in the kOhm range is sufficient. Keep in mind, that everything I said is the more valid the higher the precision of ADC is. So, what might be neglected for a 8bit converter becomes relevant when you switch to a 12bit converter, for instance. And I remember, that you told us to eventually switching to higher precision ADC, didn't you? Kai |
