With the 2019 redefinition of the SI base units, I wonder what kind of technology is needed to reproduce the meter and kilogram in practice from scratch with a tolerance of ±0.1 mm and ±0.1 g? Existing rulers can't be used. Can an individual like myself do this, or is complex and expensive equipment required or even assumed?

Some background

In ancient times, had I wanted to measure something in meters or kilograms, then the "definitional" way would have been to borrow and copy the "reference meter" or "reference kilogram" from somebody else. This would not have required very sophisticated technology. As of 2022, how would an individual roughly reproduce the definitional meter and kilogram according to the 2019 SI definition from scratch? Say, if I'd want to measure the length of a table or obtain the weight of a certain apple, but I can't trust existing rulers or labels on grocery store items. I want to do it from scratch based on the definitions. Can this be done by an individual? Or... do only very few in the world have the ability to do this? Does it implicitly assume the existence of certain modern technology?

Note: The International Bureau of Weights and Measures published documents on practical realizations, but it seems way over my head right now as a non-specialist: https://www.bipm.org/en/publications/mises-en-pratique

  • 2
    $\begingroup$ Well, there is no SI reference meter any more - the meter is defined by $c$ and time. $\endgroup$
    – Jon Custer
    Commented Mar 31, 2022 at 17:23
  • 6
    $\begingroup$ @JonCuster Yes, exactly. So how would an individual reproduce the meter in practice? What kind of technology is needed? ... Say if I wanted to measure the length of a table. Again, just going off the current SI definitions. $\endgroup$
    – Michael
    Commented Mar 31, 2022 at 17:27
  • 9
    $\begingroup$ The easiest way to reproduce a reference meter is to go to your local office products store and buy a ruler. That's already accurate to $10^{-3}$. Likewise, a reference kilogram can be bought from your grocery store, at least if you're happy with percent-level accuracy. Added benefit: You may be able to eat your reference kilogram afterwards. $\endgroup$
    – rfl
    Commented Mar 31, 2022 at 18:53
  • 1
    $\begingroup$ For the meter, it seems we simply need to be able to count the number of unperturbed ground-state hyperfine transitions of the Cs-133 atom and see how far light travels after $9,192,631,770 \pm 919,263$ transitions, with common household items. science.howstuffworks.com/atomic-clock3.htm $\endgroup$
    – jpf
    Commented Mar 31, 2022 at 23:05
  • 5
    $\begingroup$ "Existing rulers can't be used" - Ok, but what are you allowing yourself to use? Like... you find yourself naked in the middle of an uninhabited forest, armed only with knowledge, and you mean to produce everything you need from raw materials? $\endgroup$
    – J...
    Commented Apr 1, 2022 at 20:40

3 Answers 3


I don't see any particular reason why you couldn't reproduce the Foucault experiment with bronze age technology. Foucault measured the speed of light to better than 1% - not quite the 0.1% requested, but very close. If you give Ptolemy Foucault's procedure, he can probably duplicate it, although he won't know what a meter is.

One second is 1/86400 of the time between consecutive sunrises (or, for greater accuracy, between consecutive cycles of the celestial sphere, or between consecutive sunrises at the winter or summer solstice). Ptolemy can tune a pendulum and count. No atomic clocks required, the planet is a perfectly good clock at the desired precision.

Given the second and the speed of light, we have the definition of the meter and Ptolemy can convert to meters if he suddenly gets the perverse desire to define a unit of length that no-one else uses, equal to exactly 1/299792458 the distance light travels in a second.

Ptolemy can blow glass into a pipette, put some mercury in it, and graduate it 100 times between the freezing point and boiling point of water, so he has the same temperature units that we have.

Ptolemy can make another graduated pipette, this time with water in it, gradually heat the water to discover that its maximal density is found at 4 marks above freezing on his mercury pipette, and use volumes of water at maximal density for mass. One liter of water at 4 degrees is 1 kilogram to much better than 0.1%, so now Ptolemy has a unit (based on cubic light-days, which are convertible into liters filled with maximally dense water) which is convertible into kilograms.

Ptolemy now has meters, seconds, and kilograms, so he has newtons, joules, and so on.

We can give Ptolemy a measurement for volts (in terms of centimeters of dielectric breakdown of dry air) by having him experiment with triboelectricity, which should have a precision of about +/- 1%. Not great, but pretty good. Given units convertible to volts, joules, meters, and seconds, Ptolemy now has ohms, amperes, coulombs, and so on. He's good to go unless we need him to start writing about quarks.

  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$
    – Chris
    Commented Apr 2, 2022 at 23:48

Neither of the definitions pre-suppose the existence of any advanced technology in principle. But in practice advanced technology is required to make the required measurements. Here's my stab at constructing minimal "primary standards" for the existing SI system. I copy the current (April 2, 2022) definitions of the NIST SI units from https://physics.nist.gov/cuu/Units/current.html

Making a ruler that measures 1 m:

SI definitions discussion

The definiiton of the meter is

The meter, symbol m, is the SI unit of length. It is defined by taking the fixed numerical value of the speed of light in vacuum c to be 299 792 458 when expressed in the unit m s-1, where the second is defined in terms of ΔνCs.

What this tells us is that $1 \text{ m} = (c) * (1\text{ s}) / 299792458$ and $1 \text{ s}$ is defined by

The second, symbol s, is the SI unit of time. It is defined by taking the fixed numerical value of the cesium frequency ΔνCs, the unperturbed ground-state hyperfine transition frequency of the cesium 133 atom, to be 9 192 631 770 when expressed in the unit Hz, which is equal to s-1.

This means that to realize the meter it is necessary to realize the second. We can see that the second is defined by

$$ 1 \text{ s} = 1/ \Delta\nu_{\text{Cs}} $$

Where $\Delta \nu_{\text{Cs}}$ is the frequency of the Cs ground state hyperfine transition. From this discussion we see that to realize the meter we must have a realization of the second which means we must have a realization of the ground state hyperfine transition frequency in Cs. This means that if we want a primary standard for the meter we need Cs atoms and we need a way to measure it's ground state hyperfine transition frequency. This means that we need an atomic clock.

How do atomic clocks work?

You may find this answer helpful in undestanding how atomic clock work. The basic idea of an atomic clock is you have individually seperated (in a thermal or ultracold gas) Cs atoms. You shine microwaves onto the Cs atoms. As you tune the frequency of the microwaves you will see that there is one frequency where the microwaves are most efficiently absorbed by the Cs atoms. As you tune the microwaves off resonance the absorption probability decreases. The shape of the absorption as a function of frequency might look like the graph below.

enter image description here

An atomic clock works by (1) shining microwaves through a sample of atoms (2) measuring the absorption as a function of frequency and then (3) setting the microwaves to be as close to resonance as possible given the uncertainty of the measurement of the absorption function. Once this calibration of the microwave frequency has been performed, the oscillation of the microwaves serves as your "clock". According to the SI definition of the second, we would realize the second by (1) performing the above calibration on the microwave generator and then (2) counting oscillations of the microwaves until we count to 9,192,631,770. In short: we perform microwave spectroscopy of the Cs ground state hyperfine transition and then count oscillation of microwaves calibrated according to that spectroscopy. State of the art atomic clocks are state of the art because they can perform the most accurate and precise spectroscopy in the world.

Can you "DIY" an atomic clock?

Supplies: For this, you need a sample of Cs and a source of microwaves and a microwave detector. The first hit on google shows Cs vapor cells available for ~$600. If you want to DIY your own Cs vapor cell you're going to have to learn a lot of skills that I am not familiar with. Chemists might know how to make the vapor cells and source Cs but I don't. You can also buy the required microwave electronics directly, but, a dedicated electronics hobbyist could build the microwave generator, horn, and antenna/sensor from basic components themselves. I would recommend studying designs for commercial Cs atomic clocks to see the "easiest" way to do things as they probably utilize some nice tricks that make the job much easier.

Realizing the SI m

Approach (1): Time of Flight Measurement

Above we saw that $1\text{ m} = (c) * (1\text{ s})$. We also know that if an object is moving with velocity $v$ for time $t$ then the distance travelled is $d = vt$. Since electromagnetic radiation travels at constant speed with $v=c$ this means that $1 \text{ m}$ is the distance that electromagnetic radiation travels in $1 \text{ s}$.

This informs one measurement approach. We can setup some source of electromagnetic waves and setup a reflector for those waves distance $d$ away. We then setup a sensor for the radiation close to the source. We then time how long it takes the waves to travel to the reflector and back. We then know the distance between the source and the reflector is $d = ct/2$.

To measure the duration we need to use our reference atomic clock. The experiment would look something like this:

  • Run atomic spectrosocpy to calibrate the atomic clock.
  • Simultaneously trigger the generation of your electromagnetic radiation and counting of periods of the atomic clock's microwaves.
  • When the sensor senses that your electromagnetic radiation has returned stop counting the ticks from the atomic clock.

We would then have

$$ d = c t = c N / \Delta \nu_{\text{Cs}} = (N) * (c) * (1\text{ s}) = (N) * (1\text{ m}) $$

where $N$ is the number of periods measured. Depending on how good you want to get at electronics, it will probably be reasonable to measure times of flight no shorter than ~10 $\mu\text{s}$. This corresponds to about $N\approx 9000$ periods of the atomic clock and a distance of $d\approx 1.5 \text{ km}$.

You have now generated a reference kilometer (likely with pretty good accuracy). To convert this to a reference for a single meter you would likely need to use some trig and surveying techniques. My guess is that the surveying would introduce some of the largest inaccuracies of your realization of the meter.

Above I left it vague as to the nature of the electromagnetic radiation that you are actually using for your time of flight measurement. In practice that radiation could be light and a regular mirror in which case you would need a source of light such as a laser that can be electronically triggered to generate light at a precice moment in time (for synchronization with the clock) as well as a photodetector whose readout can be precisely timed. Alternatively, if you're already a microwave electronics guru from building an atomic clock, you could use a microwave horn, reflector, and receiver/antenna for your experiment.

If you want to relax the constraint of electronic triggering you could do something like the old school speed of light measurements to accurately measure a much larger distance. However, in this case you will need to do more extensive surveying to get a realization of a single meter. One the other hand, if you want to get better at electronics you can try to measure faster and faster times (perhaps down to 10-50 ns) to reduce the amount of dividing down/surverying you need to do to realize the single meter.

Supplies: Atomic clock, triggerable EM source, timeable EM sensor, EM reflector, syncrhonization electronics.

Making a 1 kg reference mass

SI definitions discussion

The definition of the kg is

The kilogram, symbol kg, is the SI unit of mass. It is defined by taking the fixed numerical value of the Planck constant h to be 6.626 070 15 × 10-34 when expressed in the unit J s, which is equal to kg m2 s-1, where the meter and the second are defined in terms of c and ΔνCs.

This tells us that $$ 1 \text{ kg} = (h) * (1 \text{ s}) * (1 \text{ m})^{-2} * (6.62607015 \times 10^{34}) $$

Here $h$ is Planck's constant and 1 s and 1 m are defined as above.

YIKES!!! this definition tells us that to realize the kg we must have a realization of Planck's constant $h$, the fundamental unit of quantum mechanics! Frankly that is just going to make things tricky experimentally because we need a measurement that refers the mass of a macroscopic object (1 kg big) to some quantum system whose dynamics depend on $h$.

How to relate mass to $h$?

This is done by NIST using the https://www.nist.gov/si-redefinition/kilogram-kibble-balance. High level, the idea is to get an electronic voltage $V$ which is somehow calibrated in a way related to Planck's constant. This voltage is then used to generate magnetic fields which levitate a mass in earth's graviational field. The voltage is tuned to find the exact voltage when the weight of the mass is balanced by the magnetic force. If accleration due to earth's gravitational field is known then the mass of the object can be related to the balancing voltage which is in turn related to $h$.

Can you "DIY" a Kibble balance?

I am much less familiar with Kibble balances than with atomic clocks, I'm learning as I go. But here is my proposal for a homemade Kibble balance.

Getting a Voltage that is Related to $h$.

I found a nice method to get a voltage related to h on youtube. There are two steps involving an LED. The LED emits light of a certain color which has some optical wavelength $\lambda$. Note that, according to the electromagnetic theory of light, $\lambda = c / \nu$ where $c$ is the speed of light and $\nu$ is the frequency of the electromagnetic radiation.

The first step is to hook up the LED to a voltmeter and adjust the voltmeter's voltage until the LED lights up. Because the LED emits individual photons with energy $E = h\nu$ in a way related to (according to the theory of solid state physics) the photoelectric effect, it will not light up in $V = e E = e h \nu$ and $e$ is the charge of the electron.

... this section is a work in progress ...

Old version

Regarding the kg: I just looked this up as a refresher so this is fresh in my brain. It might not be that bad. I watched how to measure Planck's constant. Basically you get an LED and measure its wavelength (using your special ruler you made in the previous step) using a diffraction grating (you'll need to be able to make a grating with sub-wavelength spacing; I'm not sure how this can be done easily). Then you turn up a voltage supply until the LED turns on. Because you know the wavelength of the light, Planck's constant (it is defined) and the speed of light you now know (within the accuracy of your wavelength measurement which will depend on the accuracy of your ruler) the voltage coming out of your voltage supply. You could then use some multiplying or dividing techniques to get calibrated larger voltages.

I guess you then need to figure out how to use that voltage to support a mass. This is how the Kibble balance works (https://en.wikipedia.org/wiki/Kibble_balance, https://www.nist.gov/si-redefinition/kilogram-kibble-balance). In the kibble balance I guess the voltage is used to drive electromagnets that levitate a coil that supports the mass. Yikes, it's going to be hard to get calibrated currents through all those coils. Maybe you win by making all coils out of the same material so you know the resistance is the same for all of them, then you know that all currents in the system (and magnetic fields) are directly proportional to your voltages and you can get the currents and B-fields to cancel out leaving you just with calibrated voltages.

Finally, when you balance the mass you'll have mg so you also need to measure g. But that should be pretty easy, just drop something and time how long it takes.

Phew, I think it could be done. Would be interesting to see a challenge of like, measure 1 m and 1 kg to 10% or 1% precision or something according to the SI definition and do it for under $3k or something.

Edit: I plan to clean this answer up in a bit. For now I wanted to add that the most technologically advanced part of this might be the diffraction grating. Next is probably the LED because it relies on semiconductor technology (even though it’s inexpensive and common place). Then the Cs cell or the microwave generators and sensors depending on if you are more comfortable with E&M or chemistry. Next is probably the mechanical complexity of the Kibble balance then finally all of the advanced experimental methodology/theory including the surveying techniques.

  • 1
    $\begingroup$ This is the correct answer. I would add that the availability of advanced technology is what drove the redefinition of these units. $\endgroup$
    – John Doty
    Commented Apr 1, 2022 at 13:52
  • 3
    $\begingroup$ Once you know the frequency of your microwaves, set up a Michelson interferometer. One wavelength at the Cesium transition frequency is ~32.6 mm. There’s your practical ruler. $\endgroup$
    – Gilbert
    Commented Apr 1, 2022 at 17:31
  • 2
    $\begingroup$ A Michelson with a nice translation stage and a fine screw would easily provide 0.1 mm, no surveying necessary! $\endgroup$
    – Gilbert
    Commented Apr 1, 2022 at 17:44
  • $\begingroup$ @Gilbert Let’s see yeah that’s nice. You know the frequency and speed of light so you know wavelength. You can the. Count wavelengths for your ruler. Much better. Requires similar tech, you do need a fine thread screw which you didn’t need before, but maybe not even that fine. I’ll update when I can. $\endgroup$
    – Jagerber48
    Commented Apr 1, 2022 at 18:54
  • 2
    $\begingroup$ You cannot buy a Cs atomic clock for that price: the clock you linked is a rubidium one. $\endgroup$ Commented Apr 3, 2022 at 19:12

To reproduce the meter and the kilogram from the definitions requires performing experiments that historically would be considered as measurements of Planck’s constant, the speed of light, and the unperturbed ground-state hyperfine transition frequency of the caesium 133 atom. Any device to measure the last of those constants is a caesium atomic clock, so there is no way to obtain a definition-based measurement without at least building or buying an atomic clock.

With such a clock, a speed of light measurement is not too difficult to obtain a meter.

However, Planck’s constant is more difficult to measure. Probably the easiest way to get a rough measure is via the photoelectric effect, but that will require an accurate voltmeter. There are ways to do that which are less expensive than a caesium atomic clock, but still probably more than a do-it-yourself project. However, such a measurement will give you atomic scale masses, so for macroscopic masses you would need to use something like a Kibble balance. You can build one of those to ~1% accuracy using Lego bricks and some basic DIY electronics: https://www.nist.gov/si-redefinition/kilogram/nist-do-it-yourself-kibble-balance

That said, I do object to the idea that current devices could not be used. A traceable standard can indeed be trusted, precisely because its characteristics are traceable to a reference standard. The change in the definition of the SI system does not affect the reliability of traceable standards.

  • $\begingroup$ No. In the current SI, Planck's constant and the speed of light are defined, fixed constants, not subject to measurement. You use those definitions to reduce measurements to a comparison of time to that kept by a cesium clock. $\endgroup$
    – John Doty
    Commented Apr 1, 2022 at 13:46
  • 2
    $\begingroup$ @JohnDoty that is indeed the point. You do experiments, like the photoelectric effect, black body spectra, or a Kibble balance, that previously measured Planck's constant, and now they measure the mass in kilograms instead $\endgroup$
    – Dale
    Commented Apr 1, 2022 at 18:46
  • $\begingroup$ What you're saying is quite misleading. The SI, for historical continuity, chose definitions compatible with previous measurements, but there was no physical or mathematical requirement to do so. At the present time, it is literally meaningless to measure the speed of light or Planck's constant within the SI context. So your statement that reproducing the meter and kilogram "requires a measurement of Planck’s constant [and] the speed of light" is false and profoundly misleading. $\endgroup$
    – John Doty
    Commented Apr 1, 2022 at 19:00
  • $\begingroup$ @JohnDoty said "At the present time, it is literally meaningless to measure the speed of light or Planck's constant within the SI context" That is a good point. I have revised the statement to be more clear about my intention. $\endgroup$
    – Dale
    Commented Apr 1, 2022 at 19:54
  • $\begingroup$ What historical measurement of the speed of light corresponds to using GPS to measure distance? $\endgroup$
    – John Doty
    Commented Apr 3, 2022 at 23:02

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.