High Voltage Not Gate

[rentner]

Short and simple:

I am searching for a circuit for a Not Gate with a 5V/12V Input and 500V output (Pull Up resistor 30k). So to speak an inverter circuit. Build of discrete Components. Bandwidth at least 10MEGHz or higher. Means: Rise and Fall time on the output without load, just measured at the Pull Up resistor should be less than 50ns each. 100ns for one cycle, so to speak -> 10MEGHz. Any suggestions?

[tszaboo]

Can you calculate how much capacitive loading on 30K resistor would allow you such step response?

[rentner]

That would be less than 1/2p


Well, if you look at IGBT or MOSFET High Side drivers, they do the same, but with much less energy requirement. They don't use 30k or so. How do they Make a clean signal over a few hundret volts with ultra low delay? Many drivers have less than 45ns turn on or turn off time. They are also level shifting the signal. Maybe not Inverting, Like I need it, but they do. I mean, if I saturate a Transistor, it will turn off slowly, but If I would not saturate a Transistor, shouldn't it immediately turn off? A Mosfet does not do it, since it has a rather high capacity.


Maybe I can "copy", how High Side drivers just do that, because they also do a very fast level shifting.

[tszaboo]

Mosfet drivers drive 2-3Amps easily in a gate of a mosfet. And they dont have pullup resistors, they are push-pull circuits for that speed. What I wanted to point at:
If you have a weak pullup, the edges will be weak, with any loading. If you need 300V with sharp edges, that requires a (so called) half bridge. That has less than 10 Ohm rDs inside the mosfets so capacitive loading is not an issue.
This is probably a good learning material for them:
http://www.irf.com/technical-info/appnotes/an-978.pdf

[rentner]

I know, that Mosfets are driven low impedance. But before that, the signal gets level shifted - that happens high impedance. If I would like to drive the mosfets directly over 300V, I would waste giant ammounts of power. You know, how bootstrapping drivers work, do you? And to get a clean and fast rise, I need to get a clean signal on the input. To archieve that without giant ammount of losses, I want to shift the signal with as low losses, as possible. A totem pole driver with a floating bootstrapped supply would drive a Mosfet than.

And that is my question: How do I get a signal from low to high over a very high voltage range in a few nanosecounds?

[tszaboo]

Pardon me if I dont understand you correctly, but the output voltage is only some 5-12V with 500V DC offset on it? In that case just use a digital isolator (iCoupler from analog, Digital capacitive isolator from Ti, or silicon labs). Or do you need 500V peak to peak signal?

[Marco]

How do I get a signal from low to high over a very high voltage range in a few nanosecounds?

A very fast optocoupler or a low turn GDT? (With the former the actual gate driver on the high side switch would need an isolated power supply of course, with the latter you would need some type of latching/reset circuitry if you need high duty cycles.)

The HV CMOS level shifters in integrated half bridge drivers can manage your rise/fall time requirements, but they have much more propagation delay ... trying to make your own discrete transistor level shifter seems unlikely to improve on them by 2 orders of magnitude.

PS. a quick google found this paper describing an integrated level shifter, follow the references for more.

[rentner]

GDT works just to PWM <=50%

Optical isolator? Very expensive and definitely not discrete.


Also, I need the full Voltage range. Let me explain: The 30k is tied to the Supply pin of the Bootstrap Capacitor. This pin will rise 12V over the output pin. It starts at 0 Volts and ends at the full voltage. Since the input of the isolated driver stage only can handle A voltage of a few volts, I cannot drive it with +500V, while the output still is at 0 Volts. This has to rise with the output, because the Input is floating with the output. So I need the ability to rise as fast as possible. If the Rise time is quicker than the rise time of the Mosfet, I can say, that the Mosfet will not turn off once in a while while the signal is rising. A slow pullup will have this problem. So I need a square wave with 10MEGHz bandwidth as rise+fall time. Also, If the Pull up Resistor is of low value, it will dissipate so much energy, that, the circuit will be totally inefficient. So any ideas?

[Marco]

GDT works just to PWM <=50%

There are app notes (and patents) for using GDT with high duty cycles, AFAIR the app note one charged the gate through a diode and then used a negative voltage on the GDT to activate a transistor to discharge the gate.

Optical isolator? Very expensive and definitely not discrete.

That's just defeatist ... if you want you can certainly optically couple a led to a fast PIN photodiode yourself and use the signal to activate a gate driver powered by an isolated power supply.

Also, I need the full Voltage range.

I'm just suggesting the GDT and optical isolation as a means to drive the high side of a half bridge with a low propagation delay, it's the half bridge which will give your final output voltage.

Why are you still talking in terms of pull ups when in post 2 it was established you can pull up squat at these speeds with that high a resistance?

[rentner]

Well, this is, how I thought about it. You will Propably understand it with ease. The secound image providing the graph explains, how I could use a constant current supply with turn on and turn off pin to get the Signal. 530uA will pull down the Signal just 16 Volts below the point, where the circuit is able to do anything. This provides me control after turn on and off with low losses.



If you have a circuit without a patent, I might use it, but things unter copyrights law will not be used.

Also, is there any discrete diode and phototransistor with such high speeds for low budget?

[T3sl4co1l]

So you're just doing a high side drive?  Not some crazy low power physics toy something or other?

[rentner]

So you're just doing a high side drive?  Not some crazy low power physics toy something or other?

Well, how would you drive an N-Mos high side? Actually it is crazy low power physics. The thing is, that it needs to be very, very fast! I mean, If a Transistor saturates, it does switch off pretty slow. How do I stop that from happening? If I could drive a Transistor out of saturation, I would have just fixed my problem...

[David Hess]

Well, how would you drive an N-Mos high side? Actually it is crazy low power physics. The thing is, that it needs to be very, very fast! I mean, If a Transistor saturates, it does switch off pretty slow. How do I stop that from happening? If I could drive a Transistor out of saturation, I would have just fixed my problem...

MOSFETs do not suffer from saturation induced delay time like bipolar transistors do.

If you want to use an n-channel MOSFET as a high voltage high side switch, then the high current gate drive circuit can either be floating on its input or output and there are lots of ways to accomplish that but none of them are particularly simple and you still need a gate bias supply higher than the supply voltage unless you want the additional voltage drop from the gate turn on voltage.

[rentner]

MOSFETs do not suffer from saturation induced delay time like bipolar transistors do.

They don't. But to level shift a signal with a low impedance, Mosfets will not work, because they have far to high capacities. This was tested before. Even the hardest switching does no good job with this basic inverting configuration.

If you want to use an n-channel MOSFET as a high voltage high side switch, then the high current gate drive circuit can either be floating on its input or output and there are lots of ways to accomplish that but none of them are particularly simple and you still need a gate bias supply higher than the supply voltage unless you want the additional voltage drop from the gate turn on voltage.

Yeah, Bootstrapping - that part is working for me and that without any issue.

And I noticed, that they are not simple at all. How does the IRF2110 for example work. The block diagram is kept pretty simple, yet it is looking very complex. I do not unterstand the block diagram in the datasheet. No manufacturer does release any equivalent circuits. I would love to know, how this works.

[David Hess]

MOSFETs do not suffer from saturation induced delay time like bipolar transistors do.

They don't. But to level shift a signal with a low impedance, Mosfets will not work, because they have far to high capacities. This was tested before. Even the hardest switching does no good job with this basic inverting configuration.

If the reverse transfer capacitance is a problem, and at high voltages it always is, then a two transistor cascode switch can be used to eliminate its effect.  This has a further benefit in reducing the input capacitance because the high voltage MOSFET is replaced by a low voltage one.  Of course now you would have to drive two MOSFETs in series or maybe a low voltage MOSFET and a high voltage biploar transistor which are more efficient at high voltages anyway.

If you want to use an n-channel MOSFET as a high voltage high side switch, then the high current gate drive circuit can either be floating on its input or output and there are lots of ways to accomplish that but none of them are particularly simple and you still need a gate bias supply higher than the supply voltage unless you want the additional voltage drop from the gate turn on voltage.

Yeah, Bootstrapping - that part is working for me and that without any issue.

And I noticed, that they are not simple at all. How does the IRF2110 for example work. The block diagram is kept pretty simple, yet it is looking very complex. I do not understand the block diagram in the datasheet. No manufacturer does release any equivalent circuits. I would love to know, how this works.

The functional block diagram that International Rectifier includes covers it all if you can read it.  The only real mystery is the high voltage level shift which I suspect works through capacitive coupling to achieve the specified 500 to 600 volts of isolation although they show a resistive level shift at the output of a differential pair.

[T3sl4co1l]

How long does the transistor remain on for?

If it's not much, I might just use a gate drive transformer.  Who needs voltage reference!

Tim