What is this half-bridge topology? (Ultrasonic driver)

[mdubinko]

I'm looking at a consumer-grade 35W ultrasonic cleaner that stopped working. But more importantly, I'm trying to understand how it was *supposed* to work in the first place. I've double and triple-checked, and the diagram attached is substantially correct. What topology is this? What is the theory of operation?

From what I've seen of IR2153 half-bridge drivers, the usual configuration is to switch one end of the load between +V and GND, and tie the other to V/2, usually through a capacitive divider. Often with a step-up transformer in the mix.

In this design, there is a substantially sized transformer but... (I've unsoldered it to manually pin out the connections and verify) it's connected with the primary in series with the single piezo element, and the secondary grounded on one side, the other running off to the RC timing portion of the IR2153.

One end of the piezo is also grounded. Is this supposed to be a resonant configuration? With the feedback helping the timing synchronize to give it a new push at just the right time? In that case the 8nF of the piezo + the 4.7nF of the capacitor across it doesn't seem big enough to store much energy during the 50% of time when the high-side MOSFET is off. I've seen a number of ultrasonic driver schematics randomly posted across the internet, but never seen anything like this. What would be a reasonable estimate for the p-p voltage seen across the piezo when it's working? Any clues appreciated! -m

[Berni]

Looks like that is a current sense transformer to monitor the current going into the piezzo and shift it down to ground where the chip can look at it.

Not sure exactly what it does since im not familiar with these chips. It could be a feedback loop to make the output frequency track the piezos resonance frequency (This can shift some as the piezzo warms up). Or it could be some sort of power sensing to keep driving the piezzo at the same power level as conditions change.

[Tantratron]

This is a typical resonant half-bridge topology which you find as well in fluorescent electroded tube ballast or induction (electrodeless) lamp ballast. You have a LC serie tank where you put the load (resistive, lamp or piezo) always into the C resonant part.

[2N3055]

Yes it is a resonant circuit. Second winding is for positive feedback to establish oscillations. Tantratron is right.
Voltage on piezo will be several hundred volts P-P.
Piezo transducer is highly capacitive..

Be careful not to damage scope or multimeter.
Look at the scope's probe derating frequency/voltage curve.
Most 10x passive probes will be in danger..

Ideally probe with HV differential probe is possible.

[T3sl4co1l]

Not a current transformer -- giveaway is the size, and this is supported by the lack of burden resistance.  That's a coupled inductor.  Not sure offhand what role it serves to feed back resonant voltage, but evidently it works to close the oscillator loop for whatever purpose they have.  The turns ratio must be quite low, I'd be concerned if there's more than a few volts on the feedback winding.

The Q factor appears quite low (300V in * half bridge is ~150V out, capacitor is rated 1000V so load is less than that, so maximum Q of inductor + cap + piezo is around 7), which seems kind of surprising to me offhand, but suggests they're driving it right in a valley where the piezo itself has a substantially resistive impedance (its own capacitance and inductance are balanced by mechanical coupling and resonance), and presumably that's also the primary bulk mode, at whatever couple 10s of kHz it does.

Tim

[nfmax]

Not a current transformer -- giveaway is the size, and this is supported by the lack of burden resistance.  That's a coupled inductor.  Not sure offhand what role it serves to feed back resonant voltage, but evidently it works to close the oscillator loop for whatever purpose they have.  The turns ratio must be quite low, I'd be concerned if there's more than a few volts on the feedback winding.

The Q factor appears quite low (300V in * half bridge is ~150V out, capacitor is rated 1000V so load is less than that, so maximum Q of inductor + cap + piezo is around 7), which seems kind of surprising to me offhand, but suggests they're driving it right in a valley where the piezo itself has a substantially resistive impedance (its own capacitance and inductance are balanced by mechanical coupling and resonance), and presumably that's also the primary bulk mode, at whatever couple 10s of kHz it does.

Tim

Possibly, the drive circuit senses the current taken by the transducer and derives the output voltage from it, thereby synthesising a negative output resistance. That way, the circuit will always oscillate at the mechanical resonant frequency of the transducer, even if the frequency varies with temperature and/or load, thereby getting maximum ultrasonic power out into the fluid bath.

We used a similar idea, many years ago, to drive a PZT ring element in a fibre-optic polarisation scrambler

[T3sl4co1l]

The odd thing is it's tweaking the driven frequency -- that's a rate input, not a level input.  It can tend to lock onto resonance that way, but more generally it's some weird parametric thing, right?  Putting feedback in series with the capacitor would provide direct injection locking, that I'd think would be more effective.

The traditional way is just a self-excited oscillator, either a voltage-fed half bridge (ala CFL circuits) or choke-fed push-pull (ala CCFL circuits, hah), both of which latch onto whatever the strongest mode is.  I'm not sure that they actually gain very much at all, using this topology; I guess it's an incremental improvement over a current-driven BJT half bridge, with slightly lower voltage drop and switching losses.  Truly it's incremental; the traditional circuit is impressively efficient despite its simplicity and "primitive" transistors!

Circuit is also missing a 12V supply, or at least a bypass cap and zener, which seems really suspicious.  But maybe there's a handful of other things there; aux supply, timer and LED display is pretty common fare.

Tim

[shakalnokturn]

What's the value of resistor between transformer and trimpot?

[langwadt]

Circuit is also missing a 12V supply, or at least a bypass cap and zener, which seems really suspicious.  But maybe there's a handful of other things there; aux supply, timer and LED display is pretty common fare.

the IR2153 has a buildin ~15V zener so it can be powered from the high voltage supply with just a series resistor, but it needs a bypass cap

[mdubinko]

Hey all, thanks for all the helpful knowledge. When you RE a circuit, often times you start to feel like you understand what the designer was going for. Not so much here! And while I do intend to fix this thing, I'm mainly doing it for the understanding. Much appreciation. I'm honored by all the replies.

This thing is 90% incredibly dodgy, and 10% kind of fancy. Resonant converters sound more like something in an electric car charger than a built-to-cost consumer good that burns out a few days out of warranty. Meanwhile there's not a point on here that's safe to attach a scope ground lead to. (Yes, it's on an isolation transformer...but still.) What can I say? I like a mystery.

@T3sl4co1l I measure about 12V between VCC and COM. The resistors from the high voltage really are 20K, 20K, and 33 ohms, all in the ~2W range. Go figure that one out. Did they have a warehouse of extra 33s? And there ARE 470u and 1n caps between VCC and COM that I left off the drawing.

@Tantratron and @2N3055 I measured the primary of this transformer at about 1.2mH. The piezo is about 8nF in parallel with 4.7nF. Calculator tells me that's a (electrical) resonant frequency a tad over 40kHz which checks out. Charging 12.7nF from 0 to 300 volts takes a hair over 500uJ of energy, and do that 40K times a second and we're at 20J/s = 20W. I guess that checks out. The DC loop resistance is about 0.7ohms from the transformer primary and 1.3ohms from the low-side MOSFET rds.

@nfmax and @langwadt the RC portion works kind of like a 555. The CT pin has thresholds at 1/3 and 2/3 of VCC, and flips the output at RT to keep oscillation going. The RC portion has R of about 15k to 20k depending on the variable resistor, and C of 1nF. The resistor heading off to the transformer is 330k. So it looks like it nudges the voltage over/under the threshold to help adjust the frequency.

And on to repair. When it switches on there is a metallic thunk as if the piezo is getting hit with steady DC. Sadly, I don't have a differential HV probe, and some careful measurements with a battery-powered multimeter give inconsistent readings; apparently AC and DC both present. Replacing both MOSFETs is seems like a safe bet. And probably the IR2153. Any other suggestions? -m

[T3sl4co1l]

@nfmax and @langwadt the RC portion works kind of like a 555. The CT pin has thresholds at 1/3 and 2/3 of VCC, and flips the output at RT to keep oscillation going. The RC portion has R of about 15k to 20k depending on the variable resistor, and C of 1nF. The resistor heading off to the transformer is 330k. So it looks like it nudges the voltage over/under the threshold to help adjust the frequency.

OH!

I misremembered how that timer is wired.  Yeah, so "RT" is the "555" output, so you tie a resistor back to "CT" to make it ramp up and down.  And there's no toggle divider (like in TL494 etc.), so the resistor from the transformer is indeed a, well it's still mostly a rate input because it's an RC time constant, but it's forced in a more obvious manner and can lock.

And on to repair. When it switches on there is a metallic thunk as if the piezo is getting hit with steady DC. Sadly, I don't have a differential HV probe, and some careful measurements with a battery-powered multimeter give inconsistent readings; apparently AC and DC both present. Replacing both MOSFETs is seems like a safe bet. And probably the IR2153. Any other suggestions? -m

Yeah, sounds like a short to +V, implying the top transistor is shorted, the IR2153 isn't trying to do anything, or the bottom transistor fused open.  Most likely all three are dead, a clean sweep of active components.  This should be obvious from some ohming around.  With them out, also check that resistors are nominal, and capacitors are, as well as you can check in-circuit anyway.  The bootstrap diode may actually be okay.

Tim