Mosfet switching times ?

It's the test generator impedance, usually.

Makes sense.

Don't know.

Most mosfets will switch much faster than their datasheets suggest, if you just drive them hard. The Fairchild BSS123 datasheet cites a typical turn-on rise time of 9 ns, and 17 for turnoff.

They will actually do this:

That's a 100 volt pulse into 50 ohms, transformer isolated.

Given the complications of driver and fet inductances and such, and the state of the datasheets, the best thing to do is experiment. If you want to go fast, you should probably stick to one manufacturer and not assume that identical part numbers are always interchangable. Buy a reel.

John

[snip]

I never paid much attention too data sheet switching times until recently. Mainly from the emphasis some people here and in other forums place on them. So I was beginning to think I may be missing something, but I guess not.

I've always just used Qg and my drivers sink/source ability to estimate times.

Your exceeding the 25C pulsed current rateing a tad. ;-)

Yes your right I plan on experimenting with both and maybe a couple others I have in the applications.

I generally assume that mosfet silicon is infinitely fast, and that only capacitances and wirebond inductances get in the way. Seems to work so far.

Life in the fast lane!

John

I've taken it as a device parameter, pretty much as discussed here:

... except that Rg is a term applied elsewhere, too. For example, I've seen Rg_i used for the mosfet's Rg, then Rg used for an explicit/implicit series Rg that is external to the device, and then Rg_hi and Rg_lo also for the high side and low side equivalent driving resistance of the gate driver. So I think context is important.

But on a datasheet, unless specified as part of the testing setup, I don't think those external values are included in the Rg value ascribed to the part, itself.

Of course, I'm not an expert reader, either. But that's the impression I've taken.

Few would intentionally shoot themselves in the foot.

Jon

Thanks I never found that one.

I've never heard it refereed to as intrinsic to the device specifically in a data sheet.

The ST device shows the test set-up and Rg is shown external.

To add to the confusion this AN from TI says pretty much that the intrinsic Rg is has negligible impact on switching speeds except at RF?

see pg4

That is the other confusing thing . Given how they are so loose with high current ratings.

This is true for BJTs, which tend to have switching times in the = ballpark of the device, regardless of how you drive it. This comes from = base transit time and stored charge effects. Base spreading resistance = is generally negligible, I guess.

For FETs, this is correct. There is a gate spreading resistance, and = it's occasionally specified, usually in the 1 ohm range. This suggests = that you can't really make it go faster than a 1 ohm Rg equivalent.

Now, you could drive it with a precompensated waveform, so the current = overshoots to kick the RC around, but now you're getting into territory = where you need serious power input for marginal output power and = efficiency. For most power transistors, I would guess this is in the = low VHF range.

I made a 1MHz generator with TC4420's, FDP26N40's and 1 ohm gate = resistors. The '4420s actually get hot at this frequency!

Tim

--=20 Deep Friar: a very philosophical monk. Website:

No, not missing much. Switch times aren't very meaningful in those datasheets.

Most people put gate resistors in because they are afraid to blow the EMC cert or that it could oscillate. Which spoils the power of the nice PWM chip in front of it, like driving a Porsche at 35mph. I tend to drive them hard but there's always the internal gate path resistance which you cannot overcome. So I tend to drive them with lots of gusto, preferably well north of 10V which helps in that domain. It also helps to drive the gate negative a few volts, -5V or so. That speeds up turn-off.

John likes to floor it, wonder how long his new Audi will last :-)

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

It's one of a few I've read, with what understanding I may, that seem to say similar things to me.

I'm sorry. I was thinking about Spice models when I wrote that. I believe, now that you bring it up, that I've seen it on some datasheets. But that wasn't on my mind at the time I wrote.

Yes.

Since the RC ladder is low-pass, that makes some sense to me on first blush. The areas further away from the bond point would never reach sufficient voltage, so I think that would be a real problem, as frequencies rise (no firm turn off or turn on, just finding a quiescent point and sitting there neither on nor off.) But I haven't given this much thought. But your phrasing of TI's apnote comment seems reasonable to me and not conflicting. There is still an Rg to the device. It's just that, because of the RC ladder, at high frequencies there is huge problem caused by the inability to switch the whole gate area around that fast.

But I have only just now given this two minutes thought.

TI's apnote is probably correct. If Rg were important, it would probably be on all of the datasheets rather than a few. It's not always there, though. (It obviously _is_ important for the Spice models, though.) The gate-source charge, the Miller charge, and overdrive charges are important (overdrive extracted by subtracting the other two from the total charge, I gather.) So they are spec'd.

Jon

Recently I was asking in various groups what frequency people thought I might shove through a 2N7000 in grounded gate - some people thought the limit might be around 100MHz.

Do you have an estimate on this?

Do you mean as a switch, or as an RF sinewave amplifier?

If you drive the gate hard, grounded-source, you can turn a 2N7002 on/off, with 50 volt drain swing, in about a nanosecond. I'd imagine you could get useful grounded-gate power gain at 250 MHz at least, especially if things were tuned+matched for the operating frequency. The 2N7000 version will have a little more lead inductance, but that can be tuned out.

Try this:

9 volt battery, 470 ohm resistor, LED, 2N7000. You can pull the gate up to 9 volts, or down to ground - just use your fingers - to turn the LED on or off. And then let it float in either state for hours. Or briefly touch the drain to the gate to turn the LED partly on, then float the gate again. The LED brightness will very slowly creep up or down. The gate leakage is not too many electrons per second.

In that last state, take a pencil, or some insulated/metal thing like a small screwdriver or a trimpot tool, and pump charge into or out of the gate in small steps by alternately touching +9 or ground, then the gate. Some of us are easily amused.

John

I have done it on a design. It was with the BSS123 but it's quite similar. It all depends on the gain you want and the gate-drain capacitance is your enemy there. I guess the 2N7002 would be a smidgen better. AFAIR my amp was around 30dB and linear all the way up to 40Mhz but we stopped there, didn't need more.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

My intention was to re-radiate DAB (VHF high band) from a communal aerial socket in one corner to the radio in the diagonally opposite corner.

Originally the 7000 was in cascode with a 2N3819 - it almost worked. In the end I settled for a BF998 driving a BFQ162A.

Evidently.

Those little darlington MMICs are great for stuff like this, absurdly easy to use. Like the MiniCircuits ERA series.

John

Aerial socket, now that must be a truly Bri'ish expression :-)

Don't know what a BFQ162 is but yes, why torture yourself with low frequency devices when you can get a hotrod RF transistor with tens of gigeehoitzes for under a buck? Or less than a quid in rightpondian.

Some of the bigger MMIC, the ones with a grounded tab that can be soldered onto copperclad for heatsink, should also work. Of course, re-radiation can ruffle some feathers with the authorities.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

He said "communal." Must be a Russky.

John

Hey, I just went to communion and I'm not a Russky :-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

The block of flats has a master aerial with a bunch of co-ax cables running along the wall to each flat.

Couldn't see much mention in the OP's posts about the application - if its a direct off line switcher, MOSFETs can be cascoded for a speed advantage.

The bottom MOSFET should be a high current, low voltage ultra fast device, this makes the upper MOSFET less critical in a number of criteria.