Resistor noise test setup article , interesting reading

[Overspeed]

Hello

An article with a lot of detail and results from the CERN

Regards
OS

[Jacques]

An article with a lot of detail and results from the CERN

This is the same person (Beev, N.) that designed the open-source 10V digitizer from CERN. If you search, e.g. using Google Scholar, you'll find some other interesting publications as well, for example, one that compares 1/f noise in commercially available ADCs. 

[Conrad Hoffman]

Geller has been out of business for a good few years (2013), but his JCan project is still documented, here and in the wayback machine-

https://www.nutsvolts.com/magazine/article/build_the_jcan_to_measure_resistor_noise
https://physicsopenlab.org/wp-content/uploads/2016/10/JCan-NV-article.pdf
https://web.archive.org/web/20121109190034/http://www.gellerlabs.com/JCan%20Parts%20and%20Kits.htm

I built this and it was quite enlightening.

[Alex Nikitin]

There is a difference between the Johnson (thermal) noise of a resistor of a certain value (which is measured in the second article) and the additional (excessive) noise of a particular resistor when it is passing a DC current (that excessive noise is measured in the first article). The excessive noise is fairly difficult to measure but it is one of the most useful parameters of resistors used in low noise and high precision circuits.

Cheers

Alex

[branadic]

We did investigate excess noise in parallel to Nikoai by the time: https://www.mdpi.com/1424-8220/23/3/1107

-branadic-

[Simmed]

the statistical array thingy thread

https://www.eevblog.com/forum/metrology/statistical-arrays/msg3139710/#msg3139710

[TimFox]

Since thermal (Johnson or Nyquist) noise depends only the resistance, "excess" noise is the only useful parameter that distinguishes resistors from each other in noise performance.
Approximately, the noise spectral density, in W/Hz, is proportional to 1/f, hence the common name "1/f noise" or "pink noise".
In resistors, it is only present when there is a DC current flows through them, and above a "corner" frequency, depending on current, it falls below the inherent Johnson noise and is no longer significant.
Often, it is quantified by a "noise index":  NI = 20 x log₁₀ (uV noise per V DC per decade of frequency).
Since the noise density keeps increasisng at lower frequencies, it cannot be eliminated by averaging in the same way as "white noise" from Johnson noise.
Metal foil and wirewound resistors can be better than -30 dB; depending on resistance and construction, carbon can be worse than 0 dB.
One need worry about excess noise in resistors only when there is DC across them, such as bias dividers and emitter resistors.

As is well known, active devices also show excess noise, usually near 1/f.
When I was in grad school, a lab down the hall was investigating excess noise in pure metals and found that even in pure silver, excess noise could be seen if you lowered the temperature (and therefore Johnson noise) to incredibly low values.

[Kleinstein]

The excess noise is relevant at lower frequencies. So below an 1/f corner that depends on the current. The excess noise is also relevant with AC voltage - this is the way the article from the start does the measurements. So the bias at the resistor is modulation (square wave) and this way amplifier drift and 1/f noise gets less important.

A noise index of -30 dB is alreadly relatively poor - some normal metal fim resistors can meet this. Metal foil and wire wound ones usually challange the test systems and expect them to be better than -50 dB in most cases. Even thin film can reach the -50 dBi range.
Quite some specs found for resistors are more limits of the measurement system at the manufacturer.

[TimFox]

VPG claims noise index < -40 dB for their bulk film parts:  I don’t know if they guarantee the value, since it’s expensive to measure.
The common definition of NI makes it independent of current, although the excess noise itself increases with current.

[Conrad Hoffman]

Most of my resistors have DC across them! I've only found excess noise to be an issue with carbon composition and thick film SMT resistors, but there are far more sensitive applications than mine.

[Overspeed]

Hello

My concern and I think a general concern is the case of very low voltage as resistor divider , on some volt level , noise can be perhaps a detail , when voltage level go down to uV and nV noise is real concern .

Regards
OS

[Kleinstein]

The excess noise behave like a random variation in the resistance. So with a divider and small voltages this is less an issue, as something like 1 ppm of a small voltage is very little. The excess noise is more an issue with high applied voltage and it is about good relative resolution.

I have noticed excess noise at the muli slope ADC. This is would resistors that have some 14 V applied. With a not so good (still thin film) resistor array the excess noise was more than the 1/f noise of the OP-amps.

[1audio]

After looking at the article I thought I would check my Quan-Tech 315C. As far as I know Quan-Tech was the only commercial manufacturer of resistor noise testers. Mine is their last version and they ceased to exist about 20 years ago. In any case I put a Vishay 1K foil resistor on at .1W and got an index of -43.6 dB. Given the base thermal noise you will hit a limit on noise index I would think. I think the Quan-Tech limit is from the bandwidth of the 1 KHz bandpass filter. It has an input transformer for the lowest ranges (down to 1 Ohm). I may try a 1 Ohm wirewound if I can find one in my collection.

[1audio]

OK, a 2 Ohm resistor gets a noise index of -27 and 2 different 1 Meg resistors get "overflow". I need to find a manual for this to make sense of what its limits are.

[Kleinstein]

The -43 dbI and -27 dBi for the 2 measurements are more like the meters own noise. The foil and wire wound resistors should have very low excess noise, very challanging to measure (like < -60 dBi).

Measuring the resistor excess noise is not easy. One needs to measure at quite low frequencies and the amplifier has to be suited for the resistance range. So one would generally need a different amplifier for low (e.g. 100 ohm) , medium (e.g. 10 K ) and large (e.g. 1 M) resistors.
The typical setup uses not 1 resistor but a bridge with 4 equal ones to suppress the noise of the DC source.
One would normally measure the excess noise at realtively low frequencies (like 1 Hz), so that the normal thermal noise is less of an issue.
For resistors in the right range (like 1 K to 100 K) one can get away with a relatively simple setup with a modern AZ amplifier chip, battery for the DC bias and DMM to read the output. Even the simple setup works not so bad. AFAIR I got a measurement limit (uncertainty in the compensation of the background noise) of around -50 to -60 dBi  (depends on the resistance). Excess noise is measurable for the more noisy thin film resistors, but some good ones were below my limits. It also helps if the resistors are low TC as the thermal drift of the DUT resistors can be anoying.