Design of BFO metal detectors

[mikebg]

How seems the block diagram of most sensitive discriminating BFO metal detector?
I posted this information scattered in several threads of the forum. However ham builders keep to use primitive BFO circuits. It is time to start this thread for more completed explanation of design process for BFO type MD and to make,
REVISION AND REDESIGN
of several arbitrary BFO MD circuits.

The design of an electronic appliance should start with analysis of most suitable block diagram. We need the most suitable block diagram also for revision and redesign of an existing project. Below is attached the most suitable block diagram for BFO type metal detectors, reinvented and renamed by (R)EMI group. The "new" name of BFO MD is REGEN or REGENERATIVE metal detector because transmitter and receiver are connected in postive feedback loop forming SEARCH oscillator. Google for "regen receiver" to understand why this name was very popular in 20-ties and 30-ties of 20th century.

Here is the explanation how most sensitive search oscillator works:
TX coil (L1) and amplifier (blok 1) form a transmitter for noise induction metal detector. If you can't believe this, please connect a loudspeaker instead TX coil. You will hear random NOISE because there is no regenerative feedback. To understand how an oscillator works, connect also a microfone instead RX coil. Now you have an audio locator because it can locate different objects which reflect audio waves. The audio locator can be adjusted to operate in three different modes:
A. There is always oscillation. The object causes frequency shift only.
B. There is oscillation without object. The object ceases oscillation.
C. The object causes start of oscillation.

If you have an audio amplifier with gain control and tone control, make the mentioned locator and play to understand that an oscillator is very noisy regenerative amplifier tuned to a frequency. The tuned circuit increases regenerative amplificitation of this frequency only.

[Davor]

It seem to me that noise alone is not the perfect exciter. In case the oscillations are quenched as in super regenerative rigs you could have the best of both worlds - strong excitation prior to the quenching pulse, followed by maximum undiluted amplification. That could do.
Using a PLL as a reference, you could as well have some kind of ground balance. This becomes even more interesting.

BFO has enormous gain, there is no doubt about it. However, it has enormous 1/f phase noise as well. Marrying it with regen would grant it incredible Q, and also superb gain of the faint eddy currents. Yes, I think it is feasible, and it could work well too.

[HECTOR]

I was wondering if anyone else has thought about using a stereo audio signal to help differentiate phase characteristcs? (headphones only)

I'd love to see a well designed BFO pinpointer created here.

[mikebg]

Please try this to understand how an oscillator works.

[Davor]

This part is clearly understandable, but there are some other effects that are not properly encountered by this model. The first is an unbalanced portion of a Tx signal (that's the direct path between a speaker and a mic). Next is a ground proximity signal (the reflector in a drawing). This model is non motion compensated so the "stationary" effects dominate the system. In case you reduce system sensitivity so that these do not trigger oscillation, you end up with quite a deaf rig.

IMHO the super regen approach is providing superior gain, and some level of AGC. Besides, it allows for spontaneous oscillation due to any effect there is. It's transfer characteristic is log amp like.

[Davor]

I gave this a further thought and realised that the most of the nowadays concepts are actually deeply rooted in BFOs and their shortcomings. Actually mikebg gave me a nudge to express it in my head, so I have to dispose of it here.

First off, a BFO in its nature is an equivalent of a Disc channel of a modern VLF. It reacts mostly on ferrites, and to good conducting materials such as copper, silver etc. It is mostly deaf to the foil, small gold and such low conductors. To remedy that the most of the BFOs were working on ever higher frequencies, and such rigs were able to sniff some small gold too. I've seen some very elaborate analysis of BFO behaviour with formulas and stuff, but I think there is a much easier way of presenting how and why the BFO actually works.

So, what we know is that BFO's are most sensitive to materials that have eddy current vectors slanted away from the carrier, and more so for responses closer to 90° away from the carrier. Ferrites drag the frequency to one side, while copper and such drags it to the other.

As frequency and phase are mutually related things, one may say that ferrite response vector is dragging a carrier vector to one side, and in a process it may do circles as well, hence frequency change. So far so good.

There are also ideas that the frequency change is due to the coil inductance change, however, that idea may work only for ferrite targets, not so much for the other materials. Since it works anyway, it is obvious that BFO's actually react to the targets eddy currents. I have a solution as why it happens which may shed some extra light on BFOs mechanisms.

There is a method for generating perfectly phase modulated signal according to Armstrong. It employs double sideband suppressed carrier modulation of a payload signal with a carrier frequency, and phase shift for 90° against the carrier. The carrier is added to this signal, such compound signal is clipped and amplified, and voila, you have a HiFi phase modulated signal. Seek Armstrong modulator in google for more details.

So how is Armstrong related to BFOs? Easy. Say, we have a copper coin in the ground. I pass over it by a BFO, and there are eddy currents in it at some steep angle, say 80°. This signal is picked by a BFO coil as a phase shifted one, added to the carrier, and clipped to the full oscillator swing. What also happens is that zero crossing, or whatever else crossing is used by the oscillator to keep it in action, gets phase shifted too, so that the phase change is propagated to the next cycle and so on.

Left to its own devices such BFO will be prone to be triggered by ferrites in the ground, and also by well conducting materials such as copper. It will not react much with the materials that do not skew the phase against the carrier by much, e.g. small gold.

The above mikebg's diagram actually enables us to make a real discriminating detector with much better sensitivity for small gold, and in the same time to null the sensitivity to ferrites or the ground, by simply adding a phase shift control between Rx and Tx coils. In case this rig has Rx and Tx coils signals phase shifted or delayed for the phase shift of 90° such rig would completely lose sensitivity for ferrites, same as ground balance with VLFs does. It would also gain sensitivity for small gold even at much lower frequencies.

Yeah, it makes sense

[HECTOR]

Could we have a Tx search coil producing wide spectrum "pink noise" and dual Rx circuits tied to one coil? The Rx could have active feedback peak seeking, phase detecting channels connected to a stereo audio out put that would accentuate phase differences with tone/delay. The peak seeking circuits are common in audio feedback suppression. Just some thoughts.

[Overtheedge]

Having built gennys and super-gennys back in the early 60's using triodes, I'm familiar with both. The super-genny was called a rushbox for a good reason. The noise is tiresome to flat irritating.

The triode regenerative provides an oscillator (feedback) coil and a receiving coil while the super-genny does not.

As pointed out, maximum sensitivity is right at the oscillation point which is a putt-putt sound in the headphones. This would be the ideal operating point. Any increase in regeneration due to phase relationship would throw her into full oscillation, while a decrease would drop her out of the putt-put point. I wonder about the coil shielding. I recall that my hand would affect the regeneration; sort of a proximity detector effect. Even the vari-cap regen control needed a big bakelite knob to reduce the effect.

In the triode gennys I built, the regeneration control was a vari-cap in the plate circuit to ground. The B+ line then had an RFC to keep the RF in the genny circuit, but permit the AF to go to the audio transformer. I've never tried it, but at the regen vari-cap point, one could use a gimmick capacitor to tap off some of the RF and feed it to a mixer with the stable oscillator. Shielding the stable oscillator would be required due to the regen having a problem with oscillator pulling on stronger signals.

No sense in using a PLL unless you are going to try and lock it to the regen oscillator. In that case, you will need to figure some longer time constant to permit plenty of over-shoot.

Interesting idea. I hadn't thought of re-purposing a regenerative receiver for full-blown MD.

Just remembered that Velleman makes a stud-finder kit that works on this principle: K7102

BTW it is wide-band noise generation that starts the oscillation. The tuned circuit is the filter that selectively passes or rejects the signal depending on series or parallel circuit.
eric

[Davor]

There is more to the super regenerative rigs than meets the eye. They have their shortcomings, but most of them are curable. Oscillator (antenna) loading can be cured by a proper preamp, odd self quenching is fixed by an external quenching oscillator, and odd demodulation is fixed by a proper envelope detector. Even with all these perks it will remain a simple rig.
I think the noise problem of the super regenerative Rx will not be encountered in metal detecting simply because there is always going to be some level of feedback with the Tx, and noise in super regens drops in presence of signal, even if very small. Mikebg's suggested configuration pushes it to the next level by phase shifting Rx against Tx. I think it is a brilliant idea.
When I make time I'll try making a LTspice simulation of a super regen with preamp and external quenching.

[peterr88]

I'd like to see a well designed BFO pinpointer created here.

[mikebg]

I'd like to see a well designed BFO pinpointer created here.

The regenerative principle with induction balanced search head is suitable not only for pinpointers. We can design a search oscillator which eliminates shallow targets and starts to oscillate when there is deep target buried in bad ground. For this purpose we should make analysis of regenerative block diagram:

http://www.geotech1.com/forums/showt...970#post109970

It is attached below for convenience. Note that the block 2 is simlifyed. It contains mutual inductance, ground and target. We should expand it in several blocks to show the convolution of genuine target signal made by ground.

What means CONVOLUTION?

CONVOLUTION is a Latine word used to explain what happens in Time domain. This term is widely used for analysis of signals and systems. The convolution is a complex integral operation. The father of this integral is Jean-Marie Constant Duhamel (1797 – 1872). Search the WEB for "integral of Duhamel" or "Duhamel's integral".

For visual illustration of convolution process, make an experiment. Take a white coin and a yellow coin and put them in a glass red wine. The wine convolutes the images of coins and you can not say what coin is white and what is yellow. If the wine is enough "bad" you can't say if there is a coin in the glass. The solution is to take a digital picture of the glass and to use sofware for deconvolution. It can show the genuine colors of coins.

The advantage of Frequency domain is that it transforms the convolution in a simple math operation - MULTIPLICATION. Note that Barkhausen's criterium is made in Frequency domain. We use multiplication of transfer functions of blocks:
http://www.geotech1.com/forums/showt...667#post106667

[deemon]

I performed such "regenerative" experiments with a balanced search head many years ago , and found that it's a kind of "dead-end way" ... idea looks beautiful , but completely useless in reality Too many factors of instability make the operation almost unpredictable

More interesting seems to be another way - to make a classic LC resonant oscillator , oscillating with its own positive feedback , and add another feedback loop with more amplification , going through a balanced search head - in another words , it must be a multipath feedback oscillator . Thus we'll obtain a kind of "boosted BFO" detector , working like a classic BFO , but with increased sensitivity .

[Davor]

It could work. Main theme with all metal detectors is extracting all metal response due to the eddy currents, which is at angle against the Tx. In case balanced coils are used in feedback arrangement, it is quite possible to emphasize the eddy current response, as opposed to the self-regenerative oscillation that would emphasize ground response.

Off resonance detectors do that in a way, but with balanced coils this could become more sensitive.

[mikebg]

... Note that the block 2 is simlifyed. It contains mutual inductance, ground and target. We should expand it in several blocks to show the convolution of genuine target signal made by ground.

The block 2 shown in post #11 is represented below in expanded form with several blocks.
The designations in block diagram are for Frequency domain to avoid CONVOLUTION.
There are 3 signall paths:
1. AIR signal. We obain pure AIR signal when the search head is lifted high in the air, far from ground. I think the term "AIR signal" is involved by designers of Coinmaster Pro, but I'm not sure.
The mutual inductance (represented here as block number 2) operates as an ideal derivator (differentiator). That means the AIR signal has phase led 90 deg for all frequencies. We can minimize the AIR signal with induction balance between RX and TX coil and compensate it with two controls (see block 8 in post #11).

2. TGT signal. It contains information for impulse response function of target h5(t) but convoluted with response functions of ground h4(t) and h6(t). This happens in time domain. To simplify analysis, we will use Frequency domain where the CONVOLUTION is transformed in MULTIPLICATION. That means TGT signal is product of frequency transfer functions H4. H5. H6 (point designates multiplication).

3. GND signal. This signal is suppressed (d) when we use induction balanced coil configuration and (b) when are used two RX coils connected as TWIN LOOP configuration.

Remains to analyse with Barkhausen criterium when the search regenerator will start to oscillate if there is deep buried target and how to design the target to make large change of oscillation frequency.

[Qiaozhi]

The block 2 shown in post #11 is represented below in expanded form with several blocks.
The designations in block diagram are for Frequency domain to avoid CONVOLUTION.
There are 3 signall paths:
1. AIR signal. We obain pure AIR signal when the search head is lifted high in the air, far from ground. I think the term "AIR signal" is involved by designers of Coinmaster Pro, but I'm not sure.
The mutual inductance (represented here as block number 2) operates as an ideal derivator (differentiator). That means the AIR signal has phase led 90 deg for all frequencies. We can minimize the AIR signal with induction balance between RX and TX coil and compensate it with two controls (see block 8 in post #11).

2. TGT signal. It contains information for impulse response function of target h5(t) but convoluted with response functions of ground h4(t) and h6(t). This happens in time domain. To simplify analysis, we will use Frequency domain where the CONVOLUTION is transformed in MULTIPLICATION. That means TGT signal is product of frequency transfer functions H4. H5. H6 (point designates multiplication).

3. GND signal. This signal is suppressed (d) when we use induction balanced coil configuration and (b) when are used two RX coils connected as TWIN LOOP configuration.

Remains to analyse with Barkhausen criterium when the search regenerator will start to oscillate if there is deep buried target and how to design the target to make large change of oscillation frequency.

Can you start to move forward onto the schematics, and leave block diagrams behind?
Otherwise you'll always be getting ready to get ready, and never build anything to prove/disprove the theory.