Gary Shows You How To Say Goodbye To Emi!

[Digalicious]

1 hour ago, UKD2User said:

I don't think it's possible to say that most of the EMI mimics the "signature" of a good target - a little will, most won't, I'd argue.  

I guess it comes down to one's definition of a "good target".

If a good target is considered to be a high conductor, then I would agree, because most EMI IDs randomly in the low to mid and mid high conductor range.

If a comparative algorithm was utilized, then most of the EMI signatures / IDs (the random low and mid to mid high IDs) would be eliminated, along with very deep coins, most gold jewelry, and collocated coins. 

Countless metal detector engineers over the past few decades, have been trying to solve the EMI issue. I believe it's correct to assume that a comparative algorithm would have been one of the first things they all thought of, but then quickly dismissed it as unviable. I suspect they quickly concluded that a comparative algorithm would mean the proverbial, "Throwing the baby out with the bath water" 🙂

 

[iron_buzz]

On 10/20/2023 at 4:42 AM, Shelton said:

Honestly... is not possible to say goodbye to EMI with SMF devices. They are the magnet for EMI. I have plenty of them and I'm struggling with the same problem all the time. You can reduce it a little bit but you will also lose the depth of your metal detector. Under power high voltage lines you can get a silent EMI version and if you are a happy moron like me in the past and you think:  that is good it is quiet. Yes, it is but the software is cutting noise perfectly with performance. 

 

OK, I'll bite:  how do you know when you're getting silent EMI?

[iron_buzz]

On 10/23/2023 at 11:15 AM, Digalicious said:

About a year ago, I read a post in which a metal detecting engineer was explaining why SMF is so much more suspectable to EMI than SF. It went something like this:

With SMF, multiple samples (frequencies), get accumulated during the processing. Therefore, the EMI noise also gets accumulated accordingly.

I think the bottom line, and most important aspect of EMI, is that a metal detector does not, and cannot, know if a signal is coming from the air, or coming from a target in the ground. As such, there is no true way of mitigating the EMI, without a performance loss in one way or another. For example:

EMI signals are typically weak, and that of course, is why reducing the gain on the amp, mitigates the EMI, but that means depth loss. Switching to a SF that is outside the range of the EMI is another option, but then the benefits of SMF are gone.

The difference in the channel frequencies is very small. Something on the order of a mere 0.02 khz difference between the channels. That's just too small of a shift to mitigate the various frequencies, harmonics, and random nature of EMI. Further to that, a frequency shift doesn't address the fundamental problem of noise when using SMF. That is, the exponential rise in noise due to SMF accumulating the EMI noise.

But isn't gain applied after all of the other signal processing?  If that is true (and I don't know, which is why I'm asking) then reducing it is only going to reduce everything equally, including weak signals from deep targets, so nothing really would be gained.

[CPT_GhostLight]

17 hours ago, iron_buzz said:

But isn't gain applied after all of the other signal processing?  If that is true (and I don't know, which is why I'm asking) then reducing it is only going to reduce everything equally, including weak signals from deep targets, so nothing really would be gained.

That's one of those things that detector companies keep secret. We don't know the exact circuit chain structures, but there is enough information out there (including this video) to find usable workarounds. No detector will eliminate all EMI, so we just have to adapt on the fly.

[Geotech]

On 10/23/2023 at 9:15 AM, Digalicious said:

About a year ago, I read a post in which a metal detecting engineer was explaining why SMF is so much more suspectable to EMI than SF. It went something like this:

With SMF, multiple samples (frequencies), get accumulated during the processing. Therefore, the EMI noise also gets accumulated accordingly.
...

The difference in the channel frequencies is very small. Something on the order of a mere 0.02 khz difference between the channels. That's just too small of a shift to mitigate the various frequencies, harmonics, and random nature of EMI. Further to that, a frequency shift doesn't address the fundamental problem of noise when using SMF. That is, the exponential rise in noise due to SMF accumulating the EMI noise.

All single frequency, multifrequency, and PI designs use accumulation. No real difference there. The real difference is that a SF design is narrow band and MF and PI are wide band. A narrow band design allows you to more tightly filter the incoming signals which helps suppress EMI. Not only can you bandpass-filter the preamp, you can even bandpass-filter the RX coil. Can't do that with MF and PI, they have to run more or less wide open, although the preamp is usually lowpass-filtered at around 100kHz. That helps knock down radio interference.

Frequency shifting is really only effective for a single dominant EMI, and mostly for line noise (50/60 Hz). The problem is that the demodulators (part of the accumulators) can cause, say, 60Hz to "fold" into the baseband signal to where you can hear it. A slight frequency shift (maybe even a few Hz) causes this effect to become randomized so you no longer hear it. However, even though it's randomized it can still affect the detection and cause what's called "silent EMI."

 

 

[Digalicious]

38 minutes ago, Geotech said:

All single frequency, multifrequency, and PI designs use accumulation. No real difference there. The real difference is that a SF design is narrow band and MF and PI are wide band.

 

 

Thank you for the clarification Geotech.

I wasn't sure "accumulating" was the term the engineer used, hence I started with, "went something like this".

The EMI issue began for me with the use of two different SMF detectors, from two different companies. I soon discovered that it didn't matter if I was right near large capacity power towers, or in a local park...noise reduction never reduced the EMI noise at all. Further to that, in the countless videos I watched from various YT'ers with various SMF detectors, EMI noise reduction did nothing for them as well.

I then asked on a couple of Facebook pages and forums, if anyone could post a short video showing the noise reduction on their SMF detector actually reducing the noise. Well, it was crickets all around.

Around the same time, I discovered that SMF modes have a primary frequency. For example, the Legend's M1 primary frequency is around 15 khz. I would be in high EMI using M1 and the noise was intense, yet switching to the SF of 15 khz, caused most of the noise to disappear. That's when I realized that SMF is inherently more open to EMI than SF, despite SMF modes having a primary frequency. When I asked about that, an engineer either said it's because the frequencies and their associated EMI gets "accumulated", or he might have said "they get added together along with the EMI". I can't remember the exact term he used, but it was either "accumulated" or "added together". I'll try and find that post.

 

[Sinclair]

For me, the D2 handles cellphone EMI much better than the Nox800. In fact I don't get any anymore 😄

But I have at least 2 beaches, where the D2 is as bad or even worse than the Nox800. There's nothing special there - I think the EMI comes from the lamp poles on the promenade, because I only get that EMI at night. On the D2 only lowering sens / audio response helps a bit. Shifting / limiting the freq. or any other filter doesn't do anything really.. only switching to mono 17KHz helps significantly.

[UKD2User]

On 10/29/2023 at 7:52 PM, Geotech said:

"All single frequency, multifrequency, and PI designs use accumulation. "

I'd be fascinated to know how you know this, given the level of encryption protecting most embedded code in modern detectors - have you seen the source code or reverse-engineered one/several?

Yes, I'm sure that target responses at multiple frequencies are compared/correlated in various ways, but 'accumulation' (if it means anything very much) suggests simple addition.  We know that the D2 , for example, uses more than one kind of addition and/or subtraction in processing the return signal - there are almost certainly other transformations going on.

The great beauty of processing signals digitally is that you can do all sorts of things with them - in the time domain and/or the frequency domain.

[Geotech]

12 hours ago, Digalicious said:

Around the same time, I discovered that SMF modes have a primary frequency. For example, the Legend's M1 primary frequency is around 15 khz. I would be in high EMI using M1 and the noise was intense, yet switching to the SF of 15 khz, caused most of the noise to disappear. That's when I realized that SMF is inherently more open to EMI than SF, despite SMF modes having a primary frequency. When I asked about that, an engineer either said it's because the frequencies and their associated EMI gets "accumulated", or he might have said "they get added together along with the EMI". I can't remember the exact term he used, but it was either "accumulated" or "added together". I'll try and find that post.

Metal detectors typically use "integration" as a way to improve SNR. It's basically averaging which is a form of accumulation. It can be done in analog or digital.

With MF there have been two broad approaches to processing the multiple frequencies. One way is to treat each frequency as a separate single frequency metal detector. That is, each frequency channel is individually ground balanced and individually gives a target response. The results from each channel can then be compared/correlated to determine what the target is. This is what the DFX & V3 do. We'll call this "channelized" processing.

Another way is to combine certain signals from each channel to create "composite" signals. One composite signal is typically the ground balance signal, another the salt balance signal, and target information is extracted from the composite signals. The Fisher CZ series uses this method, as do BBS/FBS. Composite processing tends to offer only a single VDI and a single graphable result, whereas with channelized processing you can get multiple graphs and multiple VDIs. This is why the V3 can give you all 3 spectrums.

With channelized processing you can treat each channel as a single narrow-band detector and tightly filter it. With composite processing, once the composite signals have been created they now have a blend of frequencies with noise mixed from different bands and you've got what you've got. It's possible that this is what is going on in the newer all-digital MF designs. It would be interesting to see if, say, a DFX or V3 struggle as much. However, another factor is that designs seem to continuously push up gain and sensitivity so maybe a DFX would be quieter but also not go nearly as deep.

3 hours ago, UKD2User said:

I'd be fascinated to know how you know this, given the level of encryption protecting most embedded code in modern detectors - have you seen the source code or reverse-engineered one/several?

I design metal detectors for a living. No, I don't have any competitor's code but I can deduce some things from how the detector operates. I also look at competitor's patents which are often revealing.

[phrunt]

4 hours ago, UKD2User said:

I'd be fascinated to know how you know this, given the level of encryption protecting most embedded code in modern detectors - have you seen the source code or reverse-engineered one/several?

He even wrote the book on it 😛