T.C. measurements on precision resistors

[MisterDiodes]

There is a lot of literature on this dating back to the 60's and even before.  This is a well known mechanism.

In a nutshell, a PWW, or any resistor of that style has the semi-conductive element "free" - as in it's not bonded directly to any substrate, and is low-stress shape: round.  So microscopically the internal wire stresses are free to reach equilibrium in a PWW resistor - at least in theory.

On a metal film resistor, the conductive element itself can be similar alloy to wire but A) It is a milled and flattened foil, which carries a higher internal stress than a round wire (even after annealing processes) and B) It is generally bonded to a substrate, which increases internal stresses again.  Whenever to different materials are bonded, that always creates a stress-riser zone, and is one of the mechanisms that disrupts smooth electron flow in the metal grain - no matter what the temperature.

That's why a bulk metal film resistor, all else being equal, will always be noisier than a PWW at the same resistance value.  "How much noisier" is generally something you measure in your own application circuit.

In general thin films are next higher noise, then thick film / metal oxide, then IC-style diffused resistors can have a pretty high excess noise and finally carbon comp resistors.  That list can be re-ordered depending on specific types, but that is what we see in ballpark measurements.

Be careful of the literature authored by Vishay / Zandman, a lot of that was a sales brochure.  As with all "datasheets" you have to separate fact and fiction for yourself sometimes.

Every precision application requires some thought as to the correct resistor to use, cost to benefit, etc - there is no right or wrong resistor for every application.

Edwin might make corrections here but this should get you a general idea of what's going on.  There is much more going on even with the resistor end lead terminations, welds etc. that can upset everything also.

[SilverSolder]

One reason to validate plastic is that they are used much more often than metal/ceramic. [...]
Lars

Have you attempted to use any kind of conformal coating to reduce the humidity sensitivity in your experiments?

Or does no coating exist that acts as a 100% humidity barrier?

[branadic]

There is no coating that leads to 100% humidity immunity. All you do is to increase time constant in an unpredictable way. Instead of a coating you can use wax also, but what is true for coating is also true for wax, even if you make a big wall of wax.
The only way to get rid of this is to use materials that are immune to humidity such as metal, glas and ceramics which ends up in hermetic packages.

But it's not humidity only, there are a lot of materials that exibit gas in some way.

[Edwin G. Pettis]

To Branadic,

Sorry for the delay, too many things going on that demanded my attention.  Most of the literature on the subject of 1/f noise in resistors was done at various Universities and also some work done for the military.  There are also several IEEE papers which also dealt with the subject.  In the case of the University papers, few of them are available on line, I obtained a couple of them through my local library’s inter-college lending program, and the IEEE papers are usually only available to members or require a significant fee to view them.  A careful search of the internet will find numerous references to the subject matter, many just mention the fact that PWW resistors don’t have 1/f noise, those are of little use per se.  You must be careful of your search terms or you will get a lot of useless junk mixed in. 
I did have an article which did discuss noise in resistors with more detailed information about why different resistors have different noise levels.  I have searched my computer’s hard drives and could not find it; apparently it is on a drive not currently in the computer.  I did come across this book which has some limited information on the subject:

https://books.google.com/books?id=avEjv8zAhQkC&pg=PA60&lpg=PA60&dq=1/f+noise+in+wire+wound+resistors&source=bl&ots=QlGHKIUXEz&sig=MK-nWIkDcxFmEJa4oC38FG91E5c&hl=en&sa=X&ved=0ahUKEwiNh5myoZvTAhUY8YMKHWtxCEs4ChDoAQgvMAQ#v=onepage&q=1%2Ff%20noise%20in%20wire%20wound%20resistors&f=false

This book from Analog Devices also mentions PWW resistor but doesn’t give too much information on the reasons why.  Small Signal Audio Design by Douglas Self.  Goggle likely has other books available in this form, try a search for resistor noise or 1/f resistor noise there.

A metallurgist would be a good source for the mechanisms that cause noise in the various alloys used in resistors.  Briefly, while the Film/foil and wire wound resistors do share the same or very similar alloys, the film/foil alloys are worked in a very different manner than the wire.  The alloy is very homogenous, and while there are very rare molecular ‘clumps’ that occur (these can produce a 1/f instance of noise on a somewhat irregular basis), they are few and far between in a well produced batch.  The working of the raw alloy into wire does not produce any of the stresses which remain in film/foil forms from the cold rolling process.  The wire naturally moves to a stress free condition after the drawing process (ask a metallurgist how that works).

In addition, in film/foil resistors, the action of chemical etching or laser cutting produces very ragged edges in the resistive material, like the rocks along a river, this disrupts the flow of electrons not only producing slightly more Johnson noise but adding a significant amount of 1/f noise to the mix.  Unfortunately there are no alternative processes available to make or ‘adjust’ the film/foil elements, there is no method which will moderate the added noise from the source, it is inherent.

The above does not take into consideration any of the additional processes which happen to resistors during manufacture, those are wholly separate issues and differ somewhat between the resistor types.  In some cases the processes of making the resistor can actually add more noise to the ‘base’ noise level.  While thermal noise can be filtered to some degree, 1/f noise cannot, at best all you can do is push the frequencies lower which still end up as low frequency wobbles.  The best way to avoid the problem of 1/f noise in resistors at least is to not use film/foil.

[babysitter]

NSFW - graphical violence against resistors - trigger warning - NSFR - FSK 18 Ohm - parental advisory recommended - explicit content - FIAL

My coworker managed to kill a Ultrohm plus Resistor during desoldering.  Edwin, obviously you really go to lengths to prevent stresses. (pun intended) - he confessed the little yellow guy had no chance. Fortunately I already ordered a replacement with my inital order.

Anyway, here is the image with details. It died for science. (I must confess I didn't measure if it is still in spec)

[Edwin G. Pettis]

Wow, send that guy back to remedial soldering class <grinning>, if the wire weld isn't broken, it should still measure good.

[Andreas]

more teardown please. (and more sharp pictures).

what is that white stuff which seems to be within the resistor?

with best regards

Andreas

[babysitter]

Wow, send that guy back to remedial soldering class <grinning>, if the wire weld isn't broken, it should still measure good.

I have to defend him, hes a theoretical physicist but a tinkerer, never had a soldering class ;-) I am on holidays for a few days, I don't know if it will survive or if it gets disposed. If its still here next wednesday, I will take more photos.

I was surprised about the "looseness" of the wire.

The white goo looks to me like a non-curing silicone grease  like Visilox.

[MisterDiodes]

One reason to validate plastic is that they are used much more often than metal/ceramic. [...]
Lars

Have you attempted to use any kind of conformal coating to reduce the humidity sensitivity in your experiments?

Or does no coating exist that acts as a 100% humidity barrier?

The better advanced vacuum deposition coatings work great on plastic parts - with virtually zero humidity influence.  The advantage over using just hermetic parts on a standard board is that the entire board after deposition: the WHOLE  is coated once and for all.  Components, solder joints, traces, everything.  To the point of it's hard to repair it.

We did a test a few years ago on a parylene coated board - and after a year sitting in a gallon of water that ~80g board gained less than 1mg of water weight after the exterior dried out.  So the absorption rate is -extremely- low, and the board ran great under water, with really no measurable side-effects.  The customer was thrilled.

This is not a hobbyist project... But the results are virtually as good as a hermetic box / hermetic packages IF you get the right coating process and thickness applied for your application.  The process is replacing the need for hermetics or hermetic boxes in some applications.

[lars]

Hello Edwin. I liked this comment from you:

"Actually the noise difference between wire wounds and Vishay can be measured in an LTZ circuit, might not be 'easy' but it is there and every bit of noise adds to the total and since 1/f isn't Gaussian, it can't be averaged out either." (Bold by me)

In one of the many papers I have read about 1/f noise I remember that the Wire Wound used was a Vishay that was compared to thin film, thick film and carbon (that might have been Vishay)!

Lars

[Edwin G. Pettis]

Any PWW resistors referred to as Vishay's was likely Ultronix (although a few other PWW houses were bought up by Vishay, they were mostly discontinued afterwards).  Ultronix was bought by Sfernice around 1980, Sfernice was then bought by Vishay later in the late mid-eighties (very big rivals).  Ultronix did not fair well under the auspices of Vishay and in about 2007, Ohmite bought Ultronix's remains as a package deal to get the rheostats they now make.  Ultronix is still more or less alive in Mexico, still making essentially the same resistors as they did decades ago.

Not just my opinion but I always take anything Vishay publishes with a large grain of salt, their test videos are laughable and results questionable.  Bob Pease and I had many a good laugh over Vishay videos.  That is not to say they don't make good resistors as such, they just have a bad habit of skewing the data in their favor.

The fact remains that 1/f noise is bad, you can't filter it out so the best thing to do in precision circuits is to minimize as much as possible.  With the newest crop of ADCs and DACs, that means using as many PWW resistors in the critical circuits as possible, other types aren't going to cut it.

[MisterDiodes]

Another thought is that Vishay does go around buying as many resistor houses as possible...For instance - it's not immediately apparent when you buy PWW resistors from Riedon that those are owned by  Vishay - and you have to look at the datasheet carefully and do your own testing.  Caddock, General Resistance and Pettis, etc. are still their own companies, last I checked...

After long experience, I would have to agree:  Vishay doesn't make "bad" resistors, but their datasheets can be a bit "optimistic" (to put it diplomatically).  Every application will call for the most correct resistor type and tolerance for that application - The decision of what resistor to use will include cost, availability, precision, tolerance, excess noise, hermetic vs. non-hermetic requirements, mechanical stability, long-term drift, DC vs AC, capacitance effects, inductive effects, size, weight, etc., etc. etc.

[MisterDiodes]

Correction on previous post: I might be wrong about Riedon about it being completely acquired by Vishay now - but the my point was that a LOT of resistor brands - BCComponents, Sprague, Dale, Huntington etc. etc. are all Vishay-owned brands that used to be separate outfits.

[branadic]

So you guys want to say, that it's no stress for a wire being stretched and compressed on a (ceramic or plastic) bobbin? And even if the wire is stressed that it can be all calmed by just a temperature treatment?
Sounds for me like the same fairy tale like micro cracks in a metal foil that has been induced by etching and the problem of sharp corners in etched metal foils. If this would have been a problem since then, why are the corners not simply chamfered?

[Edwin G. Pettis]

"So you guys want to say, that it's no stress for a wire being stretched and compressed on a (ceramic or plastic) bobbin?"

Just where the devil did we say anything like that? 

Fact: In the case of a ceramic bobbin, the ceramic expands slower than the wire with heat so that winding tension is gradually reduced with higher temperatures as the wire expands faster, the opposite happens with cold temperatures, the wire shrinks faster than the ceramic so there is actually a bit more tension with colder temperatures.  If wound correctly, there will be minimal tension, if wound with too much tension, of course that worsens the case.  Ceramics were the first bobbin of choice for many years before plastics became available and the great push for miniaturization. 

Fact: In the case of plastic bobbins, given the right material, at higher temperatures the wire will deform the plastic thus relieving the tension (if done correctly), note that there is always a small bit of residual tension that slowly decreases with age, particularly with temperature cycling (not necessarily high temperature).

Fact: Sputtering, chemical etching and lasers do in fact leave very rough microscopic edges, look at the edges of your PCB traces under magnification, they're rough too!  We're are talking about electrons being trapped, just how big do you think these rough spots have to be to trap electrons?  Get real.

"Why are the corners not simply chamfered?"  Space, a rounded corner takes more space than a square corner and when you are cramming a bit of resistance into a small place, you need every bit of available space.  Film/foil are already quite thin to begin with, this strictly restricts the size of the trace and the length of the path to get a given resistance, if you decide you're going to remove another significant chunk of available area to accommodate round corners, this is going to make the manufacture even more difficult.

This is a similar problem wire wounds face, the smaller the bobbin and the higher the resistance, smaller and smaller wire sizes must be used, this makes it more difficult to make and also limits how high a resistance you can get with a given wire size and bobbin, it is no different with film/foil.

[MisterDiodes]

There is another effect in PWW: The wire is drawn to size and annealed as a -final- step in the mill, and if you look at it even under high-mag SEM photo, you'll see that the wire surface is just about theoretically, perfectly smooth - there is no edge discontinuity like you will always have on a flat, rolled out foil with rough edges.  Again: foil resistor can't really be final-annealed the same way once the foil is bonded to a substrate, and the shape - by definition - will always have relatively very rough edges compared to a smoothly drawn wire.

Noise is always generated at surface discontinuity, or anywhere there is internal stress and strains, and any place in a conductor where there is a -change- in conductivity.  A round wire doesn't really have edge discontinuities or sharp corners that give that conductivity change - and it's a shape that can be made from very high-purity and highly consistent alloys.  Just don't kink the wire or leads!

Once you roll out foil and attach to ceramic or glass, those electron pathways are constantly changing - that isn't a design that lends itself to zero stress or virtual zero-stress -changes-.  Some of the film resistors do have noise that's pretty low, but never really achieved non-existent 1/f noise as in PWW.

You can even try binding foil to a ceramic, and then polishing the foil surface to less than 1u rms roughness, and it'll still have more 1/f noise than PWW.  Beckman tried this in the late 60's and 70's, and basically couldn't compete with PWW's at DC.  They could get the noise lowered over the standard foil rolling, but the danged things cost about 20x more than PWW's, and was a failure.

PWW have always been, and are still really the only practical resistor device design that has -zero- 1/f noise down to true DC.

Remember - this has been a problem with foil resistors that's been known since high-performance instrument amplifiers were needed - even TI was trying to make film resistors in the early 60's and could never reach true zero 1/f noise like PWW give you.  Not a lot has changed today; since the same physics still apply.

NOW: On a new PWW bobbin freshly wound, yes there is a period where you'll see internal winding stresses relaxing and getting into final position - but that is generally a change in resistance or apparent TCR - that isn't true 1/f noise.  Like all precision components there is an initial break-in time where the devices have to self-relieve assembly stresses and find their final stable running operating point and equilibrium.  PWW as well as Bulk Film resistor can benefit from an initial thermal cycling to help gain stability - the change in temperature and generally higher temperature help equalize the strain throughout the material.   No different than thermal-cycling precision mechanical components, castings, bearings etc.

You'll also see this in a new 732b:  When those are new they have to be re calibrated much more often until the Zener and it's film resistors settle in.  Compare that to something like and older 732a with PWW's inside - generally less noise & more stable to boot now that these are decades old (IF they were well taken care of).  Fluke had a real hassle getting those 732b's to behave over time...that laser-trimmed resistor network isn't all rainbows and butterflies (over time) compared to the PWW's.

Both of those models really show how hard it is to make a stable 10V from ~7V though.  There's not an easy answer to that one, no matter what kind of resistor you put in an Op-Amp feedback path.

By the time you build a DIY 10V Vref device that's truly as stable as a 732, you might as well have bought a 732... 

[branadic]

Are there public sources of what foil resistor alloys are made of? There is alloy C, K, Z and Z1 foil. My guess is, that they must use the alloys available and other resistors use as well, even though the process of how the resistive element is fabricated is different.

For what I understood the foils are cemented to alumina substrate and afterwards structured by lithography and etching, which is said to produce micro cracks at the surface and thus noise.
But what if the alloy would have been sputtered (PVD) to the surface, that is already masked with a structured coating (e.g. LDI = laser direct imaging). The coating is then removed by a lift off step and the resistive structure remains on top. The structure could be similar to how foil resistors look like and the resistive element can be trimmed to the final value using laser technology.
As far as I found this is different to how thin film resistor are fabricated, as thin film resistors are sputtered to the substrate but structured by selectively etching.

[Edwin G. Pettis]

There is only one alloy, Evanohm that exists for very low TCR/high stability resistors, there are minor variations of this alloy but this is little more than tweaking the percentages of the metals to try and get slightly lower TCRs.  There are also slight variations of the orginal Manganin alloy but again, merely tweaks to the basic composition.

The major differences is how the alloy is processed after the alloy is manufactured.  In the case of wire, it is the heat treatment, under various conditions that produces the end result TCR.  The end result becomes the inherent TCR of the wire, resistor manufacture can modify the apparent TCR by imparting various stresses on the wire plus the end terminations can also affect the apparent TCR (and noise).  Note that none of this affects the wire TCR, this requires temperatures well above the normal operating range of precisions and even power resistors which can reach a hot spot temperature of 350°C.

Film/foil resistive elements are cold worked which do impart changes into the alloy, being very thin, the resistive element becomes far more sensitive to stress, in fact stress sensors were the first application of these resistors decades ago, they were also quite non-linear.  Sputtering does not solve the 'roughness' of the element, it only changes it slightly, all forms of film/foil are rough in one form or another, there are no current processes which get around this 'problem'.  Measurements have confirmed the 1/f noise is present in all forms of film/foil.

[SilverSolder]

Just out of interest, how do you actually measure 1/f noise?  - after all, it must be very hard to separate the noise of the measuring equipment from the noise of the resistor?

Is it a case of knowing exactly how much noise your equipment has, and then assuming that any additional noise must be due to the DUT?

[Kleinstein]

The usual way to measure the excess noise of resistors is by using a bridge circuit of 4 equal resistors, so that the AC amplifier does not see an DC voltage applied to the resistors. One than compares the noise measurement without a DC bias and with an DC bias. So the equipment noise is measured together with the normal Johnson noise.

For high quality resistors the difference is very small and thus hard to measure.

One still has the problem of separating thermal and current effects. Heating due to the current also increases the Johnson noise - this is not the excess noise. Turbulent air flow due to thermal gradients can also cause variations, that can look a little like 1/f noise.

[branadic]

The usual way to measure the excess noise of resistors is by using a bridge circuit of 4 equal resistors, so that the AC amplifier does not see an DC voltage applied to the resistors.

This is only one way, called DC-measurement technique. There are more ways on how to measure the excess noise:

- double frequency ac methode
- 0°/90° subtraction methode
- 45° cross correlation technique

Source: https://dcc.ligo.org/public/0002/T0900200/001/current_noise.pdf

[doktor pyta]

Dear Colleagues,
Here are some photos of my DIY temperature controlled chamber for measuring 57 resistors at once (typical session lasts 1.5 day).
Why 57? Three scanners allow 60 channel measurements, each DUT board has 20 resistor fixings, but there is one Pt100 sensor on the surface of each DUT board to monitor the temperature on each of 3 'floors'.

I think You may be interested in way of mounting leads of the resistors. Another thing is using nine 40Pin 0.5 Pitch FFC Flexible Flat Cables to take all the 4 wire resistor connections to the scanner cards. As Datron 1271 allows guarded resistance measurement, there are guard traces almost everywhere on the DUT PCB's + there is guard wire every two wires inside FFC cable. The FFC cable is quite good here because each wire has very small cross section and even though I use total 360 wires which has total area of 3.8mm^2, the temperature inside a chamber is not highly affected.

(Info: The FFC cable has suprisingly high resistance between adjacent wires (ca. 1E+10 ohm) but if air humidity is high or dynamic effects occuring while TRUE OHMS measurement, it's just wise to have these ohm guard feature)


The setup uses also:

Datron 1271
Agilent 3499A + 3xN2260A high density scanner cards (20 x 4wire channels on each)
ILX lightwave LDT5948 TEC controller

The system uses Prologix GPIB to USB and PC for controlling and data logging.


Later I will share some measurement results and more pictures if You want.

[doktor pyta]

Pictures part 2

[doktor pyta]

Photos part 3

[doktor pyta]

Photos part 4 + sample measurements.

Now You see why I had problems with two of my LTZ 1000 boards which used 120ohm 8G16D resistors.

P.S.1.  I don't use temperature ramping. I set cretain temperature and I wait for an hour for temperatures to equalize. Then I start measurements.

P.S.2.  The project wouldn't be finished without software work done by Mr. Zbigniew K. Thank You!