TDI ground balancing scheme, twice over

[Qiaozhi]

George...maybe the Minelab SD2000: two independent transmit ,receive and ground balance channels.

Have a look at the SD2000 instruction manual; specifically section 5.3.

There may a clue here, but I'm failing to see it. If GB1 is the usual GB sample, and GB2 is a [later] sample, (presumably the main, GB1 and GB2 are fairly close together), then either GB1 or GB2 will be able to cancel the ground signal. But surely any target that looks like ground will also be rejected, regardless of whether GB1 or GB2 is used.

The answer must be in Minelab's MPS (Multi Period Sensing) method:
MPS (Multi Period Sensing) is Minelab’s advanced Pulse Induction (PI) technology that transmits pulses of different time periods. MPS also samples the receive signal at multiple time periods allowing target signals and ground signals to be separated. This effectively removes the ground signal from even the most highly mineralised ground while still being sensitive to both small and deep gold. This achieves superior depth in extremely mineralised ground.

Multi period sensing is Minelab’s patented pulse induction technology. Standard pulse induction metal detectors are limited because they use a single pulse width. MPS uses varying pulse widths that extract more information from targets, achieve better ground balance and detect to greater depths.

The main claim for MPS appears to be better ground balance.
But, unless anyone can say otherwise, I suspect the "hole" still exists with MPS.

[green]

I just trying to learn PI detectors. Don't know how the TDI works, so maybe I'm thinking some different. Gave an example in reply #7 how I think it might work with two different sampling channels. Probably doesn't if no one does it. Each channel eliminates ground, but not all of the target signal eliminated on the other channel.

I apologize for the graph not being clear. I'll try to come with a better way to explain it.

To understand graph, start by looking at two lines only. The ground curve and the thinner line labeled 20. The ground curve is 2 at 10usec and .6 at 34usec. If you multiply .6 by 2/.6 and subtract it from 2, the 10usec sample you get zero. Cancels ground signal and any target with a slope of the line labeled 20. The line labeled 20 has a TC 20 usec, parallel with the heaver TC line labeled 20.

[Bill512]

Have a look at the SD2000 instruction manual; specifically section 5.3.

There may a clue here, but I'm failing to see it. If GB1 is the usual GB sample, and GB2 is a [later] sample, (presumably the main, GB1 and GB2 are fairly close together), then either GB1 or GB2 will be able to cancel the ground signal. But surely any target that looks like ground will also be rejected, regardless of whether GB1 or GB2 is used.

The answer must be in Minelab's MPS (Multi Period Sensing) method:
MPS (Multi Period Sensing) is Minelab’s advanced Pulse Induction (PI) technology that transmits pulses of different time periods. MPS also samples the receive signal at multiple time periods allowing target signals and ground signals to be separated. This effectively removes the ground signal from even the most highly mineralised ground while still being sensitive to both small and deep gold. This achieves superior depth in extremely mineralised ground.

Multi period sensing is Minelab’s patented pulse induction technology. Standard pulse induction metal detectors are limited because they use a single pulse width. MPS uses varying pulse widths that extract more information from targets, achieve better ground balance and detect to greater depths.

The main claim for MPS appears to be better ground balance.
But, unless anyone can say otherwise, I suspect the "hole" still exists with MPS.

George, my first impression about Minelab PI detectors was the usual: yes the MPS method, different responses from metals vs ground using MPS, subtraction of channels in order to get a clean metal response, etc.
But ,reading in some depth the patents and the schematic of the SD2000, I came up on a different conclusion.
The MPS was finally something like a smoke curtain and the truth was even simpler: two independent and parallel ground balanced detectors.
Each detector has its own transmit and receive sequence-timings and by the fact that the timings are vastly different, the ground balance points (essentially the holes), will be different for each detector.
Many of the physical electronic components are shared by these two detectors and this sharing is done by timing.
Finally, these parallel detectors (channels) are combine with something like an "exclusive or " action.
As Zed says, " channel with highest amplitude is heard- both channels never heard at the same time".
Basically,when a metal target falls in to the hole on one channel, the other channel -detector , has a signal .
Now if you choose to hear only one channel,(by the channel switch),then, as the manual warns,there will be a hole in the detection and some targets may not have response.
In Minelab SD series detectors, channel 1 serves mostly for the bigger targets and channel 2 for the smaller.
From another point of view, the hole system acts as a high dynamic range ground balanced detector.
Some clues are in the SD2100 manual, " The SD2100 is essentially two detectors in one. Each "half" of the detector must be ground balanced separately...."
I never had enough time to do a real in depth analysis on these thoughts, and even better to make a physical prototype based on these thoughts, so still are just
some speculations.

[porkluvr]

Green, before I rack my brain trying to absorb the graph data (just kidding - it can't be that hard) what is meant by "log log slope -1"?

OK I get that it's a log log graph but I do not understand what the -1 is for. ?


Ignorance is NOT bliss.



edit: Just as a side note, my whole impetus for my devising this topology was because I wanted to try and clone the SD2000 from Zed's RE drawings but I could never get the timing mechanism to work in LTspice (and that for me was a deal breaker). I decided to try another route.

It's just too bad I couldn't go for something 'simple' (I stopped work on the GS4), but doh, I just wouldn't be right to press forward without some semblance of Minelab's ground balance would it? (A rhetorical question, please don't nobody answer.)

[porkluvr]

A little off-topic, but here is a previous abomination that I tossed aside a couple of years back. Maybe I should dust it off and make some refinements. Diode OR gates can be really flexible circuit components (but they can make things look sort of klunky).

I don't have the timings exactly right (comparing to the Zed's SD2000 waveform plots), but maybe close enough, and what's there can easily be changed.

Besides, after taking a good look at the post processor input summing resistors (right before the LT1057 amplifiers) with those odd-ball resistor values, I have an idea that ML maybe would maybe have been better off with some slightly different GB sample timing. Perhaps their engineers got behind schedule, and with the boss chomping at their butts... decided it would be quicker to adjust gain coefficients rather than re-work the timing circuit. (just a thought)

[Carl-NC]

Ground has a 1/t response, not exponential, so it really doesn't have a tau. Eddy responses tend to be close to exponential. Because of this, when you set up a subtractive-sample ground balance to get rid of 1/t you also end up with a small range of eddy targets that fall in the notch hole. You can adjust the GB timing/gain to move this hole almost wherever you like. In fact, when I did a full-up redesign of the TDI (which became the TD-SL) the first proto ended up with the hole right on top of US nickels. Being a common test target, that would never do, so I fiddled the timing and moved it up into pull tabs. As I recall, relic hunters noticed that the TDI would hit bullets well but not buttons, and the TDI-SL did the opposite. Or something like that.

So, yes, you can build 2 GB channels and offset the holes, then combine their responses to get rid of holes. I never did this with the TDI because I was working on a much better solution that had no holes in the first place and could ID iron.

[Carl-NC]

An exponential response plotted on a log-linear graph is a straight line. A 1/t response plotted on a log-log graph is a straight line. Here is a log-log plot of a 1/t response plus 3 different exponential responses representing 3 target taus. An early sample minus a late sample (gained up) is used to null the ground, but the right combination can also null out any one of the exponentials. But not all of them.

[Qiaozhi]

Ground has a 1/t response, not exponential, so it really doesn't have a tau. Eddy responses tend to be close to exponential. Because of this, when you set up a subtractive-sample ground balance to get rid of 1/t you also end up with a small range of eddy targets that fall in the notch hole. You can adjust the GB timing/gain to move this hole almost wherever you like. In fact, when I did a full-up redesign of the TDI (which became the TD-SL) the first proto ended up with the hole right on top of US nickels. Being a common test target, that would never do, so I fiddled the timing and moved it up into pull tabs. As I recall, relic hunters noticed that the TDI would hit bullets well but not buttons, and the TDI-SL did the opposite. Or something like that.

So, yes, you can build 2 GB channels and offset the holes, then combine their responses to get rid of holes. I never did this with the TDI because I was working on a much better solution that had no holes in the first place and could ID iron.

Thanks Carl. I was hoping you would join this discussion, as it did seem we were missing something obvious.

Dang Qiaozhi, you have busted my bubble.

Hi porkluvr ... it appears that your bubble has survived!

At least I don't now have to trawl through Minelab's confusing patents to try and figure it out.

[porkluvr]

An exponential response plotted on a log-linear graph is a straight line. A 1/t response plotted on a log-log graph is a straight line. Here is a log-log plot of a 1/t response plus 3 different exponential responses representing 3 target taus. An early sample minus a late sample (gained up) is used to null the ground, but the right combination can also null out any one of the exponentials. But not all of them.

Thanks you Carl, that is encouraging and especially welcome because after I took another look at the "wannabe" circuit in #20 I came to the conclusion that it is probably better left for dead. My approach on that idea would likely prove disappointing if I was ever to pursue that route.

[Orbit]

Hello everyone !
specific question whether Whites TDI SL there are two holes ?
Thanks !

[kt315]

people, please do not ask me to give you White's TDI schematic. SORRY BIG SORRY my archive has not it. please I have to reminder to you
ONLY YOURS to take the pains fill the archive. if you do not send me schematics - there will be nothing of schematics.

DVD with archive you will receive FREE, if order Delta Pulse board.

yours Anatloy