Resistor standard & capacitor standard

[M0BSW]

Hello everyone,
I want to build a resistor standard and a capacitor standard in a plastic box, both as a means of  reference, is there a certain resistance or capacitance, or is it all about the very best quality components, advice would be welcomed and values, or doesn't a specific value mean anything as long as you know what it is and it reads the same every time.

[quarks]

All depends on what you want to achieve.

If you are after a as good as possible standard (I am in the process of trying that, so far only for resistance), I learned it is hard as a DIY project to make it happen. But because I am really interested to find out if I can do it, I will keep on trying and probably failing, because I just still do not know enough especially about high Ohm resistance.

If you also try to do that, here are some things I found necessary to think of (besides the choice of Resistor component itself):

- For low R values you have to have 4-wire connection, read/learn about tempco and Seebeck effect,  Low EMF cables and connections and/or learn how to avoid problems with it.

- For high R values you should read/learn about shielding/guarding/wiring/leakage/contamination
and probably many more things I have not yet found out my self

- Housing maybe metal instead of plastic and maybe even environmental sealed boxes (i.e. IP67)

- avoid high currents/selfheating abuse

But in general all you need is the most stable part you can find/afford (i.e. Vishay Z-Foil). The real exact value is nice to know, but it is better to know that a given value stays within a certain tolerance in a wide enough environmental range for where you plan to use it in (temp/humidity/pressure) and that the change of value when aged is predictable.

A useful start value, I think is to begin with 10k, because most meters are most accurate in this range and you can probably live accuracy-wise with only 2-wire connection. 

My own plan is to have decade values (1m, 10m, 100m, 1, 10, 100, 1k, 10k, 100k 1M, 10M and if doable also 100M and 1G).

[M0BSW]

That's really interesting, I only want to build something to a standard , for my hobby use, so I can take a reading of all my meters , then see in a year if they have drifted, so this is the reason for wanting to build something that's reliable, it's mainly curosity for my self, I shall read the sugestions you made & the vishay film foil resistors and 10k, the money I paid for my meters, I would imagine there perfect in anyway, it's just the learning. thank you for taking the time to reply I'm really grateful.
Paul

[quarks]

oops, I am sorry I missed that this post was in "Beginners".

For a starting point just get a good 0.1% resistors with low tempco 25ppm/°C or better.   Avoid SMD versions, instead use as big as possible through hole types.

[M0BSW]

Thank you I have seen one actually 21k @ 0.1% military spec, that should be acurate enough for my needs shouldn't it

[ivan747]

21k is not a good idea. Most multimeters are either 3 1/2, 4 1/2 or 5 1/2 digits. That means their displays will only show up to 1999, 19999 or 199999, respectively (they are 2000 count, 20000, etc.). If you get a 21k resistor a 3 1/2 digit multimeter will read 021.0k. You have just lost a digit if resolution. It's better to use 10k. If you do that, your 3 1/2 digit meter will read 10.00k and you can see how much the multimeter is drifting (changing its error over time) for example.

With 21k on a 0.5% multimeter:
Ideal value: 21k
Ideal reading: 021.0
Worst case (assuming ideal resistor): 21.105/20.895
Worst case readings (assuming ideal resistor): 021.1/20.9 (thanks to rounding)
As you can see, the difference between you thinking your meter is spot on and almost out of calibration is just 1 digit. And most multimeters state there is a 2 digit error on top of the 0.5% error.

Let's try 10k:
Iedal value: 10k
Ideal reading 10.00
Worst case: 10.05/09.95
Worst case readings: 10.05/9.95
Even if you add the 2 digit error, it is much easier to tell if the multimeter is out of calibration or not, and by how much. Try to get a 10k resistor or parallel two 21k resistors. Even if all your meters are 4000 count, like the Extech EX330 is, for example, get the 10k resistors anyway. You never know if you end up getting a n 1/2 digit multimeter.

Ivan

[EEVblog]

See my video, I have a 1K and 10K precision resistors in a box. About $20ea at Digikey. Not cheap, but 0.005% with very low tempco.


Dave.

[ivan747]

I have a very accurate Keithley multimeter whose manual suggest you use 19k resistors as a reference to calibrate the unit. That's because at that level the meter reads 19.0000 and the ±2 count error isn't as significant as the 0.005% error we are trying to correct. 19k gives you a bit of margin as well.

But you cannot easily get 19k resistors, you usually use a kelvin-varley divider. It has switches to let you set an exact resistance.
http://en.wikipedia.org/wiki/Kelvin–Varley_divider

The problem is 19k is not that practical for other uses. If you have a 4000 count meter for example, 19k is no longer relevant. And if you use 10k you can tell your multimeter's error without calculations. 10.03 means 0.3% if we ignore the ±2 count error of the previous example.

[Spawn]

Farnel have some nice resistors, but pricey. Here a 10k 0.005% with 0.05ppm/°C temco, probably same as the ones Dave has:
http://uk.farnell.com/vishay-foil-resistors-vpg/145310k0000v9l/resistor-precision-10k-0-005/dp/1867937?Ntt=Y145310K0000V9L

Here couple more

These are cheaper at digikey, but I can’t find lower temco as Dave mentioned:
http://www.digikey.nl/product-search/en/resistors/through-hole-resistors/66690?k=0.005%25

I was looking for a reference for my DMM's too, I ordered DMMcheck with 10ppm/°C for time being, might get some of these resistors, keep us updated if you find something please.

[mark-r]

That's really interesting, I only want to build something to a standard , for my hobby use, so I can take a reading of all my meters , then see in a year if they have drifted

If you are only checking for drift then stability is much more important than precision. You aren't that bothered about the exact value, but you do care that whatever value it has now will be the same years from now. So tempco (ppm/degree C) and drift (ppm/year) are important parameters to look for.
 

[amspire]

Standard resistors is the easy bit, as it is easy to get 0.1% and the 0.005% Vishay foil resistor is affordable (one off).

If you have one accurate resistor (of any value), it is possible to calibrate other resistors, but be aware of the temp coefficient of cheap film resistors.  50ppm/C means that a 10 degC temp change can cause the resistance to change by 0.05% so things get difficult for cheap decade standard box at better then 0.1%. Even the current from the ohms range of a multimeter causes standard metal film resistors to drift down at the 0.0001% to 0.001% level and that is the beauty of the Vishay foil resistors - they stay rock steady.

Capacitors are much harder.

First, it is much harder to get very stable capacitors.

Secondly, a capacitors value changes with frequency and temperature, so you have to have it calibrated at a known temperature and frequency. With good LCR meters you do know the test frequency, but with multimeter capacitance meters, you do not have a clue.

Third, for capacitors in the picofarad range, it is hard to isolate stray capacitances, so if you choose a standard capacitor, over 1nF is a good idea.

Silver Mica capacitors are probably the best. I have had success buying some ex military USSR silver mica caps - particularly some hermetically sealed 0.5% ones. The USSR has some excellent quality mica, and these caps are usually very well aged by the time you get your hands on them. Polystyrene are also not bad and are available in 0.5% values. Other types will probably be 1% rated tolerence at best.  If you look at the data sheets, you will see that there can be a considerable drift over 1000 hours and that is why I like well aged caps.

Just as an example, I had a look at some 0.5% silver mica SMD caps on Digikey. They are rated to change less then 2% after a 2000hours life test so the drift with age can easily be far more then the tolerance. The data for most caps will probably not even mention drift with age.

It is possible to calibrate a capacitor from an known resistor with a bit of extra home-made hardware. It would be nice to be able to calibrate a reference capacitor to better then 0.01% from a 0.005% Vishay foil resistor. I am thinking about it, and I might do a post on it.

Richard.

[M0BSW]

Wow ,what response, all points noted, and kept, I watched your video Dave, I'm on to digikey now, that looks like a life lasting setup cheers, have a good Sunday everyone , thank you all for your responses and advice.
Paul

[PA4TIM]

About resistors all is told and I agree. I made some myself and have a seres from ESI and a GR decade and two KVs.

Capacitors are more difficult. I have a 455pF mica that was measured very accurate at 25 degrees through a resonance measurement. I used that at the same frequency to determin the value of the coil I made excact like the guy who gave me the capacitor. Then used the coil to calibrate a philips standard air capacitor.
Now I have the gear to measure things myself the silver mica is allmost ten pF off at 1KHz ( also at higher frequencys) but the Philips capacitor is less as 0.5 pF off and over a broad range of frequency.

My GR 1-10 uF standard capacitor is polyprop and still spot on. My HP capacitor decade (100 pF to 1.1uF) is also still spot on.

For quick checks I made small caps from pcb. Very easy, just pieces of pcb, two wires at the right spacing so you can put then straight in your fixture,. After a year they are still allmost the same value ( few aF deviation)  as they were.  ( I made 1 to 80 pF versions and measured them with 6 digit resolution) i have the size info so you can make the same. It will be pssible to get them probably better as 1 pF accurate this way ( FR4 is not a very constant material but as far as I tried they are pretty reproducable.)

For caps you need a farraday cup and 3 terminal or four terminal guarded measurements. The fixture I made for my GR1620 has a residu capacitance of less then 0.4 pF. I meassured that too for correction. With open cup it is 0.6 pF ( from the top of my head but that is in the ball park)

[amspire]

PA4TIM, I hadn't though about using PCB's for small reference capacitors. Do they have a significant temp coefficient? If they have a good long term stability at a fixed temperature, it is something I might experiment with. I only have one GR standard cap (2nf) and that is it. I was thinking of making up some more standards for myself, and the PCB trick could be a good way to trim capacitors up till they are spot on a standard value (like 1nf, 10nF).

I have been seriously thinking about a way to calibrate capacitors from known resistors using a method based on the quadrature bridge that all the standards labs use, but without the transformers, and a very much simplified circuit.  I think it can work and it means that I can cross check my ESI capacitor bridge (0.01% accuracy but last calibrated decades ago). Like the standards labs, the bridge would only be working at low frequencies (100Hz up to 1.591549KHz).

Also what sort of coil did you use for the resonance test - was it air wound? What sort of frequencies do you use for the resonance test?

[PA4TIM]

With one known resistor or capacitor and a good (TRMS)  voltmeter you can make a ratiomeasurement.
Connect the known resistor and unkown resistor in series to a source, measure the voltdrop over the knwn resistor, measure it over the unkown. Calculate current through known resistor and devide Voltdrop over unkown by current to calculate R.
For a capacitor you can do the same using 1KHz of enough amplitude ( calculate current from reactance)

I have not tested the pcb caps for tempco but the value written on the one on the picture was measured around 25 degrees. The picture it taken around 18 degrees. I Also took them to a friend who lives much higher and we compared them, that was at 20 degrees, and he brought his new fresh calibrated 0.1% meter when he came over here and the values measured then were all within 0.1pF
And that was somewere around 28 degrees so I guess tempco is not bad.

Do not bother to make them 1 or 10 pF exact. For that you must be able to measure them. And if you can do that, it is more easy to just note that value. Same for a resistor, better a 0.0000001% 1004 Ohm resistor as a 5% 1000 Ohm resistor. And better a 0 ppm/C 5% resistor with known value as a 0.0001% resistor with a tempco of 200 ppm/C. ( but a 0ppm will often not be 0ppm and 0.0001% will not have a big tempco. There would be not much market for unless it is a ptc/ntc ;-)

Resonance measurements are a very unreliable way of measuring capacitance or self inductance.
Many factors influence the measurenent. But if you have a known cap t 1 KHz build a good fixture and use the right coupling you can calculate L ( but it is only right for that frequency) now you can swap the cap for the adjustable one set for the same value and fine adjust it to resonance so it reads the 1KHz value, that unkown cap must be rather independent of frequency, the knwn cap too.
Caps and inductors both have self resonant frequency so theire value alters with frequency.
Beside that the surrounding plays a role, like eddy currents in nearby metal, picking up signals ( EMC) hand effect, coupling factor, the paracitics and strays from the testfixture ect.
For instance two banana jacks at the stamdard distance can be 4 to 8 pF.
Or what I have tried, i made to tanks, both excact at 50 MHz. When I then lined them up a few cm from each other, so they were coupled, and measured resonance again the frequency shifted a few hz , probably due to mutual inductance.

[M0BSW]

 Yes I would like some info on sizes exct for caps if you have the time to post them .
Thank you
Paul  M0BSW