I thought for sure you had seen this before; I guess not, in which case my sincerest apologies.
That was a curve fit to your data; it shows a simple RC.
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
Well, there are the initial inductive bumps.
Cree has the full, nonlinear device model, with inductances, but they got the substrate diode very wrong.
The TDR pulse is only 0.5 volts, so it doesn't excite the large-signal nonlinearities.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Easily attributed to mismatch -- the component leads have a much higher TL impedance for the 1-2cm length there. Which causes CM-diff mode conversion so the whole thing rings a bit.
Here's a more exaggerated case, measured with a 10cm twisted pair hanging off a BNC-binding post connector:
SiC diodes are pretty awful, but so are non-primitive SPICE models, so often...
Y'ever seen an MOV model before? The formulas they use are a nightmare.
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
The Cree model shows 15 nH of gate inductance and 9 nH of source inductance. In my HV pulser, I'm putting a couple of amps into the gate with 1-2 ns edges, and I'm switching 12 amps in a few ns. Those translate into 10s of volts of L di/dt. I'm amazed it works at all.
The limit on gate drive is blowing it up, which I have done. These people have specified abs max gate voltages right on the eve of destruction.
The chip itself is tiny and pretty far from the pins.
It's actually a terrible package for fast work. Two or three smaller parts in parallel might be better.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
That RF transistors exist, in TO-220 and 247 and other gnarly packages, is remarkable.
I remember reading the datasheet for a a low-VHF RF final in TO-220. The s-params go all the way around the circle at the intended operating frequency. Yikes.
I remember seeing a Microsemi MOSFET module (half bridge) that was rated up to _200kHz_. What's the freaking point? It cost ten times more than the equivalent rating in TO-247s, and contained as much internal stray inductance as two TO-247s in series. The mind boggles.
IGBTs are now fast enough that they _have_ to worry about that, and so most new IGBT modules are, you know, actually made right, with laminated bus bar construction.
Power GaN is almost only available in DFN and CSBGA, as it must. That anyone even thinks they want one in TO-220... yeesh!
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
My* Rubidium lamp (~70 MHz, ~1.5W) is a coil/auto-transformer, cap and 'gnarly package' RF power npn, AFAICT they are made in Russia.
George H.
*mostly copied from an Rb freq. standard company with their blessing, they stopped making the lamp with no US supplier. We were buying lamps from them, I think mostly as a favor. (We had an old atomic physicist (oap), who knew their oap.)
In an RF amp, you can tune out/resonate the parasitics. Some RF ldmos fets have deliberate wire-bond inductances that optimize for some specific RF band.
We work wideband, time domain, so we want low inductances. The tuned RF fets are useless to us.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
Right. I corrected that a few days ago.
It's 1 pF that's 160 ohms at 1 GHz.
-- John Larkin Highland Technology, Inc picosecond timing precision measurement jlarkin att highlandtechnology dott com http://www.highlandtechnology.com
That's such a nifty plot that I couldn't resist adding one of those to an outgoing DigiKey order:
Same features visible at similar detail with a 50 GHz sweeper and a (SD-32?) sampling head. Guess ol' Joseph F. knew what he was talking about after all.
-- john, KE5FX
Is that a TDR derived from a VNA frequency sweep? The right side isn't the same RC tau that I see; what was your sweep range?
TDR is DC!
Those Cree fets are great, but they do need a lot of gate drive.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
I know that SiC is very expensive oer area and next generations are supposed to be even lower RdsOn and smaller die.
So, WHY even put those tiny die into a TO-247 package ? I don't think that huge (relative) package would be any better than say, a TO-220 package ? I don't thin that the Rth J-C would be much worse if at all different for what I saw in your picture. We use D-Squared SiC packages and the die size is just slightly larger (I thiink) than in your picture of the TO-247 Cree package.
Hopefully, eventually the price of the die/wafer will be low enough where we can choose to have a larger die with super low RdsOn and Rth J-C wihtout spending $1000 per part.
The TO247 for that tiny chip is kind of silly. It does get some modest benefit from lateral heat spreading, with all that copper. I'm spacing it off a water-cooled block with an aluminum nitride insulator. I'm using a TO220 sized insulator, to keep the capacitance down, so the package may as well be TO220 size too. They maybe went with TO247 to get the clearances for 1200 volts.
There are some 1200 volt mosfets in dpaks and some 1500 volt d2paks.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Yes, it's an HP 8510C configured for 801 points at N x 62.4 MHz. The vertical scale should be the same as yours, at 200 milli-rhos or whatever per division.
One drawback with simulated TDR is that large near-end responses tend to degrade the fidelity of later responses. The RC part of the curve might be getting hosed by that phenomenon.
Are you using an SD-32 or SD-24 head?
In lowpass mode the VNA has to extrapolate down to DC when it does the math, so there's room for error there. And calibrating in SMA to 50 GHz is an exercise in wishful thinking.
The plots I get out of this setup are probably comparable to what you'd get from a properly calibrated 20 GHz rig plus some fine-structure details that might or might not actually be there. That's why I thought it would be worthwhile to compare it to a real sampler (which I don't have).
The interesting part to me was the pair of closely-spaced inductive responses near the package boundary. Those are a pretty good match considering how different the instruments are.
I have a Peak Atlas pocket-size component analyzer that's pretty hard to fool, but it only saw the body diode.
-- john, KE5FX
Any TDR has that problem. Every feature distorts the step and reflection of downstream stuff.
11802 and SD24.
The big difference is the slow, gross gate capacitance, because of the limited low frequency span of the VNA. But we can measure that other ways, so it's OK.
The body diode on the Cree fets is terrible, and their Spice model doesn't suggest how terrible.
-- John Larkin Highland Technology, Inc lunatic fringe electronics
Doesn't make much difference in stray inductance anyway; the leads are further apart but also thicker.
Also, why not D[N]PAK? Creepage is plenty, and with efficiency taking priority in many applications, heat dissipation of a THT part isn't needed.
I'd like a SMA sized SiC schottky, which would only be good for a few 100 mA, but alas the closest thing that exists is $20, what a joke.
Makes miniaturizing circuitry really hard, especially when you need a small snubber, or a high voltage output. Sigh...
Tim
-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website: https://www.seventransistorlabs.com/
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.