Beachmaster Digital

[jeep6789]

It is difficult for machines that detect AC to achieve good stability, because software filtering is not as good as the effect of op/AMP integrator, felezjoopi and other machines, stability is not as good as delta pulse,

[Olly]

It is difficult for machines that detect AC to achieve good stability, because software filtering is not as good as the effect of op/AMP integrator, felezjoopi and other machines, stability is not as good as delta pulse,

I quite agree that it's relatively straight forward to filter out noise with analogue integration and filtering. If this is what you're after then there are many very good pulse induction designs on this forum that operate on this principle and I don't want to reinvent the wheel.

This is an experimental design where I'm trying to see what is achievable with direct sampling and if at the end of the experiment I find that it's just not feasible, then so be it - but I'm going to have fun trying it out along the way. It's half the fun of engineering isn't it

[green]

Hit a bit of a testing snag I'm afraid.

I've tried to make the preamp as fast as possible but am now running into serious external noise pickup problems.
Unfortunately I have one of those radar 30MPH speed warning signs just outside my house and it seems to induce an awful lot of noise into my coil.
This is what I measure directly across the coil with nothing else connected other than the damping resistor.
[ATTACH]57175[/ATTACH]

And this is what it looks like amplified 1000 times by my preamp.
[ATTACH]57176[/ATTACH]

This has not really been too much of an issue before as previous preamps have had a low enough cutoff frequency to not be too bothered with the noise.

I don't want to slow the preamp down, so perhaps it may be time to experiment with a figure 8 coil to try to reduce the noise pickup.

I use a figure 8 coil. Does help reduce noise. Don't know if it will solve your noise problem. Be interesting to see if it does. What does a scope trace of amplifier out look like with Rx input shorted? Won't get better than shorted.

[Olly]

Very clean with input shorted, around 12mV pk-pk noise on a scope trace.
Here is the noise spectrum of the preamp output with input shorted, the blue trace is the preamp output, the red trace is unconnected.
-60dBV is 1mV.


Do you use the figure 8 as a mono coil?

[green]

Very clean with input shorted, around 12mV pk-pk noise on a scope trace.
Here is the noise spectrum of the preamp output with input shorted, the blue trace is the preamp output, the red trace is unconnected.
-60dBV is 1mV.
[ATTACH]57177[/ATTACH]

Do you use the figure 8 as a mono coil?

Figure8 Rx, separate Tx. R1=pi*L*SRF (after shielding with cable, located at circuit board). R4 and R5=pi*L*SRF (before shielding no cable, L3 and L4 not connected, located at the coil)*(1.5 to 2).

Haven't tried direct sampling with an A-D.

[Olly]

Figure8 Rx, separate Tx. R1=pi*L*SRF (after shielding with cable, located at circuit board). R4 and R5=pi*L*SRF (before shielding no cable, L3 and L4 not connected, located at the coil)*(1.5 to 2).

Haven't tried direct sampling with an A-D.

Excellent thanks, will check it out.

[Olly]

I haven't got round to trying a Fig8 coil yet, but have discovered that these opamps don't like the DC offset being adjusted via the non-inverting input.
I get far better results when employing the standard method of grounding the non-inverting inputs and applying the DC offset adjustment via the inverting input of the first stage, will amend schematic and PCB layout accordingly.

The good news is that the configuration of Diodes D2 and D3 to only give me a positive going output for the ADC works very well.

Using a 3DSS coil and timing my capture to a lull in the radar generated noise, I get the following amplified decay curve which allows me to sample as early as 6uSec.
Will assemble the ADC converter circuitry next...

[jeep6789]

You enlarged 1000 times?

[Olly]

You enlarged 1000 times?

Correct, the preamp amplifies by 1000...

[Olly]

Update: -

A-D converter up and running.
Also have the PIO code working to capture 64 samples at 1uSec intervals after each TX falling edge and transfer to a buffer via DMA. (This happens independantly of the main program). After each set of 64 samples has been captured an interrupt is raised to signal the main program that the buffer is full and ready to add to the rolling array for processing.

Am now going to spend some time experimenting with suitable filtering techniques to try to eliminate (or reduce) coil noise. A rolling median of each of the 64 sample points may be a good start. Am going to use 'R' to analyse the samples and develop the algorithm.

[jeep6789]

The people of the world cheer you on

[waltr]

Update: -

A-D converter up and running.
Also have the PIO code working to capture 64 samples at 1uSec intervals after each TX falling edge and transfer to a buffer via DMA. (This happens independantly of the main program). After each set of 64 samples has been captured an interrupt is raised to signal the main program that the buffer is full and ready to add to the rolling array for processing.

Am now going to spend some time experimenting with suitable filtering techniques to try to eliminate (or reduce) coil noise. A rolling median of each of the 64 sample points may be a good start. Am going to use 'R' to analyse the samples and develop the algorithm.

I have played with this idea using a PIC32's ADC to sample. I summed a number of TX/RX cycles to average out some noise. Other reason is TX/RX cycle happens at 625usec intervals and humans only need a change every 30msec or so to feel the response is 'real time'.
This allowed summing 32 TX/RX cycles before processing.

[jeep6789]

What kind of content will the display display?

[Olly]

What kind of content will the display display?

Not sure yet, I'm making this up as I go...
Relevant target info once I determine what that is...

[Olly]

I have played with this idea using a PIC32's ADC to sample. I summed a number of TX/RX cycles to average out some noise. Other reason is TX/RX cycle happens at 625usec intervals and humans only need a change every 30msec or so to feel the response is 'real time'.
This allowed summing 32 TX/RX cycles before processing.

I certainly needs lots of filtering in my lab to counteract the huge amounts of EMI/RFI that I have to contend with, hoping that this will be considerably reduced in the field.