For ultrasound transducers for imaging, high voltage drive pulses need to be generated, typically using a 3-level or 5-level pulser structure which resembles the image below. For a 3-level structure, two are connected together at the output, and for 5-level, three are connected.

Pulser Circuit

Image Source

The problem with this design is that discrete P-Channel MOSFETs are harder to come by with sufficiently high Vds(max) which must be at least 250V for a ±100V supply (when the output is -100V, the Vds of the PMOS is 200V). While they can be found, they generally have much higher Rds(on) and are much slower.

For one of the projects we are looking at, ideally we need to be able to drive a transducer at high frequency - 30-40MHz range - using a relatively arbitrary waveform (5-level PWM). However in order to achieve this the pulser needs to be able to switch somewhere in the region of 200MHz which means a rise time of <5ns which is beyond the range of what any of the available PMOS devices can do.

In order to get to this sort of level, the only thing I can think of is to replace the PMOS device with an NMOS device. However herein lies the challenge. Most of the high-side NMOS drivers are typically designed for applications like BLDC drives and DC-DC converter where the frequencies are in the 100kHz range at most. Additionally they and are not usually designed to cope with the high side gate sitting at negative voltages which would be the case at times in this pulser.

In order to switch from, say 0 to 100V in 5ns, the source of the high-side NMOS has to slew at more than 20kV/us which means whatever is driving the gate must be able to cope with such a high common mode transient. It must also be able to drive the gate fast enough to switch in <5ns which is quite a current spike although that will depend on the gate capacitance of the MOSFET.

I've been looking at ways to translate the signals from the logic circuitry through to the gate of the NMOS but am struggling to find anything fast enough. Optoisolators max out in the 25MHz range, and the only digital isolator I have found is the ADN4650 which can doo 600Mbps but those are not cheap and would require additional LVDS-LVCMOS circuitry which will take up space, especially when the channel count goes up (e.g. 16 element). Passive level shifters would be even more tricky to implement as the source will be going negative as well as positive.

So on to the question. Basically has anyone developed such a high frequency high-side drive? or have any ideas on circuitry that could be used to drive it?

  • \$\begingroup\$ If the output device is a piezo (a capacitor), why do you need + and - supplies? \$\endgroup\$
    – Andy aka
    Commented Dec 30, 2016 at 10:21
  • 2
    \$\begingroup\$ Its an electromechanical device it needs + and - displacement. This is probably easier than an H bridge. \$\endgroup\$
    – RoyC
    Commented Dec 30, 2016 at 10:27
  • 3
    \$\begingroup\$ @Andyaka For the 5-level variant there are actually 5 supplies: -100V, -50V, 0V, +50V, +100V, and for the 3-level the +/-50 are unused. The other side of the transducer is tied to GND so both positive and negative supplies are required to pull in either direction. \$\endgroup\$ Commented Dec 30, 2016 at 10:38
  • \$\begingroup\$ Chaps, forgive my ignorance, but if it's a piezo then it gets positive displacement from current due to a positive impulse and then it gets negative displacement due to collapsing the stored voltage with the other MOSFET (grounded and not negative). Doesn't it? \$\endgroup\$
    – Andy aka
    Commented Dec 30, 2016 at 10:45
  • 2
    \$\begingroup\$ What is the capacitance of the transducer ?This determines how hard this is . \$\endgroup\$
    – Autistic
    Commented Feb 4, 2017 at 0:21

2 Answers 2


I would consider using magnetic coupling for your gate drive. See http://www.infineon.com/dgdl/an-950.pdf?fileId=5546d462533600a40153559ee46511c2 for some idea how to design this.

This assumes that the length of time any one switch is on will be limited (which appears reasonable given the above circuit for the supertex parts), and the duty cycle is not too high (in a 5 level converter, it shouldn't be a problem). Suitable pulse transformers are easier to source than other isolators, the challenge I see is dealing with the miller effect and power handling of the pulse transformer.


Arbitrary waveform with 200Vpp amplitude and 200MHz bandwidth - it is a tall order. For much lower voltages ( 60Vpp ) I would recommend a CRT driver like LM2462. If sinusoidal output is OK, than an RF amplifier will supply 200Vpp at 30-40MHz easily. At 30 MHz there is 6-7 ns for each voltage level ( both transient and steady state portion ). Do a test - drive your transducer with just one 50V supply, N-MOSFET and gate driver IC, no optoisolation to check if you can get waveform this fast.

  • \$\begingroup\$ Arbitrary was probably the wrong term. There are 5 discrete levels in the waveform. The MOSFET structure would be what is required assuming the gates could be controlled. \$\endgroup\$ Commented Jan 1, 2017 at 17:07
  • \$\begingroup\$ Sorry to hijack, but is there a currently manufactured replacement for LM2462? (CRTs have gone the way of the dodo...) \$\endgroup\$
    – bobflux
    Commented May 9, 2017 at 11:50

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.