Measuring extremely low inductance values

Hi All,

this time this is out of pure curiosity, with no intention or need to build a real device: many parts, particularly MOSFETs are specified as "low parasitic inductance", but the values given are insanely small. So, how do they measure 1nH? Or is it the result of a FEM simulation?

Best regards, Piotr

Reply to
Piotr Wyderski
Loading thread data ...

I use TDR to measure things like that.

formatting link

formatting link

A mosfet gate is in fact complex at the gory detail level. Pin inductance, pin capacitance, wire bond, and finally the chip itself.

RF VNAs can measure 1 nH. 1 nH is 160 ohms at 1 GHz.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

Maybe something like a Agilent 4285A with 1pH resolution, plus a lot of thinking about custom test fixtures.

--Spehro Pefhany

Reply to
speff

Interesting idea, thanks!

Best regards, Piotr

Reply to
Piotr Wyderski

You say that, and yet don't realize your own waveform shows a very simple RLC network transient response!

So, they /can/ be complex, but pole-zero cancellation can also make it simple. Seems like power transistors tend to be simpler, while old transistors have "drool" (a diffusion 1/sqrt(f) property, indicative of an unoptimized gate structure, versus a fractal shape, say).

2N7002 has a dominant pole /roughly/ around 10MHz or so, but doesn't run out of power gain until some 100s MHz. The impedance is very low up there too. That Cree gate in particular seems to be nothing more than an RC (however since the R adds with the 50 ohm source resistance, it's not clear what its ultimate bandwidth is -- additional measurements are needed).

Tim

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

Basically, the inductance due to the package/body length.

As mentioned, a VNA can do that. The LF equivalent may not necessarily follow from high frequency measurements, or be very accurate if measured directly.

It can be measured in situ by determining the switching loop time constant.

It can also be measured by following the transient amplitude around the loop.

I did this on a recent power supply, where the switching loop was a 0612 current sense resistor, two 3x3mm DFN8 FETs, a 2512 +V-side jumper, and some

1206 bypass caps. I followed the switching transient peak around the loop, and measured...

I don't have any waveforms handy unfortunately.

Tim

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

But how do you know what you're measuring? Is it your pure L or some disributed parasitics? Any practical way to disentangle them?

Doesn't the probe add so many unknowns that the result is meaningless? How to do it properly?

Once upon a time my scope was able to make a TinySwitch flyback stutter, and it was merely 100kHz and some hundreds of microhenries.

Best regards, Piotr

Reply to
Piotr Wyderski

What do you want to measure? Calibrate to that standard, and there you go.

For example, you might measure the _difference_ in inductance due to the part itself, as compared to a shorting bar in the same location.

Or you might look at the inductance due to the package and length, in which case you need a stub that's shorted as closely as possible at *just before* the package leads; then, subtracting that from the measurement with the package plus its body length and the return path beneath it.

Nah, probe don't mean shit. The loop impedance is tiny, ohms. My probe is, low kohms at that frequency I think. :-)

A better question is where to ground to. In that circuit, I had multiple layers (not that that actually matters) of ground, so comparing against "infinite ground" is reasonable. The probe tip comes in perpendicular to the board, away from most of the fields, which are internal to the board and components. So, I have reasonable confidence that I can, in fact, measure the waveform progressively along the total loop inductance, at least within the, say, 20% ballpark I was interested in.

Bad scope?? I don't understand.

My scope is 350MHz, and about 200MHz would be adequate for the loop I was measuring. So, my Tek 475 would do just as well, though its trace intensity would be awful for that particular measurement, and it also doesn't have a MEAS menu...

Tim

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

I did that sort of thing measuring Uber's diode laser drivers (code name Fuji). The total inductance of the laser + FET + 0402 storage cap was under 400 pH--not bad going at all.

Cheers

Phil Hobbs

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

That is awesome, I see a long RC thing. Is the boxy thing in front a capacitance? or something else?

George H. (Who bought some fast gates to make a tdr, but is doing other stuff.)

Reply to
George Herold

The big flat part is the 50 ohm hardline cable between the sampling head and the fet.

Then a couple of inductances, then the series resistance of the gate and then the gate capacitance. SiC fets tend to have largish series gate resistances.

A longer time base shows the gross gate capacitance better.

formatting link

What I sometimes do is make a Spice TDR and fiddle a fet model until it looks like the measured TDR.

Actually, 6 ohms.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

I don't think that's very useful. Impedances, when closely looked at, are always some combination of R, L, C and delays. There is really no reasonable way to work that out with an Agilent 4285A-like instrument, however precise.

TDR or VNA is really the the right solution here. I believe VNAs yield better results. I get much cleaner TDR plots from a VNA and an inverse FFT than from a straight TDR. The VNA also makes de-embedding set-up parasitics much easier. There is no objective reason why that couldn't be done with a TDR as well, but it isn't common.

Jeroen Belleman

Reply to
Jeroen Belleman

Oh sure, silly of me.

Some day I have to get my hands on a tdr to play with. I don't have any great need though.

George H.

Reply to
George Herold

The SD-24 TDR head only puts out a few hundred millivolts. Fets start doing complicated things at a few volts.

Cheers

Phil Hobbs

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

They've been showing up on eBay for under a grand, including the mainframe and one or more SD-24 TDR heads.

Cheers

Phil Hobbs

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

OK, I can't buy one at work unless I can point to some need. And for home... well next year I'll have two kids in college. The LDP (local domestic product) will be negative for a couple of years.

George H.

Reply to
George Herold

That's insulting. And it's not really simple, unless you plan to design slow stuff.

The Cree fets have significant internal gate resistance, visible on the TDR. The approximate C is obvious too. Rg is kinda distributed, so the gate isn't exactly a lumped R-C, but adding a bit of external series inductance can peak it up usefully.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Understood. I had two in college for five years out of ten. They all graduated debt free though!

A TDR would be useful for superconducting inductance measurements. ;)

Cheers

Phil Hobbs

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

If it would just work through a

Reply to
speff

Reply to
John Larkin

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.