Verator front end

[Sergey_P]

Output synchronous detector: ground + coin ->
http://md4u.ru/download/file.php?id=9364
filter output:
blue - the classic filter frequency
red - filter by phase of ground (W6000, K18, K20, Verator…)
http://md4u.ru/download/file.php?id=9363

[Davor]

But what I see here is more deviation for ground signal in your proposed solution. Isn't that a bit odd?

I see that the ground filtered solution is compressing the ground response, which is good, but somehow the second picture is showing more deviation for ground vector angle than the initial ground + coin response. Could it be that phase noise is that much at play?

[Verator]

To evaluate the characteristics of the phase I use here is the inclusion of the circuit. The phase shift for operation should be less than 75 degrees. Generally 77-83.
The fact that the offer to apply a linear course, but it's more for pure sine wave. For my application is too broad band.Will be an opportunity to keep in mind. Allocate time to see your generator PLL.

I thought that someone deleted topic. And she is in a new place.

[Davor]

Assuming Verator Rx coil comes in a range between 25 and 85mH, and according to a table of Verator schematic frequencies can be set between 5 and 20 kHz. Assuming 200ohm being a coil resistance, it's minimum noise is ~1.8nV/sqrt(Hz)

So in case you are using some real low noise op amp in non-inverting configuration you'll benefit from series tank configuration. The other option is replacing the op amp preamp with a discrete preamp, where you can easily achieve input noise below 1.5nV/sqrt(Hz) and keep your money in your pocket.

The quasi-resonant parallel tank front end is a thing of the past when op amps with 20nV/sqrt(Hz) were considered low noise. It serves a purpose of transforming low input impedance of a coil to some higher impedance near the equivalent noise source of a bad preamp. We really do not have to do that any more.

At the moment I have some important things to do, and I play with electronics only for relaxation. Right now I'm relaxing with a better audio source for future detectors. This design is as open as it gets, so feel free to participate/use as ever you wish.

[Verator]

Take some noise values, just like numbers, it seems to me is not true. Band channel synchronous detector MD is the band 20-50Hz. The noise level normalized for audio amplifiers (op amp) is usually used to indicate to 1kHz. We need a noise level at subsonic frequencies such as 10Hz. These noises have the op-amp type OP27-OP37 Conventional audio op amp noise level ranges have a 10-100Hz. 10-100 more. Hum and Noise level ground is 100 times higher than on its own noise op amps. For proper operation of the MD must be sufficiently powerful field TX (stock) is not a high gain linear phase Coil MD
No need to get involved in low or ultra-low noise for the sake of recreation. The level of soil disturbance is still exceed

[Sergey_P]

Isn't that a bit odd?

The suppression of the signal with the phase of the soil for both methods is possible with a stable phase of the ground (red straight line drawn by you). This is achieved by a crystal oscillator and Qtx<1. Oscillator is not able to ensure the stability of the phase of the soil.
Distortion of the signal from the target method for GB caused by a distortion of the signal function: S=(G+T)*sin(dф).

[Davor]

Take some noise values, just like numbers, it seems to me is not true. Band channel synchronous detector MD is the band 20-50Hz. The noise level normalized for audio amplifiers (op amp) is usually used to indicate to 1kHz. We need a noise level at subsonic frequencies such as 10Hz. These noises have the op-amp type OP27-OP37 Conventional audio op amp noise level ranges have a 10-100Hz. 10-100 more. Hum and Noise level ground is 100 times higher than on its own noise op amps. For proper operation of the MD must be sufficiently powerful field TX (stock) is not a high gain linear phase Coil MD
No need to get involved in low or ultra-low noise for the sake of recreation. The level of soil disturbance is still exceed

True, but front end operates at kHz frequencies, and it is only natural to optimise for noise as well. Otherwise the quasi resonant approach using a parallel tank would have been abandoned long ago.

At baseband at about 10Hz the most important feature is 1/f noise performance, and it is best dealt with chopper amplifiers. I'm playing with these as well for my future LF rig.

[Davor]

The suppression of the signal with the phase of the soil for both methods is possible with a stable phase of the ground (red straight line drawn by you). This is achieved by a crystal oscillator and Qtx<1. Oscillator is not able to ensure the stability of the phase of the soil.
Distortion of the signal from the target method for GB caused by a distortion of the signal function: S=(G+T)*sin(dф).

I think it is largely overrated, as the device with sweep driving signal would not have much problems attenuating ground. Multi frequency devices show precisely that: the ground response does not change significantly over wide frequency spans.

[Verator]

Refused or rejected or that the input circuit - to me it does not matter. Now I am doing just that. Go to the sinus may use a different path. Important not outline a general approach.What type of op amp you intend to use at the entrance?

[Davor]

Maybe you'll laugh: no one.

I intend to let the channel amplifiers in form of low noise chopper amplifiers do the work for me. I realised that the only way to extend the input dynamic range is to abandon pre-amplifier completely. I can do that because the switchers' impedance is close enough to the coil resistance, and noise contribution will be minimal.

Once I have the input signal chopped by the switches, it is only natural to amplify that signal instead of the low pass filtered signal which is prone to 1/f noise. I also realised that motion compensation can be applied in a switching stage, and hence get rid of the cyclo-stationary component (e.g. "air signal" and a static part of ground signal) and amplify only what I really need.

So in effect I'd be amplifying only the dynamic components of the signal, and high pass corner can be set very low, say at 0.1Hz and the resulting behaviour will be somewhere in between motion and non-motion. Such low corner is not a problem with chopper amplifier because it is not prone to 1/f noise. The funniest thing is that switches even today are providing a chopped version of input at some phase, of course, but we supply this signal to a low pass filter and to 1/f noise prone amplifiers. It is silly - if we amplify prior to low pass there is no 1/f noise, but the frequency/phase/amplitude response is identical.

I made schematic representation (and more) in LF project topic. My intended topology is described in enough detail there.

[Carl-NC]

Take some noise values, just like numbers, it seems to me is not true. Band channel synchronous detector MD is the band 20-50Hz. The noise level normalized for audio amplifiers (op amp) is usually used to indicate to 1kHz. We need a noise level at subsonic frequencies such as 10Hz.

Are you talking about the preamp? If it is followed by demods, then the demods are frequency translators, and translate the noise as well as the signal. So if you run the demods at 10kHz, then you want to know the preamp noise level at around 10kHz.

[Verator]

Of course the noise preamplifier has a wide range. But the noise does not pass through the conversion of sync detector. It is integrated at the level of bandwidth sync detector. Tested only slow changes within the band 20-50Hz. Therefore, the density of noise is not distributed in the range of the frequency TX. When there are 2 MD unit near the same frequency, it is enough to rebuild one of them on the 200-300Hz. for normal operation. This indicates a narrow bandwidth, high selectivity circuit MD. 10kHz. never tested in the circuit channels. Not detected by the frequency itself, and its change. We are talking about simple things, understandable schoolboy.
Of course, if you work directly without sync detector and preamplifier and processed directly 10kHz. it is necessary to provide a low noise is at 10kHz.

[Dave J.]

Verator, perhaps you've failed to notice that for many years, multiple companies have manufactured pretty good metal detectors that work better than anything you have proposed so far. And some of 'em have stuff under development that goes 'way beyond what you see on the market right now.

Once the current crop of beeper engineers is gone, someone will have to replace us. Maybe the learning curve goes from us to you, and not quite so much the other way around. Your choice.

--Dave J.

[Davor]

@Verator, Sure, that's precisely the reason I'm not discussing absolute noise but spectral noise density instead. The whole point is that the ultimate noise source is the coil's resistance, and it's spectral noise density is ~0.13sqrt(R) nV/sqrt(Hz)
This way 200ohm coil produces about 1.8nV/sqrt(Hz) noise. As NE5532 has about 5nV/sqrt(Hz) it will be some 9dB noisier, and will give some 10% or so detection depth loss (unless there is some kind of impedance transformation). Of course, keeping everything else as ideal as possible.

There is also a problem with distortion. Op amps used for HiFi tend to have low but finite THD and IM distortions. Everything is fine and dandy if induction balance is keeping the "air signal" down, but rapidly deteriorates with rising amplitude of cyclo-stationary signal. 100nV signal gets overwhelmed by 100mV air signal and everything goes south. Therefore I realised that switching at very front end will extend dynamic range tremendously.

Bilateral switches come with different conduction resistance, and garden variety 74HC switches go below 100ohm. I think it will suffice. There are also switches with conduction at few ohms, but they all come in SMD packages that are not too handy for experimenting.

After switching come low noise chopper amplifiers. Choppers are often considered as expensive self-contained devices with their own oscillators and stuff, so you use them as any other op amp. Well, not in my application. Input switches provide the chopped signal, so in effect the chopping frequency is synchronous with carrier. That way I lose 1/f noise, and gain a perfect zero reference, also no drift at all. The best thing about these is that they do not have to be exceedingly linear so it will keep the cost down.

After that comes phases play, 4-quadrant discrimination, gated binaural audio ... it is all in LF project topic.

[Verator]

Verator, perhaps you've failed to notic Once the current crop of beeper engineers is gone, . Maybe Your choice.

Are you sure I'm sorry, but you're comparing companies that invest in the development and attract enough money for this technology. I do not have any education, simple lover who went to school 40 years ago, to invest hundreds of thousands of dollars, I can not. That I can even do.What you have proposed? Show discussed.
In these matters, I know about yourself. Nobody helps and does not tell. 99% did not know. I am glad that you are in that 1%