Search the Community

I contacted a friend in a high level position at very large unnamed metal detector company in the United States, but I was brushed off with a 👍 reply indicating they weren't even remotely interested. So, after a few months of letting them sit on the idea without even a call back about more information, here's the source code. Released under the MIT license. In a nutshell, it will increase your detector's discrimination, with a focus on PI & ZVT. I also have an unreleased ESP32-based embedded recording / training / classification module with rough schematics for ADC integration and DAC for optional tuning the response rather than looking at a small LCD display. I don't plan on releasing that yet. https://github.com/nickvellios/metal-detector-ai Now it belongs to the community. No sense throwing away hard work. Enjoy! 😏 I am open to work if any detector company wants to change the game with help from an engineer who understands metal detectors. This isn't my only idea or project. Overview Metal Detector AI An advanced, flexible machine learning system for classifying metal detector audio signals. Read the Readmes in the link above for more info. Completely Flexible: Train on any labels (gold, iron, banana, type_A, etc.) - zero hardcoded assumptions Advanced Audio Processing: Handles continuous detector audio with intelligent silence-based segmentation Time-Invariant: Robust to different sweep speeds (slow/fast over same target give same pattern) State-of-the-Art ML: Ensemble of CNN, Transformer, and traditional ML models Real-time Processing: Live audio streaming with WebSocket communication and visualization Multiple Formats: Supports WAV, MP3, and M4A audio files Pattern Recognition: Analyzes spectral patterns, temporal dynamics, and harmonic content Smart Segmentation: Identifies natural breaks in tone patterns (high-low or low-high) for complete event capture Optimized Performance: M1 GPU acceleration with efficient audio processing pipelines -- Audio Analysis Philosophy Unlike traditional approaches that look for silence between sounds, this system understands that metal detectors produce continuous background audio(warbling/humming). Detection events are anomaliesor pattern changes in this baseline. -- Time-Invariant Processing The system extracts features that are robust to sweep speed variations: Same pattern detected whether you sweep fast or slow over the same target Normalized temporal features (relative timing, not absolute duration) Spectral invariants (frequency relationships don't change with sweep speed) Envelope shape analysis (pattern morphology preserved) -- Advanced Feature Extraction Temporal: Onset density, relative peak positions, rise/decay ratios Spectral: Centroid, bandwidth, contrast, rolloff statistics Harmonic: Pitch stability, harmonic ratios, chroma features Wavelet: Multi-scale time-frequency analysis across 4 wavelet types Statistical: Distribution shape, entropy measures, envelope characteristics -- Ensemble Learning Architecture Combines multiple advanced models with weighted voting: CNN: 4-layer architecture with attention for spectral pattern recognition Transformer: 6-layer encoder with self-attention for temporal modeling Traditional ML: Random Forest + Gradient Boosting + SVM ensemble Final Prediction: Weighted ensemble (CNN: 40%, Transformer: 40%, Traditional: 20%) -- Anomaly Detection Algorithm Establishes baseline characteristics from initial 2 seconds of continuous detector audio Uses 10-feature frame analysis (RMS, spectral centroid, bandwidth, rolloff, ZCR, MFCCs) Calculates anomaly scores via Euclidean distance, Mahalanobis-like distance, and max deviation Adaptive thresholding: mean + (sensitivity × std) of baseline scores Filters anomalies by duration (0.1s to 3.0s) to capture realistic detection events -- Time-Invariant Feature Engineering Temporal Normalization: All timing features converted to relative scales (0-1) Onset Analysis: Density and distribution patterns independent of absolute time Envelope Characteristics: Rise/decay ratios, peak positions, symmetry measures Spectral Preservation: Frequency relationships maintained across sweep speeds Statistical Invariants: Distribution shapes, entropy measures, harmonic ratios -- Model Architecture Details CNN: 32→64→128→256 filters, attention mechanism, adaptive pooling Transformer: 256-dim embeddings, 8 attention heads, positional encoding Traditional: Random Forest (200 trees) + Gradient Boosting (200 est.) + RBF SVM Ensemble Logic: Soft voting with weighted probabilities, confidence analysis -- Performance Metrics The system provides comprehensive evaluation: Individual Model Accuracies: CNN, Transformer, Traditional ML performance Ensemble Performance: Combined weighted voting accuracy with confidence analysis Feature Importance: Rankings of most discriminative features per model Cross-Validation: 5-fold stratified validation with mean ± std scores Processing Time: Real-time classification speed (typically < 0.5s per sample) Confidence Levels: 5-tier confidence system (Very High ≥95%, High ≥80%, Medium ≥60%, Low ≥40%, Very Low <40%) -- All aspects are highly configurable: Model Architecture CNN layers, filter sizes, attention mechanisms Transformer heads, encoder layers, embedding dimensions Traditional ML estimators, depth, regularization Training Parameters Epochs, batch size, learning rates, optimizers Data augmentation, dropout rates, early stopping Cross-validation folds, test split ratios Audio Processing Sample rate, chunk sizes, buffer durations Anomaly sensitivity, baseline duration, thresholds Frame analysis parameters, feature extraction settings Real-time Streaming Classification intervals, confidence thresholds Visualization refresh rates, detection logging Audio device selection, buffer management

Apologies if this has been covered elsewhere but have come across an application in IP Australia by Nokta for a design filing titled Nokta Active Magnetic Sensor Technology lodged November 2024. Wonder if this relates to their new PI machines?

I was on Amazon just now and noticed Carl Moreland @Geotech has finished his 3rd edition of the book, "Inside the Metal Detector" I have the 2nd edition on digital kindle ebook and have been thinking of getting a paper copy. When I went to Amazon looking for it I found the new, recently finished 3rd edition. Thought I would give it a shout out and let folks know. HH Mike

Hi to all of You after a quite long time I've been silent due to tough moments in the non detecting days... In this particular moment, staying close in my humble lab means to take care of my mind health while storms are keeping me off from business and my old mother's conditions to complete with the fear of an imminent loss. My question it is...Can someone remind me of the right pinout for the coil's + and - position on the board of a TDI BeachHunter? I modified the cables and added a waterproof connector to change coil in the future but apparently made a mistake when positioned the coil cables the last time years ago. Water leaked inside before I decided to open the box and take care of the issue and however everything worked for some time, not really to my taste actually..... Now, I only remember partially from where I de-soldered the coil's thin cables and cause of the Detech 8"mono deep rebuild this will take an age to make it really waterproof in a resin mold. The faster option I have it is to use the original 12"DualField and listen again for some loved noise. I hope the pictures can helps to see the boards conditions after I cleaned as best as I can and I suspect thanks to another picture I found, that the Wp6 and WP7 are the interested pins where to connect and solder back the coil (this time the connector female)... I know, this are the battery direct pins on the board but here is my doubt... I can't find other pictures unless TDI Sl pinouts with the 5 pins/cables on the board, while the BH just has two for power and braid... Hope it isn't too much to ask and bother the experts in here, I'm in a water glass and for sure fog in mind it is making me blind to a poor error... Sorry for this.

Hi everyone, Recently, during a survey with probes, my team and I came across a structure resembling a temple with multiple burial chambers. There are indications of gold and other artifacts. We’ve attempted several times to dig deeper, but the depth is at least 15 meters, making excavation extremely difficult without specialized equipment. On top of that, the terrain is harsh—steep, muddy, and challenging to navigate. I’m looking at various detectors because of their 3D scanning and deep detection capabilities, but I’ve read mixed reviews online. What are the alternatives for this kind of environment? Thanks in advance for any insights!

A recent post on GeoTech forum. Worth a look for the technical minded. The first attached file in first post that is downloadable for members https://www.geotech1.com/forums/forum/technology/tech-forum/442274-android-phone-based-metal-detector-fcmd

Hey y'all, Let me start this by saying the opinions you read in this post are mine. Now I'll give ya some resumé: I've played in rock bands for many years, from 50s and 60s to loud heavy metal, I've even seen Styx twice, had my ears ring for three days after each show, and I still pass hearing tests with flying colors. I've been to 100s of concerts. I'm no stranger to high end audio and studio quality headphones. Basically I can hear a mouse fart in a coliseum, and no tones are lost to me. 😁 I think my eardrums are probably steel. Since I got the Manticore, I've used the ML105 headphones, and they're fine, but get in the wind on a cold day and the wind goes right through the earcups, they wheeze and whistle, and you can't hear your targets. They're great for calm days, and with an ocean breeze they aren't too bad when it's hot. They give you situational awareness which I heartily endorse. They are loud, you even find yourself turning down the volume, they have great dynamic response. They are by far the best sounding headphones for the Manticore that I have. But... Get them wet and they're toast. Go out in a high wind and you're toast. 🤔 That's where the Killer B Thresher headphones come in, Thresher was bought out by the Killer B company. They are waterproof to over 100', they're tough, well made, have a nice long cable and high quality wires and hardware. Somewhere I read that Killer B upgraded the drivers, and while they sound "canny" and thin, they do deliver some pretty decent volume especially because they are compatible with the Manticore "Underwater Volume Boost", which puts you about where the ML105 headphones are with loudness, just a much "thinner" sound. I bought them for water use. A few days ago I found out that Killer B makes a wireless set for Minelab, even uses their wireless tech so there is no lag, and no additional transmitter is required. They pair as quickly as the 105s, you hold the red button in to enter pairing mode, put the Manticore in pairing mode, and a couple seconds later they're connected. I think they are using the same circuit for wireless as Minelab does under license. They have real leather covered earcups and headband, they're comfortable, have well made components and they block out about 90% of outside noise. I bought them to use on cold windy days. The similarity to the ML105 headphones ends, however. Again, these headphones have a thin "canny" sound, their dynamic range doesn't come close to the 105s. If it did these would be perfect, as it is I'll probably only use them for cold windy days after hunting season, and maybe early beach mornings if it's windy. They are also not loud. You have to turn the Manticore volume all the way up. Another thing I find a bit trifling is the intense multicolor LED on them, it blinks like a runway beacon. 🫤 Honestly, I think I will keep all three, I like that I don't have to charge the Threshers. Each has its merit and use. For most detecting I do I'll probably use the ML105 headphones, they sound great and are extremely comfortable. For the rest of relic season I may use the Killer B's. I am also looking at a way to use my Creative Jams, they are light and great when it's really hot. I have a solution but again they aren't loud, despite their internal volume and bass boost. Most APTX-LL transmitters don't have inline volume. 🫤 🍀

I am a big fan of the White's SignaGraph display. A version 1.0 was originally developed for the Eagle Spectrum. The Eagle Spectrum underwent a complete hardware revamp, and was renamed the Spectrum XLT. A more refined version 1.1 of the software was matched up in the XLT with a much better LCD display. The SignaGraph was also used on the DFX, and was largely the same as on the XLT, with the addition of multifrequency options. The SignaGraph was later greatly enhanced on the V models (Vision, V3, V3i, VX3) and renamed the SpectraGraph. The genius of the SignaGraph/SpectraGraph is the ability to display multiple target id numbers at the same time, and even to choose how the target id number is determined. Here is the SignaGraph explained by its designer, engineer Mark Rowan. More details can be gleaned from the White's XLT User Guide. Spectrum XLT Engineering Note The SignaGraph™ "Phase Spectrum Analyzer" by Mark Rowan Some time ago, I had a conversation with an avid treasure hunter whose instrument of choice was White's Eagle II SL 90. He described to me a technique with which he could discern pull tabs from rings, nickels, and other desirable targets by listening for some subtlety in the audio response. Then he asked me, "If I can do this, why can't you program the Eagle's microprocessor to do it?" My response was, "If you can do it yourself, why would you want the microprocessor to do it for you?" I mention this as a means of illustrating what I consider to be the metal detector designer's fundamental dilemma, which is, as Prince Hamlet might have phrased it, "To beep or not to beep". More specifically, if you're faced with a target at some depth in badly mineralized ground and the detector has a hard time getting a solid reading on it, what do you do? If you design your detector to ignore the target, and then someone comes along with their El Cheapo brand detector and digs the target, which just happens to be a $10 gold piece -- you're in big trouble. If, on the other hand, your customers find that they're spending most of their time chiseling through eight inches of hardpan and finding bent nails and wads of aluminum foil, you're not much better off. The point I was trying to make with the gentleman who had devised the clever pull tab discriminating scheme was, that if you put too much of that kind of "intelligence" into your metal detector, there are always going to be those targets that you miss because the machine got fooled. Which brings me, of course, to the newest White's model, the Spectrum XLT. The Spectrum XLT has all of the features, performance, and flexibility of previous members of the Eagle series, plus a new display which makes the instrument remarkably easy to use. It also makes use of a new way of displaying information about targets -- the "SignaGraph™ or "Phase Spectrum Analyzer" -- which shows the operator everything that we currently know how to display about the characteristics of metallic objects in the ground. In this way, we have gone a long way towards addressing the dilemma I mentioned earlier. The Spectrum XLT is a very "smart" detector, but it is also an "honest" one. Having done the best it can to determine the probable identity of a target, the Spectrum XLT gives you all of the information you need to make your own decision (human beings are, despite what you might have heard, still a whole lot smarter than computers) to dig, or not to dig. Before I begin to describe in some detail what the SignaGraph™ is and how it works, I should emphasize that you don't need to know how it works in order to use it effectively, and that the best way to learn how to use it.... is in the field. In a very short time you will begin to recognize certain display patterns as being characteristic of certain types of targets. I should also point out that even if you ignore the SignaGraph™ altogether, this instrument still has the audio discriminator, V.D.I. number, that its predecessors had, plus the icons, and some significant improvements in terms of weight, physical size, and ease of operation. White's SignaGraph display For many years, White's has built detectors which identify targets based on a V.D.I. number (V.D.I. stands for Visual Discrimination Indicator) which characterizes metallic objects according to their size, shape, and composition. The V.D.I. scale on the Spectrum XLT runs from -95 to +95. Large positive numbers typically indicate objects which are good electrical conductors; for example, silver dollars will come in at 92. Smaller positive numbers usually indicate objects which, because of their size, shape, or composition, are not as conductive; nickels will read about 20 and aluminum foil may come in near 5. Large negative numbers are typical of targets which are readily magnetized, but which conduct electricity poorly or not at all. Some sands or soils which have a high concentration of ferromagnetic minerals may read -93. Metals containing iron have both magnetic and conductive properties, which causes them to spread over a wide area of the scale, although most typically iron objects will fall in the range -30 to -75. The V.D.I. reading is an excellent way to determine the identity of most commonly occurring targets, although I might mention in passing that the only 100% reliable discriminator is called a shovel. However, as a famous metal detector engineer once said, "Life is grossly unfair" (actually, there is no such thing as a famous metal detector engineer, and life really is fair, it just doesn't want anybody to know). For one thing, the signal which a detector receives back from even moderately mineralized ground is typically much stronger than the signal it receives from the targets buried in it. This makes determining an accurate V.D.I. number for a target at any substantial depth a very challenging business indeed. Furthermore, some targets will cause an abrupt change in V.D.I. response during the course of a single pass under the loop; the most notorious of these are the dreaded bottlecap and the dreaded small piece of foil near the surface in bad ground. Enter, as they say, the Spectrum XLT. The SignaGraph™ is very similar in some respects to the familiar analog V.D.I. meter. The display is calibrated from left to right in V.D.I. units, from -95 to +95. When the loop is passed over a target, a V.D.I. determination is made, and a vertical bar is placed at the appropriate place on the scale; near the right end of the scale, say, for a reading of 78. So far, this is just what an analog V.D.I. meter would do. At this point, the similarity ends. An analog meter can indicate only one value at a time; with the SignaGraph™, up to 30 readings can be displayed simultaneously. Also, the vertical height of the bars in the display has significance; the height can either be used to indicate signal strength or a running total of the number of readings at that point on the scale ( the operator may choose which of these two indications is to be used). The advantage of this type of display format becomes evident when the loop is passed over a bottlecap or some other flat, thin iron object. Although the instrument may respond with a loud, clear audio output, and the V.D.I. readout may register a value near the upper end, the SignaGraph™ will tend to "smear out"; numerous segments will appear throughout the display, many or most of them in the negative (typically iron) range. Try the same things with a coin, and you won't see the "smear"; typically you will see 1-3 bars grouped closely together near the top end of the scale. If any smearing does occur, as it might on a deep coin in bad ground, the more accurate readings will stand taller in the display and will tend to persist from sweep to sweep. Another unique advantage of the Spectrum XLT is the ability to make use of information gathered during the course of multiple sweeps of the loop. For years, clever detectorists have realized that by passing the loop over the target repeatedly and mentally keeping track of the range over which readings appear, and the most frequently occurring numbers within that range, they can achieve the highest possible accuracy on really tough targets. The Spectrum XLT performs this operation automatically. The standard mode of operation is the so-called "Graph Averaging" mode, in which a continuous count is kept of the number of readings that fall into a particular slot in the graph. This might also be a good time to mention that more than one V.D.I. determination is made during the course of a sweep; sometimes as many as 6 or 8 readings will be taken during a single pass, so it only takes a couple of sweeps for the effect of averaging to become significant. What you will see in the field will be a single bar on the display which will "grow" until it stands out prominently above the other bars on the display. Although it is not necessary to adjust them, there are a number of controls that allow you to customize the way that the graph is displayed. It can be set up to clear itself on each sweep of the loop, if you find that too much information is persisting in the display for too long. Or, you can configure it to let the vertical bars fade slowly out of view. Even the rate at which this fading takes place is adjustable. If you don't want to be bothered with any of that, then don't be. The factory preset settings should work just fine for almost anyone. For those of you who want to know an explanation of Accumulate, Average, and Fade, one is included in this Guide. If all of this sounds confusing or mysterious to you, allow me to put your mind at ease. The Spectrum XLT is one of the simplest-to-operate detectors you will ever use. I shall describe just how and why it is so easy to use momentarily; but before I finish talking about the SignaGraph™, I want to say it one more time-- you don't need to be a Nobel Prize candidate to figure out what the display is telling you. The usual response from somebody seeing it for the first time is something like: "Okay, I get it now. Now leave me alone and let me hunt!" What is it that makes the Spectrum XLT so easy to use? The key is something that is known in the software business as a "menu-driven interface". To implement that, we have used what is known in the display business as "A True Graphics Display". What all of this means to you, the user, is that all of the controls and options are listed clearly in plain English on the display. A flashing arrow appears on the screen next to one of those options; you can move the arrow up or down with the two "arrow" keys on the 5-key touchpad. When the arrow is next to the control you are interested in, you push the ENTER key. That is everything you need to know to run this machine. If you are like me and you hate reading instruction manuals, I believe I can safely guarantee that you will be able to operate the Spectrum XLT successfully your first time out without ever having to open the cover -- although the manual should be extremely helpful if you want to fine-tune the performance of your detector by adjusting any or all of a rather lengthy list of professional options. Incidentally, another name for this method of running a machine is the "point-and-shoot" method; you point at what you want, then "shoot" with the ENTER key to make it happen. Finally, for those in a hurry, there are a number of "shortcuts" designed to make accessing commonly used features as fast as possible. What makes the Spectrum XLT even easier to use are the factory preset programs (like those in previous Eagles) which you can load with just a few simple keystrokes, following the prompts in the display. These programs configure the machine automatically so that the beginner or casual treasure hunter can expect a great deal of success over a broad range of conditions. Any attempt on my part to detail all of the advanced features and controls which the Spectrum XLT has to offer would probably leave me with blisters on both of my typing fingers. Suffice it to say that all of the features we have had in previous state-of-the-art detectors are here in this one, plus several new ones. Most of the features are there because somebody asked for them -- the moral of the story being, keep those cards and letters coming, and we will continue trying our best to give you the kind of detector you really want. Mark Rowan was a Senior Engineer for White's Electronics, Inc. Mark holds degrees in General Science, and Electronics Engineering Technology, and is a graduate of the University of Oregon. His background includes satellite communications and RF test and measurement instrumentation. White's SignaGraph examples from Spectrum XLT manual

The subject came up in “Tarsacci Metal Detectors.” From experience, the Deus 1 at 4K is almost magic in trash. The enemies are steel and aluminum caps, and aluminum tabs. These are “thin and stiff” sheet metal objects. Even though gold rings respond to higher frequencies better, they have more mass than thin trash and aren’t “stiff.” The goal in trash is to try to deprecate the “thin and stiff” sheet metal objects as much as possible. Of course, there are other factors. So, I ask Google AI; “Will objects with more mass better respond to lower frequencies than thin targets?” Keep in mind that AI will lie, but it’s pretty good if it can draw from established history. Yes, objects with more mass generally respond better to lower frequencies than thin targets. The relationship is based on the principle of resonance, where an object vibrates most efficiently at its natural or resonant frequency. The relationship between mass and natural frequency For most physical objects, natural frequency is inversely proportional to mass. - Higher mass = Lower natural frequency: Heavier objects have more inertia and require more energy to accelerate. This means they oscillate or vibrate more slowly. - Lower mass = Higher natural frequency: Thinner or less massive objects vibrate more quickly. The stiffness factor While mass is a primary factor, an object's stiffness also plays a crucial role in determining its natural frequency. - More stiffness = Higher natural frequency: A stiffer object resists vibration more effectively, so it will vibrate at a higher frequency. - Less stiffness = Lower natural frequency: A softer or more compliant object has a lower natural frequency. Added note: Wadded up aluminum isn’t thin. Paper lids with aluminum backing aren’t stiff; the little bottles of milk they give to the kids at the park always get me.

The Golden Mask Altus is a real game changer. It incorporates the new Golden Mask Binary Sinus Pulse (BSP) hybrid technology to outperform any detector on today's market. The shape and cycle of emitted magnetic fields are key to achieving superior detection depth. Mass-market detectors typically use one of two field types: 1. Continuous Sinusoidal Fields (VLF-based): Efficient, lightweight, fast-response, and easy to produce with good discrimination. Drawbacks: Shallow depth, poor ground noise suppression, and reduced performance in mineralized soils. 2. Pulsed Rectangular/Trapezoidal Fields (PI-based): Greater depth, robust in diverse soils, and broader signal range. Drawbacks: High power consumption, heavy, very expensive, complex, and slow signal analysis with poor selectivity. To overcome these limitations, GM developed BSP (Binary Sinus Pulse) — a hybrid waveform combining two sine signals of different frequencies, producing a continuous bipolar pulse with sinusoidal edges and plateaus. BSP technology key benefits: • Exceptional depth performance • Full ground and cavity noise suppression • Wide signal spectrum detection • Strong magnetic field with low power use (high efficiency) • Lightweight and compact • Simple antenna system • High immunity to external interference • Superior discrimination and selectivity • Fast, clear target response (high scan rate) BSP merges the strengths of both VLF and PI systems without their flaws. It supports both deep relic hunting and fast coin shooting. BSP is already implemented in the fast and deep-seeking ALTUS detector. https://www.goldenmaskdetectors.com/detector-details-id-93-en.htm Tech Data for Golden Mask Altus Brand - Golden Mask Model - Altus Technology - BSP Frequency - Binary Sinus Pulse Ground Balance - Manual Search Mode - Motion, Bi-tonal Controls - ground balance, discrimination level, threshold, tone, sensitivity, WS/SP switch Coil Type - BSP Weight - 1.6 kg with 11 inch search coil Battery - 8.4V 2S Li-Ion Battery Life - 8 hours on speaker and max. volume, 12 hours on WS headphones Wireless Headphones - Optional, WS-ready (WS107 compatible) Charger- Dedicated charger, EU socket Warranty - 5 years - the electronics board; 2 years - all the rest, except batteries

i know Manticore does not have a true mixed mode audio, but coming from a Whites machine where I used it frequently to help me sniff for deep targets, I am wondering if anyone has come up with a setup to get the Manticore to work like a mixed mode machine

The mod I did like on the Gold Bug 2 was Gary Hudson's sensitivity modification, unfortunately though I lost contact with Gary years ago and this modified one needs fixed. I do not have the paper work either or I would try to fix it. Gerry McMullen had one when we went detecting together about 20 years ago and he did really good with it, so I finally broke down and got one. Anyone remember what all the switches did?

First off, I apologize if this is a nuisance, please delete. But I had to ask. This is kind of scary, how good this is. I have pull tabs figured out, but it took a long time. The method I use is exactly what AI gives and not commonly given In particular, the first thing I check is the “wider” using “pass your coil over the target area with a faster, tighter motion.” If it isn’t the size of a coin or ring, I walk. Checking ID is an afterthought. Added Note: Caps will also give a “wider” sound, bigger than physical size. Foil also. Question to Google AI mode: “Why do pull tabs cause problems for metal detectors?” What are the best ways to identify pull tabs? Because a gold ring can register a similar VDI number and audio tone as a pull tab, separating them reliably with a metal detector is a learned skill that requires careful signal analysis. The most effective strategy is to pay attention to the target's specific audio signal, its consistency, and its location, rather than relying solely on the visual ID number. Audio signal analysis Veteran detectorists focus on how the target sounds and feels as the coil passes over it. - Gold rings typically produce a "smooth," solid, and repeatable audio tone that is consistent from any angle. - Perform a tight, quick sweep: Pass your coil over the target area with a faster, tighter motion. A dense target like a ring will produce a tight, consistent audio response, while a less-dense pull tab may produce a wider, more broken-up signal. - Swing from multiple angles: Sweep the coil over the target from different directions. A ring will hold a solid tone, (or flat coin) while a pull tab will likely produce an inconsistent "warbling" signal as you change your sweep direction. The trade-off with discrimination settings Most metal detectors feature discrimination settings that allow the user to ignore signals from certain metals or conductivity ranges. While this feature can be used to filter out pull tabs, it comes with a significant risk. - Missing gold: If you set your detector's discrimination to ignore the signal range that pull tabs typically fall into, you will almost certainly miss gold rings and other jewelry that register in the same range.

The audio from Micks microphone is atrocious and the video'ing is less than ideal, but you'll get the gist. To be fair, Micks 5000 has an audio booster with speaker so it is a lot louder....but the benefits of the mods are obvious. See video below:

I just finished scanning and uploading the rare Engineering Guide to the White's Eagle Spectrum from 1991. A true innovation in detecting, one of the first digital models, and the first with what became the White's SignaGraph display. 28 page free download so check it out. Page 17 caught my eye as something we do not see these days. And actual factory produced air test/calibration chart for the Eagle Spectrum. It gave owners an easy way to see if their detector was performing to factory spec. It better serves here to show just how far we have come since 1991 in detector performance. Or not, depending how you look at it.

On that note, I am finding that the new version LCD readout gold bug 2 is missing something with that fine GB control gone, I miss the second fine control, it seems to smooth the machine quite a bit and quieter. This new machine is more noisy. Less tune ability. I do not think it was a good move removing the fine control on GB at all. I got to use my lcd version GB2 extensively a few weeks back at Rye Patch. I am pretty sure this new GM2000 will eliminate any need for the vintage gb2 anyway. I hope so anyway. I am pretty sure I will move my gb2 to Loaner and/or extra emergency machine only after the GM2000 is available.

I've been watching some of Crawford's videos recently about the Manticore. One of the videos has Mark Lawrie discussing the way the Manticore processes signals. At a certain point he is at a loss for words and he just says something like the onboard computer does the processing. Isn't that really it? So now the latest and greatest Gold Monster 2000 should have an even more robust computer/chip which may be reason for some of the cost. I don't think it could be said to be in the Nvidia class but it certainly could have horsepower to spare. I think the Algo has demonstrated that. So, does anyone know about these chips they are using for this GM2? Does JP or Mark discuss them yet?

I'm not sure if this has been mentioned already, but here are some new words Minelab wants to trademark. DETEXPERT - https://search.ipaustralia.gov.au/trademarks/search/view/2548983?fs=PENDING&q=Minelab MULTI-AU - https://search.ipaustralia.gov.au/trademarks/search/view/2550937?fs=PENDING&q=Minelab ECHO WAVE - https://search.ipaustralia.gov.au/trademarks/search/view/2550953?fs=PENDING&q=Minelab

I have been watching and waiting for something big to happen in detector technology development for 25 years now and am currently waiting to try out the New Goldmonster 2000 when it’s available to see how its newest technology has improved over previous technology. IMHO myself looking back 5 years I’ve seen 3 detectors that really stood out in regards to technology breakthroughs: Algoforce E1500 stands out with It’s very quiet Pulse Induction EMI handling capabilities, it’s static target ID working at almost target depth and the capability of identifying small lead shot with 00 ID’s while retaining very good sensitivity in very mineralized ground conditions is really quite a lot of technical improvements for any standalone Pulse Induction detector. Manticore although having a lot of previous technology incorporated into it really stands out to me as it’s has completely combined everything I wanted in a VLF into one detector, so kinda a tech. break through with its incorporating so many features into one detector. I now can coin hunt and find similar to Gold Bug 2 size gold all in one detector and with a decent waterproof housing. I really enjoyed all the years learning the GPX-6000, primarily with the small coil which has the ability to hear the smallest nuggets I’ve ever seen found with a Pulse Induction detector making it also a true breakthrough in technology over previous generations. I also liked the timing switch that worked well and helped with determining hot rocks and creating different metal sounds on rubbish. Mods and coil improvements have also been interesting to watch over the last decade with quite a performance boost coming from these minor modifications and accessories to make the older version detectors perform much better. Please add any detector and it’s technological advancements that I’ve missed from your perspective. 😀 Ron

I had a Garrett GTI 2500 for a while and it was a pretty interesting detector. I think it was one of the longest produced but not real popular. The ability to size targets worked very well and I don’t recall any errors but that feature didn’t move on to other models that I know of. I remember a review about a year into its production that found it detected coins deeper than about ninety other detectors at that time but I don’t recall anything else in the review. Someone gifted me the Treasure Hound Eagle Eye two coil attachment and I did a humorous air test getting a signal on my truck at about 40’ away. I also was evaluating two placer deposits and someone told me you can use the Treasure Hound to locate deep black sand deposits. I did and it worked. One at a depth of 2’ and the other at 4’ and both had fine gold. I also thought If I was looking for a metal object the mineralization would interfere. The Nel Big coil also picked up the black sand. I was doing some pipe welding in an area and used it once to trace another pipeline so we knew to be cautious trenching near it. The GTI in my opinion was actually a pretty good detector. It would air test at huge distances. Heavy by today’s standards but balanced well. I suspect most on this forum didn’t really like it. I sold it as I just wasn’t using it. My first prospecting detector was the first Fisher Gold Bug. It was a great design and worked well. A prospecting partner I had years ago also used one. He moved to the southern coast of Oregon and later passed away. His brother sent me his detector and said it didn’t work. I sent it to Fisher and it was beyond hope. Salt air corrosion had destroyed the circuits so I kept the rod and coil. I don’t really see much use in it though. Something to be said for waterproof detectors in a constant salt air environment. Thanks, interesting older technical specs. Has anyone compared the new Garret 24k with the original Whites? Since Garrett boosted coil signal (I think) did it improve performance?

There’s been a lot of discussion on detector technology and rightly so but have you ever been surprised or frustrated with broken coil bolt ears? Dropped a detector and broke some plastic or battery holder cracked? How about in general life? Car dashboard, door panels? Heavy equipment mechanic getting to wiring or hydraulic pilot lines in the cab and the plastic covers break? Of course replacement plastic is often VERY EXPENSIVE! But not as durable as throw away plastic! Anyway I’ll never forget my first Keene dredge and thinking (laughing) that plastic header box is supposed to hold up with tons of rocks slamming into it? It did! It never failed! It was nearly indestructible. I couldn’t break it with a hammer if I wanted. Why can’t engineers learn to use this plastic in situations where it might just crack or break completely. It’s been around since the 1950s. It’s called Marlex. https://en.m.wikipedia.org/wiki/Marlex I remember a friend had his detector in the back of his Jeep and a digging tool bounced and hit the control box screen and cracked it. Remember the clear plastic Lexan or Lexon (sp) you can hit it with a hammer, it won’t break very easily. Generally used in race car windshields. Seems obvious. Just wondering.

Hi Folks, Thought I would start a thread showing my first attempt at an analog detector, I've been working in the electronics field for many years mostly in the analog/rf area. Hope everyone enjoys the process, and any input or suggestions would be appreciated. I'll try and post as I go along, but this is my after hours hobby, so updates may come every few weeks. I apologize if this goes overboard on the images. So here goes..... Here is the overall unit, the display will tip up and turn on. Closeup of the display Display tipped up, it will show RSSI across the top, and material ID in the center and across the bottom. Control Panel, has gain hi lo, discrimination on off, coil balance, threshold, volume, phone jack, and on off with lo bat indicator. Coil Actual board and display turned on and balanced, I'm sorry for the vertical image I couldn't straighten it up, the coil is under the paper, it's a hand wound 3" OD test coil. Note that this board has only three adjustments and the one in the above renderings has four, this one has no threshold adjust. 1957 silver quarter Gold chain and cross. Aluminum pull tab. Coke bottle cap. Clad quarter. My steel cutters. I'm getting parts in to start putting the new one together, hopefully my updates to the circuit will work out. We have a saying, "may have to shoot the engineer to finish the project", it feels like every other day I come up with another improvement :), anyway I hope you guys enjoy this, I'll post as I go along. -Sun-Boy

Saw a post where another user scooped the availability of the Deus 2 V3.02 software, I happened to have a spare RC unit on a shaft with the 9" coil rigged up for water with the antenna, so it was a no brainer to just go ahead and update the RC, and the spare WSA II headphones I have. It took 3 attempts, but I got the RC done, and the headphones updated the first time. It wasn't hard to put my programs back in, both Beach and Relic, I didn't really notice any major differences in the menus. Nothing at all, actually. Everything was where it was before. I quickly threw everything in the truck along with my scoop and went to the beach. Beautiful day today, and ironically it was also just minutes from low tide when I got there. 🥳 I started out in the water, and immediately noticed the difference in the software - the machine was much quieter in the water, it still has the spurious blips and some bump sensitivity, but hey, sensitivity was 93. Otherwise It was silent. It was even more silent on the beach at sensitivity 95, I couldn't be more thrilled. 🙂 There is no EMI at this beach, and no black sand today. The recent rain even kept the jellyfish to a minimum, so I went in barefoot knee deep. Some of the targets were pretty deep, but the jury is out on depth. 🤔 Here's the trash: At reactivity of 2 and AR at 6, I was quite pleased at the loud response of the detector even over the tiny stuff. One piece of iron in the photo went off loudly, but I couldn't find it with the pinpointer! It really found some small stuff. 👍 I got some light iron tones in the dry, and dug a couple of toys: They're plastic except for a spring in one and screws in the other. Tiny screws. Higher on the beach I found a few coins: They were the same IDs as usual, 84 for Zincolns, and 63 for the nickel. I don't do air tests, I was going to drop a nickel, quarter, dime and penny on the sand, but I'll let y'all do that, I prefer finds in situ. 😏 I did have one issue, 3 times the detector stopped working, I had to wait for the coil to be rediscovered and reconnected. It just dropped out. At first I thought it was my antenna, but one time I saw the reconnect so this beta is definitely a beta 🙄. I have no idea what causes it. Pinpointing is better, audio is louder, I was using the 9" so I nailed the targets mostly on the first scoop. 👍 The one thing that really impressed me so far is the stability, the increased volume, and pinpointing seems more accurate. It also really finds small stuff! More to come, I'll be taking it out for a relic hunt soon. 🍀

With all the discussion of the next Minelab detector, or detectors I am interested in the opinions of gold hunting detector users as to the use of auto ground balancing. It is a controversial subject with varying opinions from the introduction of the GT16000 and later to the SD2200 and an option on all later models. As a user of the GPX6000 I find it interesting just how the tracking differs from earlier models.