New FelezjooPI o PulseMaster

[sabios]

After a lot of testing, I’ve finally sorted out the ID issue; it took a lot of work, but I’ve finally arrived at a solid configuration.

It now features true, dynamic ground tracking as well as dynamic ground balancing—meaning, in short, that you can use the coil at any height (though holding it too high does slightly reduce depth).

It also boasts more advanced discrimination compared to the Felezjoo.
It is capable of detecting a silver coin amidst construction nails—something not many pulse induction detectors can do; I actually tried that same test with the Felezjoo, and it failed.

For now, I’m still running tests to confirm its actual detection range.

Here is a sample, but remember, it's a demo.

[haylaz0036]

It’s turning out really well—thank you for your hard work. I follow the videos and tests on YouTube, but the fact that it uses a demo hex file makes me hesitant to start the project.

[sabios]

Check out the new video; it demonstrates the discrimination capabilities of both metal detectors: first the Felezjoo, and second the PulseMasterPI, which I built from scratch.

This is actually just a small sample, as the device features a more comprehensive discrimination mode capable of detecting a silver coin buried 5 to 10 cm deep amidst scrap metal (provided the trash density isn't extreme). In normal mode, it fails to detect the coin; it is only when using the most aggressive mode that the silver coin can be detected quite effectively among the scrap.

https://www.youtube.com/watch?v=7qqOSn99Aik

[sabios]

This is the board I use for the GT—feel free to check it out.
It works for the PulseMasterPI as well.
It has a small modification that might seem insignificant, but I tested it against the standard version; with this latest one, iron discrimination actually improves a bit.This is the board I use for the GT—feel free to check it out.
It works for the PulseMasterPI as well.
It has a small modification that might seem insignificant, but I tested it against the standard version; with this latest one, iron discrimination actually improves a bit.

And of course, I did that because removing and reinstalling the ATmega for so many tests is difficult given the limited space in the original design.

[nypc]

The biggest pitfall of PulseMaster: storing all strings in SRAM (consuming 60%)

[sabios]

It isn't really an issue, as the SRAM consumption is at a manageable level—and I could actually reduce it if needed—but I didn't notice anything unusual during testing; the detector works fine.

So, as long as the detector responds well, I don't see a need to adjust the SRAM usage.

[nypc]

Evolution of PI Algorithms (Three Generations)
══════════════════════════════════════════════════ ═════════════════════
Component G1 FelezJoo G2 PulseMaster G3 Hybrid/Improved
────────────────────────────────────────────────── ─────────────────────
① Tx Drive PB2 Fixed interval PB2 + Freq-controlled PRF PB2 + Freq-controlled PRF
(Freq for display only) (Tail delay: 100-Freq) ★ Uses G2
────────────────────────────────────────────────── ─────────────────────
② Sampling Equivalent-time Gated dual-point + [Dual Mode]
Scheme sampling oversampling Search → G2 Gated (Fast)
14 pulses × 5 pts = 70 pts Early & late points per pulse Classify → G1 Equivalent (Precise)
Curve reconstruction/3 windows 84 pulses, 32-bit accumulation
────────────────────────────────────────────────── ─────────────────────
③ Digital 30-frame moving 16-value sorted Trimmed mean (Primary)
Filtering average trimmed mean + Moving average (Secondary)
(Spikes averaged out) Discard 6 extremes, keep middle 10 ★ Uses G2
★ Strong spike rejection
────────────────────────────────────────────────── ─────────────────────
④ Baseline Fixed 30-frame Adaptive variable-speed Variable-speed tracking
Tracking window tracking + Automatic tracking
Slow baseline Deadband 3 / Fast jump tracking ★ G2 algorithm + G1 auto
★ Faster response
────────────────────────────────────────────────── ─────────────────────
⑤ Metal Logarithmic Linear ratio Logarithmic
Classifi- discrimination + (Early-Late)/Early × 100 discrimination +
cation Ferrous window Pure integer math (Fast) Ferrous window
ln(R) curve fitting ★ Uses G1 (High precision)
★ High precision (Integer ratio option for simplicity)
────────────────────────────────────────────────── ─────────────────────
⑥ Audio 12-tone pattern Simple high/low tones 12-tone pattern
Feedback incl. VCO Band-based incl. VCO
Intensity → Pitch ★ Used here G1 (High Information Density)
★ High information density
────────────────────────────────────────────────── ─────────────────────
⑦ Display UI 16x2 (2 lines) 20x4 (4 lines) + DIAG 20x4 (4 lines) + DIAG
3 display modes Preset modes/Flyback params ★ G2 layout + G1 display modes
★ Diagnostics/Presets
────────────────────────────────────────────────── ─────────────────────
⑧ Resource Usage Strings in Flash Strings in SRAM Strings in Flash
★ Saves SRAM Consumes 1.2KB (60%) ★ Uses G1
PWM backlight/contrast Backlight ON/OFF only PWM backlight/contrast
★ Software-adjustable ★ Uses G1

[sabios]

I imagine you're using a simulator to arrive at this data.

[nypc]

I hope this helps while you are making improvements.

[sabios]

Here is a new video I’ve uploaded to field-test the metal detector.

I took it out into the field to put it to the test on actual terrain and demonstrate that it performs well beyond just a laboratory setting; it now features an improved filter and better discrimination, along with an updated, true Dynamic Ground Tracking system.

By the way, the demo I previously uploaded was a bit different; I’ve updated it, and it is now complete. Testing has proven it to be a solid, real-world detector, not just a lab prototype.

Regarding the comment from Junior Member nypc: I had already worked everything out—the only remaining issue was the SRAM, which I’ve now fixed.

To be clear, it isn't perfect, but it outperforms many other metal detectors; even major brands can't match what the PulseMasterPI can do.

In tests, it is capable of detecting copper or silver coins buried amidst trash—such as nails, soda caps, and the like.

https://www.youtube.com/watch?v=FvA6bNLERbA