Parasitic capacitance of SMD resistors and their generated noise

I measured parasitic capacitance and inductance for 1206 resistors and got about 50fF and 500pH. I'd expect about the same for 0402, because they have the same shape. Of course, pad layout and trace width matters, as does the proximity of other conductors. I only measured 1% metal film resistors.

Common wisdom says that only metal film should be used where noise matters. I never made comparative measurements. In my field, resistor cost is relatively negligible, so I tend to use metal film almost everywhere.

Jeroen Belleman

The parasitics of various resistors are easily googled.

Johnson noise must be the same regardless of the materials.

Thick-film resistors can have excess noise if there is voltage across them, but it's hard to measure. I managed to measure some once, but it took very high value resistors and high voltages.

High tempco resistors in a voltage divider or equivalent can make low frequency (subsonic) noise from small temperature fluctuations.

Yup. Cermet, thick film, and anything with carbon in it are disastrous for 1/f noise. They're all pretty well equivalent at high frequency. ("Thin film" is SMT-speak for "metal film".)

JL made some TDR measurements of (iirc) thin film resistors and showed that the inductance was significantly reduced by mounting them upside down, so that the resistive track was next to the board.

Cheers

Phil Hobbs

The inductance or the impedance? If the resistive layer has a smaller spacing from the board it would also impact the capacitance to the ground layer greatly impacting the impedance, probably much more than just the inductance.

As Kipling's Shipwrecked Mariner said to the Whale, "Not so, but far otherwise." (At least in my business, i.e. ultrasensitive measurements, that is.)

One time I was chasing my tail for about a day, trying to figure out why my shiny new super low noise laser driver gizmo was exhibiting horrific

1/f noise on the spectrum analyzer. It turned out to be the classic Tektronix 50-ohm 2-W BNC feedthrough terminator I was using--it was made of cermet. Switching to metal film dropped the 1/f noise by about two orders of magnitude IIRC. Not subtle at all.

Yup. Random temperature drift has roughly a 1/f**2 noise PSD.

The LIGO folks published a study on noise in resistors, IIRC, but I'm not laying my hands on it at the moment.

Cheers

Phil Hobbs

Was there voltage across that terminator? Was it making unusually large amounts of Johnson noise?

Regular cermets, but the geometric effects are the same.

You can buy resistors without wrap-around end caps, intended to solder element down.

Yes, I was pumping a couple of hundred milliamps into it. 1/f noise arises from conductance fluctuations, so it doesn't appear unless there's current flowing. As you say, Johnson noise depends only on the resistance and the temperature, but of course it also assumes thermal equilibrium, which stops applying when the power gets turned on.

Cheers

Am 07.03.22 um 16:31 schrieb snipped-for-privacy@highlandsniptechnology.com:

I have a version of Win's AOE3 ribbon microphone amplifier, not differential but only single-ended with an ugly coupling cap and 16 transistors instead of 64.

That has a switchable ~6 Ohm resistor to get "calibration noise". It is easy to see.

< > and left / right arrows

My 10 * 2 * ADA4898 amplifier uses 60R as a 1nV/rtHz normal.

Cheers, Gerhard

It's probably too late, but it would be interesting to measure the 1/f noise vs current. It might be square law, namely thermal.

Thinfilms have much lower tempcos than cermets. There could be micro/localized thermal effects too, like at grain boundaries. Laser trimming can create horrors.

I found it difficult to measure excess noise in cermets, and it was in the ballpark of the Johnson noise. These were megohm range values, so self-heating was negligable.

Then it was hard to find 10M sorts of thinfilms.

Digikey had several types of resistors.

Carbon combosition Carbon film Ceramic Metal Element Metal Film Metal Foil Thich Film Thin Film Wirewound

Last time I checked, there was only thin and thick film. That is not much of a choice.

Nah, definitely linear, and definitely 1/f not 1/f**2. A constant-rate drift notionally has a spectrum that goes as 1/f**4, and a random-walk drift ('brown noise') goes as 1/f**2. That's what you usually see go away when you put styrofoam on top of the resistor.

It's quite possible that the conductivity fluctuations depend on temperature, but I don't recall hearing that mentioned.

Well, believe me, this example was the very furthest thing from subtle.

Cheers

Phil Hobbs

To an engineering approximation, all metal / 'thin film' resistors exhibit only Johnson noise.

Cheers

Phil Hobbs

Hello, Everybody I am excited to read the discussion about the influence of resistor technology on noise. It is very interesting and helpful. Though one of the original questions was about how 0402 resistor's parallel capacitance depends on the resistor's technology. Vishay's application paper (freqresp.pdf) talks about comparison of thin film resistors with different termination styles (wrap, flip ship) and their capacitance. But they are all made with thin film technology. They mentioned that for thin film "normal" 0402 resistors with wrap termination this C may be expected to be about 40fF. This is, of course, the resistor's own parasitic C, as I would like to believe, without taking into account the PCB and pads. Does anybody of you have a similar number for a comparable thick film 0402 resistor with wrap terminals? I couldn't google it, to my surprise. Thank you for your responses

Carbon resistors are semi-metals, so they get recombination noise; probably the 'cermet' is a similar material. For metal film, though, it's terribly difficult to get a thin and long continuous metal path for high resistance, AND keep it from oxidizing and changing value. So, carbon (carbon film?) still is the solution for high-ohms items.

I have some 1T ohm surface-mount cermet resistors.

Not sure what you mean by that, exactly. To get recombination noise you need minority carriers, no?

Cermets aren't homogeneous--as the name implies, they're ceramic/metal composites.

Why so? I used to make conducting films of 100 angstroms or so. Atoms are pretty small, and using sputtering as opposed to directional evaporation will make the film follow even rough substrates pretty well.

It's certainly true that it's harder to make very high resistances out of very low resistance materials, and there are lots of low-precision applications where 1/f noise is not a serious issue--overvoltage protection, for instance.

However, carbon resistors are seriously nonlinear at high voltages--the resistance of old style Allen-Bradley carbon comps was allowed to drop by a quarter at their upper voltage limit.

Cheers

Phil Hobbs

snipped-for-privacy@highlandsniptechnology.com wrote: ================================

** A simple fact overwhealmed by persistent myth to the contrary.

** No it isn't.

** Drivel.

Easy enough to characterise carbon film, carbon comp, metal glaze & MF types with values of say 100K and few DC volts. JL does not underdstand audio.

...... Phil

Phil Hobbs wrote: ==============

** Really ???? That is massively ambiguous.

CC resistors can drop in value if they run hot and over time. Mostly they slowly drift high, over a period of decades.

Crazy to crazy that " non linearity".

..... Phil