Ground Balance

[Carl-NC]

The "inversion" stage doesn't invert as in x^-1, rather it makes it negative so the output at the pot is a subtraction instead of addition.

- Carl

[greylourie]

Just the opposite... mineralization decays rather quickly. Here is how the TDI does it (this is all public knowledge, so NBD):

Turn TX on for 100us
Shut off, enter flyback
Wait 10us, take the Target sample
After the end of the Target sample, wait a short delay, take the GB sample
After the end of the GB sample, wait a long delay, take the two EF samples
Use a diff integrator to subtract Target-EF1
Use another diff integrator to subtract GB-EF2
Subtract A1*(Target-EF1) - A2*(GB-EF2)
Adjust A2 for ground balance

In TDI, all the samples have identical widths. I could tell you what to use, but you would learn more trying a range of settings, say 5us up to 25us. The delay between Target and GB samples as well; try 5us to 25us. Technically, this same delay should be used to space the two EF samples. The spacing between GB and EF1 samples should be 50us or more.

There are other ways to do GB, even simpler, but this one has particular advantages that TDI uses to produce distinct audio responses.

- Carl

Hello Carl,

I'm looking at ground balance on the goldscan IV, which appears at first sight very similar to what you have described above. I'm having a difficulty (not uncommon for me) in understanding the way the GB channel aka Iron channel value is inverted before being fed to the balance pot connecting to the output of both channels.

A1 = Target - EF1
A2 = GB signal - EF2, but becomes (GB -EF2)^-1, as it is fed through an additional inverting opamp with no gain.

So in this instance, x = A1 - (A2)^-1

Or have I got it wrong ? I have attached an image to try show what I see ?

[greylourie]

wm6 in post 46, "Probably you can help us with some PI block diagram to show how to incorporate your GB solution in some of classic analog PI design (like Surf PI par example)."

I thought you could, using the Hammerhead as an example, send the output of Q1/Q8 to IC7b. And the output of Q2/Q3 to IC7a.

And for Surf P.I, could send Q1/Q8 (aka Plus) to U7B, and Q2/Q3 (aka Minus) to U7A.

[Davor]

Yes, that's the idea. My way of designing mods is with a target device in mind, and in this case it is a minipulse. Other rigs may follow.

I figured out the way of decoupling GB and time base, so it is still on.

[WM6]

Thanks to deal with us your inventive toughs Davor.

Probably you can help us with some PI block diagram to show how to incorporate your GB solution in some of classic analog PI design (like Surf PI par example).

Is V2/555 input related to the output of pulse generation stage in PI? Some sort of how-to incorporating block diagram in classical analog PI design could be very helpful.

[Davor]

Here is the solution for GB as I see it. This would be a simple form with only one target and one GB pulse. I also designed a possibility of adjusting first pulse delay so that GB may assume some narrow range of values to compensate soils that are behaving different than 1/t law. All (other) pulses have exactly the same duration. A positive gate is supplied with pulses Q1 and Q8 via "wired OR" configuration for target and EF, while the same circuit is used for combining Q2 and Q3 for GB.
There is only one EF pulse and it should work because the gain is the same for all pulses, therefore EF duration is GB duration (2 pulses) minus target pulse duration (1 pulse) = 1pulse.
Duration of pulses is set with R1, while initial delay is not fixed to 1 exact pulse, but varied within, say, 0.8 to 1.2 pulse ratio using R2. T̶h̶i̶s̶ ̶e̶n̶a̶b̶l̶e̶s̶ ̶v̶a̶r̶i̶a̶t̶i̶o̶n̶ ̶o̶f̶ ̶p̶u̶l̶s̶e̶ ̶d̶u̶r̶a̶t̶i̶o̶n̶ ̶i̶n̶d̶e̶p̶e̶n̶d̶e̶n̶t̶ ̶f̶r̶o̶m̶ ̶v̶a̶r̶i̶a̶t̶i̶o̶n̶ ̶o̶f̶ ̶G̶B̶.̶ I realised that there is a dependence between R1 and R2 so there is no free lunch here
This is not tried yet, but should equally apply to HH and minipulse.

[Davor]

Well, don't look much further, just observe all the nice things you can make with a popular CD4017

I'll post my time base solution soon. I was doing some casual math (thank you Wolfram Alpha) and it appears one simply can't avoid a null, or better say a "hole" in a response to remain within a PI lingo, if any of the GB schemes are to be applied. The best thing one can do is to simply alternate two time bases and observe what happens.
I also checked for a bit more complex schemes with periods +(1to2)-(2to4)+(4--(8-16) and it appears to be better - hole falls at lower t/tau, and transfer function is flattened.
I also cross-examined a TDI solution, and it is superior for below half t/tau, a solution which is preferable for gold. So in a case you are after gold on tough terrains, go for a TDI. The solution I'm suggesting is working the best for initial sample starting at 0.7 x t/tau (hole is just below 0.5 x t/tau) and in such case provides a peak at 1.3 x t/tau without any holes after that.

Long story short - there are no especially bad points to this scheme, only constraints also common to the other schemes. There is a possibility to turn this approach to a real gold extracting monster by mere placing the integrator to a very front end and gate the very coil. With initial sample starting at 5us it would enable detection of targets with tau of 1.5us to 20us (and so forth). With additional taps the null moves further towards zero.

I searched for an ideal weighting function to PI receivers, and instead I found that integration works fine as well.

[greylourie]

Im looking at Johnson counters now. Im really enjoying learning, albeit very slowly with lots of bumps and headslapping moments.

[sinclairuser]

hi davor i'm also interested in your gb solution, so far i have got -5, +5 on my mpp, its coming slowly, i'm finding the solder spots small and fiddly in some places especially where it should not be connected to the ground.
some of the newcomers may need the board to be physically bigger so there can be more space around the traces, i'll mention it to george when i do an update in the mini pulse thread.

[Davor]

It is cooking.
I had an epiphany that gave me an interesting idea on employing a Johnson counter to make a whole time base with only two chips. It requires a little refinement, but it is coming.

[greylourie]

Really looking forward to your implementation. Especially the time varying solution.

[Davor]

While I'm waiting on my minipulse board, I'm also considering a GB add-on circuit for it.
There are two options, time variable sampling with constant amplification, and TDI-style time-constant samples with varying amplification for target vs GB.

Time variable is a simple solution that expects 1/t ground behaviour. I'm considering 3 samples, target, GB, and EF after some delay. Gain being constant, the T_(EF) sample is a simple T_(target)-T_(GB)
The idea behind time-variable samples at constant gain is in the fact that ground responds at power law, hence for 1/t function, an integrated sample taken after a pause from 0..1 and with duration 1 (which is between time 1 and 2) is equal to a sample taken between time 2 and 4, and so forth.



What is left to do is add an EF sample with duration 1 at some later time.
It would require replacing the timing circuitry with something else (that I already have), but the analogue part would remain the same. The downside is that by exact 1..2 and 2..4 timing only exact 1/t ground will completely cancel.
In a real life implementation the above mentioned 1..2 and 2..4 would be replaced with, say, 10..20 and 20..40 microseconds. It translates to a simple counter with a mere time increment.
E.g. 10us delay, 10us positive sample (target), 10us negative (GB), 10us negative(GB again), 4 times 10us delay, 10us positive EF sample. 90us in total.
By varying a time increment the hole also shifts, but the ground cancellation mechanism remains the same. So in a way I may have discrimination of some kind as well.

I also have a solution TDI style, which would require additional gain block, some monoflops and a slight intervention to the existing integrator. It is a sort of no-brain-no-pain well trodden solution, but I'm in favour of a time varying solution. Guess it would work with HH as well.



BTW, a sexy formula from above is entered by clicking to "Insert image" and placing the TEX formula as following (together with the http-thing):

http://www.geotech1.com/cgi-bin/mimetex.cgi?\int_1^2 \! 1/t\,dt - \int_2^4 \! 1/t\,dt=0

[Taktyk]

HI,

Mick MD, I have more problems with the preamps AD...

Here, the condition of integration in the traditional PI:


Regards
Taktyk

[Davor]

Wouldn't EF manifest exactly the same way as opamp input offset does?

[Mick-GD]

This should work. TDI applies the GB to a higher gain stage, then subtracts GB from the main channel in a subtraction diffamp. Not necessarily better, just different.

Thanks Carl.

I've done some more experimenting and the GB circuit is up and running. All that's left now is the output section to the speaker, but before I do this I'm going to rebuild it due to some instability in the output of the final diff amp and also set it up for running on 12-14 volts.

Once I have these issues sorted I'll post some more details and test results.


Mick