Amplifier frequency response

[green]

Hi Chet,

Pendulum_1 shows integrator out direct coupled with the last stage gain about x10 capacitor coupled. The trace in pendulum 3 is the capacitor coupled output. The direct coupled signal would go + only(opposite end of figure 8 coil as pendulum_1). I'm taking a 10usec sample after a 5usec delay so the 2usec amplifier filters and moves more of the signal into the sample period. I think taking a 1usec sample after the 5usec delay would be better with the .2usec amplifier. The 2usec amplifier moves what happens during the coil decay into the sample window which seems like might be a problem, maybe not. Pendulum_4 is the output of the amplifier, direct coupled, no capacitors. The upper traces, 2usec amplifier a ferrite bead was positioned on the coil to zero out the offset at 4usec and left there for the can target traces. The ferrite bead is detected during the 10usec sample with the 2usec amplifier but not the .2usec amplifier(can see the ferrite bead change in the first 4usec). I think my first problem is minimizing noise and then seeing if moving the signal during the coil decay into the target sample window is a problem.

Haven't been able to reduce the noise yet. Tried to chart amplitude vs distance for some targets with the .2 and 2usec amplifier. The 2usec amplifier looked better with the swinging target. Wasn't as stable looking at integrator out with a voltmeter and had other problems. One being a change in output when I placed my hand near coil which I didn't see with the .2usec amplifier. B and F, C and G, D and H are the same targets.

[green]

One other addition that might be very useful would be to add a photocell pointing toward the ceiling. If this is hooked to the second channel of the scope it could sense when the target blocks the light as it would when the target is directly over the coil, this would allow you see coil response delay relative to target position.

Good thought, I'll include a photocell on my next order. Been trying to increase signal to noise. Using a difference amplifier with separate Rx and Tx is working best for me. Including some scope shots. I read detecting a US nickel at distances far greater than 12 inches. I could try a larger coil and or higher peak current to increase the signal. Does someone have some scope pictures to compare or tell me how low I should be able to reduce the noise?

[Old cart]

Good thought, I'll include a photocell on my next order. Been trying to increase signal to noise. Using a difference amplifier with separate Rx and Tx is working best for me. Including some scope shots. I read detecting a US nickel at distances far greater than 12 inches. I could try a larger coil and or higher peak current to increase the signal. Does someone have some scope pictures to compare or tell me how low I should be able to reduce the noise?

Green, great job! Are you using the bandwidth limiter on the Rigol when you take these shots? Looking at your signals it looks like the actual target signal closely resembles the amplifier noise so simple filtering is not going to help you much. One thing that may be causing the noise is the supply lines. What are you using for a power supply? Highly regulated and filtered supplies are important for the first stage or two. See here for some recent suggestions:

http://www.ti.com/lit/an/slyt202/slyt202.pdf

It looks like the noise you are seeing is centered in the band at about 9-15Hz. What op amp are you using? Look for op amps that have low 1/f noise at the impedance you have with your coil and feedback resistances. Note that if your scope can do FFT's this can be very useful technique to see just what the frequencies of the noise are.

http://www.analog.com/media/en/train...als/MT-048.pdf

http://www.analog.com/media/en/train...als/MT-047.pdf

http://cds.linear.com/docs/en/design-note/dn015f.pdf

http://www.edn.com/electronics-blogs...kering-candle-


http://www.keith-snook.info/wireless...amplifiers.pdf


In an actual detector, as opposed to a test fixture, it may be possible to use long integration times to minimize this noise. Alternately it may be possible to use a micro to analyze the static noise and eliminate it using DSP.

[Monolith]

Nickel at 8 Inches

Good thought, I'll include a photocell on my next order. Been trying to increase signal to noise. Using a difference amplifier with separate Rx and Tx is working best for me. Including some scope shots. I read detecting a US nickel at distances far greater than 12 inches. I could try a larger coil and or higher peak current to increase the signal. Does someone have some scope pictures to compare or tell me how low I should be able to reduce the noise?

You are doing great.
Attached a scope shot of my signal of a Nickel at 8 Inches.

Blue trace, no target.

Yellow trace, Nickel on the pendulum, swinging at 8 Inches above the coil.

The coil is 24cm diameter, 1944uH, 60 turnTX
RX is 426uH, 19cm diameter

Gain, Preamp=68, sample window, 5us, integrator opamp gain 100, band-pass 2Hz to 15Hz, level shifter gain 45.
Total gain will depend on how you calculate the integrator gain. PPS is 2500.

[KingJL]

The pendulum setup works good. Repeatable target height and sweep speed. Used a 1 inch spacer to set target height.

Green,
What was your final pendulum dimensions?

[Old cart]

You are doing great.
Attached a scope shot of my signal of a Nickel at 8 Inches.

Blue trace, no target.

Yellow trace, Nickel on the pendulum, swinging at 8 Inches above the coil.

The coil is 24cm diameter, 1944uH, 60 turnTX
RX is 426uH, 19cm diameter

Gain, Preamp=68, sample window, 5us, integrator opamp gain 100, band-pass 2Hz to 15Hz, level shifter gain 45.
Total gain will depend on how you calculate the integrator gain. PPS is 2500.

Does your scope do FFT's? If so can you increase gain of the no target signal, decreases sweep speed to 100mS per div. Increase gain to 50 mV / div using X1probe position. You want as big of a signal as possible. Record length should be 5-10K points.
this will help with noise frequency content. If it is not to hard disable the band pass filter.

[green]

Green,
What was your final pendulum dimensions?

I have a hook screwed in the ceiling in the garage. The coil sits on some cardboard boxes 35 inches off the floor to keep pendulum length less than 8 feet. About 86 inches from coil to hook. Pendulum length 86 inches minus coil to target height. String connected to a plastic water bottle runs through hook to a tie off point. Raise or lower water bottle to adjust target height.

[green]

Green, great job! Are you using the bandwidth limiter on the Rigol when you take these shots? Looking at your signals it looks like the actual target signal closely resembles the amplifier noise so simple filtering is not going to help you much. One thing that may be causing the noise is the supply lines. What are you using for a power supply? Highly regulated and filtered supplies are important for the first stage or two. See here for some recent suggestions:

http://www.ti.com/lit/an/slyt202/slyt202.pdf

It looks like the noise you are seeing is centered in the band at about 9-15Hz. What op amp are you using? Look for op amps that have low 1/f noise at the impedance you have with your coil and feedback resistances. Note that if your scope can do FFT's this can be very useful technique to see just what the frequencies of the noise are.

http://www.analog.com/media/en/train...als/MT-048.pdf

http://www.analog.com/media/en/train...als/MT-047.pdf

http://cds.linear.com/docs/en/design-note/dn015f.pdf

http://www.edn.com/electronics-blogs...kering-candle-


http://www.keith-snook.info/wireless...amplifiers.pdf


In an actual detector, as opposed to a test fixture, it may be possible to use long integration times to minimize this noise. Alternately it may be possible to use a micro to analyze the static noise and eliminate it using DSP.

Scope on normal. Power supply_ 78L05 with a TL2426 virtual ground. Using an amplifier I posted awhile back using a dual 2N3904 for the input. Working on a instrumentation amp using a OPA1612 for the input amps. I'll post the up to date schematics after I try the OPA1612. Thanks for the suggestions.

[Monolith]

Does your scope do FFT's? If so can you increase gain of the no target signal, decreases sweep speed to 100mS per div. Increase gain to 50 mV / div using X1probe position. You want as big of a signal as possible. Record length should be 5-10K points.
this will help with noise frequency content. If it is not to hard disable the band pass filter.

Thank you for the very useful links and suggestions. I will try that.
The circuit is on a bread-board, so changes are very easy. However, the bandpass filter is inherent to the amplifier circuit.
There are 3 main noise frequencies.
1) 50Hz mains noise, normal, considering the virgin forest of air wires on the bread-board. It will disappear once the circuit is on PCB.

2) Noise at 0.5Hz, do not know where that comes from.

3) Noise at about 2Hz. No idea where from.

Reading your links now, hoping to get enlightened.

[green]

Found a photo transistor. Glued a 3 inch diameter cardboard disk painted black to the bottom of the water bottle. Added a absolute value circuit + a MAX7410 to filter the absolute value output. Looks like it might do some good, maybe more filtering than needed. Peak amplitude delayed about 1/2 second. Some high frequency glitches that came be eliminated. Still working on the OPA1612 instrument amp.

[green]

Found a photo transistor. Glued a 3 inch diameter cardboard disk painted black to the bottom of the water bottle. Added a absolute value circuit + a MAX7410 to filter the absolute value output. Looks like it might do some good, maybe more filtering than needed. Peak amplitude delayed about 1/2 second. Some high frequency glitches that came be eliminated. Still working on the OPA1612 instrument amp.

Finished and tested the OPA1612 instrument amp. Noise about the same as the dual 2N3904 input. Seems like it should be better, maybe some thing I'm doing or something else is causing the noise. The peak amplitude is delayed by about .25 seconds not .5 second with the max7410 output, must have been looking at lower scope trace for time/div.(reply #85). If I make the disk 100mm in diameter, peak velocity(mm/second) would be 100mm/photo cell pulse time(100mm/.1second=1000mm/second), good indicator of velocity for comparing output amplitude at different sweep speeds. Photo cell mounted on top of a 8 inch spacer. Thanks (Old cart) for the photo cell suggestion.

[Old cart]

Filter and photocell feedback

Finished and tested the OPA1612 instrument amp. Noise about the same as the dual 2N3904 input. Seems like it should be better, maybe some thing I'm doing or something else is causing the noise. The peak amplitude is delayed by about .25 seconds not .5 second with the max7410 output, must have been looking at lower scope trace for time/div.(reply #85). If I make the disk 100mm in diameter, peak velocity(mm/second) would be 100mm/photo cell pulse time(100mm/.1second=1000mm/second), good indicator of velocity for comparing output amplitude at different sweep speeds. Photo cell mounted on top of a 8 inch spacer. Thanks (Old cart) for the photo cell suggestion.

What cutoff frequencies are you using on the MAX7410? The 250mS delay you are experiencing is probably just due to the group delay of a very low frequency Low Pass filter or it may be due to the witched cap filter or a combination of both. If I am interpreting your graphs correctly the cap coupled trace has the photo transistor peak very near the negative going target peak. At 12" though the signal to noise is much better that the unfiltered signal.
Maybe you could design the circuit so that when doing searching the SC filter is in the loop. Once the target is found a pushbutton would take it out of the siganl path allowing more precision in pinpointing.

With the MAX7410 in the loop you might benefit from a 1st order passive low pass after the SC filter to get rid of the SC or maybe AC line noise. it is hard to tell at the sweep speeds these were taken at.

[6666]

What are you calling "the SC filter" ?

[Old cart]

What are you calling "the SC filter" ?

It is the MAX 7410 which is a switched capacitor filter.

[green]

What cutoff frequencies are you using on the MAX7410? The 250mS delay you are experiencing is probably just due to the group delay of a very low frequency Low Pass filter or it may be due to the witched cap filter or a combination of both. If I am interpreting your graphs correctly the cap coupled trace has the photo transistor peak very near the negative going target peak. At 12" though the signal to noise is much better that the unfiltered signal.
Maybe you could design the circuit so that when doing searching the SC filter is in the loop. Once the target is found a pushbutton would take it out of the siganl path allowing more precision in pinpointing.

With the MAX7410 in the loop you might benefit from a 1st order passive low pass after the SC filter to get rid of the SC or maybe AC line noise. it is hard to tell at the sweep speeds these were taken at.

Calculates 2.25 Hz(.1 and .033 in parallel). The thought has been to use the capacitor coupled output to find the target and look at the last direct coupled stage with a voltmeter to pin point. Offset shouldn't matter non motion, just looking for peak change.