Let's made a PC-base metal detector with usb interface !!!

[Aziz]

I built the following ear phone amplifier (NE5534) for driving the TX coil. I will test this circuit and will let you know, whether it makes sense to drive the coils with more current.
I found the schematic from this site:
http://www.audioattic.de/projects/small_eaton.html

Regards,
Aziz

[Aziz]

Shame on me! I have forgot to remove the short circuited input from my last lock-in experiment with one turn coil and was wondering about the strange behaviour of my amplifier on the PC scope!

Now the amplifier works perfect. Amplification factor x 5. Very low total harmonic distortion until Vout =2.4 V rms AC. Power supply at 10.5 V.
Now focusing in using the amplifier for TX coil.
Aziz

[gwzd]

My sound card has an internal noise of ~200 nV rms. ...
Aziz

Uhmm... aren't you exaggerating? A laptop? Are you kidding?
Regards

[Aziz]

Uhmm... aren't you exaggerating? A laptop? Are you kidding?
Regards

My high definition sound card in my PC. 24 Bit, 96 kHz. Short circuited line-input to ground. It's quite good. Very low harmonic distortion (can see this on the spectrum analyzer). Therefore, I can detect 0.01 degree phase changes from a far distance target to coil.

My "PCMCIA Audigy 2 ZS Notebook" was much better. But I have to wait for a new laptop. Next month, I will get a laptop for real field tests!
This is only possible, when the signal is internally fully synchronized with the lock-in amplifier. And very small signals, phase and noise can be measured really accurately.

Now I am able to say, that the digital lock-in VLF metal detector (laptop detector) can be a real competition to a good other VLF MD's. Software is much tricky but hardware is much easy.

If nobody believes me, one should make the lock-in amp by himself.


Regards,
Aziz

[gwzd]

My high definition sound card in my PC. 24 Bit, 96 kHz. Short circuited line-input to ground. It's quite good. Very low harmonic distortion (can see this on the spectrum analyzer).

What spectrum analyzer? the one made up using the same card you are measuring?

My "PCMCIA Audigy 2 ZS Notebook" was much better.

Reality check -
PCMCIA Sound Blaster Audigy 2 ZS Notebook short datasheet:
Two modes of operation 24bit/96kHz or 24bit/192kHz (playback only)
SNR = 104dB

Assuming your reference is 1.024Volts, this is a fair assumption since most sound cards try to keep the reference low, then your smallest signal to catch would be.... 6.3uV(microvolts)

So, how come you get better results than the manufacturer? Let me guess, Lock-in amplifier!! Oh, come on Aziz don't let the computer fool you.

Keep the good work Aziz
Regards

[Aziz]

Hi gwzd,

don't forget the short bandwidth of the lock-in amplifier. Only the noise on the operating frequency is measured. All other noise is not measured. There are some frequency ranges with lower noises and some with more. It depends on many factors. By increasing the time constant of the lock-in amp, the accuracy gets even better.

Look more in detail to lock-in amplifier.


regards,
Aziz

[gwzd]

Hi Aziz

don't forget the short bandwidth of the lock-in amplifier. Only the noise on the operating frequency is measured. All other noise is not measured.

I don't think so, neither does this manufacturer:

http://www.boselec.com/products/siglimnoise.html

Now, look at this data-sheet:

http://www.boselec.com/pdf/SI_430.pdf

Pay attention to the "sensitivity" parameter of their product, it is roughly 15 (fifteen) times bigger than the one you claim you have. (you can find that parameter on page two, close to the top)

So, my question remains, aren't you exaggerating? Check for errors in your calculations. Finally, have a look at the plots posted by TheWizard they seem to be more realistic than the claims you have made so far.
Regards

[Aziz]

Maybe I forgot to meantion the full dynamic range of my line-input:
~100 mV rms (with full amplification)
So the signal is amplified within the codec chip.

Aziz

[gwzd]

Maybe I forgot to meantion the full dynamic range of my line-input:
~100 mV rms (with full amplification)

So, in other words your SNR is 10dBs higher than the specified by the manufacturer of the sound card. Is that so?

So the signal is amplified within the codec chip.

Aziz

What codec chip would that be? Any part number?
Regards

[TheWizard]

About Noise and Signal....

http://www.cpm.uncc.edu/programs/tn1001.pdf

[Aziz]

So, in other words your SNR is 10dBs higher than the specified by the manufacturer of the sound card. Is that so?
What codec chip would that be? Any part number?
Regards

Hi gwzd,

the ADC part of my sound card has a SNR of 90 dB. The codec chip is a ALC888 (Realtek).

To get a fixed reference measuring system, I had to set the amplification ruler to the maximum because there is no exact amplification number display. To reproduce and measure allways same results, I use the maximum level of internal amplification. I calibrated this with two multi meters, to get not exact but reasonable values.

What about doing lock-in amp by yourself? You will see the difference.

Regards,
Aziz

[gwzd]

Hi gwzd,

the ADC part of my sound card has a SNR of 90 dB. The codec chip is a ALC888 (Realtek).

Very interesting, you have 100mV dynamic range in your line in and only 200nV noise floor and still all that with a SNR of 90dB.

To get a fixed reference measuring system, I had to set the amplification ruler to the maximum because there is no exact amplification number display. To reproduce and measure allways same results, I use the maximum level of internal amplification. I calibrated this with two multi meters, to get not exact but reasonable values.

Is the 200nV an approximated value? How many digits your multi meters have? How can you measure 200nV without even knowing your reference values? The same results doesn't mean your math is correct.

What about doing lock-in amp by yourself?

Unfortunately, most of the time I work with UWB signals where lock in amplifiers are useless. Besides, imagine driving a survey car scanning a road to measure asphalt thickness and waiting seconds for the lock in amplifier to finish locking in. I don't think so, it would leave me without customers in a record period of time.

You will see the difference.

There's no need to do stuff in order to understand how they work or what you can expect from them. In a previous post TheWizard posted a very nice introductory paper on Lock in amplifiers, but I'd personally would suggest the one posted in the Microchip's website. The application note is number AN1115.

Another thing, as someone posted earlier, you are going to have much more problems in the real life where not always, rather never, the conditions are ideal like in the lab.

Regards

[Aziz]

Hello gwzd,

You should implement your own lock-in amp and play around a bit to see the real difference. Or am I see some "ghosts" on my screen?

Aziz

[gwzd]

Hello Aziz

Hello gwzd,

You should implement your own lock-in amp and play around a bit to see the real difference.

I wish I would have enough time to "play" with many things and trust me when I say, lock in amplifiers are not in the first ten.

As I said before, I don't work with narrow band systems except for resistivity measurements and I have no need to implement lock in amplifiers for that.

Or am I see some "ghosts" on my screen?

Since you have written the software, established your references and done your math you practically see what you want to see. But that is not necessarily what it is. I asked my son the other day what would it be 7 and 3 and he answered perplexed "seventythree! doh!" I guess he is smarter than me since I didn't ask "7 plus 3". Of course, an alternative answer would have been "3" if using a bitwise AND. So, my point is that there are many ways of screwing the results in a computer. But, as pointed out earlier you will have to find out that all by yourself.
Regards

[Aziz]

Hi gwzd,

I made some numerical tests with the input data now. The noise is typically on the least 4-6 significant bits of the 24 bit input data. I tested also some software components. And they are working correct so far.
Using the noisy least significant data bits over some period time (time constant), the accuracy can be increased despite of the noise.
This is what I am using to detect very small signal changes.

Even, when the signal change is somehow correlated over time (due to drift), the software detector can handle this really very elegant.

I wish, your guys could see the real-time operation. It's difficult to demonstrate this without a camera.
Aziz