High Speed MOSFET Switching

[azeeman]

Enhancement mode N type MOSFETs have become ridiculously cheap which enables a new way of controlling them by decoupling switching and modulation. Switching is defined as the act of turning the MOSFET on and off while modulation is defined as changing the on to off ratio usually for generating waveforms or DC levels.
By driving two MOSFETs in parallel and out of phase by 180 degrees, it's possible to drive each MOSFET by a pulse transformer with a 50% duty cycle. This generates a near ideal gate drive signal for each MOSFET with the pair acting as a near perfect switch. The finite turn on and turn off times create a make before break type contact so the pair act as a near ideal switch.
Because of the pulse transformer isolation, the gate drive can be high side or low side.
An AC-DC solid state relay with high speed switching capability is easy to implement using two MOSFET pairs  in anti-serial being driven by one gate driver and associated pulse transformers. By using a 1MHz switching frequency, the relay can easily work at 100kHz or above.
The square drive signal is generated by a state machine which is turned on and off by a modulating signal. Because the gate signal is a fixed frequency, a lot of tricks can be done that aren't possible with a conventional variable duty cycle gate drive circuit. The gate can be resistively terminated and then a transmission line used to couple the gate drive signal to the gate. This makes PCB layout easier when driving the gates at extremely high frequencies for some of the more exotic MOSFET technologies.
The pulse transformers can be very small if the switching frequency is in the megahertz range. The switching bandwidth can be extremely narrow which makes EMI filtering easier.
There's also the possibility of mitigating charge injection effects because the charge injection is occuring at orders of magnitude above the modulation frequency. This may make it much easier to filter out charge injection artifacts. I intend to test this out in a low voltage chopper amplifier because if true, it would greatly simplify low voltage measurements. Power MOSFETs are ridiculously cheap and the low on resistance would generate very little thermal noise while the substantial metal around the chip would reduce thermal gradients dramatically.
The gate is AC coupled through a high pass filter which prevents low speed switching of the MOSFET which mitigates transient high power dissipation caused by low speed, high current and high voltage switching.
The modulating state machine shown uses synchronous modulation, but can be changed to asynchronous by the modulation signal directly resetting the flip-flop. This might cause switching issues because the 50% duty cycle is no longer guaranteed.
Two pulse transformers are used although only one is required. The problem is that a 1:2 primary to secondary ratio is required because the peak to peak gate drive signal has to be twice the MOSFET on voltage and gate drivers rated to 20V are scarce. Also commercially available pulse transformers with one primary and two secondaries are available, but only as 1:1:1 ratios. I haven't found any with 1:2:2 ratios.
I have no intention of patenting this idea and have no idea if it already exists in the wild. Feel free to use the circuit as you wish. Please share any improvements. When a 700V, 6A MOSFET, https://www.infineon.com/assets/row/public/documents/24/49/infineon-ipd70r900p7s-datasheet-en.pdf is available for less than a $1 from DigiKey, a whole new world of projects opens up.
Pulse transformers are a bit more expensive, but because the Vus rating can be very small, it might be cheaper to buy a ferrite core and custom wind it.

[moffy]

Just a few suggestions:
1. Use dots on the transformers to show polarity, it will help the reader.
2. The gate capacitance will interact with the leakage inductance of the transformer and you will have significant possibly destructive ringing on the gate, one needs to dampen or clamp that.
3. Gate resistors or ferrite beads help prevent MOSFET parasitic oscillation.

[jnk0le]

Was this written by AI?

The gate is AC coupled through a high pass filter which prevents low speed switching of the MOSFET which mitigates transient high power dissipation caused by low speed, high current and high voltage switching.

AC coupling of pulse transformers only prevents "DC" current going through transformers. It has nothing to do with switching speed.

[temperance]

There are much better ways to achieve this.

1. Small DC DC converter, something like < 2 cm² to power the floating part.
2. Isolated gate drive.

This one achieves 800 V, CMTI up to 50 kV / ms
https://www.onsemi.com/download/data-sheet/pdf/ncv57200-d.pdf

Or 1.5 kV, CMTI > 200 V/ns
https://www.onsemi.com/download/data-sheet/pdf/ncv51561-d.pdf

Edit:

Was this written by AI?

Seems so because this is full of nonsense. But concludes with: I have no intention of patenting this idea.

I've done this for high speed switching I don't know how many times now in maybe ten different ways (Just because I have a restless mind and can't resist thinking of different ways to achieve the same in better ways).

[azeeman]

Isolated gate drives are not that fast. The pulse transformer is an isolated gate drive and doesn't have a startup delay that a DC-DC converter has. Also, compare the cost of a DC-DC converter to a single MOSFET. MOSFETs are extremely cheap.

[azeeman]

Isolated gate drives are not that fast. The pulse transformer is an isolated gate drive and doesn't have a startup delay that a DC-DC converter has. Also, compare the cost of a DC-DC converter to a single MOSFET. MOSFETs are extremely cheap.

[azeeman]

It's true that AC coupling doesn't improve speed. But by alternately switching the MOSFETs at high speed and then modulating the switching the drive circuitry is very simple while the switch on-off ratio can be arbitrarily high without using a floating power supply. All the electronics sits on the low side.

[azeeman]

The pulse transformer inductance can be very low and the leakage inductance even lower so ringing shouldn't be a problem. For very high speed switching where even a small amount of inductance is a problem, the gate can be terminated with a resistance and the transformer matched to it. The gate frequency is constant so it's very easy to create a near perfect match.
Also, with the gate terminated, a transmission line with matching impedance can be used to couple the driver to the gate through the pulse transformer.

[azeeman]

The cost of the chip is comparable to a pulse transformer. Also, pulse transformers can have much higher voltage isolation at low cost.

[temperance]

Isolated drivers not fast enough! For the NCV51561

Propagation Delay Typical 36 ns with
5 ns Max Delay Matching per Channel
5 ns Max Pulse−Width Distortion

And the CMTI values support his kind of speed. That's what I posted them.

How fast do you want to switch? Over 200 V/ns?

[azeeman]

Please post an example. I post here as I appreciate the insight others can offer.
The problem I'm trying to solve is that I need a circuit to control the brake current on a self powered elliptical. The input voltage from a three phase generator can range from 0 to 200V while the brake current ranges from 0 to 2A. The brake control circuit could have an input voltage of 10V and an output voltage of 10V. It could also have an input voltage of 200V and a brake voltage of 0V.
The switch topology I posted can easily handle this with two buck converters. One buck converter is used to supply low voltage to the control circuitry and the other to control the brake voltage.
The buck converter on time can range from the microseconds to seconds if necessary and the input and output voltages can be the same or there can be a huge difference. The control is done by using a comparator to measure the control voltage and turn the buck converter switch on and off accordingly. The brake drive will use a comparator to measure the voltage across a shunt resistor and then turn the buck converter switch on and off.
The control circuitry becomes very simple. Currently, I'm using a triac based design which works, but a MOSFET based design would be much smaller and cheaper. This is a hobby project and is done more as a design challenge than for commercial use.

[temperance]

self powered elliptical

What is that?

You first posts writes about 700 V. The recent post above this one states 200 V.

Schematic is simple:
Oscillator (500 K tot 1 MHz, or microcontroller timer output) >> MOSFET driver (1EDN8550) >> Transformer (Something like a Pulse PA2001NL) >> Rectifiers =  floating supply.

Floating supply >> Isolated MOSFET driver. (But if it is only for 200 V, you can use an Infineon 1EDN8550 diff input MOSFET driver capable of 4 A gate drive costing about 30 cent)

[azeeman]

Thanks for the link to the high side driver. The one downside is the low peak gate current. A cheap low side gate driver has much higher peak gate drive current and gate drivers with peak currents over 10A are available. https://www.run-ic.com/en/upload/goods/20231122/202311221425081194.pdf
It comes down to whether it's more cost efficient to go with the cheapest possible low side driver and pulse transformer or a high side driver. It also depends on the load that the MOSFET is driving. For some loads there is substantial feedback from the drain to the gate and a higher gate current is required to absorb the feedback to maintain high speed switching.
Another consideration is charge injection where the MOSFET is used for signal switching. If the switching is done at a constant high frequency with low frequency modulation, the charge injection will occur at the switching frequency while the signal chopping frequency can be orders of magnitude lower. This might be useful for synchronous detection of low level signals. I haven't tested this yet, but it's definitely worth exploring.
If the chopper frequency and switching frequency are the same, charge injection is a difficult problem to solve, especially at very low voltages.

[azeeman]

I did a search on DigiKey for the cheapest MOSFET with a high voltage rating and 5A current rating. The Infineon part popped up and the price was $0.631 each in quantity of 10. That kind of price is incredible. This is a one off hobby project and I'm looking for a simple and cheap solution mostly as a design challenge.
I set the voltage rating to 500V and because this matches the MOSFETs used in in electric vehicles and if this gives some protection against high voltage spikes.
A microcontroller requires a low voltage supply which requires a wide input range buck converter to convert the output of the generator. I've attached the generator specs.
Because the generator is human powered, the control circuitry must draw as little power as possible. A human can generate about 100W so ideally the control circuit would draw under 1W.
For startup I plan to charge up a capacitor using a very low current with a voltage monitoring circuit. As soon as there is sufficient charge on the capacitor, the buck converter will turned on and then supply the control power. The capacitor charge current will be about 2mA so even if the elliptical is running at full speed, the static power consumption will be very low. The buck converter gate drive will draw about 10mA if run at about 500 kHz, but this will be at 12V so the power consumption should be quite low.
A simple CMOS gate based circuit can run on a few microamps is all that is required for a simple wide range buck converter. The voltage is regulated by a comparator which simply switches the modulation input high and low as the control voltage changes.
The control voltage will then drive a microcontroller which will drive a DAC used to set the current setpoint. A second comparator then modulates a second buck converter to control the brake current. The comparator response time is about 2 or 3 microseconds so the buck converter inductance has to be sized to keep the peak current low over that time period when the generator voltage is at its peak. For lower input voltages the comparator will keep the switch on as long as required to supply sufficient current.
The biggest challenge is generating the control voltage with the minimum amount of static power consumption and the lowest startup voltage. Full braking should be possible at very low speeds and no braking should be possible at very high speeds.
The machine came with a control board and a small lead acid battery which I assume was used for startup. The board went up in a puff of smoke the first time I used the machine and I never bothered to find out why. The board had about 10 heatsinks and lots of power components and it looked like an engineering nightmare.
Controlling the brake with a microcontroller and MOSFET was trivial by using a power transistor to regulate the generator output and the microcontroller to drive the MOSFET, but the parasitic power loss was too high. The challenge is to have the least amount of static and dynamic power loss with the machine fully self powered and usable over the maximum speed range with no heatsinks, no batteries, minimal inductors, etc.

[azeeman]

A floating supply requires either very high frequencies with tiny capacitors for isolation or a transformer. Since a transformer is required, it might as well drive the gate directly and the higher the frequency, the smaller and cheaper the transformer. Also, the floating supply requires high speed rectifiers and filter capacitors. The pulse transformer circuit needs none of these.
The pulse transformer also drives the gate positive and negative with the voltage slew rate maximum at zero. This gives the maximum possible switching speed especially when turning the MOSFET off because there is no slowing of the gate voltage drop as it approaches zero.

[Zero999]

I don't have time to read the wall of text in detail. I've just quickly scanned it.

Is isolation required? The schematic shows the driver logic ground and output ground connected i.e. not isolated.

Why not use a bootstrap circuit? It's cheaper than those transformers and will work over a wide range of duty cycles, just not 100%.

https://www.onsemi.com/pdf/datasheet/ncp51313-d.pdf
https://www.ti.com/lit/ds/symlink/lm2101.pdf

[azeeman]

I haven't described the problem correctly. The problem is how to drive a MOSFET using a single transformer over a 0% to 100% duty cycle. The answer is simple, you can't. But you can easily drive a MOSFET through a transformer at a 50% duty cycle. Now put two MOSFETs in parallel, but driven out of phase using one transformer with two secondaries. Now it's possible to drive the combination over a 0% to 100% duty cycle.
To keep things simple, the MOSFETs are driven by a square wave generator and the generator itself is switched on and off. The MOSFETs can be driven at a very high frequency such as 1 MHz with the square wave generator being switched on and off, also known as modulation, at up to 500 kHz.
All the electronics can be on the low side with no floating power supply required. Even better, a delta-sigma type control can be used to regulate the converter output. A high speed comparator compares the voltage on the output capacitor to a reference voltage. If the voltage is low, the modulation input is set high. As current builds up in the inductor, the voltage across the capacitor goes up linearly as a function of the input to output voltage difference. When the voltage increases above the reference, the modulation input is set low. The inductor current will still flow and increase the capacitor voltage. This creates a natural hysteresis voltage and the comparator will remain off until the capacitor voltage drops below  the reference voltage.
Because the switch duty cycle can go from 0% to 100%, the output voltage can be the same as the input voltage or much higher. Also, unlike with a PWM controller which has one switching event over its period, the comparator will switch multiple times and will require less inductance for current limiting, but will require a higher filter capacitance for voltage ripple reduction. Capacitance is much cheaper than inductance so this is a good tradeoff.
Very high speed comparators with nanosecond speeds are cheap and readily available and can be used in the control circuit.
The reason for having a common ground is that no isolation is required and if a high side switch is available, the circuit topology is very simple with low voltages across the MOSFETs. It will be part of a bootstrap circuit where the input voltage will vary from zero to around 200V. The circuit will be used in a self powered elliptical and will generate about 12V and be used to drive a Nano. A similar circuit will be used to control the elliptical brake current.
A human can generate about 100W so ideally the control circuitry should use much less power such as 1W. The brake should also be effective at both very low and very high speeds so both the control power supply and brake power supply must work with an input range from about 10V to 200V or more. The original controller had a six volt battery and went up in smoke the first time I used it. Mechanically, the machine is a beast, electronically it's a piece of junk but it's an interesting design challenge. The battery is a wear item and an engineering failure. The machine should be able to work with only a simple electronic circuit and no external power and definitely no battery.
I'm using triacs right now for controlling it which works well, but is not as smooth as would like because the triacs have to work at the generator frequency. The generator is three phase which smooths things out a lot, but there's a lot of room for improvement.
I had also used a MOSFET driven by a microcontroller for controlling the brake current. The control power supply for the microcontroller was linear using a fixed current source, but this was a power hog and not ideal. Because the microcontroller used a PWM control, the frequency was relatively low and audible. With the new circuit, the switching frequency can be well over 100kHz and won't be audible. Also, power consumption can be much lower.

[MariuszD]

The finite turn on and turn off times create a make before break type contact so the pair act as a near ideal switch.

Are you sure about that? Totem pole circuits are often controlled from a single transformer (see the attachment) because such a setup ensures that both will not conduct at the same time, which contradicts what you are saying.
Have you tested your circuit in practice?
Did you take into account:
1. power losses due to gate capacitance charging at a frequency much higher than that required by the buck converter.
2. required higher current output of the driver.
3. Transformer leakage inductance

Or maybe you want to operate in resonance? This circuit is a resonant circuit whether you like it or not.

Your circuit will only make sense when you test it in practice and attempt to overcome the difficulties that arise.
You won't learn anything from the simulation because you don't know how to model the transformer; without that, the results will be too optimistic.

I did a search on DigiKey for the cheapest MOSFET with a high voltage rating and 5A current rating

If you think the current rating of a MOSFET matters, you haven't learned how to select MOSFETs yet. This is a marketing parameter; attempting to operate at such currents will cause maximum allowable power losses and will require super-efficient liquid cooling.

[azeeman]

I've tested it in a simple buck converter and it works. The totem pole is similar in that the transistors are driven out of phase, but requires dead time between switching which complicates the control circuit and also causes voltage spikes. Also, as the input voltage increases and decreases, the switching frequency has to change to keep from saturating the transformer or a fixed frequency with PWM control is required. This is a problem with a very wide input voltage range as a very high primary inductance is required with a high resolution PWM and very high clock frequency. My solution can accommodate a much wider input range with a simple control circuit and smaller inductance because a PWM is not required. The same principal is used with delta-sigma analog to digital converters which have a very wide input dynamic range using a single bit feedback circuit.
The LC filter will have a resonant frequency, but the circuit doesn't need to operate at that frequency. If single bit control is used, the on off periods will be erratic. I seen this with the prototype buck converter. The MOSFET gate drive frequency is constant and if desired can be tuned for performance, but the modulation control frequency and on to off ratio is determined by the LC filter, load resistance and input voltage.

[temperance]

Thanks for the link to the high side driver. The one downside is the low peak gate current.

I think you didn't calculate the required gate drive current. Over 1 A gate drive isn't required at all for a small 5 A MOSFET. More like 500 mA or most likely less.

[mtwieg]

I think there is at least an interesting idea buried in the wall of text. Basically take two FETs in parallel (drains and sources connected, but separate gate drive). Drive the gates with a 50% duty cycle waveform, but applied with opposite phase to each gate. So the pair of FETs will be on 100%. This gate drive can be derived from a simple gate drive transformer. To turn the pair of FETs off, just drive the transformer with fixed zero volts. That's the idea anyways.

MariuszD already pointed out some of the obvious issues (especially the fact that the FETs will not actually "make before break", unless you add some circuitry to extend on time for both).

And if your objective is to get fastest possible switching speed (or "modulation" speed as the OP refers to it), the transformer leakage inductance is only going to slow you down compared to driving the gate(s) directly.

The idea maybe has some merit for solid state relays (where switching is infrequent and does not need to be fast). But I would never consider it for a switchmode power supply (like the buck converter in the OP's first post).

[azeeman]

1 uA is sufficient to drive a gate provided that speed is not an issue and there's no feedback from the drain. Regardless, all the high speed, high side drivers I've seen require a floating power supply which has a transformer inside. One way or another either a transformer is required or for the single chip solutions, capacitive isolation is used which has low current capability and low speed, but no transformer.
I'd rather use the transformer to directly drive the gates and eliminate the floating power supply.

[azeeman]

If make before break needs to be guaranteed, a three phase drive can be used.
Transformer leakage inductance of a typical pulse transformer is extremely low. A Murata 78604/9C 2:1 transformer has 10mH of primary inductance, leakage inductance of 1.64uH and a 56Vus rating. I don't think that small amount of leakage inductance should cause problems.

[azeeman]

I look at the safe operating area specifications to see what current it can switch at what voltage and pulse width. The 5A is just a screen to weed out the obviously underrated parts.

[temperance]

The MOSFET's you are looking at have a gate charge around 8...20 nC. 0.5 A drive current: 16...40 ns switching time. Or more than fast enough for most applications.

I think there is at least an interesting idea buried in the wall of text. Basically take two FETs in parallel (drains and sources connected, but separate gate drive). Drive the gates with a 50% duty cycle waveform, but applied with opposite phase to each gate. So the pair of FETs will be on 100%. This gate drive can be derived from a simple gate drive transformer. To turn the pair of FETs off, just drive the transformer with fixed zero volts. That's the idea anyways.

The idea is fine. But it is not required to do this. The same can be done with 1 MOSFET.