An ADC is a device which handles both the digital ouput data and the analog input data. though both the analog and digital have the same ground they are shorted or become one, near the ADC and not on any other part of the board. Ive heard designers saying this is a must, but i need a much satisfying reason for this.

 

Have you ever heard of Vcc sag and ground bounce? Ground bounce is the primary reason why the two grounds in ADC need to be separated.

Let us consider a ground track shared by both analogue and digital signals ground return currents. We know that any track has a finite resistance value, and from Ohm's law current flowing on the track would cause a differential voltage across the track ends. We know that most of the analogue signal measurement current loops are high resistance loops (high input resistance), what this mean is, the amount of current flowing due to the analogue signal is very very low and consequently the voltage drop due to the track resistance is usually negligible.

Now, enter the first factor, the current flow of the digital signal (ADC power supply also usually added to this) is quite significant. This current would cause a more significant differential voltage across the track ends (the ground voltage bounce). This differential voltage, because the ground track is also shared by the analogue loop, would be imposed to the analogue signal (added). Now the total measured analogue signal is the real analogue signal + voltage difference due to the shared track and digital/power current. This would lead to inaccurate measured value.

You could counter this argument by saying that if the shared track is thick enough, and not too long, and the digital + power currents are not that big then the effect would be negligible. And to some extent you are right, except that the digital part of the ADC has an associated frequency of operation (conversion, data transfer rate, etc.). What this means is that the current draw is not constant, and varies at a frequency.

The second factor is associated with this frequency. We know that any track, as well as having a finite resistance, also has a finite small inductance associated with it. If the frequency is high enough, the impedance due to the inductance and the current flow at this frequency would result in a significant differential voltage across the track ends. Similar to the above, this time the ground bounce is caused by AC characteristic of the circuit instead of DC and would also lead to inaccurate measurements.

From the explanation above, then we can see that sharing a ground track (or ground loop) is undesirable. Especially at high frequency of operation, ground separation is extremely crucial and also track length. This principle (super-imposed voltage) can be extended to almost any circuit where ground track or Vcc track are shared by different part of circuits. Now you know why we need to put a lot of those small ceramic caps across the supply rails and as close as possible to the digital ICs power pins, especially when there are a lot of ICs sharing the power rail.

 

Originally posted by n9352527@Jun 2 2006, 03:00 PM
Have you ever heard of Vcc sag and ground bounce? Ground bounce is the primary reason why the two grounds in ADC need to be separated.

Let us consider a ground track shared by both analogue and digital signals ground return currents. We know that any track has a finite resistance value, and from Ohm's law current flowing on the track would cause a differential voltage across the track ends. We know that most of the analogue signal measurement current loops are high resistance loops (high input resistance), what this mean is, the amount of current flowing due to the analogue signal is very very low and consequently the voltage drop due to the track resistance is usually negligible.

Now, enter the first factor, the current flow of the digital signal (ADC power supply also usually added to this) is quite significant. This current would cause a more significant differential voltage across the track ends (the ground voltage bounce). This differential voltage, because the ground track is also shared by the analogue loop, would be imposed to the analogue signal (added). Now the total measured analogue signal is the real analogue signal + voltage difference due to the shared track and digital/power current. This would lead to inaccurate measured value.

You could counter this argument by saying that if the shared track is thick enough, and not too long, and the digital + power currents are not that big then the effect would be negligible. And to some extent you are right, except that the digital part of the ADC has an associated frequency of operation (conversion, data transfer rate, etc.). What this means is that the current draw is not constant, and varies at a frequency.

The second factor is associated with this frequency. We know that any track, as well as having a finite resistance, also has a finite small inductance associated with it. If the frequency is high enough, the impedance due to the inductance and the current flow at this frequency would result in a significant differential voltage across the track ends. Similar to the above, this time the ground bounce is caused by AC characteristic of the circuit instead of DC and would also lead to inaccurate measurements.

From the explanation above, then we can see that sharing a ground track (or ground loop) is undesirable. Especially at high frequency of operation, ground separation is extremely crucial and also track length. This principle (super-imposed voltage) can be extended to almost any circuit where ground track or Vcc track are shared by different part of circuits. Now you know why we need to put a lot of those small ceramic caps across the supply rails and as close as possible to the digital ICs power pins, especially when there are a lot of ICs sharing the power rail.

[post=17460]Quoted post[/post]​

Thanks for giving a detail explanation about it.

I am using a optical isolator IL300D an analog optical isolator from Vishay electronics. the signal after isolation goes to the 8X1 mux and then to the input of a 16 bit ADC. After conversion the out put is varying much frequently , i could almost get stability in the MSB 2 niblles only, the lSB 2 nibblesare varying erratically. The return paths for my input signal is the ground before the optical isolator. I doubt that as the channel return path and the gnd are same, due to surges or ripples in the ground plane which is more likely in a circuit comprising of mixed signals the voltage level of the return is varying frequently. Iam enclosing the cicuit also. Iam enclosing the signal conditioning circuit which starts at the input to the card and terminates before the 8X1 MUX. could anyone solve my problem or suggest an analog optical isolator and preferably a working circuit built with it along with input and output specifications. As said the return path is the "gnd before the optical iso[attachmentid=1457]lator". I need to stop the ground bounce or the fluctuations in the Gnd plane from affecting my signal input can you suggest some way to do it. or is there any analog optical isolator with which you've worked and found to produce better results. One more additional information is that theVCC and ground are planes and not tracks.