AC-coupling cap for AC inductor drive

[ricko_uk]

Hi,
with reference to the attached schematic, I need to choose the right capacitor value. These are the requirements:
- op-amp input/output is a 200 KHz sinewave
- the amplitude of the AC signal across the inductor (i.e. the OUT signal) is fed into another op-amp (for later ADC conversion)
- the op-amp O/P sinewave is DC shifted to 2.5V and the AC amplitude vary from 0.5V to 4.5V peak-peak to adjust the current across the inductor depending on the application
- resistor R3 limits the current going through the inductor and is usually fixed to 100R
- capacitor C2 provides AC coupling so the OUT signal is centred to ground

Because of the op-amp AC peak-to-peak amplitude van vary from 0.5 to 4.5 the opamp output current (i.e. also the current through the inductor) can reach up to 45mA. That is assuming R2 is not mounted on the PCB.

Questions:
1) how do I choose C2 value? does the signal frequency and/or the op-amp output current (when driving R3 & coil) matter in the choice of the capacitor C2? How? How I don't think that if the current is several amps (just as an example) a tiny pF capacitor would work, or maybe it does?
2) does the value affect/change the phase across the inductor compared to the original signal (OP-AMP output sinewave)? If so would a small one keep it to a minimum?
3) If any value would work, any benefits in a large value (or small one for that matter)?
4) I don't see the point of using R2. It was from a reference schematic. Does it serve any practical purpose in AC coupling or otherwise?

Many thanks for all feedback.

[T3sl4co1l]

No idea.  What are you doing with it?

Tim

[ricko_uk]

Just driving the coil and check amplitude variations across it (i.e. the OUT signal) based on metals below the coil.

[T3sl4co1l]

Does it need to be sensitive, or is it more of a gross "yep there's metal literally in the coil / nope it's probably empty" sort of thing?

What about, say, supply consumption, any concern there?

Tim

[Prehistoricman]

1) how do I choose C2 value? does the signal frequency and/or the op-amp output current (when driving R3 & coil) matter in the choice of the capacitor C2? How? How I don't think that if the current is several amps (just as an example) a tiny pF capacitor would work, or maybe it does?

With your example value of 100 ohms, you won't go above 20mA even with an infinite capacitor.
The amount of capacitance you need (to get full signal conduction) does depend on frequency. Higher frequency = smaller capacitor. For example, at 1kHz with a 1µF cap, you get a peak of about 8mA. But increase that to 22µF and you get 17mA.

2) does the value affect/change the phase across the inductor compared to the original signal (OP-AMP output sinewave)? If so would a small one keep it to a minimum?
3) If any value would work, any benefits in a large value (or small one for that matter)?

Yes and no. The C, R, and L (inductor) form a high-pass filter. If your drive frequency is too low, the capacitor will be high impedance. If too high, the inductor will be high impedance.
In the case of 200kHz, this would be 'too high'. You can use a 1µF or greater cap. You aren't interested in the cap's affect on the response (it should just conduct your signal and block DC) so you might as well make it bigger like 10µF or 22µF.
Near the two poles of this filter, the phase will change. I've attached a screenshot of a simulation for 22µF. Green is the current through L1, red is the voltage at OUT. If you swap R3 and L1, you can sense the current as it will be proportional to the voltage across the resistor.

For detecting metals with a coil, you are expecting the inductance of the coil to change. To sense this, the drive frequency should be near the inductor's pole so that the output amplitude changes (phase will too). The area of interest (as seen in the graph) is around 30kHz - 100kHz. This region is where the inductor's impedance is increasing and therefore the amplitude of OUT is increasing.

If you want to play around with values, I suggest you get some circuit simulation freeware such as LTspice, TINA-TI (this is what I use and prefer, it's easier to use than LT), etc. The second screenshot shows how the frequency response changes as L1 increases from 80µH to 160µH (for example).

4) I don't see the point of using R2. It was from a reference schematic. Does it serve any practical purpose in AC coupling or otherwise?

R2 pulls the right side of the cap to ground. It is intended to be a high value - 10k and above.
You don't really need it in this circuit because R3 and L1 already pull it to ground.

[ricko_uk]

Thank you T3sl4co1l,
I know it's not a scientific definition, but it does need to be as "sensitive as possible" so we can detect (in some cases) the distance between the coil and the metal.
Any ideas/suggestions?

As a separate but related topic, we tried choosing C so that they resonate at a specific frequency we want to use. The issue is that we noticed that at resonance the phase shift changes/swaps to the opposite value extremely abruptly, practically instantaneously. Sometimes we want to detect the phase, is there a way to make the phase swap more gradually? I tried playing around with the series resistor's value but does not seem to make a difference. Any suggestions?

[Zero999]

What sort of metal is it? Ferrous metals will increase the inductance and non-ferrous decrease it.

How about building an LC oscillator and monitoring the output frequency? If it goes, down (up), the inductance has, increased (decreased).

[T3sl4co1l]

Thank you T3sl4co1l,
I know it's not a scientific definition, but it does need to be as "sensitive as possible" so we can detect (in some cases) the distance between the coil and the metal.
Any ideas/suggestions?

Hi,

Not much sensitivity is needed just to detect a change in distance.  Things get a bit more tricky if you're asking about identifying metals, or their condition, even.  (I've seen one case where the metallurgical hardness of a steel item was deduced using an eddy current tester.)

Start with your requirements.  What materials are you sensing, in what geometry, to what precision and accuracy?  How often does this reading need to be performed?

I will give you this food for thought: "as sensitive as possible" brings to my mind, measuring proton precession in water.  NMR.  Which isn't really even all that sensitive an application, but when it's done with dilute chemical samples it gets a bit challenging.  But in any case -- did you know water has an RF resonance?  It depends on DC magnetic field.  In Earth's field, it's pretty weak, and down in the kHz.  Normally the sample is placed in a nice strong (~1T+) magnet, which raises the resonance to 30MHz or so (and also gets stronger).  You put a test tube of water, inside a coil, inside a magnet, and measure the coil impedance.  It's got an imperceptible dip at the resonant frequency, very narrow (a few ppm across).  The equivalent circuit is a notch filter, with a very shallow notch, and a Q of >10^5.  Well, with a little care, the background can be nulled out, greatly magnifying the notch, showing the resonance in plain form.  When applied to chemical compounds, multiple peaks are seen (spaced a few ppm apart, i.e., at frequencies a few 10s of Hz apart, relative to the ~30MHz center frequency), elucidating atomic structure of the compounds(!).

It's not clear what you're doing, but it doesn't seem likely at this point that "as sensitive as possible" will be your biggest problem.

Tim

[ricko_uk]

Thank you Tim and Zero999,
replies to each below.

Tim,
It is only in some cases that we need to detect distance. In other cases the same system/circuit is used to detect cracks in the metal. In yet other applications the system is used for checking variations in thickness of a weld or of extruded metal. So I agree, in comparison to your example it is definitely not that extreme.
To answer your other question, materials are only common stainless steels (304, 316, etc). The shape is always sheets. The defects depth to be detected vary from 0.05mm to 3mm. The "horizontal" size can be anything from a thin line 0.1mm long/wide to several cm long. All we need to do is measure the depth not the horizontal shape or size.
The linear motion of the device over the steel is relatively low 300 mm/sec. Having a peak detector allow to store the deepest depth detected which is all we care about. The peak detector is reset 20 times per second or more (we can set that).

Zero999,
the material to be sensed is stainless steel, the various common grades (304, 316 etc). What type of LC oscillator would you suggest to have the largest possible output frequency variation for any given "microscopic" variation in inductance? Your suggestion is very interesting for a different but similar project we are working on!!

Thank you again both of you!!
Modify message

[Zero999]

Zero999,
the material to be sensed is stainless steel, the various common grades (304, 316 etc). What type of LC oscillator would you suggest to have the largest possible output frequency variation for any given "microscopic" variation in inductance? Your suggestion is very interesting for a different but similar project we are working on!!

There are many different types of LC oscillator designs. In theory, it doesn't matter what frequency you choose, the percent change in frequency will be the same, given the same percentage change in inductance. In reality, the magnetic properties of the metal will be frequency dependant. 200kHz probably isn't a bad frequency to choose. A colpits oscillator can be made using a single transistor, such as a BJT., a CMOS logic inverter IC, or an op-amp.
https://upload.wikimedia.org/wikipedia/commons/b/bf/NPN_Colpitts_oscillator_collector_coil.svg
https://en.wikipedia.org/wiki/Colpitts_oscillator
http://www.learningaboutelectronics.com/Articles/Colpitts-oscillator-calculator.php

[ricko_uk]

PrehistoricMan a while back wrote in a reply above in this post:
"For detecting metals with a coil, you are expecting the inductance of the coil to change. To sense this, the drive frequency should be near the inductor's pole so that the output amplitude changes (phase will too). The area of interest (as seen in the graph) is around 30kHz - 100kHz. This region is where the inductor's impedance is increasing and therefore the amplitude of OUT is increasing."

Questions:
1) What does it mean to be "near the inductor's pole"?
2) How do I find where the inductor's pole (frequency?) is?
3) in his reply's attached graphs, where can I see the inductor's pole?

Thank you

[Prehistoricman]

1)
In layman's terms, a pole is a frequency where the characteristic has greatest rate of change. Trust me, that's actually simplified compared to what a pole really is.
That's not a foolproof definition but it's something to use to identify a pole.
To be "near the pole" is to have a frequency close to the pole frequency.

2)
Calculations, experiments, or simulations. I would suggest the latter two

3)
If you look at the second image, upper graph, you see three poles. One is made by the capacitor and can be seen at 80Hz. You can see how the gain curve is changing from a diagonal line up to a horizontal line. The part of the curve where that change is 'fastest' is the pole frequency.
The second pole is due to the inductor and is about 6kHz. The third is also from the inductor, and is at about 200kHz.

However now I'm simulating this again, I can't reproduce those graphs. That's weird...
Also I wasn't really right in saying that you want your frequency close to the pole. What you want is the frequency to be in the middle of a steep part of the graph. A steep part that is made steep because of the inductor.

I've attached a new sim + results that's a bit easier to understand. It's also very intuitive when you think about how capacitors and inductors behave (in general) across frequency.
The idea frequency in this one would be something like 200kHz. High enough such that changes in inductance will result in large changes in amplitude, but small enough such that there is a big enough signal to observe.

It's also worth mentioning that my idea for this project was very different to Zero999's. I don't have enough practical knowledge to evaluate the two methods against each other. However mine's very easy to build and observe with a signal gen + scope in a few minutes.

[ricko_uk]

Thank you PrehistoricMan,
could it be that the simulation does not come out the same because you inverted the inductor and the resistor compared to my initial schematic?

I can see the horizontal scale is logarithmic but not the values. I tried to work them out but if the first one is 1 then the 200KHz is outside the graph. What are they?

Thank you

[Prehistoricman]

Thank you PrehistoricMan,
could it be that the simulation does not come out the same because you inverted the inductor and the resistor compared to my initial schematic?

Yes, I know. I couldn't reproduce the original results for the original schematic. I tried for about half an hour.

I can see the horizontal scale is logarithmic but not the values. I tried to work them out but if the first one is 1 then the 200KHz is outside the graph. What are they?

Thank you

Whoops haha! It starts at 10Hz, ends at 1MHz, just like the other graphs.

[ricko_uk]

Thank you PrehistoricMan,
so when you say "What you want is the frequency to be in the middle of a steep part of the graph. A steep part that is made steep because of the inductor." you mean better frequencies would be slightly higher, perhaps 500KHz with reference to your graph?

How do I shift the inductor's down-going slope further to the left so that 200KHz is on the steep part of the downward curve like you suggested?

Thank you

[ricko_uk]

Also...
I just simulated the same circuit in Altium (see attached pictures) and ran a AC analysis to play with the values but the output comes out a weird shape with ringing and totally wrong/different frequencies...?
Any idea what's going on?

[Prehistoricman]

I don't know why your simulation looks odd, but I have figured out why my original graph looks so different.
The inductor's series resistance matters A LOT to the behaviour of this circuit. I seem to have made it 8 ohms in the original post, probably by mistake.

Parasitics matter!

you mean better frequencies would be slightly higher, perhaps 500KHz with reference to your graph?

How do I shift the inductor's down-going slope further to the left so that 200KHz is on the steep part of the downward curve like you suggested?

Yes. Between 200 and 500.

Resistor value and inductance affects the location of the ideal region. Series resistance to the inductor seems to reduce the size of the ideal region.

[Zero999]

How about an RL oscillator, using the 555 timer? The inductor is an air core coil with a large enough surface area. There are various calculators for the inductance of an aircoil on the Internet. Ferrous metals will reduce the frequency and non-ferrous increase it.

[ricko_uk]

Thank you PrehistoricMan and Zero999!