All sorts of noise at the preamp.

[Tinkerer]

I am trying to build a detector for small nuggets. I have wound a coil with a resonant frequency of about 750 khz. (310 uH, 11" diameter, 2 Ohm)
On the output of the preamp I get all sorts of noise, but the strongest is at about 750 khz. I take this to be "white noise" that shows at the resonant frequency because of the enhancement caused by the resonance. Is there a way to get rid of it?

Another problem at the input of the pramp (5534) is a "kink" at the point where the limiting diode stops conducting. No matter more or less damping resistor, the kink stays.

At present I can sample at 5 us, If I could solve the above problems, I could gain one more us at least.

Any good advice is welcome.

Tinkerer

[pml]

Hi Tinkerer,

Carl's the expert on this stuff, but I'll just say what I know and hopefully no calculation mistakes.

I am trying to build a detector for small nuggets. I have wound a coil with a resonant frequency of about 750 khz. (310 uH, 11" diameter, 2 Ohm)

So there's about 145 pF

On the output of the preamp I get all sorts of noise, but the strongest is at about 750 khz. I take this to be "white noise" that shows at the resonant frequency because of the enhancement caused by the resonance. Is there a way to get rid of it?

You can dampen it out with a resistor across the coil or you can redo the coil so it has less capacitance-- better spacing and winding technique.


BTW, what's the exact part number of your op-amp?

Paul

[pml]

You should be able to significantly lower the coils capacitance, but if not then just use less coil turns. In this case the op-amp noise is not the problem. So you can simply increase the op-amp gain.

Paul

[Tinkerer]

coil capacitance?

pml,

thanks for the reply. My first reaction was to blame coil capacitance. However, when I increase the power by 300% I get very little difference. Could you explain why that is?
Because of the above I have started looking at the limiting Diode for the fault. The fact is, changing the limiting Diode for different types, does show large differences. What is the explanation for that?

Thanks for the hepl

Tinkerer

[pml]

Hi Tinkerer,
You're talking about the preamp, right? Can you post the circuit diagram of this?

I like using LTSpice/SwCAD III to simulate my circuits
http://www.linear.com/designtools/so...gistration.jsp
It's free and works great.

Paul

[Carl-NC]

I am trying to build a detector for small nuggets. I have wound a coil with a resonant frequency of about 750 khz. (310 uH, 11" diameter, 2 Ohm)
On the output of the preamp I get all sorts of noise, but the strongest is at about 750 khz. I take this to be "white noise" that shows at the resonant frequency because of the enhancement caused by the resonance. Is there a way to get rid of it?

If it's primarily thermal noise, then the integrator will average it out and it shouldn't be a factor. If it's not purely random (say, an AM radio station) then you probably need to do something about it. You might be able to low-pass filter some of it after the preamp.

Another problem at the input of the pramp (5534) is a "kink" at the point where the limiting diode stops conducting. No matter more or less damping resistor, the kink stays.

You'll need to post a plot.

- Carl

[pml]

I think Carls right. My first thought was to first get rid of the resonance because I don't think you want that. If you want to go beyond 5 us then you'll need to lower that capacitance unless you want to get into writing a PIC program which could calibrate all the frequencies via software.

If you want broad bandwidth then the only thing I know is what Carl said, use signal average.

Paul

[Tinkerer]

Scope pix

The coil:11" diameter, 310 uH, 2 Ohm + 10 Ohm resistor with 6ft shielded audio cable.
The preamp: NE5534 inverted, gain 100, (1) - 33, (2) - 22,(3) - 15,(4) and 15 with 10 pf feedback.(5) upper trace, 2V div and 1uS div. is the output.

Lower trace, 200mV div. is the input

Pic (6) is gain 15, upper trace and flyback lower trace 50V div about 175 V peak.


Tx is 60uS, 12 V

I hope this makes sense

Trying to find very small nuggets it is important to take an early sample. Somewhere is a tradeoff between gain and microSeconds of delay. Where is the best spot? I am trying to take a 1 uS sample to the S+H amplifier where I can average and store it. Take several samples and then chose pairs like first and last or first and 5th etc. (PIC) to amplify the difference.

Thanks for the help

Tinkerer

I see there are no numbers with the photos. The sequence is 1,2,3, top left to right, and 4,5,6, bottom, left to right.

[Tinkerer]

Thanks pml, I downloaded the Spice software, now I have to learn how to use it.

Tinkerer

[Carl-NC]

I'm wondering if this kink is caused by the opamp... during the flyback, even with the clamping diodes, it is being seriously over-driven, and right at the end of the decay it comes back into its linear region. So it might have some kickback current into the diodes, which are in the process of turning off, and going high-impedence. Try removing the opamp, and looking at the diode clamp settling.

- Carl

[Tinkerer]

Opamp kikback

Thanks Carl,
I will try that.

By the way, the inverting and the noninverting configuration give similar results. I tried Germanium diodes, 914 and 4158 and the 5817 Schottky diodes. Everyone gave significantly different results.

I also feel that the NE 5534 opamp is getting near its limits.

Tinkerer

[Tinkerer]

Kickback of opamp

Thanks Carl,

below the traces of flyback and limiting diode before and after removing the opamp. You are right. It is the opamp. How do I fix that?

Tinkerer

[pml]

Hi Tinkerer,

Have you looked at the Hammerhead detector circuits:
http://www.thunting.com/cgi-bin/geot...head/index.dat

You wanted 1us detecting, but won't that will be pushing the NE5534 op-amp, which has a slew rate of 13V/us? I think it's possible with the 5534 if you keep the gain down. Your graphs show that as you lower the gain then the 750KHz signal becomes more evident.

Personally I prefer fet amps (not fet op-amp). They have far higher bandwidth, less capacitance, less noise, and extremely high input resistance.

I'm definitely not a big follower of the typical circuits using clamping diodes and such, but once you get that solved then I think you really need to solve the capacitance problem. First, 6 feet of audio cable isn't helping. Also the coil has a lot of capacitance. If you would cut the windings down to perhaps one fourth the amount and just make sure your windings continue forward and never going backwards. In other words, a big no-no for example is if turn #30 is wrapped near turn #1. This adds huge amounts of capacitance. Also, as you know, the more amount of space you add between turns does wonders for lowering capacitance.

Also, if your Rx isn't too close to Tx then you could add some Rx counter turns as close as possible to the Tx to help cancel out the Tx signal on the Rx coil.

Paul

[Ferric Toes]

Hi All,

The 5534 opamp already has back to back diodes internally, so by connecting external ones, you end up with four diodes. I have used the internal diodes successfully with a low power TX i.e. 300uH coil, 4 ohms winding resistance and 27 ohms external. I am not sure how far you can push it, except to say that a "normal" TX with 2 ohm coil and 2 ohms in series, blew the 5534. You must take pin 3 of the opamp direct to ground, rather than through a resistor, for this to work. You can use just one external diode to catch the positive high voltage spike, which is the damaging one. I recently tried 1N4448 diodes as these are lower capacitance, but we are only talking of a few pf, and as pml says, the audio cable is doing you no good in the capacitance stakes. Never use low noise microphone cable. This has a semiconducting layer under the lapped shield and, strangely enough, is noisy when the cable is flexed. Use at least RG58 coax, or try some of the SVHS
video cables, which are of low capacitance.

I usually get a slight positive overshoot, followed by a smaller negative one, before the response goes flat. This also is true for very much faster opamps, such as the AD8055. What you could try is a cermet preset in the bottom end of the damping resistor. If your present damping R is 1k, try a 820 ohm fixed and a 200 ohm preset. You have to watch the spike voltage across the preset, otherwise you can burn the contact, and end up with some real noise. This is best for low power TXs, and it is surprising how critical damping can become, when you are looking for that last microsecond.

Another thing to try - increase the series resistance and put up the repetition rate, so that you end up by drawing the same current from the supply. You can switch off a small current faster than a large one, and you get less back emf. It seems counterintuitive, but all my most sensitive designs (in terms of small objects or poorer conductors) use low power TX pulses at a high repetition rate i.e. 0.1A peak current at 10,000 pulses per sec. You can then use a smaller, low capacitance Mosfet.

Be aware that the coil can detect the coax shield where it enters the coil shell, and give a superimposed decay. SVHS cable is better in this respect as it has a lapped shield of fine wire.

Also, avoid the use of wirewound resistors for the series resistor. They are not completely non inductive and are usually wound with a magnetic wire. Another source of odd responses. Use thick film or metal oxide resistors of appropriate wattage.

Lots of things to try - have fun,

Eric.

[Carl-NC]

Good tips, Eric. Most IC's have 2 ESD protection diodes on every pin, one backwards to supply and one backwards to ground. Although they are usually brute devices that can handle large fast transient currents, I think you would only want to use them in a low-power design, like you said. I know I have fried a couple of opamps when I pulled the external diodes. The bond wires used to attach the chip to the package also have a current limit.

Also, because there is one diode to ground and one to supply, you might not get the clamping level you really want, which could increase the recovery time.

- Carl