Which integrator

[green]

Including schematic for #1 and #2 integrators. Have been using #2 integrator. Looked good except for some output offset. Think switch charge injection was causing some of it. Tried #1 integrator. Matched caps and resistors. everything looked good until I wiped a magnet across the coil. For some reason I can't get integrator #1 to cancel the magnet even close to integrator#2. Is there something about the two integrators that would cause integrator #2 to cancel the magnet better or is there something I'm missing?

[eclipse]

This was posted by Eric. The same principle is used in Minipulse, Superscan, etc.
Notice the 200R (100R) trimmer, it may help balancing the signal. I would try diff. op amps, different switches.
Try different delay for EF sample.

[green]

This was posted by Eric. The same principle is used in Minipulse, Superscan, etc.
Notice the 200R (100R) trimmer, it may help balancing the signal. I would try diff. op amps, different switches.
Try different delay for EF sample.

http://www.geotech1.com/forums/showt...421#post171421
Might be the reason I've been using the #2 integrator. He didn't get to (more on that latter). Reply #1 is the same as your image.

http://www.geotech1.com/forums/showt...781#post195781
Was trying to calculate gain and frequency response for the track and hold integrator. Not sure the calculations are correct. I've been trying to add GEB and I think changing sample time doesn't change gain so I think track-hold integrator is not a good choice.

Some of the PI's use integrator #1. Matching caps and resistors wasn't a problem. I still don't see why it doesn't cancel the magnet better.

[6666]

http://www.geotech1.com/forums/showt...421#post171421
Might be the reason I've been using the #2 integrator. He didn't get to (more on that latter). Reply #1 is the same as your image.

http://www.geotech1.com/forums/showt...781#post195781
Was trying to calculate gain and frequency response for the track and hold integrator. Not sure the calculations are correct. I've been trying to add GEB and I think changing sample time doesn't change gain so I think track-hold integrator is not a good choice.

Some of the PI's use integrator #1. Matching caps and resistors wasn't a problem. I still don't see why it doesn't cancel the magnet better.

Is it a metallic magnet ?

[ODM]

You could use just one input resistor in the single input integrator. Are they matched?

One thing may be uneven gain recovery after opamp overload, even if a scope plot shows steady decay, HF gain may not have recovered yet. This can be seen when step driving an opamp into overload with a small HF signal riding on top, the HF gain recovers slower than bias point in some opamps. - Does it get better if you increase target sample delay?

When having more than one integrator in the design it starts paying off to make them single input versions, as they take up less board space especially in SMD and matching channels needs half the matched capacitors.
Charge injection is different on the dual input one, as its input are resistive and the single version is virtual ground, but in practice the difference is minimal.

[green]

Is it a metallic magnet ?

No. The #2 integrator does a good job of reducing the magnet signal.

[green]

You could use just one input resistor in the single input integrator. Are they matched?

One thing may be uneven gain recovery after opamp overload, even if a scope plot shows steady decay, HF gain may not have recovered yet. This can be seen when step driving an opamp into overload with a small HF signal riding on top, the HF gain recovers slower than bias point in some opamps. - Does it get better if you increase target sample delay?

When having more than one integrator in the design it starts paying off to make them single input versions, as they take up less board space especially in SMD and matching channels needs half the matched capacitors.
Charge injection is different on the dual input one, as its input are resistive and the single version is virtual ground, but in practice the difference is minimal.

The single input integrator #2 works with one or two input resistors. The first time I had one. When I rebuilt it after trying the differential integrator I left the other one in the circuit for two input resistors. All 1% metal film resistors. Needed to add a offset trim pot to the preamp to minimize the integrator output ripple.


Target and EF sample 10usec, GEB and EF sample 100usec. The differential sampling integrator #1 has the same problem with just the target and EF sample or all four samples.


The differential input integrator is simpler. Add one resistor and capacitor. The resistors and capacitors were matched. Have to add an op amp and some resistors for the inverter to the single input integrator. The problem is I can't get the differential input integrator to work as good as the single input integrator.

[Teleno]

If you use integrator #1 you must invert the second sample before feeding it to the integrator or else the samples won't subtract. Are you doing that?

[green]

If you use integrator #1 you must invert the second sample before feeding it to the integrator or else the samples won't subtract. Are you doing that?

The target sample goes to the inverting input switch. The EF sample goes to the non inverting input switch. Preamp out, no inverter. You need an inverter for the single input integrator #2.

[Orbit]

For some reason below integrator works well but creates quite noise I write from experience !

[Teleno]

The target sample goes to the inverting input switch. The EF sample goes to the non inverting input switch. Preamp out, no inverter. You need an inverter for the single input integrator #2.

I was referring to #1 in Eric's image. Now I see it's #2 in your image and there's an inverting buffer.

in circuit #1 the output is the integral of the diference in the input currents (see here). Assuming both samples have the same amplitude V, the current at the non-inverting input is V/3K because the resistor is connected to virtual ground. However, the current at the non-inverting input is not V/3K because the other end of the resistor is connected to a floating potential that depends on the level of charge of the lower capacitor. That's why you can't achieve voltage cancelling.

The botton line: #1 is a current differential integrator suited for current sources, while #2 is a voltage diferential integrator. You should stick to #2 because your signal is voltage.

[Carl-NC]

A small-signal AC analysis suggests both opamp approaches should give the same results, but because they are used in a large-signal voltage sampling application they do not. Teleno got it right, it's the voltage sampling on the non-inverting side that screws up #1. I use option #2 (or variations of it) because it is a true integrator for both polarities, does a better job of reducing thermal and flicker noise, has better EF rejection, and doesn't require cap matching (unless you have a ground channel). The only downside is the extra opamp for signal inversion.

[green]

I was referring fo #1 in Eric's image. Now I see it's #2 in your image and there's an inverting bufer.

in circuit #1 the output is the integral of the diference in the input currents (see here). Assuming both samples have the same amplitude V, the current at the non-inverting input is V/3K because the resistor is connected to virtual ground. However, the current at the non-inverting input is not V/3K because the other end of the resistor is connected to a floating potential that depends on the level of charge of the lower capacitor. That's why you can't achieve voltage cancelling.

The botton line: #1 is a current differential integrator suited for current sources, while #2 is a voltage diferential integrator. You should stick to #2 because your signal is voltage.

I don't use either arrangement now as matching the two sets of RC components was always a pain and you get a far better performance by summing + and - signals into a single TC integrator. More on that later.

Eric.

I agree. Eric's statement says the same.

[Orbit]

Should be a beautiful works !

[mickstv]

The single input integrator #2 works with one or two input resistors. The first time I had one. When I rebuilt it after trying the differential integrator I left the other one in the circuit for two input resistors. All 1% metal film resistors. Needed to add a offset trim pot to the preamp to minimize the integrator output ripple.


Target and EF sample 10usec, GEB and EF sample 100usec. The differential sampling integrator #1 has the same problem with just the target and EF sample or all four samples.

I have two detectors both PI GB one is using a single input integrator and the other is using a differential integrator, both circuits needed a small value multiturn pot to smooth out EF.

My target sample is 5us and GB is 130us minus EF 125us. These wide samples cause ripple on the output of both integrator types especially when a magnet of magnetic soil is waved over the coil, so I added a 2 pole LPF filter to the outputs of both integrators to smooth the output.

Final note the GB point between both integrator types is different. The differential integrator one has a GB delay from target sample of 18us, whereas the single input integrator has a GB delay of about 27us.