Turn-on and off times in a FET, weird data

Comes up as password protected for me.

Cheers

Phil Hobbs

P.S. This would be a more normal set of data and I can't imagine that IR's process is so much different from Fairchild's:

Okay, Octopart had an unprotected datasheet.

If you're moving 50 nC of gate charge in 150 ns, that's 300 mA. Maybe their gate drive circuit only put out 10V through 25 ohms?

Cheers

Phil Hobbs

that is what it says ;) VGS = 10V, RG = 25R

-Lasse

Yeah, we routinely get mosfet switching times that are a fraction, or a decade, faster than the data sheet numbers. Just drive the gate hard. I've switched 50 volts, 1 amp, with 2N7002s, in under 1 ns, well over 20x faster than the data sheet suggests.

Source inductance starts to be a problem at high currents. Paralleling smaller fets can help there.

Another issue can be the fet's internal series gate resistance, which is not often specified. Some are fractions of an ohm, some are multiple ohms. The kilovolt SiC fets look great until you see the Rg numbers.

Ignore those pessimistic Vgs breakdown voltages specs!

How fast do you want to switch? I have one potential application where I'll need about 100 amps of gate drive. Somewhere in the 2 ns range, peak gate current equals switched drain current!

I TDR mosfets to measure their source inductance and gate R+L values. Apparently similar parts can be very different.

That's weird. I can load that even with cookies disabled (standard procedure here) and no password and such.

If the datasheet link in here (same document) doesn't work then there's a more serious problem:

That would be like testing the acceleration of a new just introduced car on a sheet of ice and write down 0-60mph in 30 seconds. Normally marketeers want the products to look like "With our socks you can fly to the moon!" and this datasheet looks like the performance is the pits.

In my case 20nsec 10-90% in either direction would be fine. 10nsec would be better, of course. Thing is, I need to keep the bottom FET in a class D amp away from where its body diode comes on and that's why I need an Rdson in the 50mOhm range.

With datasheets like the one above it remains unknown how much gate resistance the part has. No idea why they did that. Why on earth would anyone make their parts look this bad?

Yeah, which is where my question comes from:

Why?

Speccing a big FET with 25ohms RG makes the part look real bad. Unless it is bad because of high internal gate resistance and IME this one isn't. However, eventually before this goes into production the whole design needs to be validated and such data in a datasheet can quickly make the part a pariah and requires it to be replaced by something with better "formal" data.

Currently I am looking for a part with better formal data but not easy because I need isolated Full-Pak or similar.

Oh, I can think of a REALLY stupid reason. Use a pulse generator with a 50 Ohm series resistor in the output, and a 50 Ohm terminating resistor at the DUT. So, looking back from the transistor, the driving impedance looks like 50 Ohms. But, that 50 Ohm series resistance at the pulse generator gets massively affected by the Miller capacitance, especially with Vdd of 100 V or more.

Who did such stupid testing? Is this because all transistor data sheets are tested like this? Makes a big difference whether you are testing small signal devices or 100+ V, 30+ A power transistors.

Jon

-Lasse

Den onsdag den 1. oktober 2014 22.29.00 UTC+2 skrev John Larkin:

I've seen an app-note somewhere that used the kelvin connection on a curren t

-Lasse

Any chance someone at the company would tell you, so you can tell us? Mikek

PS. When I see this kind of spec stuff or all the negative talk about Maxim, I don't know why the companies aren't responsive. Although I'm not in a purchase dept. for manufacturing, I have heard comments about Maxim that I would never look at a Maxim part for use in a product. Maybe this a niche, a company that just spec's parts and sells the data.. Disclaimer: These spec's are only valid for the parts we tested. :-(

I just sim'd with the IRFI4227 and came away disappointed. Huge gate charge. Will do the 4229. STF40NF20 could be good as well but ST's web site is screwed up. Can't get a SPICE model and so far they've typically not even answered requests, so maybe not.

No Spice model for the 4229! Just great. Hurumph.

ster when driven hard. But even with 25ohms driving resistance which any sa ne engineer would never do it should be faster than this. I tend to drive w ith zero additional gate resistance, pedal to the metal. Why is such unfavo rable data published? I need them in a class D driver and want to nail the dead time well before turning it on.

decade, faster than the data sheet numbers. Just drive the gate hard. I've switched 50 volts, 1 amp, with 2N7002s, in under 1 ns, well over 20x faste r than the data sheet suggests.

smaller fets can help there.

s not often specified. Some are fractions of an ohm, some are multiple ohms . The kilovolt SiC fets look great until you see the Rg numbers.

I'll need about 100 amps of gate drive. Somewhere in the 2 ns range, peak g ate current equals switched drain current!

be better, of course. Thing is, I need to keep the bottom FET in a class D amp away from where its body diode comes on and that's why I need an Rdson in the 50mOhm range.

Perhaps just ignore the timing data and just look at the input cap and gate charge. The EPC2010 (digikey it) don't even say anything about timings. But with 500pf cap and 5nC gate charge, i am sure you can switch it in less than 10ns.

yeh, the irfi4227 and irfi4229 specs about twice the Ciss but the datasheet for those and STF40NF20 shows a much more sensible RG of a few ohm

-Lasse

ST does have some nice looking mosfets, very low Qg.

Infineon often puts Rg on their data sheets, and some of their dpaks are, like, 0.2 ohms.

I wrote to them and will let you know. But I am not holding my breath.

Beats me. Some like Linear Technology are very responsive and have earned fat design-in profits on my part. Others such as ST are often rather unresponsive and, therefore, hardly get any design-ins from me. Design decisions by consultants are usually final and designs are typically not touched again for decades unless there is some pressing need. Sometimes they'll never know how much they've left on the table. Other times they notice and a sales guy bangs on the door. Too late.

I do not use their products in my designs. But that has mostly to do with numerous occasions where clients had been left without timely production quantity shipments. OTOH that has brought me several "design-out" assignments.