Phase delay for different targets

[mikebg]

Pebe,
Thank you for answering my posts, this is very nice, but you lose precious time to argue rather than use it to improve or to create something. For the participants of this forum is interesting to obtain useful information, to learn things that they do not know and to apply them in practice.
Here are examples:
From my post # 12 they learned that the balance of search head can easily be improved if using two cores: ferrite and aluminium piece fixed on coil housing.
From your post # 13 they learned that a better balance of the search head to be done with two potentiometers and complex electronic circuit, with two cores instead, because this is "pretty crude".
I think that these two postings are useful because they carry information.

However, it is useless a terminology that is not inherent of radio technology.
My hobby is to design RXs, TXs and antennas for QRP amateur radio. Despite I'm not professional designer, I know how should be designed a sensitive metal detector because it contains RX and TX with their antennas. A metal detector is radio equipment that can operate in LF, VLF or ELF frequency band. Audible frequencies are in ELF and VLF band.
As ham designer I can see that metal detectors are not designed properly:

The transmitter of conventional MD is incompetent designed because:
- It generates very low power,
- Operates at low efficiency wasting battery energy to heat transistors and emitter resistors,
- TX antenna is not impedance matched and
- There is unwanted amplitude and angle modulation caused by changing the distance TX coil - earth.

The receiver of conventional MD is incompetent designed because:
- Sensitivity is not limited by thermal noise, but by GND & AIR signal.
- Received is very large frequency band despite target signal needs a narrow band below 16Hz,
- No AGC (Automatic Gain Control) as in every conventional radio receiver, - No ABC (Automatic Balance Control) to compensate GND & AIR signal in RX input,
- Used are primitive nonsymmetric sinchronous demodulators, that can't suppress noise,
- As reference for demodulation is used voltage across TX coil, instead the received GND & AIR signal.

Pebe, You're very helpful on this forum with your experience and professional knowledge. Help us to solve these problems. Do you see any other design problems in known circuit diagrams?

[simonbaker]

Pebe,
- Used are primitive nonsymmetric sinchronous demodulators, that can't suppress noise,
- As reference for demodulation is used voltage across TX coil, instead the received GND & AIR signal.

Hi mikebg:

Would you explain some more what you saying there?

Regards,

-SB

[mikebg]

Hi mikebg:

Would you explain some more what you saying there?

Simon, it is good that you have only two questions.
Here are the answers:

1. As shown below, the primitive synchronous demodulator uses a simple electronic switch S and integrates only half period of signal (operates as halfwave rectifier). Note that in this case the capacitor C is charged by noise, ie the demodulated signal depends on interference.
To suppress noise, should be used "change-over" electronic switch and difference integrator which subtracts interference. A widely used circuit diagram with CD4053 is shown below.

2. To extract a faint signal masked by large interference, we should use the interference as reference voltage for synchronous demodulation. Then demodulator eliminates interference because operates as "lockin amplifier" or "synchronous correlator". Conventional metal detectors also use GND signal for reference, but this reference is irreal because is not received by RX coil. It is created artificial by using phase shifted voltage across TX coil. This is bad design solution because when tuned circuit of RX coil changes the phase of received signal, the phase of artificial reference remains the same. In this case, to increase phase stability of RX, we should tune RX tank to resonance frequency different than radiated by TX. This is made in search heads of White's; RX tank is tuned to lower than TX frequency.

[pebe]

Mike

Pebe,
Thank you for answering my posts, this is very nice, but you lose precious time to argue rather than use it to improve or to create something…….

I don’t intend to argue and I’m sorry if that’s the impression I give. But when someone makes a statement that I don’t agree with or don't understand, I challenge it. Not to score points, but to try to better understand the point that is being made. More often that not, I’ll learn something from it in the process, as I hope others do.

I don’t quite know what you mean by

However, it is useless a terminology that is not inherent of radio technology.

I agree with you on the need to improve detectors. A lot seems to be being done already. Simon and dflowers are doing excellent work on improving sensitivity, and Tinkerer is working on getting PI to get discrimination with good depth.

It seems that what is required is higher Tx power and/or increased Rx sensitivity. I don’t think we need to have permanently oscillating Tx coils, and Tinkerer has already given me a good idea to get a high powered pulse that will ‘ring’ a tuned circuit long enough to measure the phase of the Rx signal, and I want to get working on that. The other possibility to increase sensitivity is controlled regeneration on the Rx amplifier. Regeneration (reaction) was used on early TRF receivers to improve sensitivity and was very effective. But of course, when the gain goes up so does the noise but by using a microcontroller to measure phase and integrate readings over many cycles, then phase jitter due to noise could be eliminated.

I think that is the way to go and I hope to ‘have a go’.

[simonbaker]

Simon, it is good that you have only two questions.
Here are the answers:

1. As shown below, the primitive synchronous demodulator uses a simple electronic switch S and integrates only half period of signal (operates as halfwave rectifier). Note that in this case the capacitor C is charged by noise, ie the demodulated signal depends on interference.
To suppress noise, should be used "change-over" electronic switch and difference integrator which subtracts interference. A widely used circuit diagram with CD4053 is shown below.

2. To extract a faint signal masked by large interference, we should use the interference as reference voltage for synchronous demodulation. Then demodulator eliminates interference because operates as "lockin amplifier" or "synchronous correlator". Conventional metal detectors also use GND signal for reference, but this reference is irreal because is not received by RX coil. It is created artificial by using phase shifted voltage across TX coil. This is bad design solution because when tuned circuit of RX coil changes the phase of received signal, the phase of artificial reference remains the same. In this case, to increase phase stability of RX, we should tune RX tank to resonance frequency different than radiated by TX. This is made in search heads of White's; RX tank is tuned to lower than TX frequency.

Thanks! I'm not sure I get all of your points, maybe terminology not clear to me.

1. Good point that the TGSL synchronous detector (SD) only uses half of the sync pulse. Theoretically you are throwing away some noise reduction, I think maybe 1 / sqrt(2). But otherwise, the SD seems well designed to me. Is this your point, showing a circuit to use both halves of the sync pulse?

BTW I have wondered about ways to use the other half of the sync pulse without adding too many extra parts. Most ideas I have played with end up adding parts that disrupt the nice simplicity of the TGSL, and may add more noise of their own, so is it practical? Always worth experimenting with though to find out! So good point to mention.

2. Sorry, still don't quite get what you are saying there in your second point. Can you explain with examples?


Regards,

-SB

[simonbaker]

Mike
I don’t intend to argue and I’m sorry if that’s the impression I give. But when someone makes a statement that I don’t agree with or don't understand, I challenge it. Not to score points, but to try to better understand the point that is being made. More often that not, I’ll learn something from it in the process, as I hope others do.

I don’t quite know what you mean by

I agree with you on the need to improve detectors. A lot seems to be being done already. Simon and dflowers are doing excellent work on improving sensitivity, and Tinkerer is working on getting PI to get discrimination with good depth.

It seems that what is required is higher Tx power and/or increased Rx sensitivity. I don’t think we need to have permanently oscillating Tx coils, and Tinkerer has already given me a good idea to get a high powered pulse that will ‘ring’ a tuned circuit long enough to measure the phase of the Rx signal, and I want to get working on that. The other possibility to increase sensitivity is controlled regeneration on the Rx amplifier. Regeneration (reaction) was used on early TRF receivers to improve sensitivity and was very effective. But of course, when the gain goes up so does the noise but by using a microcontroller to measure phase and integrate readings over many cycles, then phase jitter due to noise could be eliminated.

I think that is the way to go and I hope to ‘have a go’.

We'll look forward to whatever you try out and any results you report. Always useful information.

If we can ever find a place where EMI noise is absent and ground is perfectly smooth, then we can use that extra gain.

Regards,

-SB

[mikebg]

NORMALIZED TRANSFER IMPEDANCE

The transfer impedance of a nonferrous target is shown in inpedance plane above in posting #6 . It can be shown in Cartesian (rectangular) coordinates as two frequency functions; Re (real) and Im (imaginery), but designers prefer two functions in polar coordinates: Magnitude and Phase.

Below are given these polar functions in normalized form. Note that the phase angle is measured relative to TX current, not to voltage across TX coil. If you wish to express phase lag relative to voltage across self-inductance of TX coil, you should add 90 deg.

The frequency in this diagram is normalized relative to cutoff frequency of target fc. At fc, the target signal induced in RX coil has phase lag 45 deg relative to TX current, but amplitude drop differs from 3 dB because transfer impedance is formed by several timeconstants. Despite this, the cutoff frequency is called "3dB frequency" as in a first order network.

I measured cutoff frequencies fc of euro coins by a primitive method giving not exact values. The values are approximately:
€1 ____ fc=5600Hz
50c ___ fc=3600Hz
20c ___ fc=6200Hz
10c ___ fc=6400Hz

[mikebg]

Thanks! I'm not sure I get all of your points, maybe terminology not clear to me.

1. Good point that the TGSL synchronous detector (SD) only uses half of the sync pulse. Theoretically you are throwing away some noise reduction, I think maybe 1 / sqrt(2). But otherwise, the SD seems well designed to me. Is this your point, showing a circuit to use both halves of the sync pulse?

BTW I have wondered about ways to use the other half of the sync pulse without adding too many extra parts. Most ideas I have played with end up adding parts that disrupt the nice simplicity of the TGSL, and may add more noise of their own, so is it practical? Always worth experimenting with though to find out! So good point to mention.

2. Sorry, still don't quite get what you are saying there in your second point. Can you explain with examples?

Hi Simon,
1. It is not correct to say that the primitive SD uses half of sync pulse (reference voltage). The sinc only closes and opens the switch. The primitive SD uses the information of only half period of useful signal when the switch S is closed by sync. To use the information containing in the other half of period, we need an additional switch S2, closed when the first one is open and vice versa, ie we need a changeover switch for synchronous demodulation and complicate signal processing. The primitive SD needs an integrator (R and C). The changeover switch needs two integrators and subtraction of their signals. The shown above circuit diagram with changeover switch is used by White's in TM808, Coinmaster, Eagle spectrum.
TGSL is an "el cheapo" design. The prices of electronic components are now so reduced, that to design "el cheapo" is nonsense for competition in the market. This is valid also for builders (homebrew amateurs).
2. For example I will use the circuit diagram shown above in posting #13 (because Pebe has no copyrights:-). This circuitry must generate for synchronous demodulation two signals, let call them RX (received by RX coil) and REF (reference voltage). The RX signal is combination of AIR, GND and TGT (target) signals. The TGT is very small relative to AIR and GND. The gain of amplifier is limited by AIR and GND because the RFA can not operate saturated.
Will be continued in the next my postings because I should modify the drawing in #13.

[simonbaker]

Hi Simon,
1. It is not correct to say that the primitive SD uses half of sync pulse (reference voltage).

I didn't use good terminology I agree -- I meant same as you are saying, that we're only using half of the RX signal information for detection.

I'm all for trying new circuits -- after I feel I have milked the most performance out of the simplest circuits I can find. I like to start simple and add complexity only as needed; which means thoroughly investigating how well the simplest circuits can be made to work.

It would be interesting to see if a "full-width" SD would make a measurable improvement that we could see in real tests.

Of course this stuff could be done so much more rigorously in the digital domain, but it seems more fun trying to coax analog parts to do the job.

The changeover switch needs two integrators and subtraction of their signals...

Does the circuit you showed actually use one integrator for both?

Are you sure there is subtraction of the signals involved?

Cheers,

-SB

[mikebg]

Does the circuit you showed actually use one integrator for both?

Are you sure there is subtraction of the signals involved?

Simon, the attached circuit contains 4 integrating capacitors x 1nF. Note that they are connected to inverting and noninverting inputs of opamps.

[Tinkerer]

SIMULATION PROBLEM

Hi Mikebg,

This is going back to the VLF TX posts above.

I am working on a hybrid VLF-PI detector.

In the simulation attached, I tried to produce a high power sine wave. At a certain point, with the TX ON and OFF times at the proportion shown, the circuit gets resonant and produces an extremely powerful sine wave.

This only happens with IGBT's, every one of the different simulation models of the MULTISIM software.

The same circuit with Mosfets does not get resonant.

So I built a PCB to test the circuit in real life.

The results are totally different.

Interestingly, I do get very good sensitivity with the attached coil of 450mm diameter. I also get FE discrimination.

Could you help me understanding the problem?

Tinkerer

[Tinkerer]

Here are the Zip file and a picture

Tinkerer

[simonbaker]

Simon, the attached circuit contains 4 integrating capacitors x 1nF. Note that they are connected to inverting and noninverting inputs of opamps.

I'm not sure where the subtraction is. The circuit looks like a quadrature-based SD with two separate channels that would normally run at 90 deg phase difference. Correct me if I'm wrong.

Even though each channel has inverting and non-inverting inputs, the purpose I think it to add, not subtract, the demodulated signal from the two half-cycles. The inverting input serves to "rectify" the other half of the cycle so that it is analogous to a "lock-in amplifier" which multiplies by a sine wave, causing two in-phase signals to make a positive output because the two negative halves multiply to be positive also.

I'm sure we're probably saying the same thing, just a different interpretation of the terminology "subtract".

Regards,

-SB

[mikebg]

Simon, with the below circuit I will try to explain why tuned RX coil creates phase problem in conventional metal detectors.
The conventinal blok diagram uses voltage across TX coil to create by phase shifting one or more reference signals for synchronous demodulators and for compensationn (balance) of AIR and GND signal in RX front end as shown. However the AIR and GND signals are generated by TX current because it creates magnetic field. When we use a ferrite core and a piece aluminium foil for balance, the phase problem of tuned RX coil is solved correct and easy. Remains to use the same principle to obtain reference voltages for synchronous demodulation. Vaino Ronka used for this purpose additional RX channel and automatic balance.

The voltage across self-inductance is in exatly 90 deg phase lead to TX current, however we can not connect to it two wires for reference:-(
The voltage across TX coil is different from voltage across its self-inductance L because contains in addition voltage drop across coil resistance r and imported by ground impedance Z. Note that imported impedance is variable because TX coil moves relative to "earth core".
The same happens with RX coil because near it is moves te same earth core. The earth imports an impedance Z which changes when we alter the height of search coil and when soil properties alter.

But let we see how arises the problem of phase instability in RX coil tank.
To detect deep targets, we should amplify TGT (target) signal much more times than in a conventional metal detector. For high gain, in this circuit are used two stages for amplification of received frequency. The first stage is low noise preamp and preselector. The TGT signal exists in too narrow bandwidth - no more than 16Hz. When front end receives narrow band of frequencies, it receives less interferences and generates less thermal noise. To narrow the band we simply tune capacitance of C1 for resonance to radiated by TX frequency. However, the so tuned resonance of tank circuit shifts the phase of induced EMV by a very steep phase characteristic. Let SPICE display the phase and amplitude characteristic of tuned circuit.

[mikebg]

I'm not sure where the subtraction is.
-SB

Simon, subtraction means difference. For example, the difference between 5 and 3 is 2, ie we subtract 3 from 5 to find the result.
I searched web for the demodulating circuit diagram of White's and for educational explanation of device "difference integrator". Google found several books where this term may be is explained comprehensive, however because of copyrights they are not readable online.

Since I'm not familiar with English terminology in electronics, I used the correct translated term "difference" for subtracting procedure. However, as often occurs in English literature, instead "difference" or subtraction is used the incorrect term "differential". So the name of above circuit is not only "difference integrator", but also "differential integrator". Translated in other European languages, the term "differential integrator" means the nonsense "derivating integrator". When we integrate a function and then differentiate the result, the function remains unchanged, ie when "differential integrator" operates, it makes nothing.
I found the circuit diagram shown above in Application note AN1515 of National:
http://www.national.com/an/AN/AN-1515.pdf
Here is the circuit with formula that shows that is calculated integral of the subtraction: