Transmit amplification mod

[wirelessguy]

I am new to metal detecting; not new to electronics or magnetic fields.

What undesirable by-products will occur if you merely increase transmitter amplification for the goals of increased depth capability? Application is gold detection, particularly gold in quartz.

To an IB machine?
To a PI machine?

It seems simple to me to insert an amplifier to the tx coil even if I had to provide separate DC power for it. However, I see hardly any reference to people doing this, so I must be missing something. I understand the d^6 decrease in field strength, so a mere doubling of tx coil voltage will buy you nothing in terms of increased depth range. I believe PI monocoil systems would need a tx/rx switch and that adds a little complexity.

What am I missing?

[Davor]

Not much, just Rx dynamic range window, and a cart for lugging batteries.
While extra weight is a discipline practised by a market leader, and is in essence trivial, the Rx dynamic range is worth some extra attention.
Somewhat less obvious, but nevertheless there, are the configurations that break free from d^6 rule. These come at a cost of separate Tx and some exotic configurations, and are seldom discussed here.

So Rx then? You have noise on lower side, and saturation on the upper side.
In IB the Tx creeps in on both sides. You may expect Tx to be reduced by factor of 1000 by means of induction balance, so you have both residual Tx injection (air signal) and noise reduced by factor of 1000.

The noise in Tx consists of voltage regulator AM component, and 1/f (1/f² and 1/f³) noise of an active component. Because of a factor 1000 reduction even a noisy LT1086 with 150uV of noise in audio band (~1uV/sqrt(Hz)) will be OK even with a very quiet front end. Unfortunately there are no specifications of noise performance in 1/f region for these, but it seem as if there is no real problem there because pure AM noise is in phase with a carrier, and thus removed in GB circuitry.
1/f noise of an active component is a real culprit because it transfers to both PM and AM components, thus creating SSB noise which is in quadrature against carrier. In poorly designed oscillators it goes up to ~1mV/sqrt(Hz) around 10Hz, which is important for detection. Even when reduced by factor 1000 it is still waaaay too much. This noise can be significantly reduced by emitter degeneration, keeping the loop gain low, and high Q tanks. This noise can be kept very low by applying external signal source to a coil, provided that the source has a superior noise performance.

Although not obvious, PI suffers from exactly the same maladies, except that the phase noise is of negligible importance. There is avalanche noise, and various exotic principles that ruin input noise performance, and of course the Rx front end noise performance which can go down to ~1nV/sqrt(Hz). If everything else is done properly, only the battery-lugging cart maximum load is the limit for PI

[Davor]

I missed the THD influence as a limiting factor. THD stands for total harmonic distortion, which is the ratio of the RMS amplitude of higher harmonics against the RMS amplitude of the fundamental. The mechanism involved is that a harmonic, say 3rd, will spoil the amplitude in quadrature to the carrier, and thus gains free pass at the GB stage.
This influence is not straightforward as THD is a number representing multiple harmonics. Also, the influence depends on each harmonic's phase, but all in all, you may estimate that the noise floor is raised to, say 1/10 x THD x input air signal.

[Qiaozhi]

What undesirable by-products will occur if you merely increase transmitter amplification for the goals of increased depth capability? Application is gold detection, particularly gold in quartz.

To an IB machine?
To a PI machine?

Simply increasing the Tx power of an IB machine will more than likely overload the preamp, plus there are undesirable side-effects due to saturation of the ground matrix. Most IB detectors are already unable to run at maximum sensitivity, except in the most benign ground conditions, so increasing the TX power will achieve nothing apart from increased instability.

For the PI machine, increasing the TX power will also increase the coil decay time. But, for gold detection, you need to use early sampling in order to capture the eddy currents from the target before they disappear. Increasing the TX power will achieve the opposite effect for these small targets. There is little point anyway, as the target will already be fully stimulated.

[wirelessguy]

Davor,

Many thanks for the thorough reply. While reading your post I had a couple of "I shoulda thought of that" events. It's OK, happens to me all the time. Couple of thoughts/Qs below to insure I understand your points thoroughly.
DC Power. I understand the DC power required part and in my application I actually could tote a car battery.
Dyn Rng top end. I understand your points on Dynamic range in an IB MD. The saturation end seems very problematic, only resolved by improving / changing the MD receiver. More than I want to tackle.
Dyn Rng low end. I understand I'll be amplifying MD oscillator noise as well as MD amplifier signal. (Assume one would run a high power amp backed off adequately so the amp does not add to this problem.) What I'm wrestling with is a notion (a self-manufactured fantasy? ) that this noise stimuli on the target would not result in a significant eddy current. That is, on a strong return target, the noise will be so small relative to the signal such that the MD detection circuitry would process the return adequately. On a weak return signal (these deeper targets I am after!!!) the noise will get so d^6 diminished that it will not be seen by the receiver. Only the signal will make it back.

thanks again!

[wirelessguy]

Q,

Thanks for adding to Davor's points, particularly the increased coil decay time issue on PI MDs. I hadn't thought of that and am enough of a newbie that I was not aware that gold eddy currents are more rapidly settling.

As noted above to Davor, I get the overload of the preamp point. However, can you elaborate on "saturation of the ground matrix." I don't understand that phrase.

Many thanks for running a great website and forum. I've learned a ton from reading them. Your book is scheduled to show up at my door later this week. Maybe I won't ask these Qs after I read the book! ... More likely, I'll have many more.

[Qiaozhi]

However, can you elaborate on "saturation of the ground matrix." I don't understand that phrase.

It is a common misconception that simply increasing the TX power will naturally result in greater depth of detection. Although this may be the case in an air test, the reality is quite different when the coil is above the ground. There are often small ferrous particles in the soil that are usually too small to detect, because either the returned signal is miniscule when compared to a "good" target, or (for a PI) the eddy currents have died away before sampling takes place. If you increase TX power, the ground itself will respond more strongly, and in some cases can even become magnetically polarized. Effectively, you end up with a situation where you cannot see the wood for the trees (so to speak). It's as if the ground has become saturated, and it doesn't matter how much energy you pump into it, you will never go any deeper. Using massive TX power is not the answer.

Another analogy would be driving in foggy conditions. Trying to see further in the fog, by putting your headlights on full beam, only makes things worse.

[WM6]

Crude reality.

[Gwil]

Re post #5:

In an IB detector, AM noise in the the Tx signal coupled via the target to the Rx will, as you say, be negligible. However, unless coil nulling is absolutely perfect (which it never is), there will be coupling between the Tx and Rx coils without any target being present, and AM noise in this signal will go straight into the receiver.

[Davor]

... What I'm wrestling with is a notion (a self-manufactured fantasy? ) that this noise stimuli on the target would not result in a significant eddy current. That is, on a strong return target, the noise will be so small relative to the signal such that the MD detection circuitry would process the return adequately. On a weak return signal (these deeper targets I am after!!!) the noise will get so d^6 diminished that it will not be seen by the receiver. Only the signal will make it back.

While you are right about propagation of noise through ground, and its attenuation there, you are missing an elephant in the room. The noise creeps in by other mechanisms, and soil is usually not that much important there.
In IB something that is usually called "air signal" is a residual Tx signal that remains after balancing a coil. Let's assume it is at 10mV at input (1/1000 of full Tx swing), and it will be our reference.
First for THD contribution, we'll assume Tx is a pure sinus. With THD of 0.01% the RMS of various harmonics will be at 1uV. Let's assume 10dB lower the value of random dynamic contribution (noise) because harmonics are very lively (~300nV). Half of it resides in-phase, and half in quadrature (-3dB -> ~200nV). Useful bandwidth is ~25Hz, so sqrt(25)=5, hence 200/5=40nV/sqrt(Hz). Perfect match to a LF353 generation of opamps.
So to go below 50nV/sqrt(HZ) you need an op amp with better than 0.01% THD. With a modest NE5534 you get ~0.002% (8nV/sqrt(Hz)), and with NJM2068 you get 0.001% (4nV/sqrt(Hz)) and only in that case you get noise performance you paid for.

Now, by the same mechanism, and considering the Tx noise, but with no THD influence ...
Every Tx observed in frequency domain has a skirt of SSB noise that logarithmically falls by linear frequency. The slope depends on various mechanisms, but for all practical purposes, we may assume an average of, say -100dBc at around 10Hz from carrier, or 1/100 000 in voltage. So 10mV of air signal gives 100nV of noise that is equally distributed in-phase and quadrature, so the effective noise is ~71nV. For simplicity I can assume this noise equally distributed over 25Hz and it falls to ~14nV/sqrt(Hz). Because this noise is stronger at lower frequencies, motion detectors that reduce those make sense. It is also obvious that by only reducing 10dB of this noise at carrier, we reach very decent performance. By rising Tx voltage and keeping the same noise performance, the noise floor is simply dragged along, and you have zero performance boost.

Mechanisms in PI are somewhat different, but eventually it all gets there: the performance hit the wall where no single parameter can be "increased" to get a better result. Also the battlefield is more in the refinement of both Rx and Tx, rather than in increasing single parameters.

E.g. you increase gain, and you get a chatterbox of a machine, with exactly the same depth.

[wirelessguy]

Q (post 7), thanks for the concise and clear explanation. I get saturation of ground matrix now.

Gwil (post 9), thanks for turning on the lightbulb to another of my [duh!] moments. Now I am thinking about real world issues like imperfect coil balance and air coupling. and, you helped me better understand Davor's post 2 much more fully. His 1/1000 reference is a rule of thumb about what to expect on tx to rx air coupling in an IB!

How lucky I am for such a great forum!

[wirelessguy]

Davor (post 10), Thanks again for the time to provide an in-depth quantitative reply. You and Gwil got me to see an elephant in the room! ahhhh.

Your comments on THD noise have awakened some more Qs on MD receiver design in general. In an IB receiver, what is the typical front end filtering? Is there need to move corner frequency octaves above the fundamental tx freq? I do not understand why and there must be need hence your concern with THD.

Thanks for the input on phase noise. I am very familiar with the concept and its effect in higher order modulation communication systems. Hadn't thought about it in the MD context.

[Davor]

The beauty of MD electronics is that you have a full 3-course meal of everything you know about signal processing in easily accessible package that you may tackle even with a PC soundcard based oscilloscope. Bottom line is that you can see first hand what works.

Harmonics produced by harmonic distortion enter receiver chain by means of subharmonic mixing in switching mixers. These harmonics may assume just about any phase against the fundamental, which also may walk wildly and randomly. The most affecting are the components in quadrature, as they are confused for targets. So because the air signal amplitude is changed by some percent, a 2nd harmonic is changed twice as much in percent, 3rd three times etc. That makes them very lively. Because they may assume any arbitrary phase, the GB circuitry can't eliminate them, and you get a chatterbox. You can't fix these with filtering, but you can buy an opamp with lower THD.

Regarding front-end filtering, the strongest contributions worth eliminating are from mains, and from spark plugs. Also from other MD-s but that may bother people in detectoring contests, and perhaps when detectoring in groups. In any case, a high pass filter at front end will fix mains to some extent.
Now, mains may be of a bit more interest here, because it will inevitably be picked by our coils in urban environment or near powerlines. Because of the very nature of the switching mixers, some portion of mains voltage passes through. Single ended switchers pass mains directly, and a switcher acts merely as a chopper. Balanced mixers are much better in a sense that they significantly attenuate mains contribution.

I guess the spark plugs impulse noise can be reduced by some noise blanker, but so far I haven't seen any MD with it.

[wirelessguy]

Thanks Davor. I follow the subharmonic mixing in active mixers thought and the mains and spark plug noise filtering.

Great that this site exists for audio soundcard DSP geeks. We need more of 'em to leave the crowded space of more and more compact voice and music coding and move on over to metal detection!

[Sean_Goddard]

Possibly why Fisher used a high purity oscillator in a lot of their machines?

As regards DSP (mis)used in metal detectors one word..CORRELATION! There is not a SINGLE company out there who have truly exploited what DSP can offer in a metal detector and I think Aziz is the only one on this forum who knows and he doesn't let on much ;-)