Tweet
Teleno ... I ran your temperture profile on the model. Top trace is Input ( 3 temps x 4 input levels ). Bottom trace is output. Looks pretty stable.
-
-
... putting it all together we have the functional loop for the complete detector. Showing TX waveform and RX lock waveform.
I have added in a 3rd order bandpass filter. The operation frequency is 25 Khz. The waveform is true bipolar.
You will be able to see how this can have a modular front end with channels. ( Pi or VLF ) though PI is not shown in this scheme.
Comment
-
What is the transient lock time for the RX?Comment
-
Good Question ( not the answer to the challenge question though ) ... the lock time for a transient phase burst say the coil passing over a target is proportional to the bandwidth or "lock range" of the circuit.Originally posted by Carl-NC View Post
The lock range or BW is proportional to the incoming signal strength as the signal gets stronger the lock range gets wider and faster.
With weak signals the bandwidth is only 50 hertz or so the lock speed is proportional to this eg 20 milliseconds. The amount of noise present will also influence this ( as phase and amplitude noise in the recovered signal ).
My measurements indicate a lock speed of around 40 milliseconds for a transient signal of 1 millivolt in 1 volt of mathematically generated true white noise.
I would not expect to see this type of signal in the field.
if there are strong CW waveforms near the wanted tone then all bets are off of course for such a simple cct.
Comment
-
Whether it was the answer to the challenge question or not depended on your answer.Originally posted by moodz View Post
One thing I would look closely at is a multi-target environment, namely a mineralized soil matrix combined with a target. In this case, you will have a nominal locked signal (to the soil) that then needs to get pulled slightly by a transient target. I would assume you then I/Q demodulate that signal, apply motion filters, and extract the target amplitude/phase. It seems (at a glance) that it should work, but I've seen too many TGTBT (too good to be true) circuits that look great on the bench, and fail utterly on real ground.Comment
-
I will freely admit I have never built a VLF detector before ... so I am just working on what I see as the problems to be solved and trying to implement methods that might produce a good result.Originally posted by Carl-NC View Post
Whether it was the answer to the challenge question or not depended on your answer.
One thing I would look closely at is a multi-target environment, namely a mineralized soil matrix combined with a target. In this case, you will have a nominal locked signal (to the soil) that then needs to get pulled slightly by a transient target. I would assume you then I/Q demodulate that signal, apply motion filters, and extract the target amplitude/phase. It seems (at a glance) that it should work, but I've seen too many TGTBT (too good to be true) circuits that look great on the bench, and fail utterly on real ground.
So if i was just using an amplifier in front of the DSP I would have to scale and subtract the ground / soil signal. Because the output is an approximate log to do the scaling I do A + G where scaling would be a x g ( amplitude by gain ) normally ... A is now log(a) and G is log ( a constant ) = another constant. To cut along story short the log of the input places all the targets and soil in the same range ( eg 400 mV to 3.3 volts ) which suits the ADC and addition an subtraction of log values is the same as multiplication and division of the unlogged values.
Bottom line it is now really easy to calculate the unwanted phase / amplitude reference and feed it back from the DSP via a DAC and subtract it from the input to the PHLOCKER ( phase log locker cct .. trademark
) say using a prior opamp.
This leaves only wanted target phases of the signal. Of course to ground track this is a continuous process.
Doing some roughed in calculations and models the cancellation only has to be better than 75% to get a good lock on wanted target phases. The DSP can do considerably better than that.
Comment
-
-
It looks like super regenerative receiver, what is used in modern time and show good resultsComment
-
May a challenge problem will be a lowest limit of signal, on coil we
have microvolts of interest and need preamplifier.Comment
-
You may be thinking of Pulse Induction. In VLF the transmit signal is "on" all the time ( sine wave ).Originally posted by solv View Post
There is always a signal due to imbalance of the RX/TX IB coil and / soil / some targets so the signal at the RX is not microvolt level.
More likely to be 10's of millivolts.
What you are looking to do is resolve phase and amplitude of wanted target.
For a 12 bit system we sample at a high frequency and decimate the signal.
For every power of two decimation we gain 3 dB of dynamic range so
The 12 bit ADC with 74 db of dynamic range can be improved to 92 db with a 64x decimation.
The bandwidth also reduces by 1/64 ... not a prob for metal detectors if sampling ADC is fast enough.
This is the processing gain.
In this way amps are not needed.
Comment
-
Signal for synhronisation in VLF variant is from residual of imbalance in coil, clear. How it in PI case, and what form and level better, sinusoidal ? Or not mater.Comment
-
Here is transmitter simulation file for helpful program "circuitjs" Transmitter 1.zip
Mosfets-current sourses after battery is forming square shape of current.Comment
-
Well probably the best way to think about it is that the VLF is processed in the frequency domain and the PI is processed in the time domain.Originally posted by solv View Post
For VLF we will consider Amplitude and Phase. For PI we will consider Amplitude and Time.
Synchronisation is important in both methods ( however we know the TX signal in both cases so we dont need to regenerate a sync from the RX signal ).
Its more about when to take samples from the RX.Comment
-
Solv wins the challenge question in post #99 CHALLENGE QUESTION : This circuit seems almost too good to be true ... What is the the downside ( there is one)Originally posted by solv View Post
???
Yes the circuit is very similiar to a super regen RX from previous times ( but now used in simple RF chips like garage door openers etc )
The reason the super regen never went further for a number of reasons
1. For optimal reception you have to keep tweaking the regen control if you change freqeuncy.
2. You cant digitally lock the oscillator as that will upset the recovered RX phase. ( the osc needs to lock to the RX signal )
3. The superhet arrived and was simpler to drive. ( but not necessarily better in performance )
So what this means is that the RX oscillator is now the master oscillator .. not the TX oscillator.
To solve this problem the ARMD includes a reciprocal frequency counter that measures the RX oscillator and adjusts the TX oscillator as a FLL ( Frequency Locked Loop ) NOT a PLL ( phase locked loop ).
This way the FLL keeps the TX frequency in the middle of the "passband" of the FLOCKER which for weak signals is very narrow ( eg 50 Hz ).
More complex but there is no free lunch.
Comment
-
And why is it that the classic full H-bridge is not used in branded metal detectors, and in homemade ones too? Two assemblies of P/N mosfets (for example, IRF7105, 7309), one TC4428 type driver are enough. This, of course, applies to the VLF, for a pulsed metal detector, such a bridge circuit is not suitable. What is the advantage of the TX cascade on 6 transistors over the TX cascade on 4 transistors?
With a TX cascade of 4 transistors, the metal detector can use a lower voltage (3-9V) TX and a coil with fewer turns. The TX coil with low inductance will be less affected by the Earth's magnetic field, which in turn will provide a more stable phase characteristic of the TX circuitComment
Comment