Target ID with simple F.E.T analogue multiplier?

[dbanner]

A little circuit.

https://www.edn.com/analog-divider-uses-few-components/

[Skippy]

"A little circuit"
I've seen what I think is a fancier variant of that 555 circuit elsewhere ... probably in one of those massive Sheets/Graf 'Encyclopedia of electronic circuits' books. I'll try and find it, for interest....
Here is , pulled from Volume 3 of Graf:



I think the issue with these oscillator-based circuits is they are generally slow-responding. Good for DC work, but they all have plenty of low-pass filtering on the output to remove the AC component.

[Altra]

Here is another pre-micro ID circuit.

[moodz]

... you guys do realise that the phase response of a low pass filter is :




... so a simple phase detector ( AKA mixer ) and low pass filter
properly implemented calculates the arctan without all this falaffel about multipliers and dividers.

Reminds me of days sitting in the tea room listening to the new engineers discussing how to make products more complicated.

The main reason I am mentioning this though is that your opamp solutions will barely crack 60 dB of dynamic range ... whereas a good solution ( not using opamp trig operators ) will crack at least 90 dB and exceptional implementations will crack 120 db or more

Just sayin ... moodz.

[dbanner]

Moodz, you are way more advanced than I can hope of being.

On the other hand, give me about 6 months intense study with some hard weed, I think I might get there.

[KingJL]

On the other hand, give me about 6 months intense study with some hard weed, I think I might get there.

On the other hand, you might not even care!

[dbanner]

On the other hand, you might not even care!

Why, whatever do you mean?

Carl Sagan was a "wonderjunkie" astronomer and planetary scientist. It seemed to have worked well for him. He coined the phrase "wonderjunkie".

Curious he used the word "junkie". He might have well said "weedjunkie" instead.

I have no use for such dreadful herbs.

[dbanner]

A useful sim maybe for playing around.

[Skippy]

".. so a simple phase detector and low pass filter ... calculates the arctan"

If what we ultimately require is a TID ( analogue or digital ), then evaluating the angle using arctan isn't really that useful. It produces a value that's too squashed-up at the top and bottom end, and too stretched-out in the mid-range.
So working out a division of X and R has worth, if it's then followed by a "ratio-to-TID" generator.
In one of the "TID / phase angle / amplitude" threads, it was stated that the phase angle can have uses for further maths calculations, for example ground-tracking algorithms. So there are some uses for a true arctan-based calculation.

[dbanner]

i suppose that's why one elegant solution was to use log amps in the analog implementation of target ID.

• Each log amplifier:
• Converts the AC amplitude of the input signal into a logarithmically compressed DC voltage.
• This allows the circuit to handle signals from small, deep targets (microvolts) and large, shallow targets (volts) without saturation.

• The log-amped X and Y signals are combined in a summing amplifier.
• Output: A composite signal where the ratio of X/Y indicates target type.

• The summed signal is rectified to convert the AC component to DC.
• producing a DC voltage representing the logarithmically scaled, phase-discriminated target signal.

Why Logarithmic Amplification?

1. Dynamic Range Compression
   • Metal detectors encounter signals varying over 6+ orders of magnitude (e.g., a coin at 1cm vs. a soda can at 1m).
   • Log amps compress this into a linear voltage range (e.g., 0–5V).
2. Improved Target Discrimination
   • Ferrous and non-ferrous targets produce phase and amplitude differences in X/Y channels.
   • Log scaling makes these differences easier to compare.
3. Reduced Saturation Risk
   • Linear amplifiers would saturate on strong signals, masking weaker ones.

[moodz]

...you dont need log amps or rocket scientists or herb gardens.

Years ago log amps were essential items in things like spectrum analysers etc.
However since the advent of high resolution ADCs ... log amps are no longer used as the dynamic range of even mediocre ADCs exceeds the best log amps. ( except maybe at GHZ frequencies )
Eg a common as dirt 12 bit ADC can resolve better than 70 dB of dynamic range in a 3.3 volt span input signal.

However the real meaning of post #19 is that the so called "state of the art" use of IQ demodulators in metal detectors is actually wrong and a misapplication and actually renders the VLF detector to be less sensitive than it could be.

One of the fundamental reasons for this stems from the fact that IQ demodulation was invented to overcome the problem that radio receivers had where the phase of the transmit carrier ( and to some extent the frequency ) was not known.
To overcomes this deficiency a "fake" transmit carrier is generated at the reciever and mixed with the incoming RX signal to demodulate the signal and recover the modulation information ( audio, data etc ).
The use of a quadrature signal for the "fake" carrier or local oscillator ensures an each way bet on the demodulaiton so that information is not lost due to phase cancellation.

The use of arctan and root sum of squares to calculate magnitude and phase compounds the problem even further because the magnitude can never be zero ( thanks to the quadrature mixing ) and the arctan has a discombobulation at zero.

And here are the kickers that make the IQ strategy a less than opportune choice for metal detection.

1. The IB coil can never be perfectly balanced ... there is always a leak through carrier. ( and ground / targets upset this further ).
2. Because of point 1 the leak through carrier cannot be perfectly cancelled thus target signals ( phases and amplitudes ) are swamped by the residual carrier.
3. The IQ strategy was implemented because the transmit carrier phase was unknown in a radio context ... however in a metal detector the transmit phase is known ... so there is no need for quadrature demodulation.
4. The use of arctan and root sum of squares to calculate magnitude and phase compounds the problem even further because the magnitude can never be zero ( thanks to the quadrature mixing ) and the arctan has a discombobulation at zero.
5. Phase noise with low level signals.

Before they ( bell labs I think ) came up with IQ demodulation the industry used " homodyne " and " synchronous homodyne" receivers that used a single mixer but tried to lock onto the transmit carrier using phase locked loops or synchronous oscillators.
However what was old is new again .... "homodyne" detectors are back in use in quantum computer circuits because IQ detectors cant cut the mustard due to phase noise and problems I mentioned above.

What is forgotten today is that the homodyne reciever is actually superior to the IQ reciever ( if you can access the transmit carrier oscillator ).

In summary the IQ demodulation technique is the wrong solution for metal detectors .it works fine for relatively strong targets but lacks the sensitivity of PI for instance.

I have been able to show that in a single mixer ( homodyne ) type metal detector that the "leakage" carrier can be balanced out ( and the ground ) leaving only target phases and amplitudes and because a VLF detector demodulates the whole signal ( ie not just a point in time like the PI ) the sensitivity exceeds that of a PI detector.

It can also be shown that once the leakage carrier phase / amplitude is cancelled then the target signal phases and amplitudes are discernable beyond 100 db which translates as full depth discrimination of targets.

So you dont need to do polar math on X and R signals..... You only need to measure amplitude ( peak detector ) and phase. ( phase detector ) same as in a phase lock loop ( which dont use IQ demods. ).

moodz.

[Carl-NC]

I don't understand... VLF detectors since the 70s have pretty much all used homodyne receivers.

[moodz]

I don't understand... VLF detectors since the 70s have pretty much all used homodyne receivers.

Absolulely right ( about the homodyne type demods in previous detectors ) .... I didnt mention that because that is another kettle of fish.

I am having a poke at the IQ type demodulation scheme that is more needless than necessary.
IMHO it works but not as well as a homodyne.

Now getting back to the fish. The homodyne type mixers in previous detectors as you correctly point out Carl are OK but generally consist of a chopper switch ( mosfet / gate ) and some integration.

IMHO they have a defects ...

1. I talked about the importance of cancelling the leakthrough carrier. This leakage comes primarily ( in a well built detector ) from the imbalance of the IB coil.
A simple chopper switch scheme is maybe good for 30 db of carrier isolation on a good day ?? and I am probably being generous. Even a custom demod mixer chip may get 40 dB ...

2. Also a phase detector should not be sensistive to changes in amplitude of the signal .....

The leakage signal from the IB imbalance has a phase shift on it due to the path from TX to Rx etc.

Imagine a target signal that is -90dB at say 10 degrees. I need to get the carrier leakage to at least -96 dB @ 2.5 degrees. ( through cancellation ) ... the chopper switch scheme wont make the chop here literally !

Anyway bottom line is that the homodyne type detector I am talking about needs ...

1. Phase X is measured by precision mixer demodulator. ( and here we are using the mixer as a phase demodulator /detector ).

2. A precision phase shifter because the leakage carrier from the IB coil is not actually in phase ( close but not close enough ) with the TX signal.
The transmit phase correction is needed to phase cancel the leagage at the demod and it requires nanosecond type accuracy.
( eg for 100 db of phase comparision accuracy you are at the 100 to 10 nanosecond level )
At the precise balance point amplitude changes in the signal will cause no phase error. ( this is a homodyne detector feature )
... and this feature alone makes the ground balance a piece of cake.

3. Amplitude R is performed by a precision rectifier.

Using a precision 24 bit ADC and the FPGA I was able to implement both of these things.

I dunno if there is "cheaper" way to do it. The FPGA is a $10 dollar item ... but the ADC is pricey.

moodz

[dbanner]

Here is a critique by deepseek:

[Carl-NC]

I am having a poke at the IQ type demodulation scheme that is more needless than necessary.
IMHO it works but not as well as a homodyne.

OK, but IQ demodulation widely used is homodyne. Analog demods tend to use square waves which produces wideband demodulation, while the newer direct sampling designs use sine waves which produces narrowband demodulation. The null error ends up as a DC offset which can be cal'd out or just ignored. But even if you perfectly cal out the null error, ground loss angle can still induce a target phase error and it gets worse with depth.