Power MOSET gate capacitance variations

I've just measured ten MOSFETs from the same reel and their gate capacitance is between 1450 and 1900pF, with ~1620 being the most common. Since capacitance = area*permittivity/distance and the dielectric is the same SiO2, only area and distance (thickness) can vary. I don't think the manufacturer has no control over the area, as it is some form of litography, and thickness is proportional to the oxide growth time, which is exactly the same within a batch. So what is going on here?

Best regards, Piotr

Reply to
Piotr Wyderski
Loading thread data ...

Thicknesses and etching processes are not completely uniform across the whole wafer. This gets worse as companies move to larger wafers. You can guarantee ratios to be very close on the same device but not when compared with a device two inches from that location on the wafer.

Jim Thompson could shed more light on this.

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

Sadly Jim doesn't seem to be around today (so far at least) which is unfortunate. I recall many enjoyable discourses with him on this particular day going back 20+ years.

--
This message may be freely reproduced without limit or charge only via  
the Usenet protocol. Reproduction in whole or part through other  
 Click to see the full signature
Reply to
Cursitor Doom

Just let off a political rant and he'll be back, prontissimo :-)

He's got a big family and Christmas is probably not the time for him to be on Usenet much. I am only here over lunch, got to work later :-(

--
Regards, Joerg 

http://www.analogconsultants.com/
Reply to
Joerg

I rarely concern myself with any matching beyond a single die.

For multiple die matching issues I apply Thompson's "Secret Sauce" (*) methods.

(*) Coming soon to my website: A load dump of ideas that worked, ideas that didn't, ideas still in fermentation... no explanations... just "stuff" that I've let my mind ramble over. I have a copy of just about everything I've ever tried... some dates to BC, before computers, just on paper... If I last long enough I'll scan that it as well... ENJOY... or fume, your choice >:-} ...Jim Thompson

--
| James E.Thompson                                 |    mens     | 
| Analog Innovations                               |     et      | 
 Click to see the full signature
Reply to
Jim Thompson

Are you measuring at the same DC gate voltage? You will see a huge difference between the capacitance below and above the threshold voltage, as one of the capacitor plates is mostly the channel, which turns non-conducting when the FET is off. Mosfets are used as the varactors in a lot of RF VCOs for e.g. cellphones.

Maybe your parts came from different wafers.

Is there a correlation between the capacitance you measured and the threshold voltage or gm of the same parts? This might help you to figure out whether the thickness of the oxide is responsible.

Reply to
Chris Jones

Hope you get your ideas out on the net Jim. I had a physicist friend that worked in acoustics all his life, sonar and such. How had a lot of secrets that through working at a series of companies he found others didn't know. He could build efficient transducers, low loss vessels for research. At one company he reported to his manager he had built a transducer that was 110% efficient, the manager thought that was great, he said he just walked away shaking his head and fixed their measurement system. I told him several times he needed to write a book about his subject to pass the info along. He never did, Once he closed the business he just went dark, didn't make contact with any one that worked with him. He died a couple years ago and his education went with him. I had a lot off fun working with Henry, I miss him. Mikek

Reply to
amdx

Not gate voltage actually, but drain voltage -- which could very easily be drifting around if it's not grounded explicitly.

The gate has very little capacitance to the channel, at least in modern VDMOS parts -- I can't detect a Ciss(Vgs) variation, admittedly this was from a crude setup. That is to say, the linearity is better than about 10%, but how much better, I cannot say.

Cgs itself doesn't seem to vary with Vds, either. Since Ciss = Cgs + Crss, the only varying component is due to Crss, which varies quite dramatically with Vds.

Yes, it's quite possible that parts end up reeled from different wafers. I don't know any process reason why they'd do that (except where one wafer ends and the other begins), but they certainly have no reason not to, either.

For sure, if the ID code is different, they're from different lots or production dates. Any closer than that, who knows; you'd have to ask the manufacturer very nicely to get them to tell you what dies, from what fabs, were packaged on what lines, when.

Good point!

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

Doping also matters.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Of course then C_DS wouldn't contribute to the measurement either. ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

I haven't measured it on a power mosfet, I have only looked into it on lateral mosfets in e.g. 0.35um and 0.18um CMOS. For those, there is a very large variation with Vgs - more than 2:1 but I can't remember how much. (In the VCOs used in cellphone chips, usually the RF was applied to the gate and the tuning voltage was applied to the drain and source tied together. Normally it is a differential oscillator circuit. The varactors vary in capacitance greatly over one cycle of RF, going from a flat region of low capacitance through a steep transition to another flat region of high capacitance and back again, so it is more like pulse-width-modulating the capacitance than a linear varactor. Usually the varactors make up a small proportion of the tank capacitance and the rest is provided by a bunch of different sized pairs of capacitors with digitally controlled MOS switches in series, between each pair of capacitors. The switchable capacitors have less conversion of AM noise to phase noise than the MOS varactors, and smaller varactors also makes the tuning voltage less important as a source of phase noise.)

Reply to
Chris Jones

With such a high capacitance, it appears you're talking about a power MOSFET (it'd be nice to know the type). As you know, these are made with a vertical structure with a repeating fine-pitched laterally-stacked pattern. The exact height of the vertical structure is probably not very well controlled, and the lateral (not vertical) gate oxide thicknesses may also not be well controlled. From an Infineon note, "laterally stacked fine-pitched pattern of alternating arranged p- and n-areas. The finer the pitch can be made, the lower the on-state resistance of the device will be". Manufacturers are optimizing for density, not dimensional uniformity.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

Yes, as the title says. The part is IPD60R400CEATMA1

-- maybe not the best in its class, but dirt cheap and sufficient.

The origin of that spread would be obvious for parts from different wafers, but for a given wafer the oxide thickness should be almost constant, even if one is not willing to control it tightly. I have no evidence that the parts are from the same wafer, but why should the factory mix them before encapsulation in the reel's compartments?

Best regards, Piotr

Reply to
Piotr Wyderski

That's not a bad part, one of the new SuperJunction MOSFETs, with nice low drain capacitances. These have a severe horizontally-compressed architecture, with everything turned 90 degrees, changing all the de-facto rules we're used to.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

That's the kind I measured; and lower voltage (30 to 200V, say) parts seem to be grossly similar designs (though with not as much historical improvement as SJ did in one fell swoop!). SJ also has a current limiting characteristic unconnected to Vgs -- transconductance tanks above about 7 or

8V. Reminds me of toobs, with limited cathode emission... :^)

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

Not bad for sure, but not that shining either. 400mOhm in 2016 is not a jaw-dropper, the 12 years older 20N60 has just 150mOhm, but was *much* more expensive back then.

Best regards, Piotr

Reply to
Piotr Wyderski

Well, if you just want low resistance at minimum cost, there are IRFPs for you. But there's about three other critically important parameters to a switching circuit that you're missing there, and the total improvement is almost an order of magnitude between them. All that, at nearly the same price.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Contract Design 
 Click to see the full signature
Reply to
Tim Williams

My continually-updated MOSFET spreadsheet (a portion of which appears in a few places in AoE III) now has over 1300 items, 380 in the 500 to 650-volt range. 180 of these have lower Rds(on) than the IPD60R400CE. So it's entire a matter of choosing your tradeoffs. There are a many reasons against selecting a lower Rds(on): high capacitance, poor distributor stock, high cost, larger package, no SMT package avail...

--
 Thanks, 
    - Win
Reply to
Winfield Hill

That sounds very handy. Is it possible to purchase a copy?

Thank you,

--
Don Kuenz, KB7RPU
Reply to
Don Kuenz

Did you get my email? Has it changed since your usenet log-on information? Mine has, now snipped-for-privacy@yahoo.com

--
 Thanks, 
    - Win
Reply to
Winfield Hill

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.