Tweet
The problem with magnetic ground is that it decays with a 1/t law. This means that the initial decay is fast, but then slows down at later times. This can be seen if you use a log amplifier for the front end. The output of the log amplifier shows an exponential decay, while that from a gold ring shows a linear straight line decay. To get magnetic ground to give a linear decay, you have to have a logarithmic timescale as well as a log compression of the amplitude.
The total measured decay time of a 5oz (160gm) nugget is 700uS to the point where it disappears into the noise. This is not "time constant" in the accepted sense. The total decay of a 30oz (940gm) nugget is >2000uS by comparison. Taking the 5oz nugget the approximate true time constant would be 700/5 (disregarding unseen signal in the noise) = 140uS. The log amplifier (and theory) shows that the signal only becomes a single exponential after 1 time constant. Prior to that it is a sum of exponentials. To get best S/N ratio, early sampling is employed and this part of the decay is faster and higher amplitude than the later invariant decay. Except in the case of a ring. Maths get tricky, but do the experiments. Maybe on a simulator program.
This doesn't answer the capacitor v dual sampling and subtraction debate, but is relevant to it. I always avoid capacitors early in the signal chain as there are always unwanted overshoots and undershoots when dealing with pulses, and a lack of preserving a true dc base level.
Nothing is straightforward, and there is bound to be some loss of the wanted signal in getting rid of the ground response.
Eric.
The total measured decay time of a 5oz (160gm) nugget is 700uS to the point where it disappears into the noise. This is not "time constant" in the accepted sense. The total decay of a 30oz (940gm) nugget is >2000uS by comparison. Taking the 5oz nugget the approximate true time constant would be 700/5 (disregarding unseen signal in the noise) = 140uS. The log amplifier (and theory) shows that the signal only becomes a single exponential after 1 time constant. Prior to that it is a sum of exponentials. To get best S/N ratio, early sampling is employed and this part of the decay is faster and higher amplitude than the later invariant decay. Except in the case of a ring. Maths get tricky, but do the experiments. Maybe on a simulator program.
This doesn't answer the capacitor v dual sampling and subtraction debate, but is relevant to it. I always avoid capacitors early in the signal chain as there are always unwanted overshoots and undershoots when dealing with pulses, and a lack of preserving a true dc base level.
Nothing is straightforward, and there is bound to be some loss of the wanted signal in getting rid of the ground response.
Eric.
Go back and read my post again.
Comment