Drive a piezo at 6.78mhz

[ryba7]

Greetings all !

I want to use a 15mm , 6.78mhz piezo.
I have a class E amplifier ( 400 w)  , output at 50 ohms.
But the capacitance of the piezo is very high ( 10 n F), so when i try to send power, the current is too high and no voltage act to make the piezo move. Usually there is a coil parallel to the piezo to raise the impedance , but here a calculation give a way too small value for it (=-= 50nH), and the cable (1m)  has has a capacity and inductance that act.

How can I make my crystal oscillate?

Thanks.

[Whales]

Size/position of piezo = voltage across piezo

Driving AC at high frequencies into capacitors -> very low Z (dead short), ridiculous amounts of current needed to get a target V waveform.

Piezos are hard.  I made a high-voltage amplifier (180V?) to run one a while back for optics positioning, but I could only do a fraction of the full voltage range peak-peak once above a few kHz due to the currents needed & difficulty keeping it stable in a closed feedback loop.  Albeit this was a massive piezo (~1-2uF?)

EDIT: Adding a coil in parallel might allow you to do some clever resonant stuff to avoid having to push/pull all of the current yourself?  But you'll need a really low R coil, a really low R piezo and really low R wires between them.  Your tiny 50nH might make sense, not sure about the feeding cable (maybe worth simulating the piezo C, parallel L and cable L in a spice sim).

[ryba7]

I tried LT, only with the cable inductance ;it works theoretically but it is how to make such a small inductance that bother me.
I have a small VNA and tried measuring what should be a theoretical infinite impedance piezo/one turn coil joined to the cable but this is not the case.

What bother me is that you can find 2.4mhz piezo driver by the score on alibaba, so 6.78 should not be so hard...

[Whales]

I tried LT, only with the cable inductance ;it works theoretically but it is how to make such a small inductance that bother me.

Indeed it doesn't seem very practical.  You would probably have to build the amplifier ontop of the piezo, then make the inductor using a single turn of thick wire in free air.  Then don't put it near anything other than air and wood.

What bother me is that you can find 2.4mhz piezo driver by the score on alibaba, so 6.78 should not be so hard...

What search terms do you use?  I can only find KHz drivers for ultrasonic cleaners.

Putting MHz into a piezo is not easy unless you can live with very little travel distance.

[dmills]

15mm diameter? That is fucking massive for that frequency, not surprised that you are finding it basically undrivable.

When I was doing sonar stuff the usual approach was a fifth order network using a parallel inductor to tune out the fixed capacitance, and often using the transformer leakage inductance to form one of the inductive elements, generally we needed series resistance to kill some of the Q and get a reasonable bandwidth.

Can you get a more reasonably sized element?

[ryba7]

What search terms do you use?  I can only find KHz drivers for ultrasonic cleaners.

https://www.alibaba.com/product-detail/2-4MHZ-atomizer-piezo-transducer-with_60369552443.html?spm=a2700.details.maylikeexp.7.513d733csvGLiI for exemple, but sometimes alibaba is a little creative with the written description...

When I was doing sonar stuff the usual approach was a fifth order network using a parallel inductor to tune out the fixed capacitance, and often using the transformer leakage inductance to form one of the inductive elements, generally we needed series resistance to kill some of the Q and get a reasonable bandwidth.

Can you get a more reasonably sized element?

I will look about this fifth order network, i don't know what it is now. The resistance to kill the Q will absorb most of the power no ? leaving only a small voltage to make the piezo oscillate ?
The thing is that i need it as large as possible, for the largest amplitude. 15mm diameter was good for that.

[JohnG]

A high quality 50 nH inductor is not that hard to get. 50 nH and 10 nF gives a characteristic impedance of about 2 ohms or so, meaning that is also the reactance of the inductor at the resonant frequency.

A Coilcraft 132-05L air core inductor has a Q of 75 at 50 MHz, and since Q for air core varies roughly as the square root of frequency, the Q of the inductor will be about 28 at 6.78 MHz. Not stellar, but not bad. It's about 10 mm long solenoid. You can wind your own bigger one and get higher Q, plus you can adjust the inductance to your needs.

You will still most likely need to do some impedance matching, but that requires having some idea of the piezo impedance to do it intelligently. Otherwise you could be guessing for a while.

John

[Marco]

How about putting a 50 nH in series, measuring the resistance at resonance without cable and then adding a L/Pi/T impedance matching network in between the cable and the 50 nH inductor? The cable is presumably 50 Ohm coax, so that way everything should be matched.

PS. voltage is of course going to swing up a lot, dangerous voltages will be present and if you try to put 400 Watts into it something will go wrong.

[Marco]

A Coilcraft 132-05L air core inductor has a Q of 75 at 50 MHz, and since Q for air core varies roughly as the square root of frequency, the Q of the inductor will be about 28 at 6.78 MHz. Not stellar, but not bad. It's about 10 mm long solenoid. You can wind your own bigger one and get higher Q, plus you can adjust the inductance to your needs.

Does he really need to pass 100 Ampere? A simple small SMD inductor will do 10 amps easy.

[ryba7]

How about putting a 50 nH in series, measuring the resistance at resonance without cable and then adding a L/Pi/T impedance matching network in between the cable and the 50 nH inductor? The cable is presumably 50 Ohm coax, so that way everything should be matched.

I have a tuner that i use for that, all cables are 50 ohms, are you suggesting to place the coil in serial with the piezo directly after the tuner ? The impedance of the serial should be very low, no ?

I always try first at low power.

A Coilcraft 132-05L air core inductor has a Q of 75 at 50 MHz, and since Q for air core varies roughly as the square root of frequency, the Q of the inductor will be about 28 at 6.78 MHz. Not stellar, but not bad. It's about 10 mm long solenoid. You can wind your own bigger one and get higher Q, plus you can adjust the inductance to your needs.

I will try to make such a coil, but when i already tried, the VNA didnt show a great improvement, it always "made real" the impedance of the cable and piezo+coil in parallel, but only at low frequency (6mhz) and it made the resistance low (20ohms) when if the piezo and coil where successfully resonant, their impedance should be high.

[radiolistener]

How can I make my crystal oscillate?

First, measure it's impedance at resonant frequency, then you can decide how to deliver power into it.

You can use ham radio HF Power Amplifier with matching circuit on the output.

What search terms do you use?  I can only find KHz drivers for ultrasonic cleaners.

Chinese ultrasonic water vaporizers (humidifiers) working at 2.4 MHz, but they all usually very low power, about 5-10 W. Here is piezo transducer for such devices: https://www.aliexpress.com/item/32850544059.html

They usually use high voltage power supply (about 24-60 V) with power mosfet and inductor. Here is example of output circuit:



Left inputs connected to a controller 5V TTL IO. Upper one is used to control mosfet with PWM and bottom one is used to sense feedback (it is used to find the exact resonant frequency of the piezo transducer)

But note this circuit is designed for low power 10 W output. If you're needs 400 W, it will needs much higher voltage, at a glance about 160 V and more powerful mosfet. So there is a sense to add HF transformer or matching circuit for amplitude gain.

the VNA didnt show a great improvement, it always "made real" the impedance of the cable and piezo+coil in parallel

it looks that you're forgot to calibrate VNA with the cable to see impedance on the end of cable exclude the cable influence.

[ryba7]

Thanks for all infos. I will perhaps not need all the 400w, i will see if i can obtain my results at lower power, but it would still be nice if it could work.

it looks that you're forgot to calibrate VNA with the cable to see impedance on the end of cable exclude the cable influence.

But the oscillator see all the cable + piezo/coil impedance, no ? So Then if i only measure the coil/piezo i will have a result that do not correspond to what the power source actually see, with the 1 meter cable in place, producing missmatch ?

[Marco]

I have a tuner that i use for that

As you say, the resistance of the piezo at series resonance with an inductor will be rather small. I doubt your antenna tuner has that kind of range.

I think using say a pi network to match the piezo to 50 Ohm (don't even need the original 50 nH inductor then) will prevent any voltage ring up though ... so the source then will need higher voltages to put some power in the Piezo. Using a long cable with a 50 Ohm source might not be the best idea.

[Marco]

How about you make a fast half bridge driver (for instance by modding EPC9506 development board) to drive a square wave, you put a L-match on one side of the coax and a Pi-match on the other side. In simulation that seems to work to keep the circulating current through the amplifier somewhat under control and dump decent power into the Piezo. Equivalent circuit for the Piezo guessed based on this.

This is not very broadband and the component value errors will pull the resonance a couple 100 kHz away from 6.78 ISM centre frequency (in this simulation 6.5 MHz).

PS. oh, that pi match is wrong ... if I put in better values the resonance is doesn't get shifted so much, but time domain simulations stop working and power to the piezo (ie. R1) plummets. So take the exact circuit and values with a giant heaping of salt, though I still think you will need a matching circuit on both side of the coax.

[RoV]

You never talk about bandwidth, it seems you need to emit a fixed frequency. In this case, normally, the piezo is carefully optimized to make it resonate mechanically very close to the frequency of interest, where it is no more a simple capacitor. This is done with a careful choice of the crystal and, often, with a suitable resonating mass on the side opposite to propagation. You also never talk about the medium of propagation, I assume it is oil or water or a solid (through a suitable couplant), because  I believe it is impossible to produce a high power sound wave in air at this frequency. The impedance presented by the crystal varies radically according to the environment it is put in, so the impedance should be measured in the actual environment.
Only at this point matching can be taken under consideration.
Keep also in mind that, when the piezo is resonated (mechanically, electrically or both), the voltage on it may become quite high as you increase power. Piezos, if excited above a critical voltage, permanently loose their properties (they are polarized during manufacturing, it is the electrical equivalent of magnetic hysteresis). You must avoid overcoming that critical value, typically 300-400 V for a single crystal, more for a sandwich.

[ryba7]

I think using say a pi network to match the piezo to 50 Ohm

I tried that first, but couldn't find any valid configuration to match, i think i always over or undershoot.

You never talk about bandwidth, it seems you need to emit a fixed frequency. In this case, normally, the piezo is carefully optimized to make it resonate mechanically very close to the frequency of interest, where it is no more a simple capacitor.

Yes , single frequency, but even then the impedance is too low, so i cannot put any voltage inside.

[ryba7]

I tried to measure directly the coil/piezo in parallel with the scope. I obtain resonance for large value of the coil, at 6mhz, but when i reduce the inductance to increase the frequency, the resonance became weaker. I use a 100X probe.
With a cable and the VNA, a large coil give a real impedance at low frequency with a SWR of 1, but with a smaller coil, the frequency rise but the impedance remain complex and the SWR is higher.
Why does the resonance become weaker with a smaller coil, it should stay the same , No ? 

[Marco]

Have you actually verified the resonance frequency without the coil? (ie. just attach it to the VNA with very short wires.)

[ryba7]

I made the test , who do you understand this result ?

[Marco]

Your cursor is at anti-resonance, resonance is on the other side of the Smith diagram.

Your start frequency is too high. Are you sure it's really connected by properly short wires BTW? A standard BNC to alligator clip converter is not short enough. Just use a coax cable with BNC to connect it to the VNA, cut the coax and strip as little of it as possible so you can connect the shield and core to the Piezo.

[ryba7]

What do you mean by other side ? Minimal real impedance ?

Yes, i connected as short as i could, only 2cm of sma coax, the core to one side, the shield soldered to a 1cm copper wire to the other side.


I measured again, The lowest real impedance is 97mhom and 1.6nH at 6.535mhz, So this is the resonnance frequency ? i didnt know that you could measure it that way.
Thanks for your help !

[Marco]

What do you mean by other side ? Minimal real impedance ?

Yah, the resonance was below your start point, then it goes high impedance at anti-resonance and then at higher frequencies because of the 10nF capacitance it goes back to low impedance.

I measured again, The lowest real impedance is 97mhom and 1.6nH at 6.535mhz, So this is the resonnance frequency ?

Sounds about right, though I'd expect the impedance to be slightly capacitive ... but ehh, the electrical model is just a model.

Now that's something you can plug into a Pi-matching calculator. Using this with 0.065 reactance and Q=20 I get 62 nH, 9.7 nF source capacitance and 178 nF load capacitance. Doable values. Not an endorsement of that exact calculator, I just googled it. With your RF amplifier which is presumably 50 Ohm matched already, that's the only network you need at the Piezo side. If you were ever to build a non matched driver it would probably need matching at the generator side too with the coax.

With 48V peak to peak this should get you a little over a Watt of ultrasonic power (this is an inherent limitation of Pi matching I think, with a transformer you could probably get much more).

[ryba7]

Thanks for breaking it down to me, so i could check if i understood correctly !
I build the generator for 50ohms  indeed.

The value are very low, no wonder i was unable to tune to this...

[N7PJ]

Assuming your transducer is air backed and the front in water -

Attached is a measurement of a 7MHz transducer made with an Agilent 4395B. This one is a 19mm focused bowl rather than a disk, but it should be roughly similar to yours. The back of the piezoelectric element is air in a sealed chamber. The front load is water in a cup including something like tissue paper or other lossy material to suppress reflections. The connection to the network analyzer is a short (75mm) length of twisted pair. The pair connects to the analyzer at the point where calibration is performed.

At a chosen frequency, an L-network (near the transducer) can transform the impedance to 50 Ohms. For this example, I'd choose to operate at or near parallel resonance to minimize losses in the matching network. At 7.08 MHz, the transducer resistance is 13.7 Ohms and reactance -1.41 Ohms. An L network to make this 50 Ohms is 732 pF shunt on the 50 Ohm side and 533 nH in series.

Depending on the PZT, metallization, mounting of the element, inductor losses etc., electro-acoustic efficiency is typically better than 80%. Voltages at 400W are highest at 50 Ohms, 200Vp. Even running short bursts I'd be a little uneasy about the long-term health of the piezoelectric element.

Hope this helps,
PJ

[ryba7]

Very interesting, i would like to do the same graphs in my nanovna !