Digital TGSL

[aft_72005]

Hi wild_desert

Block diagram Ok, be better if you begin setup hardware ,write software ,and
Then if having trouble at act, correction again it .
Best regards.

[wild_desert]

Hi aft
nowadays try and error is not acceptable method. Simulation softwares and advanced mathematics available. block diagrams are starting points of new ideas.
regards

[aft_72005]

Hi wild_desert
Yes, simulator software’s help us .specially when writing program for
Microcontrollers or Debugging it.
But circuit simulation having with problems sometimes. Especially when
working with oscillators. For example I simulated tgsl oscillator With
protues 7.2 . I cannot running it yet, but in practice Working well.

Best regards.

[Sean_Goddard]

I'm a bit dubious as to whether the internal ADC's of the Atmel will give the resolution and therefor the sensitivity you need. Maybe an external ADC would be better, or maybe switched gain range amplifiers feeding the Atmel ADC's?

[KingJL]

I'm a bit dubious as to whether the internal ADC's of the Atmel will give the resolution and therefor the sensitivity you need. Maybe an external ADC would be better, or maybe switched gain range amplifiers feeding the Atmel ADC's?

Totally agree with Sean!! Ten bit resolution not enough to give sensitivity or dynamic range needed. Even Whites Spectrum series with an effective 14 bit resolution (8 bit resolution ADC fed by a series of scaling amps) was marginal to get the job done. Either use an external ADC with adequate resolution or a scaling amp circuit (switched gain or switched amp) to get the required resolution. Personally I would opt for an external ADC with 20-24 bits (my preferrence is 24 bit) resolution (they are inexpensive and available with either parallel or serial SPI, I2C interfaces); They result in better accuacy and are easier to implement in terms of circuit complexity, real estate (PCB), and interfacing. Scaling amps appear simple, but you have to control the scale to keep the signal in the middle of the usable dynamic range of the ADC, get the current 'scale' into the uPC, and correctly recompute the actual signal level.

Regards,
J. L. King

[Sean_Goddard]

Don't forget switchable inverters! Signal polarity can be an issue if you want to have an all metal mode, UNLESS you bias the incoming values to centre range as JL King states. Then it doesn't matter which way the signal goes.

I used a "9 bit" ADC a long time ago in a sound sampler. Bit 9 was a polarity detector which was used to switch an inverter on the output, the rest was just fed through a precision rectifier to give positive input to the ADC. Bit 9 = 1 was positive 1/2 cycle and vice versa. I did the polarity processing in software (16bit) to give me a better dynamic range.

So, with scaling amps AND the above system, you could use the internal ADC's in the AVT.

Still better, use 24 bit AND scaling amps!!

ALSO check our Robert Hoolkos superb article on wide range inputs (do a search for his name) I posted ALL his documents as one BIG file on here three or four years ago.

[Sean_Goddard]

Had anyone read this>> http://www.dspguide.com/ ? The WHOLE book is FREE from the net. DSP without the brain ache maths AND sample programs in BASIC

Check out section 2, and the section about convolution. I think statistical analysis or Rx signal might yield some good detection system results if tried.

Anyone??

[simonbaker]

Had anyone read this>> http://www.dspguide.com/ ? The WHOLE book is FREE from the net. DSP without the brain ache maths AND sample programs in BASIC

Check out section 2, and the section about convolution. I think statistical analysis or Rx signal might yield some good detection system results if tried.

Anyone??

Basically lock-in amp (Aziz) seems to boil down those statistical methods best. Lock-in is basically a convolution at two discrete phase angles instead of continuous. That way you can efficiently time-average to reduce noise. You'd need tons of computation to do full convolution over a time window I think. Convolution more useful when signal is broad spectrum and you want to find multiple frequencies, like FFT. Convolution is related to FFT. Lock-in amp is basically a super efficient convolution for determining phase at one frequency I think.

My 2 cents.

-SB

[Andy_F]

...
Anyone??

Project of the device with "direct processing":
http://alteh.narod.ru/MDs.html



Variants of shematics:
http://alteh.narod.ru/Gif/MD_DS_0.GIF
http://alteh.narod.ru/Gif/MDMT.GIF
http://alteh.narod.ru/Gif/MDAUT.GIF

Used Discrete Fourier Transform (DFT), not FFT.

[wild_desert]

go further

Hi friends
Thanks for your good ideas.
But go further and see the world. When i started to design a perfect metal detector, i saw all previous methods. ultimately, i found Software Defined Radio (SDR). http://www.sdrforum.org/pages/aboutS.../whatIsSdr.asp

This is latest technology in signal processing. Many companies are trying to make a single chip SDR. for example 3G mobile phones use this technology and the price is decreasing more and more.

Now i want to make a broadband MD using SDR chips. it sames the GPR (radar).

Dont afraid of SDR. New chips have ARM processors, simple to programming.
Regards

[Sean_Goddard]

Not to put too fine a point on it but a software defined metal detector was my idea back in 1989, but processing power was not compact enough to do this.

I recently sent Carl all my documantation on this subject for Whites to use so he will bear me out on this claim. You are right though, it IS the way forward. Imagine being to emulate ANY machine PI, VLF, etc