Transformer with integrated current sense?

[KaneTW]

Long story short, I'm planning to use a fullbridge topology to drive ultrasonic transducers at various frequencies (around 40-100kHz, with a switching frequency of ~500kHz).

Due to the metal-bodied transducers being in contact with the case, I need to run them through a transformer so that the side that touches the case can be grounded.

My original plan for a driver PCB involved a current sense transformer and direct drive, but since I'm now isolating the entire output: is there an off-the-shelf way to combine the isolation and current sense transformer?

A shunt instead of the current transformer could work, but that'd replace it with an opamp.

Ideas?

[NiHaoMike]

Put a current sense resistor in series with the low side of the secondary. Or swap to a half bridge design with twice the supply voltage to eliminate the need for an output transformer.

[T3sl4co1l]

Nah, you need two magnetic paths to do that.

But current transformers are readily available, have more bandwidth than Hall effect sensors or shunt amps, and better CMRR than amps.

Huh, is that 40-100kHz fundamental with a PWM at 500k?  What's that for?

Tim

[KaneTW]

I need to keep track of phase for resonance tracking, and that becomes pretty difficult with a square-wave drive according to simulations. 

So I use a modulated PWM signal that, after an output filter, appears to give a pretty decent sinewave into a (series resonant) piezo load.

[T3sl4co1l]

Weird...

I'm literally playing with a (same frequency) resonant controller at the very moment; it seems to track phase just fine (it's using a type 2 phase detector, between driven voltage and load current, of a series resonant LC tank).

You would probably want to do the inverse, since a piezo is a majority capacitive load.  That is, current-sourcing inverter, sense voltage for feedback.

Heh, I have a few control boards to spare, if you're interested.

Tim

[KaneTW]

Hmm.
Basically, my current plan is:

Each piezo gets its own fullbridge (via DRV84x2, around 60W with impedances between 5-50Ohm depending on load) and gets supplied a PWM waveform from a MCU (TI C2000 series).

Driver output gets passed through an LC lowpass filter (some issues with varying impedances there) and fed into the output transformer.

Current transformer provides output current, divider gives the output voltage. Both get sampled by the MCU's ADC and converted to phase info.

--

In simulations, with a direct square drive (PWM at fundamental, no filter, just some ~15u output inductors) I'm seeing a lot of harmonics reflected from the piezo (modelled as a C || series RLC) and a lower power efficiency.

With drive at fundamental, an output filter (15uH/69nf) and 100kHz, fairly good efficiency and clean waveforms. But at 40kHz the voltage waveform becomes really ugly (

). You can still detect the phase, but I feel like it'd be more reliable if the voltage didn't have the additional harmonics.

When driving with an optimal PWM (also called programmed PWM) waveform and the same output filter, I get a very clean sinewave at all frequencies of interest. That's a bit harder to simulate, though, since I need to precompute the waveforms.

--

My next course of action is to get a DRV8412 evaluation board and just see how it behaves in real life, since I know people drive ultrasonic transducers with square waves. Either they don't have feedback requirements or I'm missing something.

What control boards do you have? Maybe they'll be suitable.

[KaneTW]

I got the optimal PWM idea from https://purehost.bath.ac.uk/ws/portalfiles/portal/134383936/Power_Converters_for_Power_Ultrasonic_Transducers.pdf btw, if you want to take a look.

[T3sl4co1l]

Ah, that'll just be a matter of using the right filter design then.  Make a model of the piezo if you can, and tweak the filter in SPICE until it behaves.

You can use a filter to invert impedance, to a certain extent.  Basically make an LC tank with that series inductance and the piezo as the capacitor, maybe with some inductance and/or capacitance in parallel with the piezo to tweak its impedance around resonance.

Otherwise, a current-sourcing inverter would drive it nicely, with an inductor in parallel to tune out the bulk capacitance.  You do this usually by say, taking a regular MOSFET full bridge, put a diode in series with each MOSFET (to isolate the body diode; not necessary for devices that don't conduct in reverse: RB-IGBTs, SCRs, vacuum tubes), and instead of bypassing the supply with capacitors, use no capacitors, and supply it with a constant current instead (usually a large series inductor from a current-mode buck regulator, since we don't have general current-source supplies and need to create one from a CV source).

Note that because of the diodes, the high side CMV can be below GND (in the same way that drain current can be negative in a voltage-sourcing inverter), which sucks for bootstrap drivers.  You may need transformer or opto isolated gate drive to do this, or go with a common-ground topology like push-pull (current source into the CT).

Finally, note that a current-sourcing inverter has interleaved dead time, i.e., output power is zero when all switches are on.  All switches turning off means destroying the inverter, in the same way that turning all switches on in a CV inv. destroys it.

Tim