I would strongly recommend against attempting any sort of audio output transformer to boost the output of the op amp like is being suggested in the comments. Don't get me wrong, it is a perfectly valid and frequently used technique. In reality, it's just a kind of impedance matching, and impedance matching transformers work great for this, and let low impedance amplifiers drive much higher impedance voice coils by effectively stepping up the output wave form.
The problem is that none of that works at all in your application because you're not driving a voice coil, you're driving a piezoelectric ultrasonic transducer.
This is a fairly important distinction that you should be aware of, because many of the normal techniques that work to drive traditional speakers are not applicable to your situation. Voice coils are deflected by current flow and their impedance is resistive and inductive in nature. Your ultrasonic transducer is none of those things. Its impedance is entirely capacitive and open circuit to DC. It's actuated by charge difference between its plates, not current. So effectively, it is deflected by voltage directly. You actually need polarity, and coupling it via a transformer will cause you to lose this, so you'll also have to use a synchronous rectification stage to restore the polarity on the output of the transformer. This isn't a problem if driving speakers, but it matters with this kind of load. You need to drive it by directly pulling plate charge out or into it, and the way you should do this is the way you were planning to do it from the start.
You have a couple options there though:
- You can use a high voltage op amp with wide supply rails, OR
- Use a low voltage op amp controlling push-pull output stage tied to high voltage rails.
I think 1 is going to have the best performance as well as being simpler and easier, especially considering the nature of the signals you will be working with. Chips, clicks, various other things need to be amplified and (at least, I would assume) with minimal distortion. Option 2 is really only when you need the performance you can only find from a lower voltage op amp, but since there is an op amp that is specifically designed for your application (amongst others) and we are really only talking about maybe 80kHz tops, I don't see a reason to attempt option 2. I suggest option 1. The LTC6090 works very well for the exact thing you need to use it for. Down to the thermal pad area, everything is accounted for.
As for obtaining the rails, the datasheet for the LTC6090 actually contains a couple of schematics dedicated entirely to producing ±65V or ±70V rails for it. You can use them or not, but the topology is the point. In my opinion, a flyback converter is definitely your best bet. 5V to 70V is really too high of a ratio for a boost converter. 135V is pushing it even more - the LTC8331 is great part, and it can probably make 135V from 5V, but it won't be efficient or pretty. And I don't think tapped inductors are all that available for making the virtual ground. It will be simpler, cheaper, and much more practical to simply use a flyback converter with two output windings.
Don't be intimidated by flyback converters - they're just inverting buck boost converters. I kid you not. Look at the circuits. They're the same circuit, the same converter, the same thing. The only difference is the flyback has two or more windings on the same core for its inductor. And the diode might be one way or the other depending on the polarity you want. Honestly, I would recommend simply using the LTC3411 as suggested in the LTC6090's datasheet. You might want more current though, which is what the LTC3412 is for - it's just an LTC3411 with twice the current capability. Unfortunately, you should probably go with the somewhat pricey ($6 in single quantities) Würth transformer Linear suggests in their datasheet. Flyback transformers don't really need a proper transformer and just a coupled inductor, but there are few coupled inductors with 3 windings, and I doubt any of them are rated for 100+V (you want some headroom). The Würth transformer is small, surface mount, and can safely handle 150V on each secondary winding. You can't go wrong with Würth, but they'll make you pay for the privilege.
I know you want to probably keep the BOM down, but trust me, it will be worth the time saved and lack of headaches and complications to just kind of 'copy and paste' an application schematic from a converter like the LTC3411/2. Linear's switching parts are also the best and easiest to use (and also the most expensive sadly) and I usually suggest starting with a linear part. Once you're more experienced and comfortable with switching converters, TI is good too and much cheaper, though often will not be quite so easy to use.
You should be worrying about those ultrasonic chirps and hoots and what not, I think this is one of those situations where it makes sense to pay a little more to just not have to deal with the power supply so you can deal with the thing you're actually doing. If you can, just grab that Würth transformer, follow the schematic on page 22 of the LTC3412 datasheet, or there is a table of transformers and their potential outputs on page 10. The circuit will work fine with 5V input but with reduced output power, you only need to adjust the EVLO divider for 5V (or just tie it directly to VIN).