T.C. measurements on precision resistors

[TiN]

Let's continue, shall we? This time 120 ohm resistor test on meter 6 (Keithley 2002, calibrated to <15 day SR104/732B via 5700A in August 2016).
Meter 4 (K2002 too, same calibration) measured Fluke 759753 9K+90K+900K+9M?, which equals to total resistance 9.999M?.

Two tests with 2.5K/hour ramp speed performed, both up and down direction. Same conditions as before on previous page.



Fluke hermetic behaved predictable, with near zero TCR up to 30-32°C, with max TCR +0.6ppm/K in first run, and +0.5 ppm/K in second run.

Edwin's 120R behaved much better than previous data on 1K/12.5K, much more linear vs temperature curve, with max TCR +4.3(+4.6 ramp down) ppm/K in first run, and +4.2 (+4.55 ramp down) ppm/K in second run.
Resistance change still present here, but not as big. About -7..-8 ppm after the first run.

DSV datafile.

[MisterDiodes]

TiN:
What were the specs on these resistors when you ordered them - what is the spec'd TCR?  Did you get these -directly- from Edwin or third party?  I know there was a short time where there was an epoxy supply problem at his facility, but I think he's changed the process with a new epoxy supplier.  That might be interesting to find out if these units came from that troublesome batch.

We've got Edwin's 120's (3ppm TCR rated) in stock that show MUCH lower apparent TCR, usually around 1~-2.7 ppm TCR Same for the other 13k/1k and 70k.  All within spec, and usually around the same or a little less than GR's of same value.

It's odd that you're showing such a high apparent TCR on these guys.  We don't usually see it that high, at least on the units we have that are spec'd for 3ppm TCR max.  They are usually better than that.

One thing I do see:  The fact that the PWW's don't return to the same value on ramp-down shows they maybe aren't thermal cycled quite enough - this sometimes happens during shipping especially if they get very cold for some time (cargo airplane trip?).  Usually after some stress-relieving thermal cycles that effect will diminish quite a bit and they settle down and get more repeatable.   That might be something for you to look at - say 5 or 10 trips to 60~80 C with a good cool down in between.  Edwin might have a better recommendation.

[TiN]

I've got these from a volt-nut. They were manufactured last year, so fairly recent. Surely resistors were subjected to shipping stress, but same as any other stuff I have/test. I did not see such changes in any of plastic VPG S102's/SFERNICES so far. This 120 ohm was sitting two weeks on top of hot K2400, so maybe that's why it got relaxed well enough not to show any curvy TCR.

I'll need to check how high I can go safely on my TEC box, as YSI 44031 thermistor I use with K2510 is rated/specified only up to +70°C. Might swap to Honeywell HEL-705 Pt1000 RTD to extend range bit higher, for ramps with 20C - 60C span..

Also to check on meter/setup I've turned AC to drop room ambient to +23C from +28C, and resistance curve from 120 ohm didn't even noticed. Live data in progress. Humidity due to AC dropped to ~45 RH% instead of 50ish too.

[Kleinstein]

If there is trouble with epoxy curing / aging, the critical temperatures are more around 100-120 C. This is the typical glass temperature. So to get a stable, well relaxed state one would heat it up to about 100-120 C and than slowly cool to something like 60 C.  At least this is what is recommended for high strength epoxy glue.

Tempering might not need such tight temperature control - I once did such a thing on the hot plate of a cheap coffee machine.

[Edwin G. Pettis]

First, measurement equipment is not necessarily the problem, method could be to some degree, that is not how we test for TCR or any other PWW manufacturer does it.  Secondly, you are trying to compare apples and oranges, film/foil and PWW resistors are not the same by any definition except that they are both resistors.  You cannot compare a Vishay against a PWW for thermal behavior.  I am not having any problems with the epoxy I have been using for months now (and it is not epoxy glue, not the same thing).  I have not received any reports from any of my customers around the world as to any problems with my resistors.  I have not seen or heard of any problems with resistors outside of their TCR specs, they are all close to the TCR of the wire used.

I do not know why you are seeing higher than normal TCRs but I see in your charts that neither mine nor the Fluke resistors are achieving a stable TCR plateau, while it more or less follows the ramp, you cannot tell if the internal temperature of the resistors have actually matched the chamber temperature.  While you are seeing a indication of TCR linearity during the ramp, there is nothing to indicate thermal equilibrium.  Once a resistor achieves thermal equilibrium, the measurement should flatten until the external temperature changes again, then after a time lag, the resistance should start to follow the temperature.  In your charts I do not see any indications of thermal equilibrium in any of the resistor plots, I believe MisterDiodes also pointed this out.

The question of stress, built up by whatever factors, is complicated.  Since the wire wound resistor inherently has some stress built into it simply by the winding process, ignoring any other source of stress, this 'barrel' stress takes some time to relieve.  Since the 'stress' on the wire during winding is a variable that cannot be completely controlled, it cannot be stated with certainty how long it can take to relieve the bulk of it.  Military grade resistors and standards are put through a fairly long process to relieve those stresses to a high degree.

Throw in whatever other external sources of stress may be present and things get more complicated, in many cases, some of the outside stresses are rather small and temporary, others can present more of a problem, but in all, these stresses generally add up to very small perturbations in the resistance.

In case you were wondering, those specs about 'long' term drift, so many PPM per year, that isn't a linear spec, it doesn't mean that a resistor will linearly drift from the initial value to 35 or 50 PPM or whatever at the end of the year.  It means that the resistor can show changes within that limit at any time during the year, it might jump 35 PPM within days or even at a later time during the year, it can go up or down or both.  This applies to all resistors, a spec for change only defines the end limits, it doesn't define how or when the change can occur.  In most cases, but not all by any means, the lion's share of change will usually occur in the youth of the resistor, just what that will be depends on the individual resistor.

The same spools of wire, the same bobbins and shells are still being used, the only changes that have changed over time is the sealant and that has been improved as time has passed.  The faulty epoxy sealant happened over two years ago and only affected two customers, it is time to stop fussing over that.  The current sealant has been working quite well and there have been no complaints of resistors being out of spec for a long time.  The epoxies used in end sealing are room temperature curing, they do not require elevated temperatures.

If you believe you have a resistor that is out of spec, I suggest you send it to me for evaluation and if found to be out of spec, it will be replaced. 

[TiN]

Thank you, Edwin. Let me revise the method , and retest with step response. I'd like to get to bottom of this, as it just shows required efforts to correctly measure and interpret the numbers, and it's interesting topic. I've used current gradual ramp method for over a hundred different resistors, PWWs, foils, metal film's, so it'd be interesting to see if step method would match. Now that I have K6221, once it's output tested to confirm stability, I can use it for new method, and rely on base 10(20) VDC range performance of meters, instead of current source/ocomp function.

I might also finally build an small oil bath to allow measurement of the resistor temperature more or less directly. Current temperature control is using sensor sitting in the box, but not physically coupled to resistor body. If sensor or temperature controller behave non-linear way, we wouldn't see that. I've tested this once with Pt1000 sensor in box few months ago, and other then offset, temperature curves followed, but that might also be point of error.

Let me do all this (which likely to take few months), and if data still incorrect afterwards , then I can send resistors for your evaluation.

[MisterDiodes]

TiN:  It sounds like you're putting some good thought into this.

Another interesting part of the story:  We expect the PWW resistor to have a -higher- TCR and a fairly long thermal response time lag compared to film, but a couple of other effects we look for also besides the usual Temp. / Humidity / Pressure effects:

1.  Noise.  The film resistor datasheets are a bit sketchy on this one; it's better to measure on your own.  The PWW will have the lowest noise down to DC - usually right at theoretical thermal minimum for resistance value.  The metal film and chip-scale diffused resistors are doomed to have more noise - even the foil-to-lead attach points can be a problem..  It is interesting to look at how much more noise comes from certain film designs, and to what substrate the foil has been bonded.  If the foil has been laser-trimmed we've seen various effects from the foil edge thermal damage and slag debris that's left over.   And so on.

2.  Mechanical effects: Another effect those flat "blade" resistors have is relatively very fast thermal response to ambient - they are all "fin".  Which sometimes you want. Sometimes that will bite you though as any air draft around the package shows up as a small amount of noise.  This is also related to mechanical flexure:  Depending on how the foil is bonded to the substrate and how they are mounted, these can act as a mechanical microphone and pick up small vibrations from nearby motors, fans, voices, etc.

One more than one occasion we have solved a stubborn noise issue by simply switching off a small vane vacuum pump or broken-bearing cooling fan  running at a workstation several few feet away from the test rig.  You couldn't even feel the vibration to touch.  Blade-style flat film resistor paks, ceramic-mounted hybrid networks and ceramic caps (esp. surface mount!) are usually the pickup point.

[Edwin G. Pettis]

While we are on the subject of things that can affect resistance, I don't want to sound like there isn't any possibility of one of my resistors going a bit off track, particularly with lower ohms.  Usually a low ohms per foot wire size is selected, this gives the 'biggest' target area for welding (and even then it is still quite small), normally for the 1K resistors I would have chosen a larger wire size but since we're trying to minimize the drift factor in a ratio, the lower value resistor will end up with a finer wire size to closer match the higher value resistor in the ratio (more on this in a minute).  This makes the welding target even smaller and more difficult to hit on the head (I generally do it all the time) but because the target is so small, if the wire even touches the weld joint between the weld peg and the copper lead in the slightest, it can raise the apparent TCR of the resistor.  The weld interface between the weld peg and the copper lead is seldom smooth so thee is plenty of chances of a wayward sliver from that joint extending into the weld peg area, this weld peg area is only about 0.030" wide.  It is fairly rare but sometimes microscopic slivers can touch the wire weld joint and cause a deviation in the normal TCR.  Sometimes this 'touching' is visible to the eye, other times it takes magnification to see it.  This is one possible suspect when the TCR appears to be off.

Sometimes I fail to mention this to customers but the standard lead spacing for measurements is 0.375" from the body of the resistor (never minding any slight bulge from the end epoxy), in the case of the 802 bobbin, this is 1.250" point-to-point.  More than that can create additional reading errors.

Ratio matching can get very sticky when the resistors have values that are relatively far apart.  In fact, with the assumption that the same current is flowing through both resistors and both resistors are at the same ambient temperature, having the exact same TCR in both resistors will result in TCR mis-tracking for the simple reason that the resistors are dissipating different power levels (same thing happens with film/foil).  This is where things get tricky, the general solution is to try and equalize the two resistor internal temperatures by external means, tight thermal coupling and this takes several forms.  In the case of film/foils on the same substrate, the temperatures are spread out to some degree because of the substrate thermal coupling but this isn't perfect either, however because of the small sizes involved, it tends to be easier, there are still drawbacks.

With wire wounds, it is theoretically possible to find two TCRS that will track very well given the two different power levels, that is more difficult than it sounds since you cannot order wire to a specific TCR as such, you can ask for a certain TCR but there is always a range limit, I can ask for -1.5 PPM/°C but I may end up with only something close.  To a similar degree, the film/foil resistors are within a certain TCR tolerance and it is what you get.  I am limited to the TCRs and wire sizes on hand, I can make a ratio set to track within a reasonably close value but not zero PPM by any means.  The Linear Tech designers obviously took this information about tracking and designed the LTZ1000/A to nearly eliminate the effect of ratio mis-tracking.

The LTZ circuits (or most any other high performance reference circuit) are going to change over time, that much is clear from years of past performance data and while a tiny tweak here and there over the years have made slight improvements, it is still a fact that anything you tweak for today is going to change later with time.  The main question is whether it is worth the time and effect to chase such small effects and are your instruments capable of even measuring those tiny drifts accurately over time.  I think at this time, the answer is no unless you might be a NIST lab.

Anytime you start digging into the PPM garden, you're going to find a lot bugs that are happy to cause all kinds of trouble, some you might not even notice.  Be careful not to fall into the digital trap and assume every reading is good let alone accurate there are too many artifacts that must be accounted for.
 

[lars]

I have hesitated to write this but I think the yellow resistors TiN has tested are very HUMIDITY sensitive. What I see in the graphs is a resistor that bakes out with higher temperatures. One reason that TiN gets more shift than Andreas might be the longer measurement time, another might be that it is a spread in humidity sensitivity due to spread in the manufacturing process. Another reason is how much moisture the resistors have acquired before test. With 57%RH and maybe a long time exposure it might be more.

In December I finally got replacement resistors from Edwin as I was one of the two customers that got the “spongy” epoxy. As the new resistors deviated from the measured value from Edwin, Ok I only claim an uncertainty around 10ppm but this was worse, I put some in a bag with 150g silica gel with about 10%RH and some in another with about 90%RH.  An interesting side note was that one of the marked values from Edwin was 96.485ppm (quite of lot of precision) I just wonder what the uncertainty was as it wasn’t specified  .
?
All new resistors in the 90%RH bag went up and in the 10%RH bag all went down. For example, a 10kohm (size 0807) went up about 75 ppm in the 90%RH bag. I had about 30-40%RH in my lab before I put the resistors in the bag. The 10kohm in the 10%RH bag went down about 45 ppm. So the difference was 120ppm for an 80%RH difference. So I can guess the humidity sensitivity was about 1.5 ppm/%RH. The time constant for the humidity sensitivity was slightly shorter on the new resistor than the “spongy” ones. About 2 weeks instead of about 3-4 weeks. Remember that the time constant probably is very dependent on the air circulation that in the bags (or a tight box) is low if the temperature gradients are small.

For all those that now fear that the resistors that they got are useless for their LTZ1000 project I can say it will be no problem! As probably all your resistors are affected in a similar way and the most critical resistors 1k and around 13k has a sensitivity that cancels each other if they drift in the same way. Even if you have a large humidity variation over the year you will probably have less than a ppm of drift on the LTZ1000 output due to humidity even if your resistors happen to be sensitive to humidity as mine.

It is still so that ordering resistors from Edwin is easy and cheap for WW.

What concerns me is that I wanted to use some of the ordered resistors in a simple boxed resistor reference with four binding posts. I have done several of this type with VHP-type resistors and they have worked fine for me. My guess is that you might get down to 10ppm uncertainty at 23C for a year for these VHP and I had hoped that it could have been possible to get 20ppm for Edwin’s resistors for a much lower price as they were claimed to be humidity insensitive and have very low drift over time. Note: I have no intention to sell any of these it is just out of curiosity.

Just a note to Edwin: if you wonder why I haven´t told you this it is because I found it hard to get the replies from you saying I am the one doing wrong so I decided to forget these resistors, but with the comments on TiN’s work I had to write something here.
 
Edwin: I also would like to know how you do your humidity tests and what the result was?

Lars

[Kleinstein]

The time it takes to reach equilibrium for humidity depends on the temperature and likely only a little on air movement. The slow part is usually the diffusion in the plastics.

In the LTZ1000 circuit there is a good chance that the circuit part with the resistors get slightly warmer than room temperature. A 10 C temperature rise corresponds to about a reduction of RH by a factor of 2. This also reduces the possible range of humidity variations.

[razberik]

Hello, I have a question, I believe it is appropriate for this topic. Excuse me if it is duplicate, however the topic has already 30+ pages.

We use in our products resistors from Nicrom, parts are from HVC BT series and 200.3 BA series. Product catalog should explain it.
The reason is 0.1% accuracy and 10/15ppm/°C.
The problem is that Nicrom is not responding for weeks/months and it is not possible to source parts from them.

Can anybody advise me alternative source of high-resistance resistors with such low tempco ? We can omit accuracy by parts selection or calibration.
My colleague found out Ohmcraft manufacturer and their SM HVC series looks good, but not exactly as good as Nicrom.

Any advice on high resistance resistor manufacturers please ?

[Alex Nikitin]

Hello, I have a question, I believe it is appropriate for this topic. Excuse me if it is duplicate, however the topic has already 30+ pages.

We use in our products resistors from Nicrom, parts are from HVC BT series and 200.3 BA series. Product catalog should explain it.
The reason is 0.1% accuracy and 10/15ppm/°C.
The problem is that Nicrom is not responding for weeks/months and it is not possible to source parts from them.

Can anybody advise me alternative source of high-resistance resistors with such low tempco ? We can omit accuracy by parts selection or calibration.
My colleague found out Ohmcraft manufacturer and their SM HVC series looks good, but not exactly as good as Nicrom.

Any advice on high resistance resistor manufacturers please ?

What values/power/voltage rating are you looking for?

Cheers

Alex

[razberik]

There are two:
HVC-2010-20M-BT .... 20M, 2.2kV, 0.5W, 0.1%, 10ppm/°C, -5ppm/V
200.3-100M-BA-R .... 100M, 20kV, 3W, 0.1%, 15ppm/°C, 0.3ppm/V

[Alex Nikitin]

There are two:
HVC-2010-20M-BT .... 20M, 2.2kV, 0.5W, 0.1%, 10ppm/°C, -5ppm/V
200.3-100M-BA-R .... 100M, 20kV, 3W, 0.1%, 15ppm/°C, 0.3ppm/V

For the 100M / 20kV replacement have a look at Ohmite Super MOX series and Caddock USG.

If you need top stability at 10M 2.5kV then Caddock USF300/USF200 is probably the best I know, but large. If you definitely need SMD, check also Rhopoint offers.

Cheers

Alex

[razberik]

Thank you !
Ohmcraft manufacturer was chosen using Rhopoint distributor catalog. That is for SMD res.
100Meg resistor was replaced by Slim-mox from Ohmite.

It seems that Nicrom is already defunct for year, it is blacklisted:
http://www.supplierblacklist.com/2016/06/26/nicrom-electronic/

This leads to a theory. We used Nicrom's products for such a low tempco. But what if these parameters were fake and it is the reason Nicrom stopped ?
Is it possible to make these resistors with these parameters ?

[Alex Nikitin]

Is it possible to make these resistors with these parameters ?

Quite a lot depends on the way the temperature coefficient is specified (or even understood). For example Ohmite specifies it for 100C change from 25C to 125C. So for 10ppm/C tempco all you can be sure about is that the change between two points (25C and 125C) is less than 1000ppm or 0.1%. There is no actual guarantee that somewhere at, say, 75C, the value won't be off by 0.2%  . Nicrom does not even specify the way they define the tempco in their catalog. Caddock specifies some of their best resistors at 5ppm/C or even 2ppm/C "absolute maximum" TC in -40C to +85C range - potentially the TC slope should not exceed that value anywhere in the range (though I might expect too much  ).

Cheers

Alex

[lars]

The time it takes to reach equilibrium for humidity depends on the temperature and likely only a little on air movement. The slow part is usually the diffusion in the plastics.

You are correct, my bad.

In the LTZ1000 circuit there is a good chance that the circuit part with the resistors get slightly warmer than room temperature. A 10 C temperature rise corresponds to about a reduction of RH by a factor of 2. This also reduces the possible range of humidity variations.

This is also my findings. But my LTZ1000 design in an alubox with a well-insulated LTZ have just a couple of degrees C rise on the resistors. If the temperature is higher one drawback is that you might get more drift and if the unit is off a long time it will absorb moisture that during the first day or weeks after start will desorb giving a long “warm-up” period. But as said before this will normally be a minor problem for the resistors for the LTZ.

For 10V refs in plastic packages I have somewhere else showed this with data from boxes with average internal temperatures of about 23C respectively 36C. For a seasonal variation of about 40%RH I had for REF102 about 20ppm at 23C and 10ppm at 36C and for LT1236 about 24ppm at 23C and 12ppm at 36C. For the AD587 I had only 4-6ppm at 23C for four samples and around 1-3ppm at 36C for five samples but one sample had 9ppm (at 36C). In another test with six AD587LN with different date codes I got very different humidity sensitivities. Think I have shown this before also but don´t remember so well. All these were DIP-8 in sockets (on FR4 boards).

I have also tested a few SMD versions at the same time. For the SMD they were mounted on SMD to DIP adapters from Aries (Farnell). For the REF102 the DIP8 and SMD results were very similar. For the LT1236 the SMD was much better, only about half of the drift. For the AD587 the SMD was about 6ppm at 36C for two samples so worse than most of the DIP8’s. That Maxim says that SMD is worse than DIP’s might be because of MAX674. I had one sample in SMD (no DIP) in the 36C box and the seasonal variation for that was more than 20ppm! After a power down (controlled) for three weeks it started 40ppm higher (measured after 1 and 3hours) but after a couple of weeks it was back on track again.

Lars

[ap]

For 10V refs in plastic packages I have somewhere else showed this with data from boxes with average internal temperatures of about 23C respectively 36C. For a seasonal variation of about 40%RH I had for REF102 about 20ppm at 23C and 10ppm at 36C and for LT1236 about 24ppm at 23C and 12ppm at 36C. For the AD587 I had only 4-6ppm at 23C for four samples and around 1-3ppm at 36C for five samples but one sample had 9ppm (at 36C). In another test with six AD587LN with different date codes I got very different humidity sensitivities. Think I have shown this before also but don´t remember so well. All these were DIP-8 in sockets (on FR4 boards).

I have also tested a few SMD versions at the same time. For the SMD they were mounted on SMD to DIP adapters from Aries (Farnell). For the REF102 the DIP8 and SMD results were very similar. For the LT1236 the SMD was much better, only about half of the drift. For the AD587 the SMD was about 6ppm at 36C for two samples so worse than most of the DIP8’s. That Maxim says that SMD is worse than DIP’s might be because of MAX674. I had one sample in SMD (no DIP) in the 36C box and the seasonal variation for that was more than 20ppm! After a power down (controlled) for three weeks it started 40ppm higher (measured after 1 and 3hours) but after a couple of weeks it was back on track again.

For any precision reference, you need hermetically sealed devices. Metal can or ceramic. Plastic is useless to even validate, they will drift considerably due to humidity/environmentals.
And next, if you have the choice, burried ceners.

[Alex Nikitin]

Here are the results of my recent measurement of LT450C 10K resistor in one of my mini-references, using my "Happy Reptile" box and Keysight 34465A meter, for the temperature range 18C-38C. Unfortunately, the room temperature did change by few degrees during the day and the apparent hysteresis (about 1ppm) after coming down from 38C back to 28C and 23C could be also a reflection of the meter TC. I am quite surprised by the stability and the low tempco of this resistor. The drift over last half a year was less than I could reliably measure (<5ppm, possibly <3ppm), and TC between 18C and 28C is less than 0.4ppm/C .

Cheers

Alex

[lars]

For any precision reference, you need hermetically sealed devices. Metal can or ceramic. Plastic is useless to even validate, they will drift considerably due to humidity/environmentals.
And next, if you have the choice, burried ceners.

One reason to validate plastic is that they are used much more often than metal/ceramic. Another is that it is difficult to get metal /ceramic that are RoHs. Of course, if someone else that you trust has validated the circuit you want to use it might be ok or if it is in the data sheet but except some notes in the datasheets from LT that even the hermetic LTCC packages are affected by humidity, if mounted on a board, it seems difficult to get how sensitive Vrefs are to humidity.

One reason I started to test some Vrefs was comments on the Geller Labs SVR boards (that was before I had one) and obsolesces due to RoHs. But most of my data comes from tests that are not specific humidity tests but only regression analysis of humidity data that I normally store as well as temperature and pressure. Long long ago I started to always note humidity as I found some resistors to be very humidity sensitive! Nowadays I have a weather station that I store the data from so I can go back if needed. I like the weather station but using a RPi+BME280 would of course work as well .

Except for data for humidity sensitivities it also seems difficult to get data for atmospheric pressure sensitivities for Vrefs (and resistors). With the about maximum 60-70mbar difference I have in my lab I cannot conclude anything about pressure sensitivities of the common AD587 and REF102 (that I have both in metal/ceramic and plastic and tested against LTZ1000 ref).  I guess the LTZ1000 is fairly insensitive against pressure as it has been used in Fluke and Datron refs? Anyone know data? Maybe the sensitivities are so low that a movement from a low altitude to a high doesn´t matter?

Lars

[lars]

Another interesting part of the story:  We expect the PWW resistor to have a -higher- TCR and a fairly long thermal response time lag compared to film, but a couple of other effects we look for also besides the usual Temp. / Humidity / Pressure effects:

1.  Noise.  The film resistor datasheets are a bit sketchy on this one; it's better to measure on your own.  The PWW will have the lowest noise down to DC - usually right at theoretical thermal minimum for resistance value.  The metal film and chip-scale diffused resistors are doomed to have more noise - even the foil-to-lead attach points can be a problem..  It is interesting to look at how much more noise comes from certain film designs, and to what substrate the foil has been bonded.  If the foil has been laser-trimmed we've seen various effects from the foil edge thermal damage and slag debris that's left over.   And so on.

I have to ask how much noisier at low frequencies (0.1-10Hz) a standard LTZ design will be if I use thin film resistors as SMD RN73, PCF0805 or ERA6 (Digikey, Mouser etc) or through hole like PTF56 or other similar to the ones HP uses in the 3458A?

Humidity and long term tests I have done on 1-100kohm SMD0603 and 0805 as above shows about 0.5-1ppm/%RH so less or equal to WW I have tested. For long term I have never seen more than 100ppm the first 6 months and after that 5-20ppm/year. This with components mounted on FR4 boards. So my conclusion is that they will give much less than a ppm drift over a year after the first 6months so for 1USD resistors it will be below my measurement capability. If they don' contribute significant noise in the LTZ design  is it any reason not to use them for the LTZ? Except that you don’t get the ultimate performance that you still cannot measure.

Lars

[Andreas]

I have to ask how much noisier at low frequencies (0.1-10Hz) a standard LTZ design will be if I use thin film resistors as SMD RN73, PCF0805 or ERA6 (Digikey, Mouser etc) or through hole like PTF56 or other similar to the ones HP uses in the 3458A?

Hello,

Good thin film resistors are specced with -40 dB
Against the magical limit of -42 dB of bare metal resistors.

So it will be hard to measure within the LTZ noise.
But which of the resistors will be more sensitive to magnetic pickup of mains line frequency?

With best regards

Andreas

[Edwin G. Pettis]

Re: Reply 757:

To Lars; Re: Humidity

My reply here is based on common practice and military standards pertaining to precision wire wound resistors.  The main reason for humidity testing is not for resistance shifts, it is to see if there is a problem with the physical resistor which is spelled out in the standards I will quote below.  Commercial grade resistors are not tested for humidity by anyone I know in the industry.  Humidity is tested under controlled conditions in a humidity chamber; anything outside of a controlled environment is unacceptable.  Since humidity is a constantly varying parameter in normal environments, it is impossible to accurately track actual humidity inside of a sealed resistor as it has a significant time lag absorbing or evaporating humidity and depends on several factors, is it powered or not, the dew point, temperature, and humidity and the absorption rate of the epoxy (or epoxies).

Extracted from applicable MIL-STD and MIL-PRF specifications:

6.13 Low tolerance resistors.  Low tolerance resistors, exhibiting resistance shifts due to high
humidity is normal to precision, fixed resistors.  Before being considered out of tolerance, resistors should be
conditioned in a dry oven.  Users of said resistors should contact suppliers for temperature and drying time.
Resistors which continue to be out of tolerance after the above conditioning process should be considered rejects.

Standard procedure for moisture resistance is given in 4.7.10, initial measurement conditions in 4.7.5 and MIL-STD 202 Method 106F.  While the procedure explanation is longer than I will post here, simplified initial is DC measurement at 25°C±2°C.  Resistors are conditioned at 40°C ±5°C for 24 hours, then put into the humidity chamber and given 6 cycles of temperature/humidity.

Final measurements: Upon completion of step 6 of the final cycle, the resistors shall be conditioned at a
temperature of 25?C ?2?C, and at a relative humidity of 90 percent to 95 percent for a period of 1 hour 30
minutes to 3 hours 30 minutes.  After conditioning, the resistors shall be removed from the chamber and
within 8 hours, the dc resistance and insulation resistance shall be measured as specified in 4.7.10b.
Wiping and forced air drying prior to measurement is not allowed. The subsequent 4-hour to 24-hour
conditioning period and measurements do not apply.

Examinations after test: Resistors shall be examined for evidence of breaking, cracking, spalling, and
loosening of terminals and mounting hardware.
 
The object of humidity testing is primarily physical damage as noted in the above sentence, only in the case of a resistor which does not return to within specified tolerance as noted in paragraph 6.13 above is considered a reject.

To answer Lar’s question directly, I do not do any humidity testing in general, unless there is an actual defect caused by humidity as noted above, in which case I would gladly examine the resistor and replace it if defective, such minor temporary drifting caused by humidity is not considered a defect by any manufacturer.  I invite anyone to examine the data sheets of any of the major precision wire wound manufacturers; it is very unlikely that you will find any humidity specifications for commercial resistors.

The best and cheapest solution to highly varying humidity in your environment is to put a supply of silicate in with your circuits and change it out (or dry it) every so often.  Resorting unnecessarily to hermetic resistors will only cost you heavily and likely give no improvement in circuit performance.

Under accepted industry practices, I really do not regard your humidity variations as a genuine problem or would any other manufacturer for that matter.  I am sorry you believe you have ‘bad’ resistors but you admitted that the resistors had in fact returned to normal readings after a short stint in the oven.  Powered resistors are going to react differently to humidity than resistors laying about unpowered, I suggest you just put them in the circuit, power it up, put a bag of silicate in there and button it up….you’ll find it works just fine.

[Edwin G. Pettis]

All resistors, with the exception of wire wound, have 1/f noise in addition to the usual thermal noise and it can be very significant depending on the type of resistor.  Resistors made in the fashion of ICs are not known for having low noise or low TCR (except that the resistors can be TCR matched pretty closely since they are one a single substrate).  Many film resistors have significant 1/f noise and even some foil resistors can be noisy, just because the data sheet says one thing, don't take that as a guarantee.  Precision wire wounds are the only resistor technology that can very close to the theoretical thermal noise limits as calculated, all other resistors are higher, not just my opinion, it has been proven by Universities independently.

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.

Since you mentioned it Andreas, if you are picking up any line noise in your Vref circuit, don't blame the resistors, you've got interference you need to eliminate since it can be picked up by other components including active ones.  Why do we put Vrefs inside metal shielded enclosures......

On the subject of "if you can't measure it", does it matter?  I quite agree, if you can't repeatedly measure a parameter with a low enough uncertainty and high enough accuracy, then chasing after anything you can't reliably measure is a waste of time.

[branadic]

Precision wire wounds are the only resistor technology that can very close to the theoretical thermal noise limits as calculated, all other resistors are higher, not just my opinion, it has been proven by Universities independently.

Edwin, would you please refer to the appropriate publicatons coming to this conclusion? Especially I would like to understand the mechanism why WW are off of 1/f noise comparing to other resistor technologies. Thanks.