Dynamic Temporal Sampling Detector

[dbanner]

The massive delay between every DAC step was needed because the external pin output of the DAC goes into 3-state when switching from one value to another, causing the opamp output = ADC input to clip massively.
That is why I suggested to use the PWM output + filtering when doing a variable opamp offset voltage.
This is also why the final DAC value after calibration is stored in EEPROM - to be used as a starting value for the next calibration, which normally takes only 1 or 2 steps then.

Pinpoint in Pickini V4 = do not update the slow-moving average value anymore. Push the button outside a target - when going over the target, the value is not added to the averaging array anymore and sticks out on the spot.

YES, this was a very nice solution. However this design is using Bi-polar pulsing(VOP and VON) It uses a true symmetrical ±5V dual-rail regulated power supply, not a single rail power supply. So the mechanics of the framework is going to be different. Who knows, maybe it will not work in the end, but I am letting AI lead the way, let me see if it can succeed!

Utilizing a bipolar H-bridge design provides a fundamental advantage when dealing with ground balance in a time-domain pulse induction system. Not to mention powerline EMI and Earth field cancellation.

In older microcontrollers like the PIC16F1824, the internal R-2R ladder architecture of the 5-bit DAC momentarily disconnects its internal analog switch network during a register value transition. This puts the output pin into a high-impedance tri-state (floating) condition.

The PICKINI preamplifier design used that DAC value to bias the input stage, a floating pin causes the feedback loop to break instantly. The op-amp output clips against its power rails. requiring a lengthy 1-second software delay loop to let the AC-coupling capacitors recover from the electrical shock. Storing the final value in EEPROM was an elegant workaround to prevent this blinding glitch during active field use.

This design as it stands, does not use a DAC to bias its input stage. Its non-inverting reference path is tied directly and permanently to a clean 0.00V true system ground. The analog baseline never floats or breaks its feedback loops during hunting. It uses a true symmetrical ±5V dual-rail regulated power supply, which is completely independent of the main raw high-voltage battery rail. There is no virtual ground bias to wander, meaning the baseline rests naturally near zero, very stable.

[Detectorist#1]

Hi dbanner,
How to be connected the outputs of the RX stages generating VOP and VON signals to the ADC inputs of uP supplied between +5V and naturally zero (GND) rail? If you use true symmetrical +/-2.5V regulated power supply for uP and RX amplifiers, these connections will be direct, without need of any level translations. Also, after sampling of the RX signal will have relatively long pause to the next TX pulse. Maybe this time will be enough to calculate VON value without using of inverting stage in the RX tract. Sorry if my remarks are totally not useful.

[dbanner]

VOP and VON feed directly into the two internal hardware comparators, not the ADCs.

I added the two 1K series resistors in the pathways of (VOP) and (VON) as discussed before. Also added GND(coarse) and GND(fine) pot-meters.

[Detectorist#1]

Hi dbanner,
Sorry but you are FAR from the exact working connection for the signals VOP and VON to the uP unit. It is not possible for the uP (supplied from +5V and GND) to accept and converts (or compare) the input signals below the negative rail of the supply of uP unit!

[boilcoil]

........ accept and converts (or compare) the input signals below the negative rail of the supply of uP unit!

It uses only the positive signals from U2a and U2b (depending on the coil supply—positive or negative), utilizing only one of the comparators depending on the current cycle.

[dbanner]

It uses only the positive signals from U2a and U2b (depending on the coil supply—positive or negative), utilizing only one of the comparators depending on the current cycle.

Yes, one at a time!

Cycle 1 - Positive pulse decay - U2a - VOP - routed to comparator 1 - captured by SMT1 (comparator 2 not used and is clamped by internal diode and current protected by series input 1K resistor)

Cycle 2 - Negative pulse decay - U2b(flips) - VON - routed to comparator 2 - captured by SMT1 (comparator 1 not used and is clamped by internal diode and current protected by series input 1K resistor)

Maybe adding external clamp diodes on VON and VOP as extra protection, a single Schottky diode(BAT85) after each of the 1 K resistors. Only need to protect against the negative voltage swing, because the MAX412's positive saturation already sits safely underneath the PIC18's +5.3V damage ceiling.

[dbanner]

The exact microsecond that exponential voltage decay tail drops and crosses that DAC threshold line, the comparator trips. Mathematically, the time delay (dt) at which that crossing event physically occurs is the natural logarithm of the signal.
Because U2B cleanly inverted the negative tail, the mathematical slope and amplitude characteristics of Cycle 2 remain an exact mirror image of Cycle 1.

Because these two time intervals (dt positive) and (dt negative) are perfectly mirrored in magnitude but opposite in native magnetic orientation, adding them together in code allows any slow-moving external EMI or Earth Field Effect (EFE) noise to completely cross-cancel and drop out to absolute zero.

[Detectorist#1]

OK. Now, the idea is clear to me. The both signals VOP and VON will be used.

[dbanner]

Optimal Pulse Width Selection

50 µs

The Physics: Keeping the pulse short ensures that the larger superparamagnetic (SPM) iron domains in black sand do not have enough time to saturate. This forces the resulting ground VRM signature to decay much faster than the standard (1/t) law (approaching (1/t^2) at late times), allowing the ground noise to clear out rapidly. Long pulses (e.g., 500 microsecond or more) would deeply saturate the soil matrix, blinding the early listening windows.

Pulse Repetition Frequency (PRF) Selection

2.5 kHz

Interestingly, this the same as the pulse5 (ITMD3) if I am not mistaken.

THE TEMPORAL SAMPLING DESIGN uses voltage height to establish the time boundary.

15 microsecond sample delay in the traditional PI, let's say 12 microseconds translate to:

[BC547]

SMT1 starts counting here

??
You just gate the internal timer with the pulse output of the comparator, right ?

With temporal sampling,"sample delay" is just a matter of shifting the slice level ?
Everything that is determined by "voltage" in a classic PI needs to be translated to "time" and vice versa ?
Shifting the sample delay in a classic PI = changing the slice level ?

[Carl-NC]

The Physics: Keeping the pulse short ensures that the larger superparamagnetic (SPM) iron domains in black sand do not have enough time to saturate. This forces the resulting ground VRM signature to decay much faster than the standard (1/t) law (approaching (1/t^2) at late times), allowing the ground noise to clear out rapidly. Long pulses (e.g., 500 microsecond or more) would deeply saturate the soil matrix, blinding the early listening windows.

It all depends on what you are designing for. If the goal is sub-gram nuggets, then a pulse width of 10-20us is sufficient and would allow a 10kHz pulse rate and a sub-5us sample delay. If the goal is Atocha bars, then you need a pulse width of several milliseconds which pushes the pulse rate below 100Hz and limits the lowest sample delay to 10s of us.

[dbanner]

??
You just gate the internal timer with the pulse output of the comparator, right ?

With temporal sampling,"sample delay" is just a matter of shifting the slice level ?
Everything that is determined by "voltage" in a classic PI needs to be translated to "time" and vice versa ?
Shifting the sample delay in a classic PI = changing the slice level ?

Yes, this is the general idea, every single one of your points is 100% correct.

Because the coil's decay tail is an exponential curve descending over time, changing the vertical height of the DAC slicing level automatically shifts the horizontal point in time where the curve crosses it.

If a classic PI machine wants to ignore wet beach saltwater(conductive noise), it stretches its sample delay further out in time (e.g., waiting from 8µs out to 15µs) to let the saltwater voltage collapse to zero. This is why I chose a slicing level equivalent to 12µs. The remaining beach sand matrix will be solved by the timing-vector subtraction and filtering methods which are well known.

So my very basic understanding of this method is if the temporal machine wants to ignore wet beach saltwater, it simply raises the internal DAC slicing level higher up in voltage. Because the saltwater voltage decay curve is small and collapses rapidly, an elevated DAC threshold ensures that the saltwater signal falls beneath the tripwire almost instantly. The SMT1 timer won't log any saltwater extension ticks, achieving perfect beach stability purely by adjusting a voltage reference register.

[dbanner]

It all depends on what you are designing for. If the goal is sub-gram nuggets, then a pulse width of 10-20us is sufficient and would allow a 10kHz pulse rate and a sub-5us sample delay. If the goal is Atocha bars, then you need a pulse width of several milliseconds which pushes the pulse rate below 100Hz and limits the lowest sample delay to 10s of us.

This design is aimed at beach hunting.

I am far from expert, but reading up on the topic, I understood that wet salt eddy currents have an extremely short electrical time constant (τ) and collapse entirely within the first 6 to 8 microseconds. So the chosen minimum slicing level of 12 microseconds stays well clear of this.

[dbanner]

Black sand (VRM)




Infinite Impulse Response (IIR) Digital High-Pass Motion Filter.
1. The Exponential Moving Average (EMA) Ground Baseline
2. The High-Pass Transient Target Signal Vector

• (Traw): The fresh, raw time tick value coming out of the 24-bit SMT1 hardware stopwatch after the EFE and VRM ground balance subtractions have completed.
• (T(n): The current, updated running background ground baseline value.
• (T(n-1)): The historical background ground baseline value saved from the previous pulse cycle.
• (Δ dt): The final high-pass target signal vector passed directly to the 20-bit NCO1 audio synthesizer to map the variable-pitch VCO speaker output

[Carl-NC]

This design is aimed at beach hunting.

Most black sand I see at beaches is magnetite and does not have a viscous response, it has a more instantaneous response. But, yes, salt response is pretty fast but maybe not 6-8us; I've successfully run a PI in salt at 7-8us but 12us is safer bet.