Ground balance

[waltr]

Hi Waltr
you say in your thread that the sample timing.sequence is:

EFE sample : delay3 : TX pulse : delay1 : Target sample : delay2 : GEB sample

Would you mind saying which sample goes to which input of the intergrater (inverting-non inverting) thanks

That is the correct sequence.
EFE sample -> inverting
Target sample -> inverting
GEB sample -> non-inverting

The pre-amp in inverting and SAT is non-inverting.
The SAT output should then be high going for long TC targets and low going for short TC targets.

If you find this the other way then simply reverse which sampling switch is pulsed, swaps which integrator input the samples go to.
When I first put this GEB code in the output polarity gave a low tone for long TC targets. I just swapped which sampling switch (PIC output pin) was used in the PORT defines.

[6666]

That is the correct sequence.
EFE sample -> inverting
Target sample -> inverting
GEB sample -> non-inverting

The pre-amp in inverting and SAT is non-inverting.
The SAT output should then be high going for long TC targets and low going for short TC targets.

If you find this the other way then simply reverse which sampling switch is pulsed, swaps which integrator input the samples go to.
When I first put this GEB code in the output polarity gave a low tone for long TC targets. I just swapped which sampling switch (PIC output pin) was used in the PORT defines.

Thanks Waltr, much appreciated.

[green]

I?m currently working on designing a ground balancing Pi machine. My initial research identified that the standard Ground Balancing formula for a four sample design is:

X = A1 (S1 ? S4) ? A2 (S3 ? S2) [Formula 1]

Where X = the resultant signal to be integrated
A1 and A2 are the gains for different sample groups
S1 = Target sample
S2 = Ground earth field effect (EFE) sample
S3 = Ground sample
S4 = Target EFE sample

As the EFE samples are reasonably static, you are meant to be able to take two different (albeit consecutive) EFE samples, and they should be the same, therefore the S2 and S4 will cancel out, leaving just the target and ground samples to process.

When A1 (S1 ? S4) > A2 (S3 ? S2) it gives a positive response (i.e. When the target has a high time constant (TC), or in other words, the target response decays slower than the ground response).

From a practical implementation perspective, under formula 1, when you change the gain (A2) in order to ground balance the machine, it also affects the gain of S2, resulting in an EFE imbalance. Additionally, there are resistance tolerances in real world circuits which means that one of the EFE samples would ideally be adjustable separately (and permanently locked in) to fully cancel out the other.

To overcome these issues, I?m proposing a more usable/practical formula for consideration. I?m also proposing to flip around the formula so that High TC targets (relative to the Ground Balance set point) give a negative pulse response (like on the Whites TDI etc.), and Low TC targets give a positive pulse response, as follows:

X = (A2 (S3) ? A3 (S2)) ? A1 (S1 ? S4) [Formula 2]

The gain of A3 would be independently adjusted to fully cancel out EFE from the equation, regardless of the GB gain setting (A2).

An example schematic of formula 3 is as follows: [ATTACH]53162[/ATTACH]

Elliot

The gain of A3 would be independently adjusted to fully cancel out EFE from the equation, regardless of the GB gain setting (A2).
I'm thinking if first delay is changed, GB(A2)needs to be readjusted. Think if A2 is changed A3 would also need to be adjusted. Maybe I'm wrong.

Your thread started me wondering again if changing gain or sample time to GB is better. Thinking it might depend on the target. Been using a 1C integrator and adjusting sample time to GB. Tried a 2C integrator awhile back, didn't work as good. Tried 2C circuit in spice and can't see why it wouldn't work. Thinking of making another integrator. Make delay times #1 and #2 adjustable. Make target sample time adjustable. Make GB sample time and gain adjustable. Test circuit to see if I can tell if adjusting GB sample time or gain is better. Any suggestions appreciated.

[Elliot]

Hi Green,

"... wondering again if changing gain or sample time to GB is better."

In my design, I will be able to Ground Balance by either adjusting the gain on the GB sample or by adjusting the ground sample width. That way I will have the best of both worlds, and can pick and choose at will

In your 2C integrator example, whenever you change the GB gain pot to SW4 (Ground sample), you upset the gain to SW2 (Ground EFE sample). My idea gets around that very issue very simply.

[green]

Tried 2C circuit in spice and can't see why it wouldn't work. Yesterday I connected both inputs to the same signal, no switches(cancelled input signal).

Tried again today with switches, doesn't cancel. Don't see why, maybe the reason I had problems trying the 2C integrator before or something wrong with my spice circuit? Does cancel if both switch commands are the same(make both first delay times 10us).

1C integrator cancels.

[Carl-NC]

I’m currently working on designing a ground balancing Pi machine. My initial research identified that the standard Ground Balancing formula for a four sample design is:

X = A1 (S1 – S4) – A2 (S3 – S2) [Formula 1]

Where X = the resultant signal to be integrated
A1 and A2 are the gains for different sample groups
S1 = Target sample
S2 = Ground earth field effect (EFE) sample
S3 = Ground sample
S4 = Target EFE sample

As the EFE samples are reasonably static, you are meant to be able to take two different (albeit consecutive) EFE samples, and they should be the same, therefore the S2 and S4 will cancel out, leaving just the target and ground samples to process.

I'll relabel this with the pulses in sequence:

.....__.....____................__.....____.......
____|S1|___| S2 |______________|S3|___| S4 |______

S1 is the target pulse and is paired with the target EFE pulse S3. S2 is the ground pulse and is paired with the ground EFE pulse S4. The math is:

X = A1 (S1 – S3) – A2 (S2 – S4)

This is the same end result as Elliot has but maybe less confusing.

The purpose of S3 is to cancel the EF signal in S1 (not S4) and is completely independent of anything happening in the Ground channel. Likewise, S4 cancels the EF in S2 and is independent of the target channel. As you adjust A2 to ground balance, EF cancellation is maintained. That is, you want the gain of S2 and S4 to be identical.

Note that I drew S2 and S4 with a wider pulse width than S1/S3. You can also achieve GB by controlling the pulse width of S2 instead of the gain A2, but the pulse width of S4 should also be adjusted to match S2. You can also do a hybrid approach whereby you set S2/S4 to a wider fixed pulse width and then adjust A2 for GB. Running a wider Ground pulse reduces the amount of A2 required and will be less noisy.

Yes, you can merge S3 and S4 into a single pulse. If you were to run identical gains (A1=A2) and use only pulse width to GB, then it would be:

.....__....._______................_____.........
____|S1|___|..S2...|______________|S2-S1|________

If the Target and Ground channels have different gains then it becomes difficult really fast, and you will likely need a separate EFE cancellation control, which IMO is just overly complicated. Better to use separate EFE samples and keep it simple.

[waltr]

Yes, you can merge S3 and S4 into a single pulse. If you were to run identical gains (A1=A2) and use only pulse width to GB, then it would be:

.....__....._______................_____.........
____|S1|___|..S2...|______________|S2-S1|________

If the Target and Ground channels have different gains then it becomes difficult really fast, and you will likely need a separate EFE cancellation control, which IMO is just overly complicated. Better to use separate EFE samples and keep it simple.

This is how I do the three sample method. This uses a standard two input integrator circuit (Hammer Head) so the two gains are ideally the same (as good as component tolerances).

[green]

Anyone have a thought why my spice simulation https://www.geotech1.com/forums/atta...8&d=1599836935 for the 2C integrator doesn't cancel the sine input?

Changing the value of R2 from 10k to 12.24k causes it to cancel. (Attached Image 2C integrator_sine2)

[Elliot]

I'll relabel this with the pulses in sequence:

.....__.....____................__.....____.......
____|S1|___| S2 |______________|S3|___| S4 |______

S1 is the target pulse and is paired with the target EFE pulse S3. S2 is the ground pulse and is paired with the ground EFE pulse S4. The math is:

X = A1 (S1 – S3) – A2 (S2 – S4)

This is the same end result as Elliot has but maybe less confusing.

The purpose of S3 is to cancel the EF signal in S1 (not S4) and is completely independent of anything happening in the Ground channel. Likewise, S4 cancels the EF in S2 and is independent of the target channel. As you adjust A2 to ground balance, EF cancellation is maintained. That is, you want the gain of S2 and S4 to be identical.

Note that I drew S2 and S4 with a wider pulse width than S1/S3. You can also achieve GB by controlling the pulse width of S2 instead of the gain A2, but the pulse width of S4 should also be adjusted to match S2. You can also do a hybrid approach whereby you set S2/S4 to a wider fixed pulse width and then adjust A2 for GB. Running a wider Ground pulse reduces the amount of A2 required and will be less noisy.

Yes, you can merge S3 and S4 into a single pulse. If you were to run identical gains (A1=A2) and use only pulse width to GB, then it would be:

.....__....._______................_____.........
____|S1|___|..S2...|______________|S2-S1|________

If the Target and Ground channels have different gains then it becomes difficult really fast, and you will likely need a separate EFE cancellation control, which IMO is just overly complicated. Better to use separate EFE samples and keep it simple.

Very Interesting! Thanks for your comments. I can kind of see what you are saying, however, my thinking is this - If we lived in a world where there were no magnetic poles, then we wouldn't need to sample EFE at all. The equation to ground balance (in the prefect world) would therefore just be X = A1(S1) - A2(S2), and ground balance would occur when X = 0 (achieved either by adjusting A2, or by adjusting the S2 pulse width (noting a non-linear nature).

The fact that we don't live in a perfect world therefore means that we need to effectively knock the EFE component out of the equation of X = A1 (S1 – S3) – A2 (S2 – S4).

Given that EFE and ground/target signals are independently driven (assuming no major interplay here), from a purists perspective, I would have therefore thought that it's more straight forward to simply have a circuit that permanently makes A1(S3) - A3(S4) = 0 (with A3 being a new independent gain setting for the GB EFE sample), thereby avoiding any change to the ideal A1(S1)- A2(S2) (i.e. perfect world) equation. Otherwise, when you change A2 in A2(S2 - S4), the gain of the S4 sample will also be changed, and adverse consequences may arise.

In thinking this through further as I type (always a dangerous thing!), I'm sure your methodology will work, but it would be reliant upon the ground (i.e mineral) samples being more consistent when sampling, which kind of counteracts the effectiveness of the need to Ground Balance in the first place. Whereas, I suggest that my proposed approach better separates the impact of the two truly independent variables (i.e. EFE and ground/target samples). I'm always open to be shown otherwise though.

Separately, if one was to adopt the changing GB target sample width approach to ground balancing, wouldn't this enlarge the target "hole"?

[waltr]

Anyone have a thought why my spice simulation https://www.geotech1.com/forums/atta...8&d=1599836935 for the 2C integrator doesn't cancel the sine input?

Changing the value of R2 from 10k to 12.24k causes it to cancel. (Attached Image 2C integrator_sine2)

What op-amp is U3 & U4?
Possibly it is the input current causing the imbalance.
How is the signal at 'b'? Is the imbalance in U3 or in U4?
Try putting on U4's non-inverting input a series resistor to ground equal the the Thevenin equivalent of the input & feed-back resistance.

[green]

What op-amp is U3 & U4?
Possibly it is the input current causing the imbalance.
How is the signal at 'b'? Is the imbalance in U3 or in U4?
Try putting on U4's non-inverting input a series resistor to ground equal the the Thevenin equivalent of the input & feed-back resistance.

Changed U3 and U4 to LTC6244HV, no change in waveform. b is 1/10 of a. Changing R11 makes small changes in amplitude. If switch delay times are the same, cancels. Changing 800us delay to 590us almost cancels. Still don't see why the simulation doesn't cancel. Are we sure the real 2C integrator cancels?

[green]

Separately, if one was to adopt the changing GB target sample width approach to ground balancing, wouldn't this enlarge the target "hole"? from reply #39

I have an Excel program that simulates a PI. First attempt(_9) I changed A2 gain, left second delay at 4us and changed ground sample width to ground balance with a -1.3 slope. Moved the hole. Second attempt(_10) I changed A2 gain, reduced ground sample time by 100us/A2 gain and changed second sample time to ground balance. Hole stayed near 20us. Hole size doesn't change much. Guessing the method with the best S/N ratio is best.

Can attach the program if anyone is interested.

[Elliot]

Can attach the program if anyone is interested

Yes please.

[green]

Can attach the program if anyone is interested

Yes please.

Instructions:
1_enter ground slope into D2, I find it can vary between -1 and -1.4. Long constant current Tx closest to -1. Shorter Tx or constant rate Tx causes steeper slope.
2_enter delay time(I2), target sample time(I3), delay time(I4), GB sample time(I5) and A2 multiplier(I7). Vary one of them until M3 is minimum(ground balanced)
3_change target TC(F2)until )O3 is minimum(hole bottom)
4_copy and paste I2 thru I6 under working chart. Enter ground slope under working chart. Insert text box(A2 gain= )into chart.
5_enter M3 into GEB off, enter absolute value of P3 into GEB on. (enter hole first row)
6_change F2 to other TC's and fill in the blanks.

Thicker long TC targets decay straight line log X log Y at start of decay. Can enter slope in B2 for those targets. Ball park, real target decay straight line log X log Y in the beginning then decay straight line linear X log Y. Depends when samples are taken.

Program kept changing so maybe not as easy to use as should be. Any questions or suggestions to improve?

[green]

No thoughts why the 2C integrator cancels only at certain target and EF delay times? Tried disabling target sample switch or GB sample switch with 1C and 2C integrator https://www.geotech1.com/forums/atta...8&d=1599836935 1C integrator same signal opposite polarity. 2C integrator: similar to 1C when +input switch disabled, about 1/10 the signal when -input switch disabled. Maybe could be the cause?