So, where is these 100nH you speak of? (Dead power resistor microshot)

[daqq]

Hi guys,

I've been killing experimenting with various snubber resistors. I thought you might like a photo of a dead PWR163 series resistor ( http://www.farnell.com/datasheets/1896410.pdf ) from the inside.

Basically, it's a resistive square on a ceramic substrate with leads on it... the trimming seems to be done on the inside of the square. The package is a D2PAK-ish device

The datasheet specifies:

Inductance: 0.1uH maximum

I'm wondering how it would be even possible to get such a high inductance on such a small space as is the resistor. A centimeter of wire should have around 10nH. What would you say is the actual inductance value for such a resistor?

Resistor:

Plastic body:


David

[Kleinstein]

The 100 nH is a maximum value. So the actual value can be considerably smaller.

 At such low Inductance values in the sub 100 nH range - there is limited value in assigning an inductance to a single part with pins wide apart. The inductance depends on the whole circuit: even if you say each cm of wire has 10 nH, one just can't add them piece by piece like resistance. The sections will interact: in a large loop or even with several turns inductance can be higher and with a return path close by it can be lower.

[T3sl4co1l]

They don't specify inductance versus value.

Probably the 10k+ parts have a sinuous etch pattern, which would easily give that much inductance.  (Though it would be difficult to notice, and at any frequencies where it would matter, capacitance would tend to dominate anyway.)

The low value pictured above is probably little more than the package inductance itself, i.e., around 6nH.  This should be fairly consistent from part to part (within the same value), so if you'd like to measure it yourself, it would be okayish to rely on it being roughly consistent.

They don't mention capacitance, either (they at least say the tab is isolated).  It would be quite reasonable to suppose the resistor has an L-C-L model at very high frequencies.  Which means you can maximize its bandwidth, and minimize peaking or whatever, and all that, by matching these parameters to the circuit.

You'd have to measure C to figure that out, but otherwise, both L and C should be fairly consistent between parts of a given value.

Example: suppose the tab is 5pF (just guessing).  Then 3nH + (5pF)/2 (*) gives a lowpass cutoff frequency of 1.8GHz and a characteristic impedance of 35 ohms.

*The capacitance is divided in half here, by symmetry: half associates with the first 3nH, half with the other 3nH.  When a network is symmetrical, you can divide it in half and analyze each half-network, then put them back together by basically squaring the result.  (So for example, the overall -3dB frequency is actually the half network's -1.5dB point.)  Alternately, we can take the totals of L and C (6nH and 5pF), and we'll get the right result for Zo still, but Fo will be underestimated by 2x.  Which is a conservative figure, and a good answer if you want to ask: at what frequency might I have to worry about attenuation and phase shift, or transmission line effects?

That means, if you were to use this resistor as a current sensor, it would be best suited to monitor the current flowing in a 35 ohm transmission line: the attenuation of that line will be minimum, and flat bandwidth will be maximum, at that impedance.  And for impedances less than that, it will look largely inductive, or if higher, capacitive.


FWIW, I once built a 20dB power attenuator, using a similar part (a 75.0 ohm TO-220 version).  It gives quite reasonable step response out to at least 400MHz bandwidth, despite the somewhat haphazard wiring.  Note that an average value resistor (like 75 ohms here) will tend to be better-behaved than a very low or very high value, since R is closer to Zo.

Tim