Learning Mosfets...RdsON Warm Up?

have a look at the datasheet for a random mosfet, look for tj vs. Ron looks like it generally doubles from 20C to 100C

-Lasse

Which MOSFET, and how are you driving the gate? Please describe the entire circuit.

And what do you mean when you say that "Vdson is still increasing in voltage (Ids still increasing) after a 50uS mosfet turn on time"? Do you mean they hit bottom and are now moving up again? Can you post a shot of the waveforms?

-- Joe

Waves coming up.. Have to set up scope+PC interface and try out the software for the first time.

Hey, don't make too many waves up there in Canada :-)

Take a look at figure 4:

--
Regards, Joerg

http://www.analogconsultants.com/

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

To bad they dont have 500 or 600V Fets like that. Low gate charge and Rdson.

Using function generator to drive the mosfet.

Using obsolete mosfet:

Dead bug picture.

603KB JPEG

A 12ohm resistor from drain to V+ = 15V. Source to ground. Gate to function gen. Scope 10:1 probe on drain. Ceramic cap is 0.1uF.

Yup. I'm watching Vds after the Vds negative edge. Upon gate +pulse, Vds drops down, goes a bit negative, rises with a steep slope (circuit inductance) then morphs to a lesser slope(from heat???). (Scratching head..)

2.4MB bmp... Can you post a

I was expecting Vds to drop to Ids*Rds during the on period. All nice a square like seen on a simulator. A little puzzling.

Yup. Rdson is sensitive to temperature. I'm just surprised to Rdson variation within a 100uS on time. If that's what I'm seeing. That or something goofy somewhere.

For really good devices check out Zetex.

Purchased by Diodes Inc. Funny, I would have predicted the other way around.

John

The thermal mass of a drain channel isn't all that high. Eventually the heat distribution acros the transistor evolves into a nice smooth thermal gradient from channel to heat sink, but initially you've got a lot of room temperature silicon to take up the first few micro-joules of dissipation. Jim might know representative dimensions for drain channels.

Silicon has a heat capacity of about 700 joules per kilogram at room temperature.

The data sheet suggests limiting avalanche energy to 70 mJ, and stops allowing you to increase the single pulse avalanche power for pulse- widths below 6usec - which has probably got more to do with the onset of channeling than anything specific about the thermal mass of the channel.

Figure 13 of the data sheet shows a three-componenet thermal model, with the shortest time constant at 260usec, followed by 1.228msec and

8.12msec. A more physically realistic model, with concentric shells of silicon getting warmed up in succession, could probably be resolved into exponential components including some with shorter time constants.

-- Bill Sloman, Nijmegen

Yikes! You really need some sort of ground structure. Even if it's just a piece of copper clad.

Try using the PNG format next time, much better, much smaller. My scope uses just 5k-10k for a pic of similar detail.

Well, it has no heat sink at all. The channel will get hot very fast. But without a ground structure it's hard to say how much of the measurement is for real.

--
Regards, Joerg

http://www.analogconsultants.com/

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

IRF is usually the leader of the pack. But IIRC they have sold rights to some devices to others (Vishay?). Disinvested or whatever the MBA speak might be.

--
Regards, Joerg

http://www.analogconsultants.com/

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

Aside from that twiddle at the beginning, it looks like two exponentials, one fairly fast (tau ~ 2us?), the other a lot slower, closer to 15-20us.

You're clipping off a lot of waveform, do you know what the recovery time is on your scope? At that scale, the ADCs should be overloaded, I wonder how well they do.

Pfffbt, simulator, schmimulator. ;-)

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

b17n20d.pdf

New 'scope right? Try adjusting the little screw on the probe.

-- Joe

This 'octopus circuit' is glued to a pcb copper side down. Grounding the plane doesn't make any changes to Vds.

I've yet to read up on the PNG format. I don't know yet if it's supported by net browsers.

I move all the wires around to check for E field and M field coupling. No Vds signal change. afaik parasitics to the ground plane reduce E-field coupling and I don't think that was needed for this circuit. I'm not familiar with conductor inductance difference with and without a ground plane. I was playing with a strip line calculator and the conductor inductance went up as the conductor got closer to the ground plane???

mosfet:

frick! That's it! Adjusting the probe comp screw is the only thing to affect the area of interest.

Vds is yellow trace. Green trace is my other probe deliberately put out of calibration(more than what is was) to emulate out of cal probe 1. Green is on the cali pulse out on the scope and due to f diff I waited for the signal to float by for stopping the acquisition.

40k PNG and in color!

I suppose if I had a fet probe/active probe, Vds would look more like as seen on a simulator. That or I learn how to discern what's probe signal and what's circuit signal.

groan.. :(

Thanks

I dunno.. In this case, the simulator was more right than my testing. It was my probe changing the signal. :( When I adjusted the probe comp screw, I get other curves in this area.

Yeah, but it makes it difficult to predict exactly what you are measuring. When hunting very subtle or strange effects the first thing to do should be establishing a well defined environment. A piece of copper clad ain't that expensive and can be re-used :-)

They do. TIFF can be a problem, or usually is. I use PNG with all my clients and never had an issue. A click on the file and it either opens right up in the browser or a viewer.

You might get lucky sometimes but I'd still do this on a piece of copperclad. Source solder smack to the plane, coax (with divider if needed) straight to the scope with it's shield also soldered to the plane. I have a bunch of BNC with solder tabs in the drawer for that so I can just plug it in. That also takes out any weird effects a probe can have.

--
Regards, Joerg

http://www.analogconsultants.com/

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

mosfet:

Neato! Notice how the lettering is crisp and clear while it was fuzzy in your huge bitmap before.

That's one reason why I prefer a direct coax link to the scope for stuff like this. I've once scratched my head for a few minutes, knowing I had cal'd the probe an hour ago. My mistake was that I answered a nature call, couldn't find the bathroom right away because it was a client building and later when I found the probe cal out of whack one of the guys said "Oh, maybe it's because I just dropped it?"

--
Regards, Joerg

http://www.analogconsultants.com/

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