Project Problems! Need guidance.

You will need to post a circuit diagram before anyone can help diagnose, or all we will be doing is guessing.

 

Also post some photos of your coil assembly.

 

Yeah, post a schematic with labeled components (so we can make specific comments) and pictures are always appreciated. For specialty parts (the transducers maybe?) identify by manufacturer and model or part number.

Is your tank circuit 470 nF capacitor resonant with the 135 μH inductor at the driving frequency? If so, there will a LOT of circulating current to heat up the 22 Ω resistor in series with the capacitor. What IS the driving frequency?

This looks like a really FUN project. I hope you make a lot of money if you can get it working! NDT (non-destructive testing) is a "big thing" in aerospace applications.

 

Interesting. What op-amp are you using for the output driver in the voltage-to-current converter?

What happens to the voltage-to-current converter inputs when not driven by the multiplexer? Do they just float?

Do you switch the multiplexer only on zero-crossings of the input signal?

Do you fade the input signal to zero before switching the multiplexer, then fade it back up?

How many cycles of input signal go through on each multiplex selection?

 

According to the schematic, there are four voltage-to-current converters, and the mux only drives one at a time. The other three will be undriven. Their input voltages will be undefined unless you add a resistor to pull them to 0V.

You need to add a zero crossing detector that monitors the 20 kHz signal, and only change the multiplexer setting at the instant of the zero crossing. This isn't enough to give clean transitions with no disturbances; the best way would be to fade the signal in and out for each magnet over a period of several cycles. The quicker the fade-up or fade-out, the more modulation products are generated. Look up amplitude modulation sidebands.

You could do this using four VCAs (voltage controlled amplifiers or attenuators) instead of a multiplexer. That would also remove the need to synchronise with zero crossings. You can make VCAs with the LM13700. They require an analogue control signal.

 

Changing the capacitor value will indeed affect the resonance frequency, f = 1 / (2pi√(LC)). For the values you listed (0.47 μF and 135 μH) resonance occurs at about 19,980 Hz. Close enough to 20 kHz, but small variations in frequency around the resonant tank frequency can have a large effect on the circulating current in the tank, depending of the Q of the tank. The coil resistance broadens the resonance response by lowering the Q.

Since you are pulse-exciting the coil by switching its excitation through the CMOS analog multiplexer, things get more complicated. It would be interesting to know how much the coil "rings" between excitation pulses. Is there a particular pulse repetition rate that the Arduino UNO produces when stepping through the addresses of the multiplexor? Just curious. I don't think this is related to your resistor and capacitor overheating.

Capacitors in parallel-resonant tank circuits can have large circulating currents at resonance. With an oscilloscope you should measure the voltage waveform across the 22 Ω resistor, or with a clamp-on oscilloscope current probe measure the current through the resistor in series with the LC parallel-connected tank circuit. Do this just to get some idea of what is going on. The 22 Ω resistor should be temporarily moved from its present location to the ground end of the coil for a voltage measurement. Do this to avoid having to make a differential voltage measurement across the resistor in the presence of a large common-mode AC potential.

There are specialized non-polarized capacitors, called "snubber" capacitors, that are specifically designed to handle low frequency, high-amperage, currents without overheating. Check an on-line distributor for part numbers and available values. These caps generally have at least 600 VDC rating. A spec sheet for a typical snubber can be found here. I really doubt the capacitor is heating from circulating currents at the power levels you are providing. Any good-quality polyester or polypropylene capacitor should be okay. Make sure its voltage rating is at least twice the peak-to-peak voltage you measure across the tank.

Your 11X-gain non-inverting op-amp circuit (10 kΩ feedback to 1 kΩ inverting input to ground) driving the paralleled OPA522 voltage-to-current converters is probably going into rail-to-rail saturation with a 2 V RMS sinusoidal input. The peak voltage of 2 V RMS is 2.82 V which becomes 31 V peak (62 volts peak-to-peak) after an 11X gain. The op-amp is powered from +/- 15 VDC supply rails, so assuming it is even capable of linear rail-to-rail output, that still leaves you lacking enough headroom to pass a 2V sinusoidal drive signal. When this op-amp saturates you will be essentially applying a DC pulse instead of a sinusoid to the coil. The current will rise appreciably. I would suggest decreasing the feedback resistor from 10 kΩ to 3.9 kΩ, thereby decreasing the gain of this driver stage to 4.9.

Why are you not exciting the coil as a series resonant circuit instead of a parallel resonant tank? Parallel resonant tanks are high impedance to external exicitation, and it is difficult to control the circulating current in a parallel resonant tank. Just asking. I know you said the design worked fine before you added the voltage-to-current conversion.

.

 

Attached pdf analyzes the load current through the 22Ω resistor and parallel LC tank from DC to 100 KHz. Note that the power dissipation through the resistor at 60 KHZ is almost the same as at DC. As hevans1944 observed, due to the amplifier gain there would be a highly clipped waveform driving the load, rich in harmonics. With a 15V signal amplitude the third harmonic could be as high as 5V which equates to approximately 1W power dissipation in the resistor. That is certainly enough to burn a quarter watt resistor, and enough to warm up a 5W resistor. Also, the accuracy of the LC values may cause additional power dissipation in the resistor if resonance is not exactly at the driving frequency. Does the metal sample change the inductance of the coil?

 

Wow! What a can of worms we have opened up here!

The first problem you need to solve is the non-linear performance of the transmitter circuit following the CMOS MUX, caused by the 11X gain stage going into saturation. As I mentioned in post #13 you MUST lower the gain so the amplified pulse output of the MUX channel is a sinusoidal waveform NOT a square pulse. At maximum output current of 200 mA, the OPA551 can only swing to within 3V of the supply rails. With ±15 V supplies, your sinusoid will be limited to ±12 V peak-to-peak amplitude, which means a maximum gain of only about 4 for a 2V RMS input can be accommodated without clipping the waveform.

Note that your circuit configuration is NOT a voltage-to-current converter. It is a high input impedance voltage buffer with higher current capability to the load than a garden-variety op-amp. There is no negative feedback to make the output current proportional to the input voltage, which is what a voltage-to-current converter circuit does. The load current for any particular output voltage is a function of the load resistance, mostly 22 ohms in your circuit.

Getting the current you need from the so-called voltage-to-current amplifiers (the paralleled OPA551 op-amps) is determined by their output current capability driving the 22 Ω resistor and the LC tank. The Texas Instruments datasheet says the OPA551 is limited to ±200 mA, so two in parallel will max out at ±400 mA. This is the peak-to-peak current; the RMS current from your sinusoidal pulses will be lower. The datasheet offers a solution capable of providing 1 A load current: a pair of external NPN and PNP power transistors connected in a classic "totem pole" circuit. There is a 100 Ω resistor from the op-amp output to the load to allow the op-amp to source current to the load when the output is less than the forward bias junction voltage (about 0.7 V), which occurs as the sinusoid waveform goes through zero. With this approach you only need one OPA551 per coil because most of the load current is provided by the totem-pole transistors. The load current will be determined by the load impedance and the output voltage per Ohm's law. Note that negative feedback causes the totem-pole output voltage to be proportional to the input voltage.

Ringing occurs in a tank circuit when the excitation promoting oscillation in the tank is abruptly removed. The oscillations continue but at decreasing amplitude as the tank energy is absorbed by ohmic resistance losses. How long this occurs depends on the losses and is related to the Q of the tank circuit.

 

It should be easy to drive a tank circuit with a sinusoid at the resonance frequency since the driving current will be zero. Of course, the current will be non-zero to initially energize the resonant tank, and to overcome any losses especially in a lower-Q tank. So why should a high-current driver circuit be necessary?

However, if the driving signal is at a fixed frequency it will likely not be exactly the same as the resonance frequency of the tank and that will increase the drive current. Which leads to the question, what is the magnitude of the signal current in the coil? Would it be better to drive a simple coil directly or drive a tank off-resonance?

With that in mind this analysis looks not just at the driving current through the load resistor but the inductive and capacitive currents as well, including several values of wire resistance in the coil. Note that in terms of the coil current there is nothing special about the resonance point, but including the capacitor is a way to decrease the driving current in the vicinity of resonance.