Creating Standards for Measurement

Well, having been a designer of measuring instruments for over 30 years I think I can quite happily say that none of us will have a primary standard to hand!

There are of course only a couple of primary standards in each country in the UK the National Physics Laboratory (NPL).

I have my own standards which hve been calibrated against secondary standards traceable to national standards...

For voltage we use a standard Weston cell housed in a temperature controlled instrument, this is used to check our Schlumberger voltage calibrator to within 20 ppm accuracy.

For Resistance we have several oil filled manganin wire wound resistors.

For current we use the above resistors and voltage standards.

For capacitance we have an old (but still superb!) General Radio 1620A capacitance bridge fitted with standard capacitors of less than 5 ppm drift with a ratio transformer giving 1 ppm resolution, down to 10 attoFarads, and an absolute accuracy of about 70 ppm or +/-0.01%.

Time measurement is perhaps the easiest to obtain, with the readily availability of decent OCXO or even TCXO.

For length we have a set of grade '0' slip gauges.

I would suggest that a.c. standards are perhaps the most difficult to have available...

Due to the general way that a.c. is measured... i.e. the sine wave must be pure, low distortion etc... So that readings peak to peak are related accuratly to RMS voltages...

I would be interested in what others use as an a.c. reference? We use a Gertsch 6 decade ratio transformer to provide the very accurate attenuation accurate to 0.001% of a.c. voltage... But we don't have a recognised standard a.c. set voltage....

John.

Use of a Fluke model 5790A AC reference measurement standard is widely used in the Metrology Community....You set up your synthesized signal generator or oscillator or whatever your source is to a known ACV level using this standard. Of course using an accurate counter/timer as well for the freq. accuracy might be required..Fluke PM6681R comes to mind. The main thing to remember is that it is not neccessary to have your "own" standards as long as you have traceability via calibration certification back to ISO or local National standards.

I agree with Your comment about primary standards..

Regarding AC measurements. I am planning also to buy this kind of inductive divider. General Radio was making very good dividers of this kind: resolution 10^-7, uncertainty a fraction of ppm. Similar uncertainty had dividers made in England by Tinsley. But this accuracy is for frequency up to 2kHz only. A good (price/performance) source of ac voltage could be used Fluke 5200 calibrator. For measurement could be used Fluke 792A AC/DC Transfer (10ppm/year) +DC Voltmeter, but 792A is not cheap.

I have a question to You. I am completing small DC (Volts, Ohms) laboratory; goal: 1ppm accuracy.I am wondering to buy or not a group of Muirhead K-231-A Temperature controled Standard cells. Do You have experience with this type? Or Technical Data? I have no access to specification. I would like to know their annual drift and at what temperature cells are kept (nominal temperature of thermostat?). And may be You know the current price of used units of this/similar type? If You have any info, send to my email czgut@vp.pl, (because I m a guest till now here in CR4).

Regards!

czgut

1ppm/C reference sources are not a problem. You can write to Maxim Inc www.maxim-ic.com and can get two free samples for 2.5V, 4.096V and 5V. Problem is the way you make circuit and how you place it in the test system. Best thing to do is to put the circuit in large mass of brass filled with highly insulating and thermally conducting oil and this makes sure that temperature will not change even for months if the entire setup is isolated using polyurethane or PUF material. One can easily create thermocouples on the PCB and that becomes an added noise.

I am planning to do some research on materials that are proper for such handling. I was first exposed to primary standard for measurement of ionizing radiation, for which we needed something which will hold the heat and we can measure the temperature of 1C change for about 1M Rad dose given to the sample. By heating the surrounding to almost same level, we could stop the flow of heat and heat lose was set to minimum. Similar to active shied, an active temperature shield works well.

You are right in general concerning fighting against changing of Value with temperature. I can stabilise temp. to 0.01deg, and can measure it with 0.0001deg resolution. But my goal is <1ppm/year, not <1ppm/deg. And this is not so easy. Mentioned Maxim will drift at least an order of magnitude more per year. Much better is LTZ1000 having 1.5ppm/year typical drift. Fluke selects them for their 7000 series Voltage standards. But still this is more than 1ppm/year. Only linear regression of drift taken into account as correction and averaging of four units gives accuracy about 0.5ppm/year. I have 732A drifting about 1.6ppm/year, and Solartron 7081 drifting about 0.5ppm/year at 10V, but for 95% confidence I need more standards. I know also Resistance standards drifting<=1ppm/year, but still I do not have good one (with checked drift).

czgut

LTZ1000 is perhaps a better choice. I used LM199A with heater. This was old design by National Semiconductor.

Temperature stress and then restructuring of semiconductor causes some stress based drift which shows in time. Also it is much in those instruments that are often switched ON/OFF. If instrument remain ON for years, then drift is different. Annealing effect is better at some elevated temperature rather than at room temperature. All temperature changes, small or big cause some amount of problem and hence thermal inertia helps a bit. Slow up and slow down means greater chance of stability. This problem comes from defect formation and these defects are also caused by natural cosmic radiation in the semiconductor. These defects can be removed by taking the semiconductor to about 120C and then keeping at somewhere near 60C for many hours. Migration of atoms by diffusion is also a problem. Worst problem was noticed was a whisker formation, which is a nano-wire formed by slow evaporation of metal and forming a crystal under vapor pressure within the cavity. This can change the property of the device drastically, and sometime can cause spark due to high electrical field generated even at low voltages. Gold, Silver and Tin whiskers are known to form in semiconductor where wires are bonded to chips. You can also see much of the dirt that remain scattered on bonding the wires. Diffusion of metal, oxidation and surface change does a lot of harm and some of it is faster in surface semiconductor technology. Buried semiconductors are much better and their energy levels are almost stable. Surface structure has hundreds of energy levels for the electrons and some of them very close to room temperature energies and such levels get affected by room temperature changes. When you buy a part then you also have to look into which technology it belongs to and not just some part number to be picked. Better to talk to the manufacturer and get the right parts. It sometime comes with price. I pay normally US$12 for a JFET while the same may be available in the market for US$0.02. The reason is that the manufacturer becomes responsible to provide me what I want and what one claims. I get results that are an order of magnitude better. It is highly pains taking job to get to perfections and we all learn this in time as we mature in electronics engineering. Simple copy will not work all that easily.

I agree with You. When I was testing usability of semiconductors as temperature sensors (diodes, transistors) I observed hysteresis in U(T). Due to temperarure excursions over 40deg there was a shift in U(T) caused by stress (not the same expansion coefficients with temperature of Si chip, sealant and metal case). Small temperature excursions had no effect and traceability was better than 0.001deg.

czgut

IMHO, Datron 4910 is a much better choice than Fluke 732A/732B/734A because I have all of them and have tested them for several years. 4910 is LTZ1000 based but employ PWM for 7V to 10V conversion which is superior to those resistor divider used in fluke systems.

7000 series origin from Wavetek and utilize statistical divider for step up. Performs nice but hard to find on the market.

Lymex/bg2vo

But I am planning to get better stability, about 0.001deg, so may be we could do some work together, or exchange experience/materials/solutions. But this goes out of main topic, so I think we should move this discussion to direct e-mailing. So, please, let me know Your e-mail address. My is czgut@vp.pl

Let me know, how you intend to measure absolute temperature with 0.001C accuracy. The known way to have the temperature reference is the melting point of pure metals and perhaps some references for water phase changes.

Do you have any other way to know exact amount of temperature in absolute terms. There may be some Fluorescence Point as well but I am not aware of their accuracy down to 0.001C.

Perhaps John can shade greater light on absolute temperature measurements down to this limit and also the way to estimate error. I am not sure if we can easily know if metal started melting unless there is shape change, which is observable. What are sharp changes in temperature scale and also reproducible to 0.001C level?

I have written that I am planning to have 0.001deg stability/year, not uncertainty.

If I will select ultra stable thermistors, from having 0.01deg/100months stability in 70'C, I hope I will get 0.001deg stability in 30'C, needed for me to stabilise Weston cell's temperature.

But replying to Your next question: It is possible to have 0.001'C uncertainty. Today's commercially available Melting point of Gallium cells, with purity 99.99999% have uncertainty of 0.0001deg. Best SPRT (Standard Platinum Resistance Thermometer) have hysteresis below 0.0001deg up to 500'C. Best measuring devices have linearity of 0.02ppm and uncertainty of 0.1ppm. Resistance standards have stability of <1ppm/year. 1ppm in resistance corresponds to 0,00025deg. But measurement takes less than 1 year, about one to two weeks, so stability at that time allows for measurements with repeatability better than 0.0001deg in best laboratories. My equipment allows me to have 4ppm/2 weeks stability which is equivalent to 0,001deg.

czgut

Unfortunately we cannot create our own primary measurement standards. What we can do is to compare the measuring instrument against a better one (called standard) which in turn is checked against an even better one, and so on up to the national level standards, compared themself with the international standards in an uniterrupted chain.

Assuming we are able to create our own primary measurement standards we still need to be part of this international system, which is called just that, abbreviated SI. I hope I wasn't too didactic.

Dear John

I think you have given a very good list and it is nice to see that.

NPL New Delhi is near to my place (300km) and will like to see how they are maintaining standards in India. I think NPL UK and NIST and other agencies have regaular inter comparision scheme. I may have to do the same. I can get some financial support from Govenment of India for this purpose as long term loan or grant. I will also hold a visit to NPL UK and NIST USA. I think it will help if I spend some time with them and follow their basic ideas.

For time I will have Cs clock and will compare with SRS Cs Clock module and NPL standard and NIST standard. How long doe sit take to cross check on this time standard? Do I have to leave the instrument for a week?

Can I use beta radiation source as current standard? There will be fluctuation in the statistical counting due to decay rate distribution. Th-205 and Ni-63 I can easily get. Can we discuss on this? Beta can be counted in 360 degree geometry in liquid scintillator and I can do this on continuous bases to get good statistics. Or do you think the voltage and resistance method is better for current?

Temperature and pressure I can use Vaishala sensors as most of the people use them.

can you look into these web site and give me more links

http://www.scitechresources.gov/

www.sandia.gov/psl

What we are talking is to use BASIC law of Physics and Chemistry and use highrity materials and then set up controlled environment, and over a period if we keep following others also in parallel then that primary local standard is possible. No one can do it without taking pains and spending time and money all together. That is the plan. Idea is to extend the standard to many other things and we need something to compare with. Secondary standards I have been using for long now. What my engineers supposed to do is to eliminate sources of errors in measurement over long time bases extending to say 5-10 years from now.

Dear Shyam, I think for most practical purposes I've ever encountered, the accuracies of industrial laboratory standards are perfectly acceptable.

If you are talking about creating your own in-house primary standards there is a problem...

First the cost to do this and maintain them.

Secondly what will you calibrate them to? A normal calibration lab will only check them to their own secondary transfer standards.

Thirdly, and perhaps most important, is the increase in accuracy will be small compared to the huge cost of operating these primary standards.

I think its best to leave the primary standards to the government's laboratories to maintain and pay for.

John.

Dear John

I agree on that. Secondary standards may be good enough to create new primary standards also where there is no other information possible.

Plan is to use Government funding for the project in the form of Grant and create a facility for the purpose and also to use it for the same. Usually, Government can pump sizable amount in this and expects only return on profit sharing bases if we at all make profit in long run. Each country requires people to do some work so more or less the Government will own the facility. I am working on this plan.

Idea is to give support to the local industries in many areas and build instruments that can be secondary standard. Lab standard is what all do now.

I generally find problem in getting standard for photons or light intensity measurement. Gamma radiation measurement, mono energy 10eV to 10keV photons. However, now that we have electron storage ring and other facilities are getting equipped to better standards of measurements, I need to build much better instruments.

I can get any amount of funds for a purpose or zero when I do nothing. I am trusted by the Government and locals for sincerety. I need to create an environment around me and generate more work force used to build special technology.

I am planning to get some quantum devices also as one of my group is working on low temperature Squid and other Quantum devices. We have one Liquid Helium plant and Liquid Nitrogen plant. We also have a diamond high pressure lab. Other equipments are SIMS and ESCA to get the chemical contents information to atomic level. I also have vacuum high temperature furnace, single crystal growth facility under vacuum or atmosphere. There are lots of electron microscopes and field ion microscopes, ESR, Atomic spectrometers, ICP, FL and other equipments.

I have access to 1m thick wall steel room for shielding from cosmic radiation. We also have EM shield and EMP test room. Few more such special facilities were created in long run for same desire.

It is not difficult for me to analyze the quality of any material at atomic level. Time has come to put all resources to some better use. I will make a 10-year plan to be reasonable to get some good results. I do not expect any fast results but work can start now. I had meeting with scientists and discussed this and they are thinking now. I think I will hold many meetings in coming months to get things into action plan.

Primary standards are computed and compared and not calibrated to anything. Only support instruments are calibrated to secondary level. Primary is just a primary.

Apparently you would like to create an independent measurement system, but no matter how independent will be you will need to compare (calibrate) your primary standards with the rest of the world, for compatibility. There are also technological limitations. It took the world 131 years to create the existing measuring system and we still working on it. Metrology is by definition conservative. Also John pointed other financial prohibitive limitations, i.e. the cost of calibrations at the primary level could be ten times more than next lower level.

I agree to what John has said. The main purpose of taking trouble is to help the nation and build technology in more areas. Funds is something country spends for research in anyway. Few million dollars yearly can flow for it on regular bases for perhaps 10 years plan.

As I said before, only instruments that we use and are others make, need to be secondary level calibrated. Our instruments supposed to be primary and are only to be compared and quality of information to be maintained. Whatever, we do, we also notify with whatever accuracy that is possible now and update results in time. We create our own standard models and maintain for other to make relative measurement. Not everything is measured by all. Something that is our interest to be taken up seriously.

This requires debating on each standard we work on and improve on the plans regularly. That is what each of us try to do in maintaining standards and not that some one else does it for us.

NIST was very small in 1974 when I just started my research and they now get huge Government backup. I am not sure if I will get very high government support but sure will get reasonable support.

My immediate priority is to get some standard for radiation energy sources and dose estimation in human body. This will be some great help in cancer treatment and planning. I am associated to that field and also manufacture radiation detectors and instruments. If we can get 0.1% accurate measurement then it is closer to what we want. Secondary instruments can be +/-0.5% accurate and lab standard within +/-1% accuracy. Right now we operate at +/-5% error in radiation dose estimates, which is high.

Other requirements are in Photon energy sources and emission and absorption measurement for material testing. This is a high tech research area.

I understand that money is not the issue to creating your own local standard. I would contact NIST and ask for advice on particular areas you are considering for your own standards. You don't have to go thru the same trial and error process that ISO and NIST have done. I am sure that lesson learned information will be helpful to you. They might be able to get you started in the correct direction and avoid costly delays and experiment failures. Perhaps this might help for a place to start.

Public Inquiries Unit
NIST, 100 Bureau Drive, Stop 1070, Gaithersburg, MD 20899-1070
Email: inquiries@nist.gov
Phone: (301) 975-NIST (6478) or TTY (301) 975-8295

Standard Reference Data
100 Bureau Drive, Stop 2310, Gaithersburg, MD 20899-2310
Contact Us Phone: (301) 975-2208

I will plan to visit NIST and will like to hold a talk with people. They have small industry program and they also allow short visits. For ionizing radiation, there is a group in NIST which has done some good work. I am actually aware of some of NIST programs for the last 20 years now. Our local standards are maintained by NPL, at New Delhi. I have to meet these people also. For radiation, BARC used to have standard using ionization chamber. I have used that instrument some 32 years ago when I took some training under WHO program in Radiological Physics and Health Physics Certification. I am not sure if they are still keeping the standard, but for Indian hospitals they are doing the instrument certification through AERB or Atomic Energy Regulatory Board. www.barc.ernet.in and www.aerb.gov.in

Where is the World...: turn-key Quantized Hall Resistance Standard ,(accuracy 0.02ppm) can be bought from Measurements International:

http://www.mintl.com/media/pdfs/quantohm.dsrev.3_000.pdf

More and more National Laboratories buys/installs this kind standards, therefore their prices should go down :-)

czgut,

Looks good, but where do we get He-4 for this instrument to maintain standard. This also tells that there is no good material found yet that sustains in resistance value near room temperature. This also ensures that all ADC DAC and current sources have drift in time if you are using at room temperature. At best they can be compensated and again re-compensated at some regular intervals by calibration and in between by extrapolation of drift trend if that is predictable.

What is the most stable resistance source and its stability at room temperature with predictable trend in changes per year? I think there is no standard material as such but some materials like bulk metal films may show better performance of about 5-10ppm/year at best. It is really a tough thing to work with when people casually demand 1ppm stability as if it is very simple to do it.

Life isn't easy for mastering the measurement technology to perfection. It is a real hard work, lots of money, and dedication that gets to something to count on.

I wish to thank all who participated in this discussion.

Probably the most stable with time is Manganin, first developed by Isabellenhute Germany. Best selected Special type Standard Resistors made of Manganin, for example Thomas Type Leeds & Northrup 4210 have drift below 0.1ppm/year, and are still used at National Laboratories, even those having QHR Standards. Manganin Resistors have relatively high temp coefficient, therefore are used in temp bath (T=const +-0.01deg). Best with temperature is Evanohm. Special Type 10k Standards made of Evanohm for example ESI SR104 have drift below 0.2ppm/year and <0.2ppm/deg around nominal temperature.

Quite many commercially available Resistor Standards have declared typical drift <1 ...3ppm/deg. From those some selected can have drift <0.2 .. 0.5ppm/year. To avoid problems with thermal hysteresis and long stabilisation time after transportation some firms are putting resistors in thermostats with battery backup.

Process Instruments is using approach You have mentioned. Their SmartResistor is temperature controlled and it's drift with time is corrected by software.

Bulk Metal Foil as You mentioned is not as stable with time as wirewound, but is much cheaper.

In general Your conclusion is very right. To have 1ppm accuracy one need much money (to buy proper Standard) or some money and much of work to do (to select stable resistor from many).

czgut,

I think for practical use, Vishay metal film may be more usable on circuits. They do not come very cheap and cost $10 to $20 range. For standards I think ST104 is a very good idea.

Indeed L&N 4210 are very old and nice standard resistors. I have three and like them very much, although I'm still struggling for how to test and verify them properly. However, drift of <0.1ppm/year is probably the tested actual value from selected samples but not specified. The only specification I can find is from Tucker catalog saying its <1ppm/year.

If that is the case, SR104 is better specified at <0.5ppm/year. I've seen several SR104s with less than 0.1ppm/year drift. Better still, type 9210 from Measurement International specified at <0.2ppm/year. These standard resistors are made from evanohm due to very small beta temp-co.

In recent years, several types of hermetic resistors (such as 202Z) from Vishay have emerged with 2ppm/10years shelf life, that equivalent to 0.5ppm/year. The bad news is that the drift accelerates with time. Vishay also modified the shelf life to 2ppm/6year for new DS.

Some supporting photos can be found at: http://www.38hot.net/bbs/read.php?tid=142

Lymex/bg2vo

bg2vo

Very informative. It also needs standard voltage or current source to monitor the resistance. How far we can go lowest in this direction? What are primary standards for current and voltage and how stable they are in terms of electron charge and statistical deviation?

Of course there are many ways to measure one ohm standard resistors, but I cannot afford to use expensive ones such as a DCC. What I'm going to do is to compare instead of measure. The plan looks like this:

Build a battery-powered 100mA current source with very good short term stability and low noise. This current runs through all the one ohm resistors (including the reference) in series to be compared. By employing a low EMF scanner, a voltmeter such as Agilent 34420 or 3458A will be connected to these resistors in turn, and iterate for a long time. Owing to short time switch, the drift of voltmeter can be largely canceled. Process these data and the difference of each resistor with the reference will be obtained with good confidence.

Agilent 34420 is a very good nano voltmeter; its absolute noise is as low as 10nV for 100mV range.

Agilent 3458A is a very good voltmeter; its relative noise is as low as 0.01ppm.

lymex/bg2vo