Minelab H1-24 Results

[phrunt]

I'm a little baffled by the anticipation of new ground breaking machines, baby steps are more my mindset, maybe as I'm from an IT background where we hit a wall a long time ago with processing power, instead of faster and better CPU's they put multiple CPU's onto one core, so in detecting terms that's like sending you and a bunch of your buddies out swinging detectors at once.  They hit a wall, CPU's could not get much faster a decade ago, processing power could not improve.  So, instead let's use lots of processors to combine together to increase performance.     You had dual core, hyper threading 😅 quad core and so on. 

I think with my very limited knowledge of detecting technology and it is very limited this same limitation has been reached, albeit somewhat slowly with the much smaller market and demand for the technology so it wasn't exactly prime picking for big buck's investments.  There are a small handful of people that have dominated detecting technology, if it were a big money maker this wouldn't be the case and likely our "Heroes" now would not be so significant.   

In saying that I'm not under appreciating the investment that has gone into making great detectors and many companies have fallen doing so, I also have an appreciation of how little to expect from new models, and expect at some point all brands will be doing "paint jobs", a key termed by Fisher, one of the bigger early innovators and somewhat of a fallen hero.

[jasong]

Like with computers though, eventually different design approaches minimize the need for the ever increasing clock speed race. 

Same approach can be used for improvement in new detectors and people insisting we've reached some wall because TX power and sensitivity is at it's limit of usefulness - that isn't everything, there is plenty of room to improve by looking at different design considerations. Those ARM chips are $20 each now. Toss a 2nd one in to channel surf every channel and constantly keep the detector on the quietest channel - no down time stopping and pressing the noise cancel button plus reduced noise. Similarly, keep the 2nd chip running EMI reduction algorithms. Have it do auto ground tracking/ground signal processing too. This stuff makes a massive difference in time savings when you are exploring, which really is the strength of these lightweight detectors like the 6000 (and presumably the E1500). It's hard to understand this for people that just work old patches slowly though as the need isn't as great as it is in situations where you aim to cover lots of ground and that ground (and EMI environment) changes and alters greatly. Saving time = more ground covered. 

There is a lot of processing power that can be thrown at detectors cheaply now. Make them faster, quieter.

In terms of power, you don't need more raw power to get more sensitivity, another approach is to lower the noise floor another magnitude and boost the RX gain. RX gain is limitless (well, essentially). I've been saying this for a decade. I think modern detector design is starting to show this idea now that a ton of power can be put to signal processing.

[Steve Herschbach]

You are one of the optimists Jason. Just because you have been saying something for a decade does not make it true. I’m more in the Simon camp when it comes to a new detector actually having enough of a leap in performance to make an actual real world difference in the gold being found. The GPZ 7000 was a true advancement over what came before and made dead patches come alive again. The 6000 picked up the crumbs. This next go I think we will see improvement more in the ergonomic side than anything else. I place my bet that there will not be genuine performance that outperforms substantially on what we can currently get with the 6000 and 7000 combined, or even just a 7000 with a proper set of aftermarket coils.

Metal detectors have a basic limitation in how far they can detect gold items. From http://www.talkingelectronics.com/projects/200TrCcts/MetalDetectors/MetalDetectors-1.html 

“the sensitivity is roughly proportional to the cube of the object diameter (as expressed as a function of the search coil diameter). Sensitivity is also inversely proportional to the sixth power of the distance between the coil and the object. All this means is that if the object size is halved the sensitivity is reduced to one-eighth. Also, if the depth is doubled the sensitivity is reduced to one sixty-fourth. It’s easy to see why all metal detectors which are designed to pick up small objects use small coils, (150 to 300 mm diameter) and really only skim the soil surface. If the search coil is doubled in diameter for greater penetration the sensitivity to small objects falls to one-eighth. You rapidly encounter the law of diminishing returns.”

Famed metal detector engineer Dave Johnson reiterates this in a different way at 

https://web.archive.org/web/20230719232930/http://fisherlab.com/hobby/davejohnson/davejohnsonjohngardinerinterview.htm

“Getting extra depth out of a VLF, multifrequency, or PI machine is very difficult, because these machines follow an inverse 6th power law relationship between signal voltage and depth. If everything else is maintained equal, doubling the depth requires 64 times as much signal. If this is done by increasing transmitter power, doubling depth requires 4,096 times as much battery drain. That’s the basic reason why depth increases come so slowly in this industry.”

 

That is where I think we are now and why the GPZ 8000 has been slow in coming. Much to Minelabs credit they don’t release a machine unless the engineers can point to data showing some real performance improvement differences - you know, those 30% things. But there is actual real world data in hand to back up those claims when they make them. I think in this case the most they might eke out will be a marginal gain they can point to in detection depth on multi ounce nuggets. That’s enough to sell lots of detectors but in the end I think it will be greatly debated whether this new machine is any better in big gold at depth than a modded GP/GPX or 7000 with a big X-Coil. And for anyone but those few still finding the big ones deep on a regular basis nothing that will change anything. Most U.S. patches in particular simply don't have those monster nuggets at depth that people dream of. The 7000 and 6000 have already bled them close to dry. We have hit the wall not only in electronic terms but even more importantly in geological terms.

Peter Charlesworth picked a very good time to retire. Go out at the top of your game. I have a ton of respect for Peter and his retirement is a message of sorts for those who follow the business of business.

Long story short, next generation my prediction is a small gain at best, so small it won’t make any real difference over returns being seen with machines that we already have at our disposal. I actually hope I am wrong, and I hope I am eating crow and apologizing to you, and acknowledging you were right and I was wrong. Nothing would make me happier! 

[jasong]

The x^6 problem is well known, it's not related to the point I'm making though except tangentially (I'll touch on that in a moment).

What I'm saying is that ignoring depth gains entirely, exploration prospecting really can benefit greatly from faster, quieter detectors. This is achievable by throwing more cheap signal processing power at it. Namely - EMI and ground noise reduction. These two things slow a detector down greatly and the faster you can work, the more ground you can cover, the more gold you can find. Noise of any sort massively slows exploration down.

Now back to the x^6 limitation. First off - yes, it's true. But understand that it's an ideal model. In other words - it's like looking at air tests in a Faraday cage with no noise. We don't detect in a Faraday cage, thus we don't use these machines in environments where we get close to that theoretical asymptote/limit. We have to deal with EMI and ground noise. The first thing we do in these cases is bump back RX gain. Then maybe some audio smoothing/stabilizer. Then maybe change timings. All of these reduce us even further from this x^6 limit.

The 2nd thing I'm saying is this: by working on reducing EMI and ground noise we don't just make exploration faster, we can also bump RX gain back up and thus actually, we do gain depth. We can not use smoothing, and gain depth. We can use more sensitive timings, and gain depth. And thus, get closer to that x^6 limit each time we eliminate more and more noise. 

 

[Digalicious]

42 minutes ago, jasong said:

What I'm saying is that ignoring depth gains entirely, exploration prospecting really can benefit greatly from faster, detectors. This is achievable by throwing more cheap signal processing power at it.

 

I doubt processing power is some sort of bottleneck or impediment to metal detector performance. I mean, even cheap processors can easily handle the meager computational requirements of a metal detector. 

[jasong]

30 minutes ago, Digalicious said:

I doubt processing power is some sort of bottleneck or impediment to metal detector performance. I mean, even cheap processors can easily handle the meager computational requirements of a metal detector. 

It is when you do complex signal processing, multiple things in parallel like the examples I gave. This is nothing specific to detectors, just electronics in general. A lot of this is done in the MCU now in general. That takes clock cycles up. 

Now, running multiple Fourier transforms simultaneously, doing stuff like monitoring all the channels for the lowest noise constantly, analyzing ground constantly, you'd benefit by either multiple cores or faster clock speed when you add a bunch of parallel tasks together that are always operating in the background if you moved to a real time, constant noise reduction and ground monitoring system. The 6000 does this with ground I think, to some degree. 

Add to the signal processing all the routine functions of the detector like writing to the screen, operating audio, bluetooth, user interface/buttons, stuff like discrim/TID calculations, configuring operation of the analog circuitry (changing timing patterns/pulse configuration/etc). It all takes clock cycles up. 

Even if technically you have enough speed to complete all the tasks individually, electronics get "sticky" or laggy when too many processes all compete for those cycles during the same time frame. So you have to slow the operation down or reduce functionality in those cases, even though you have enough speed technically speaking. This is why multiple cores benefit applications running many things in parallel even if you aren't using all the capabilities all the time. Or multiple processors. 

I'm not suggesting throwing more processing power at current or obsolete (IMO) detector designs. It'd be pointless for a 5000 for instance. I'm suggesting changing the way detectors are designed to take advantage of all this cheap processing power today that wasn't available cheaply 15 years ago. 

[Digalicious]

Jason,

Multi core / thread processors are pretty cheap. Even fairly basic CPUs in computers have the computational "power" to easily run tasks, that are far more complex and resource hungry than what a metal detector needs.

If all it takes to get much better performance out of a metal detector, is to add $200 to the price to get a "better" processor, wouldn't the engineers have done that by now?

In addition, I'm not so sure about your EMI suggestion. I mean, the bottom line is that a metal detector doesn't know if the signal it receives is coming from EMI or metal in the ground. As such, there is no viable way to mitigate EMI in SMF modes, without losing performance in one way or another. Channel hopping with tiny 0.2 khz increments, isn't going to mitigate EMI, given that the EMI and its harmonics are wideband, and so is the receiver (SMF).



 

[phrunt]

I hope Jasons onto something, and I'm sure if he's onto it the engineers at Minelab and a few other places have been thinking along those lines too.  

Regardless of the lack of competition for the GPZ, if anything in recent years was going to inspire them into releasing a new GPZ it would have been the birth of X-coils, they don't want them around, a way to get rid of them or severely distrupt them is discontinue the detector and bring out the new model, the crowds generally move onto the new model and of course the new model would be harder to modify for aftermarket coils, yet they didn't do this, and I believe the reason is unless they want a massive flop and a lot of customer backlash the detector has to show reasonable real world improvements to justify people buying it.   If people go out and buy it and it's doing little over their current equipment that's not going to go down well, social media will spread the fact like wildfire and Minelab will be in trouble.

I tend to think any performance gains will be in relation to ground handling which in a way the 6000 lacks with its inability to handle many hot rocks, perhaps some depth can be gained with better ground handling on the GPZ too.  EMI is a tricky one to overcome, as the detector can't tell EMI from signals, maybe it could use processing power to check for patterns and try filter it out, but then you're adding another pretty risky filter.  Better shielding on the detector itself can help, and that's an area Minelab for some reason is a bit lacking, so they could get minor improvements there.

A little bit of an improvement here and there is hardly enough to justify a new model of such a high-priced detector and if the detector is not a hit, a good seller and popular it's going to make it very difficult for them to release another new model in the future, the market for these detectors is shrinking which doesn't help especially investing big money into making new ones when you're already at the top of the market so dominate the sales anyway.

Recently I saw Nenad had a GPZ 19" Coil with carry bag and lower shaft for sale for $650 Aud, so I brought it up for people wanting a cheap coil to turn to an X-coil adapter, in response to that I had another guy say he'll sell his for $500 AUD if anyone wants one, another guy saying he's got two he can sell for $400 each.  All these basically new barely used coils for such cheap prices for a coil that was $1900 AUD not long ago and price reduced to $1500 now, if there was such a big demand from the big deep gold hunters for these larger coils they probably wouldn't be selling for such low prices.  Even with X-coils doing massive coils sales of these very large sizes are low compared to the 15" Concentric, the most popular coil by far.  I don't think releasing a detector geared at big deep gold is going to be a great success for them, some of the really old models with a few modifications seem to be really close to if not at GPZ with large coil level already for that task and the Quad Bikers towing large coils prefer the older models too.

I'm like Steve, I'd love to be completely wrong and they shock us with some wonderful machine, I'd like nothing more than to be proven a fool.  Fingers crossed :)

 

[jasong]

57 minutes ago, Digalicious said:

Jason,

Multi core / thread processors are pretty cheap. Even fairly basic CPUs in computers have the computational "power" to easily run tasks, that are far more complex and resource hungry than what a metal detector needs.

If all it takes to get much better performance out of a metal detector, is to add $200 to the price to get a "better" processor, wouldn't the engineers have done that by now?

In addition, I'm not so sure about your EMI suggestion. I mean, the bottom line is that a metal detector doesn't know if the signal it receives is coming from EMI or metal in the ground. As such, there is no viable way to mitigate EMI in SMF modes, without losing performance in one way or another. Channel hopping with tiny 0.2 khz increments, isn't going to mitigate EMI, given that the EMI and its harmonics are wideband, and so is the receiver (SMF).

I'm not talking about throwing more power at obsolete design paradigms. I'm talking about adapting new design principals to the power we available today. Until very recently, most detectors were still designed like it was 1995, even when they did use more modern components. Like procedural style programming instead of object oriented. One thing at a time type stuff, it's ultra inefficient. Throwing more processing power at old designs and ideas accomplishes little and is not what I'm suggesting. 

EMI is absolutely something that can be dealt with more effectively. It's done already in scientific instrumentation via numerous different methods, though these are complex. I suggested one simple way that I know already works in electronics other than detectors - real time noise monitoring and channel selection. That alone would save the 30 seconds of noise cancelling plus would keep you on average on a much quieter channel. 

I'm just repeating myself here though. I guess people either understand what I'm saying or not. 

 

[jasong]

16 minutes ago, phrunt said:

EMI is a tricky one to overcome, as the detector can't tell EMI from signals, maybe it could use processing power to check for patterns and try filter it out, but then you're adding another pretty risky filter.  Better shielding on the detector itself can help, and that's an area Minelab for some reason is a bit lacking, so they could get minor improvements there.

Right, this is one commonly used method in instrumentation. I've posted about this years ago, but the way it's done is by doing something like a gradiometric analysis. In other words - you sense noise from two different spots in real time, and then look at the rate of change and direction when possible of the E field. Vector gradient analysis. (fancy terms for looking at rates of change and direction)

The difference between the signal from the ground to the coil is much higher than the difference between taking that same reading somewhere else, say the control box.

Now the same analysis can be applied to EMI. Since it originates from the sun, lightning, planes, cell towers, etc the difference in magnitude is far less as it's picked up between those same two points. So you can more easily determine those types of signals are likely to be EMI. Now - with enough CPU speed you can do further analysis on the signals to differentiate them, and still have results in "real time" before the brain hears anything like Fourier analysis on top of the gradiometry. You can even take gradiometry further and do vector (directional) analysis with a fast enough CPU and some clever circuitry to determine which direction the signal is coming from.

This is a massive oversimplification but I'm trying to keep it understandable for all.