1
\$\begingroup\$

In my setup, I have a VCO CD4046 Phase Locked Loop IC generating a square wave, and an ultrasonic transducer set that transmits and receives that setup. You can imagine it looking something like this.

enter image description here

The problem I am currently facing is that the VCO itself (not the square wave that the VCO is generating, but the VCO IC itself) is generating a lot of noise, picked up by the ultrasonic receiver. This is what my receiver input looks like without the ultrasonic transmitter connected to the system; the square wave from the VCO's signal output is not transmitted through. This is all coming from the VCO IC itself.

Disconnecting the VCO does remove the problem, so I am sure it is a VCO problem. The noise is at multiples of the VCO frequency. I am running it at 28kHz here, and the noise is detected as 56kHz or 140kHz.

enter image description here

This is a problem because my expected signal should have a receiving magnitude around 50mV. As you can see, the VCO is generating over 100mV of noise for the receiver. It doesn't help that my full circuit has two of these things.

Any suggestions to minimizing this noise would be greatly appreciated. If there's any more information needed to solve this issue please let me know.

Edit: (suggested by Kuba) The schematic that I'm working with right now. Yes, the VCO is disconnected from the transmitter. All noise originates from the VCO IC.

enter image description here

The VCO connection. The VCO is currently generating a 28kHz square wave.

enter image description here

The US receiver. Thing at the bottom is an Op-Amp that gives the signal gain.

enter image description here

The full schematic. The one at the middle is the transmitter, one at the right is the receiver. One at the bottom left is the VCO. There is an unconnected PLL to the top left (but also causes the noise problem when connected) and an unconnected Diff. Amplifier to the top right. And yes, the VCO is not connected to the transmitter.

The cable that the transmitter is connected to is an external waveform generator I would use to create a sample wave. At the time I took a picture of the oscilloscope, it was inactive.

Yes, I am using one of the power rails on the right as a ground. This is because I set up a floating voltage of 7.5 volts using the ground rail and couldn't be bothered to swap the two. That set of power rail is connected to the receiver ground, though I'm not sure how that affects anything.

I have not tested the effects of lowering the supply voltage on the VCO. It's certainly testable, I'm guessing it will lower the effects. Though a method of eliminating the electrical noise is still a lot more preferable.

enter image description here

\$\endgroup\$
9
  • 1
    \$\begingroup\$ Shared power rails sound likely to be the issue. \$\endgroup\$
    – Andy aka
    Commented Mar 24, 2022 at 12:26
  • \$\begingroup\$ @Andyaka My receiver isn't using a power rail, though they are using connected ground rails. Though I am curious, how would shared power rails affect the ultrasonic receiver? \$\endgroup\$ Commented Mar 24, 2022 at 12:45
  • \$\begingroup\$ Show all design details and photos of power,gnd,signal , freq. and filters with jitter and piezo specs \$\endgroup\$ Commented Mar 24, 2022 at 13:07
  • 3
    \$\begingroup\$ Breadboards is an EMC nightmare by definition. Also there is notably almost a complete absence of decoupling caps. Please post the schematic. \$\endgroup\$
    – Lundin
    Commented Mar 24, 2022 at 15:29
  • 1
    \$\begingroup\$ Ceramic capacitors can generate sound because they are piezoelectric \$\endgroup\$ Commented Mar 24, 2022 at 15:47

1 Answer 1

3
\$\begingroup\$

TL;DR

This circuit will only work sensibly with shielding and a tight layout on the breadboard. With high impedance sensors like piezos, layout and shielding are 90% of the job.

Almost any circuit on the TX and RX end will work as long as there's proper shielding between TX and RX. The can of the RX transceiver needs to be grounded. With high gain RX amplifiers, separation of the first 1-2 stages from the rest is desirable to prevent oscillations.

The receiver needs to be selective, and the more selective the better.

The (Ultra)Sound and the Transceiver

The wavelength at 29kHz in "room air" is about 12mm. There'll be lots of variation in amplitude due to resonant peaks when checking out short TX-RX paths. When using e.g. a hand as a reflector, minuscule movements on the order of millimeters cause large magnitude changes. The phase tracks reflector position very well, and lends itself to demodulation using the PLL :)

I've had a visually identical transducer laying around. Just for kicks I've measured its equivalent circuit to be 2.0nF in series with 120 Ohms. The transducer is tending to be resistive above 0.66MHz, and capacitive below.

If we wanted just a dummy load for the transducer, it'd be a 2.0nF ceramic cap in series with 120Ohm resistor.

The Big Picture

I presume that your eventual goal is to close the PLL loop with part of the path being acoustic. For such experiments, we'd want a circuit where a barrier, or lots of open space, between the transducers opens the loop and makes the VCO free-run. When you remove the barrier, the loop should close and the PLL should capture the return signal and lock its frequency and phase to the reference.

As soon as that works, we've got ourselves a nice doppler velocimeter, with the doppler signal being the demodulator output. This would be a good base for whatever ultimate application you have in mind. Even if you want the VCO to free-run, observing the PLL behavior as you open and close the loop ensures that the receiver is receiving acoustic signal and not just picking up electrical interference from the transmitter.

As a starting point, we can put together just the circuit you want: a VCO, TX driver, acoustic path, and RX amplifier.

The Transmitter

The transmitter can look as follows:

schematic

simulate this circuit – Schematic created using CircuitLab

The 4041 may be overkill for this application, but it certainly does the job, and will support lower impedance transducers too.

R1 is the ballast resistor for the 5.1V shunt Zener regulator in U1. About 1mA flows through it. C1 sources the AC current for U1. C2||R2 provides the frequency setpoint voltage. The frequency is in the right ballpark if you start with the wiper centered (in the middle of the range). C3 is the timing capacitor.

R7 slightly decouples the VCO output from the load presented by the push-pull drivers. Ideally, you'd want an additional buffer gate (can be an inverter, NOR, etc.) between the U1's output and the rather heavy capacitive load of U2's inputs.

Since U2's inputs have threshold at 4.5V, vs. U1's output switching around 2.25V, an RC shifter is used to make the input symmetrical with respect to U2's threshold.

U2, a CD4041 quad true/complement driver, provides a relatively low impedance (a couple Ohms) 18Vpp drive across the transmit transducer. C4 provides the AC current for U2. R5+L1 block the driver's AC current from escaping and polluting the rest of the circuit.

I've laid out the transmitter on the breadboard as shown below. The power supply rail is quite clean, with at most a couple mV of remnant square wave present. This can be adequately coped with by the receiver circuit.

The transmitter section, including the VCO (U1) and PA (U2) as laid out on the breadboard

U1 is on the left, U2 is on the right.

Note the minimal area of loops where the AC supply currents for each of the chips flow, as annotated on the photo below. The loop areas are comparable to the areas of the chip footprints themselves. That's about as good as you can get it on a breadboard without having capacitors straddle the chips. I purposefully didn't do that for clarity's sake, as the performance is adequate as-is.

Detail around U1 and U2, with AC supply current loops marked up.

Shielding is critical. In production, the PLL and the power amplifier need to be individually shielded. For solderless breadboard, there at the very least needs to be a metal shield under the transmitter section. This shield can be separate from the receiver's shield.

The Receiver

We need about 1000x gain, evenly split between three 10x stages. The bandwidth of each stage, using "jellybean" 0.7MHz GBW amplifiers will be about 70kHz - plenty enough for this application.

In simple free-space object reflection setups, the gain of the acoustic path varies considerably with small changes in object distance and orientation, due to constructive and destructive interference.

Multiple receivers could be used to work around that, but an automatic gain circuit seems necessary as well.

I'm using TL072 since I've got lots of them laying around, but essentially any op-amp that'll work from 9-10V will be fine for this application.

Just as fine would be a common-emitter amplifiers, but let's say we'll stick to ICs for this design.

The first 3 stages of the amplifier can look as below.

The amplifier should be selective. I didn't bother with a 2nd order bandpass for each stage, although that would be ideal.

schematic

simulate this circuit

At minimum, a shield under the breadboard and a shield between the transmitter and the receiver are critical.

That should take care of the interference you're seeing in your circuit. It simply will not ever work satisfactorily when assembled on a table that acts as a capacitive coupler between everything. Neither will a wide

Again, just to make sure no one wastes their time: This circuit has no chance of working properly without shielding. On a PCB, physically smaller shields will do a fine job. On a solderless breadboard - nothing succeeds like excess.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.