High 1/f noise in SM resistors

[bcarso]

[quote author="AMZ-FX"]
He is talking about carbon comp but the principle applies to film resistors to some extent. I would expand that to say that 1/4w are more noisy than 1/2w and 1/8w more noisy still.... size does make a difference.

regards, Jack[/quote]

That is definitely borne out in the experiments.

I did finally find a Matsushit* document on SM resistors wherein they show some excess noise data. The English is a bit garbled. See their pg 11 (pdf page 12) and following:

http://industrial.panasonic.com/www-data/pdf/AOA0000/AOA0000PE8.pdf

There was also a paper about noise within the resistive material, as opposed to simply the contacts, and methods of reducing it, that came up on a search, but I'll be damned if I can get back to it. It was also the search that yielded up the outrageous flamer so I suspect my subconscious is intervening :grin:

 

[Rossi]

BTW. There's a German online shop that offers metal film resistors ("Metallfilmwiderstand") and metal layer resistors ("Metallschichtwiderstand"). Is there a difference between the two? Or is it the same stuff from different manufacturers?

 

[bcarso]

[quote author="Rossi"]BTW. There's a German online shop that offers metal film resistors ("Metallfilmwiderstand") and metal layer resistors ("Metallschichtwiderstand"). Is there a difference between the two? Or is it the same stuff from different manufacturers?[/quote]

Could be their version of thick film vs. thin film. The thin films have a much lower excess noise according to the Matsushit* reference.

 

[Rossi]

I was just browsing the Vischay site, because I think the ones they sell could be Vishay (I don't really know). But they offer so many types, I'm confused. There's metal film, metal foil, metal oxide, bulk metal foil. I guess I just stay with the cheapest ones I can get. Curiously enough metal film 1% is almost the same price as 5% carbon film around here.

Anyway, I found this note on resistor noise in audio applications. Of course it pimps Vishay's "noiseless" bulk metal foil stuff: http://www.vishay.com/docs/49414/vse-an00.pdf

 

[bcarso]

The Vish*y bulk metal foil is where they got started, circa 1968. They have this amazing tempco which has to do with matching piezoresistive effects with more conventional temperature coefficients if I recall correctly.

They hand-trim the parts to tolerances as low as .001%, and the tempco can be as low over a narrow range as <<1ppm/degree C. They are also very low inductance. I have a few samples of an older process that look like super-expensive capacitors, in metal cylinders filled with oil and glass-metal hermetic seals, from when UCLA Physics was scrapping a brute-force precision D-A converter which used an R-2R ladder made out of them. Each bit switch was a whole circuit board. They used the beast for driving deflection coils in some scanning data analysis system looking at bubble chamber photos.

I never expected a boutique outifit like them to become the resistor conglomerate they are now, but evidently they are doing a pretty good job of it. They have branched out into semiconductors too.

 

[PRR]

> document on SM resistors wherein they show some excess noise data. The English is a bit garbled.

A bit.

But if they wrote what I am reading, 1/f noise is M/V, where M is a quasi-constant for the material (magic foil or sooty clay) and V is the Volume of the resistor. (They throw in some LOG to get a result in dB re: -120dB.)

So you can get some advantage with a better material.

But you can also (if I am reading this right) get a similar advantage with a BIG resistor.

Taking Vishay's self-pimping "data", magic-foil 1/f noise is 40dB or 100 times better than carbon-comp. So we could use a c-c resistor of 100 times the volume (~5 times the linear dimensions, maybe 25 times the watts) of a magic-foil resistor, and expect similar results.

Or if you have a design that shows 20dB more 1/f noise than you want, use the same type resistor but 10 times the volume (about 2.2X linear dimensions, maybe 5X power rating).

That's no help if you have a 50sq.mm. project that has to fit a 30sq.mm. board.

Is there really a fundamental reason a physically small resistor is noisier?

As a counter-argument: the noise is real power output. Therefore as the generator gets smaller, its power output rises, and an infinitely small resistor would make infinitely large noise power. Ignoring Quantum physics (maybe we can't?) that seems to defy thermodynamics, though I can't cite which law it breaks.

The noise they are talking about only happens when there is a DC bias. Simple termination resistors would have very low noise (though some noise modulation with signal). Maybe the interpretation is that a large DC voltage in a small resistor, high volts per micron or per atom, excite random processes? If we scale physical size, we must scale voltage? That does make intuitive sense if you don't ask "Why?" too much.

 

[PRR]

Practical Issues in Electrical Measurements on Nanoscale Materials, Structures and Devices may be relevant; if so, it is over my head.

----

Incidentally, I hit some low-level pages that may be helpful for beginners:
http://www.pc-control.co.uk/
http://www.pc-control.co.uk/resistor-types.htm
http://www.pc-control.co.uk/capacitors.htm
http://www.pc-control.co.uk/toc.htm
Not everything written is perfectly correct, but a good overview.
Might be good in a Meta.

 

[bcarso]

If I read Professor Martini (reminds me of the movie Back To School, where the Dean is named Dean Martin) correctly, based so far on the abstract that comes up with the link from Crusty2, the stochastic process that goes on is revealed when the d.c. bias is present, but not necessarily driven by it. That is quite intriguing.

A while back I argued with a company who were out presenting to H*rman, promoting their programmable switched-capacitor structures to make analog function blocks to sell to the legions of people who can't design anything (almost everybody now), but can maybe follow a GUI. Mr. Chief Technical Officer, appropriately doctorate-ized, took issue with my contention that their MOS structures suffered from 1/f noise due to various trapping phenomena. He said "If I recall correctly, 1/f noise only occurs when there is a current flowing."

I knew better from experience with self-scanned photodiode arrays, but I didn't rub it in. Do you want an olive or an onion in that, Mr. CTO?

Unfortunately I couldn't convince the client in the present situation to do one more board turn to accomodate some bigger parts, but the use of the P or Y chip resistors plus a reduction in currents picked up enough improvement to be just bearable.

 

[12afael]

great topic.

I suppose that internal resistor on ic suffer from the same problem.

one point more for discrete stuff...

 

[Winston OBoogie]

...

 

[dale116dot7]

If this is correct, then the TLM103 microphone should (by simple observation) have a very low amplifier noise, and a clone constructed out of SMD parts should be much noisier. I haven't measured the difference, but I have definitely heard the noise difference - for a while I thought my mic capsules were bad or noisy. At least a 15 dB difference, judging by my preamp's attenuator settings and my ears. Some of the printed resistors in the Neumann mic are very large, and that mic is particularly quiet.

For low-noise analog parts, where resistor performance is critical, I can imagine that a ceramic hybrid circuit might offer a good solution, provided the quantities are high enough to justify tooling up for it. You can optimize the resistor sizes in your circuit for optimal noise - the most critical resistor can be made to the largest area, for example.

All of my engine controller designs are SMD, but they have to be for reasons that include cost, ease of assembly, packing density, and reliability with that packing density. Noise doesn't come into these designs at all.

 

[bcarso]

Evidently the thin-film parts are pretty quiet but quite a bit more expensive. DigiK*y now has some available, and they are the only SM resistors that mention anything about noise. MELFs, which are essentially metal-film axial parts without leads, are apparently comparable to the same size axials. But they are so despised by manufacturing engineers that they are hard to get.

The thick-film parts are quite a bit noisier than anything else. I laughed a few days ago when I found that I had a book on the shelf all this time with a paper about thick-film resistor noise that I'd paid no attention to---I got the book because of another article about 1/f noise and number theory.

The design that brought all of this to a head for me has a lot of gain at low frequencies, and an input stage with substantial d.c. present. When designed initially the gain requested was lower and the potential input voltage range higher. Once I got hold of the actual signal source sample I realized that at least some of the input stage currents could be scaled down without any downside.

Had I known how bad the cheap thick films were going to be I would have just designed in a few standard 1/8W or 1/4W axials, even if they had to be hand-inserted (also hated by manufacturing though). Time ran out.

 

[waltzingbear]

a little late,but.... next time.

You forgot the useful technique of paralelling things, lower the noise by 3db with two resistors, 6db 4 resistors.....

and the manufacture guys don't complain as much. Its still automated.

and I can never really remember if its 3 or 6 db per paralell either, I'd have to think it thru and I'm much too lazy and its time for a pint at the pub with my wife and some live Irish music.

Alan

 

[alk509]

[quote author="waltzingbear"]You forgot the useful technique of paralelling things, lower the noise by 3db with two resistors, 6db 4 resistors....[/quote]

Yeah, Brad, just parallel 250 SM resistors and the noise will be comparable to that of a single axial device. Simple!

:roll:

Peace,
Al.

 

[mikeg]

The process of making resistors has a huge impact on resistor noise. SMT parts can be sputter'd, silked inks or made other ways. These methods are all noisy compared to a radial part.

And don't forget En=(4*K*T*R)1/2.
THis means that if you're using SMT parts and thinking that they can dissipate as much heat as a radial part. No way. Temp goes up, noise goes up. One of the big reasons why large resistors sound better in guitar amps. They can get hot.

 

[bcarso]

When I said I ran out of time, this meant that the customer said they had to go forward without further board changes---hence the parallel/series approach was out of the question. The only option was finding quieter parts that fit in the same footprint, and revising the biasing of the stages to reduce currents driving the low freq noise.

The subject of the temperature effect per se on low frequency excess noise is a tricky one.

Certainly, for good old Johnsonian thermal noise, the square root dependence on absolute temperature is significant if you are getting way hot. But the difference in pure thermal noise between 200 C ( a pretty hot resistor) and 25 C is only that of the ratio of square roots of 473 K and 298 K, or about 1.26 (2 dB).

The effect on excess noise can be quite a bit worse, and probably related to intricate details of the construction.

EDIT: PS, welcome waltzingbear and mikeg to the forum.

 

[Gus]

Any updates on SMD resistors with low noise

This is a very interesting thread and I forgot it started in 2005

 

[JohnRoberts]

Any updates on SMD resistors with low noise

This is a very interesting thread and I forgot it started in 2005

Dead thread walking...  Brad is missed (RIP).

I reported issues elsewhere about small SMD resistors having issues. In the course of my outlet tester design, I experienced small SMD resistors not delivering their rated resistance. Alternately I am aware of linearity issues (voltage coefficient) from too small SMD packages. I never put this on the bench myself but a well respected console designer friend (also RIP) did and confirmed it is worth understandings.

I didn't review the entire (very old) thread but Brad noted this was excess noise, different from other resistor noise sources.

JR

 

[volker]

Here is a nice overview:  https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf

 

[john12ax7]

Any updates on SMD resistors with low noise

Avoid thick film for audio applications.  Use thin film and the largest physical size you can fit, 0805 is good,  1206 is better.