NEW !!! Full differential PI front end.

[moodz]

NEWS FLASH ..--.. ..--. FPGA code crammed into DSPIC chip .. read all about it.

Field testing proceeds on a "Model T" version of the differential frontend as per previously published schema. The processing code has been ported to a DSPIC 30F4012. This chip is "barely" fast enough to run the core algorithm as implemented in the FPGA code. ( Curiously the PIC chip alone draws more current from the 5volt supply than my FPGA board which is hundreds of times faster .... )

As I have previously stated ( claimed ?? ) the differential front end has significant benefits over the single ended coil designs.

Field testing is now proceeding however early indications are that EMI ingress is hugely reduced over comparable PI designs. Coil sheilding is NOT required except maybe in ultrasensitive equipment like the FPGA design and that is a maybe.

The "MODEL T" version as shown below is designed to demonstrate the benefits of differential front end whilst achieving a price point below even simple PI detectors like the surf etc.
It does this by utilising a minimum of front end components feeding into a cheap DSP CPU that auto compensates for coil decay and ground balance etc. There are no displays and the audio is driven directly from a pin on the PIC via a 220 ohm resistor.

It can run on 7.5 to 20 volts and the bias generator is a simple capacitor / diode network that is driven directly from the same mosfet driver driving the main TX.

If we dont count the cost of the case / hardware the component cost of this circuit is less than $30.

AMP = $7
DSPIC30F4012 = $5.40
FET = $5
DRIVER = $2.50
Misc diodes, resistors, caps $5


The TX pulse repetition is slightly more than 5 Khz.

I will be conducting extensive field trials over the coming weeks to verify the design .... away from the bench and the armchair.

Regards and happy holidays .... Moodz.

PS Carl ... how about a project page for this puppy ??

[goldfinder]

Stay with pic please

I wish. My latest thought on HH3 is to just set up a 2-stage preamp followed by an 18-bit ADC and micro. Very simple circuit. Then everyone can play with pure processing methods to see what works and what doesn't. The Big Question is what micro? PIC may be under-powered.

- Carl

Carl,
I know the PIC backwards and forwards so hope you stick w/ the PIC. Go w/ the larger PIC interfaced to a good ADC is one way. Another is to use a PIC w/ a larger A2D. I worked many years in DSP for military and we could get very good signals out of signals buried in noise. Cross correlation filters one good way but compute intensive.

An mostly digital HH3 sounds great. Small frontend analog thence into the digital domain and we can play w/ filters to our hearts content. Output doesn't have to be fancy. LED lights good enough to start.
Goldfinder

[moodz]

Hi Moodz,I never have holidays! last one I had was 2006,6 weeks thru Asia to check up on Sat.TV.n ,couple of small providers and installers for Astro,an CBN thruout Asia/Pacific.
n a bit of inquirey to Yashimira Gold thru Asia,in the Phlilipines.
All a bit off topic Sorry?
Digi Scope on LCD u using I take it? what softtwares package u using,for captures,saves?
regards Rov

Not using Digi Scope ... the display is part of the detector code but not required for the detector to work. The FPGA generates the VGA output directly ..which is rather nice as I can display any signals inside the chip ..can also display text. This will be of great assistance for tuning coils etc without carrying around alot of test equipment in the field trials. To capture the screen I just take a Photo and resize with Gimp.
Off topic ... are you in the broadcast / satellite business ?
Moodz.

[Rov]

Hi Moodz,I never have holidays! last one I had was 2006,6 weeks thru Asia to check up on Sat.TV.n ,couple of small providers and installers for Astro,an CBN thruout Asia/Pacific.
n a bit of inquirey to Yashimira Gold thru Asia,in the Phlilipines.
All a bit off topic Sorry?
Digi Scope on LCD u using I take it? what softtwares package u using,for captures,saves?
regards Rov

[moodz]

Ha Ha

Nice! U been on Holiday!Cheers!
appreciate the new scopes!
Rov

Hi Rov ... no I have been at work and now I am going on holidays so I have time to finish this FPGA project even if it kills me. I am very very happy with the sensitivity and ground balance results so far though .. thanks for your encouragement.

moodz.

[Rov]

Nice! U been on Holiday!Cheers!
appreciate the new scopes!
Rov

[moodz]

ground balance on FPGA moodz detector achieves long range detection

A picture says a 1000 words.

FPGA processing waveforms 100 uv division approx ... thus the top displays are off the scale as they are several volts ... pink = output of synchronous filter 1 . white = reference baseline filter yellow = output of synchronous 2nd stage filter. Red is the output of all processing contains target info ( VGA waveform display generated by FPGA chip ) These are not 'analogue' signals they are samples from the processing chain inside the FPGA chip.

THE GROUND signal .. is there a target here or not ... hard to say


After pressing the BALANCE button ( manual or autotrack available )


And the target signal a Aus 20 cent coin at 90 cm from 18 inch coil.
The polarity of this signal will depend on metal type !!!!
This signal will be alot higher in magnitude and width for larger and closer targets.

[moodz]

fpga update take 2.

Its workin ... the receiver locks onto the signal even though it is disconnected by 3 feet of air and 100mv pp of noise at the input. I must finish the frontend electronics ......

[moodz]

FPGA update

Differential front end pulse driver completed. Pic shows a quad pulse burst with long follower. Alternate pulses drive the two sides of the diff coil switches. Burst repetition is set to 512 uS. Pulse widths adjustable by 10 ns steps ... individually. All settings mapped to IO ports on the onchip CPU as 16 bit words.

moodz.

[Tinkerer]

Traditional oversampling and decimation is done on a Nyquist-bandwidth periodic signal. In PI, we are normally sampling a roughly DC point, then averaging it to remove noise. So essentially, we are oversampling and using noise as a dither signal to get a higher filtered SNR.

- Carl

There is one point I have wanting to be clarified for some time. That is all the different significant frequencies involved.

We have a no target, roughly DC point that is a steady as we can make it.
Then, when we move the coil over a target, this DC point varies at a rate of about 3 to 20 Hz, depending on the size of the target, the size of the coil and the sweep speed.

Now when we filter the initial signal, high pass and low pass, how much is the 3 to 20Hz target signal influenced?

When I filter my initial coil signal, I notice that the signal of short TC targets, gets attenuated with a 100kHz low pass filter.

A 150kHz high pass filter would help reducing 50-60Hz power line noise picked up by the coil, but the 3 to 20 Hz target signal remains mostly. I say mostly, because I have noticed that there are some frequency interactions that influence the target signal.

The pulse repetition rate.
The integration time.
And the general interactions of all these frequencies. Could someone explain to me how all this together works?

Tinkerer

[Carl-NC]

Traditional oversampling and decimation is done on a Nyquist-bandwidth periodic signal. In PI, we are normally sampling a roughly DC point, then averaging it to remove noise. So essentially, we are oversampling and using noise as a dither signal to get a higher filtered SNR.

- Carl

[Altra]

Carl,

I have been waiting for the Pic12f1822 also, 32mhz internal osc! They say it will be for sale in august. What do you think of increasing the AD resolution using over sampling? See attached article

Regards
Mark

[greybeard]

I wish. My latest thought on HH3 is to just set up a 2-stage preamp followed by an 18-bit ADC and micro. Very simple circuit. Then everyone can play with pure processing methods to see what works and what doesn't. The Big Question is what micro? PIC may be under-powered.

- Carl

I'm currently starting to play with the HH2. As part of what I'm implementing is a fancy front panel ( 128x64 ) with a rotary encoder and menuing system. Obviously the 8 pin pic's don't have enough I/O to do this without addons so I'm considering a seperate processor for the 'front panel / HMI' interface. What would be nice is if a suitable processor can be found with a few spare i/o for the base board so it can be easily interfaced into a HMI interface. ie instead of ann 8 pin PIC use an 18 pin which would then make a serial interface between the two pretty simple.
It also give you an option of making the base processor configurable via software rather than requiring a new program.
By seperating the functionality, the incremental cost for an extra pic is only $4-$5, you still have the full power of the pic available to run the metal detector without the code having to cope with real time processing of data and human's.
I know code can be written to do anything, I code for a living, but it could be much simpler for others to play and experiment.
Rod

[Krzysztof]

Hi Tinkerer.
My system magnetoinductive fluxgate, the coil current 1 - 2mA.
Repetition frequency of 20 - 100kHz.
Best regards Chris.

[Carl-NC]

Your HH II, is a very good way to show how an analog PI detectors can be changed into a digital one.

At how many PPS are you running it in the example above?
How does the HH II compare with the HH I in sensitivity and depth?
Is the response faster? slower? the same?

Actually, my digital experiment wasn't done on HH2, though it can be. But I need MicroChip to release the 12F1822 so I can do that. Supposedly they are sending me samples now.

I think I was running 600Hz. Depth was not good (but still better than I expected), and I'm sure the response speed was not optimized.

Is there any chance you get time to work on the HH III?

I wish. My latest thought on HH3 is to just set up a 2-stage preamp followed by an 18-bit ADC and micro. Very simple circuit. Then everyone can play with pure processing methods to see what works and what doesn't. The Big Question is what micro? PIC may be under-powered.

- Carl