Talk:Proposed redefinition of SI base units

From Wikipedia, the free encyclopedia
Jump to: navigation, search
Good article Proposed redefinition of SI base units has been listed as one of the Natural sciences good articles under the good article criteria. If you can improve it further, please do so. If it no longer meets these criteria, you can reassess it.
Article milestones
Date Process Result
April 11, 2013 Peer review Reviewed
July 5, 2013 Good article nominee Listed
Current status: Good article
WikiProject Measurement (Rated GA-class, High-importance)
WikiProject icon This article is within the scope of WikiProject Measurement, a collaborative effort to improve the coverage of Measurement on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
 GA  This article has been rated as GA-Class on the project's quality scale.
 High  This article has been rated as High-importance on the project's importance scale.
WikiProject Physics (Rated GA-class, Mid-importance)
WikiProject icon This article is within the scope of WikiProject Physics, a collaborative effort to improve the coverage of Physics on Wikipedia. If you would like to participate, please visit the project page, where you can join the discussion and see a list of open tasks.
 GA  This article has been rated as GA-Class on the project's quality scale.
 Mid  This article has been rated as Mid-importance on the project's importance scale.

Revocation of 31-Jan-2011[edit]

It is a moot point whether one writes 2 x 10-8 or 20 x 10-9. The former keeps the mantissa in the range 1.0 to 10.0 while the latter maps onto μg. In such circumstances, I think it appropriate to keep the article in line with the original text. Martinvl (talk) 20:17, 30 January 2011 (UTC)

Rollback of 29 March 2011[edit]

I have rolled back a series of changes that were made earlier today. While I agreed with some of the changes, I did not agree with them all. In particular,

  • The CCU wrote to the CIPM - see the title given in the reference.
  • Hard spaces between numbers and symbols were removed. The hard spaces were put there to ensure that the value and symbol were not separated by line breaks. This is standard practice.
  • The way in which negative superscripts was handled - for example changing J·K−1 to J/K. While I agree that these two have the same meaning, I used J·K−1 for the sake of consistency with the rest of the article.
  • The symbol "sec" is not the recognised symbol for "seconds" - "s" is.

Martinvl (talk) 21:08, 29 March 2011 (UTC)

Impact on reproducibility[edit]

The current text contains a table, introduced with:

"The following table catalogues the improvements"

They don't look like they are all improvements, as in half the cases the change is from an exact value to an uncertain value. I suggest the table catalogues the changes.
—DIV ( (talk) 07:41, 16 April 2012 (UTC))

Following your observation, I am revising this section to include all constants listd in the reference. Martinvl (talk) 10:12, 16 April 2012 (UTC)

GA Review[edit]

This review is transcluded from Talk:Proposed redefinition of SI base units/GA1. The edit link for this section can be used to add comments to the review.

Reviewer: Adabow (talk · contribs) 06:52, 21 June 2013 (UTC)

  1. Perhaps link the background section to the History of the metric system using {{main}}
  2. Can a citation be added for the paragraph concerning the setting-up of the CGPM, CIPM and BIPM?
  3. "In 1889 the CGPM took delivery" - what does "took delivery" mean?
  4. "the mandate of the CGPM was extended to provide standards for all units of measure, not just mass and length" - before reading this, it is not clear that the three organisations only dealt with mass and length; please make this explicit earlier on.
  5. "In the ensuing years" - vague. Is there a known end date?
  6. Can you explain how the conditions set by the General Conference were not fully met?

Adabow (talk) 07:51, 21 June 2013 (UTC)


I have taken on board your proposals and have implemented them. I have also ensured that the date passed to the accessdate parameter is consistent across all citations. Martinvl (talk) 10:49, 21 June 2013 (UTC)


  1. It is not stated in the background section that/when the speed of light was fixed.
  2. Is "an elementary charge" correct, or should it be the elementary charge?
  3. Please provide a citation for the current definitions of the base units.
  4. There are some great analyses about the consequences of the redefinitions (eg effect of amp redefinition on vacuum permeability, vacuum permittivity and impedance of free space), but it is not all referenced
  5. Ditto with the example of potentially defining kg from G - WP:OR
  6. Who is Leonard?
GA review (see here for what the criteria are, and here for what they are not)
  1. It is reasonably well written.
    a (prose): b (MoS for lead, layout, word choice, fiction, and lists):
  2. It is factually accurate and verifiable.
    a (reference section): b (citations to reliable sources): c (OR):
  3. It is broad in its coverage.
    a (major aspects): b (focused):
  4. It follows the neutral point of view policy.
    Fair representation without bias:
  5. It is stable.
    No edit wars, etc.:
  6. It is illustrated by images and other media, where possible and appropriate.
    a (images are tagged and non-free content have fair use rationales): b (appropriate use with suitable captions):
  7. Overall:

Placing review on hold now. Adabow (talk) 07:07, 22 June 2013 (UTC)


I have taken the liberty of replacing bullet points with numbers to simplify cross-referencing of comments.

The outstanding issues in the first set of comments have been handled as follows:

  • Item 4: In this change, the text "under which three bodies were set up to regulate units of measure that were to be used internationally." was replaced by "under which three bodies were set up to take custody of the international prototype kilogram and metre and to regulate comparisons with national prototypes.[1],[2]" Martinvl (talk) 15:00, 24 June 2013 (UTC)
  • Item 5: The updated text reads: "In 1921 the Convention of the Metre was revised and the mandate of the CGPM was extended to provide standards for all units of measure, not just mass and length. In the ensuing years the CGPM took on responsibility for providing standards of time (1956), electric current (1946), temperature (1948), molar mass (1971) and luminosity (1946)." Martinvl (talk) 15:00, 24 June 2013 (UTC)

The issues raised in the second set of comments have been handled as follows:

  • Item 1: Done - diffs here. Martinvl (talk) 10:39, 25 June 2013 (UTC)
  • Item 2: Done. Martinvl (talk) 16:16, 24 June 2013 (UTC)
  • Item 3: The intro paragraph to the section now reads "The current (2008)1 and proposed (2011)2 definitions are given below." Martinvl (talk) 16:16, 24 June 2013 (UTC)
  • Item 4: Discussion on the impact of the changes on the ampere has been overhauled and citations provided. Martinvl (talk) 16:16, 24 June 2013 (UTC)
  • Item 5: Section Proposed redefinition of SI base units#Impact on reproducibility has been modified - citations were found that paralleled work which could be held to be WP:OR. Martinvl (talk) 10:42, 25 June 2013 (UTC)
  • Item 6: Final section has been rewritten - the comments made by the reviewer no longer apply, but the whole section needs to be reviewed.

As each item is addressed, it will be signed.

Martinvl (talk) 14:45, 24 June 2013 (UTC)

Also note the contradiction "In 1960 the metre was redefined in terms of the speed of light, making it derivable from nature" in the lead vs "Similarly, the 17th CGPM (1983) replaced the 1960 definition of the metre[Note 2] with one where the metre is derived from the speed of light" in the background section. Adabow (talk) 06:12, 26 June 2013 (UTC)

Thank you for spotting this - I have corrected it.
When trying to find out more about Leonard, I discovered that a lot more discussion material had become available, so I am rewriting the final section which I am calling "Discussion" rather than "Criticism". I hope to have it in place in a few days. Martinvl (talk) 06:44, 26 June 2013 (UTC)
Sounds great! Take your time - no rush. Adabow (talk) 07:38, 26 June 2013 (UTC)
Hi Adabow,
I have now completed my revisions following your initial review and when you are ready, I would be grateful if you could assess my latest offerings. Martinvl (talk) 20:49, 3 July 2013 (UTC)
There is still no explanation for who Leonard is. Ditto for Chyla. When someone is mentioned, there first name should be included in the first instance. If the first names are not known, then please at least give initials. Affiliation should be given as well. Adabow (talk) 23:27, 4 July 2013 (UTC)
Done. Reading between the lines Chyla appears to be a freelancer so his affiliation carries no weight, however the journal where this paper was published referees its papers, so that has been included in the text (complete with Wikilink).

OK, everything looks good to me now. Passing. Well done. Adabow (talk) 10:45, 5 July 2013 (UTC)

Diagram description[edit]

the diagram for current definitions of SI units claims in its description that a meter is defined as the distance traveled by light in 1 second. according to the article on the meter it is equal to 1/299,792,458 of this value, and the claim is sourced. I suggest the source be reused here and the information corrected. — Preceding unsigned comment added by (talk) 20:13, 14 September 2014 (UTC)

The caption doesn't say that the metre is defined as the distance traveled by light in a second, it says "in term of". The point is that it's defined in terms of distance traveled. But you're right, it's clumsy enough to be potentially misleading. I'll try tweaking it (though without putting in the precise value because that isn't the point) - see what you think. NebY (talk) 23:17, 14 September 2014 (UTC)
Fortunately the difference is big enough that it is hard to get wrong in actual use or discussion. It is nice to get it right, though. Gah4 (talk) 23:49, 5 December 2015 (UTC)


maybe it's time to change kg to gram? — Preceding unsigned comment added by (talk) 10:26, 23 June 2015 (UTC)


I noticed that the article has the word artefact, and was about to change to artifact. But then again artefact seems to be the British spelling, and so maybe consistent with the rest of the article. But then I find that there are many artifacts, too. Seems to me that we should be consistent, one way or the other. I don't know which way, though. Gah4 (talk) 23:52, 5 December 2015 (UTC)

I would use the word "prototype" anyway. (talk) 04:04, 7 December 2015 (UTC)

unit order[edit]

Is there a consistent order that units should be when appended to a quantity? I am mostly used to kg - m - s order, for example usually used in explaining force or energy. I saw some quantities with a different order, though about changing them, then decided to ask here. Gah4 (talk) 23:54, 5 December 2015 (UTC)

Figures are confusing, how to fix?[edit]

SI base unit.svg
Relations between New SI units definitions.svg

The diagrams ("directed graphs" to a mathematician) of the current and proposed definitions seem unnecessarily hard to compare. Each has the units in a different order and a different colour code. The m is blue in both, K orange, and the kg red, but s/A swap green/purple and cd/mol swap yellow/turquoise.

This graphic that the first is based on shows the units in the same positions, but I'm not sure

Assuming that having arrows mostly "down" would be most legible, A, mol, K and cd have only in-arrows in both diagrams and so should be placed lower down. s has only out-arrows. kg and m have both.

The other place to look is the changes, which are the arrows we'd like to be noticed most.

  • 5 links are unchanged: s→m, s→cd, s→A, m→cd, kg→cd
  • 3 links are deleted: m→A, kg→A, kg→mol (A is now based only on the second, mol is standalone)
  • 5 links are added: s→kg, s→K, m→kg, m→K, kg→K (kg is no longer primitive, K is no longer standalone)

So perhaps the following positions (expressed in terns of a 12-hour clock)

  • 9:26: A
  • 11:09: s
  • 12:51: m
  • 2:34: kg
  • 4:17: mol
  • 6:00: cd
  • 7:43: K

That's like the first diagram, but with the kg-A-cd triangle rotated counterclockwise.

Any other ideas? More importantly, any volunteers to actually draw it? (talk) 15:17, 28 December 2015 (UTC)

Trying to fix the problem, I’ve just replaced the first diagram (current SI) with another one which is color-consistent, position-consistent and size-consistent with the second diagram (new SI), which I have created back in 2011 and which was color-consistent at that time (the color inconsistency dates frome 2015 when the creator of the current-SI diagram updated it and changed its colors for some unknown reason).
As for the optimal unit order in the diagrams, you mention valid points, and one advantage of your proposal would be the absence of upward arrows in both diagrams. Still, I have sticked to my original unit order. Here are some (debatable) reasons :
  1. At least, the new-SI diagram has no upward arrow between units.
  2. The highest, 12 o’clock position of the unit second emphasizes its major "donor" status in the new-SI diagram, with 5 arrows flowing down from it. (In hindsight, I find it looks like the rays of the sun-god Aten!)
  3. This unit order allows a certain vertical symmetry of the arrows in the new-SI diagram.
  4. The heptagon figure is visually more "stable" when oriented with two vertices on the base line and one vertex at 12 o’clock position. In your proposal (only one vertex on the base line at 6 o’clock position), the figure would look in an unstable equilibrium. In my new-SI diagram, the stability of the figure is further "weighted" with each of its two lower-edge units receiving three downward arrows.
Those two diagrams are PNG files converted in SVG format, but Wikimedia Commons warns me that "This SVG image contains embedded raster graphics.[1] Such images are liable to produce inferior results when scaled to different sizes". Anyone knows how to convert them in "pure" vector graphics format, without having to redraw them from scratch?
--Wikipetzi (talk) 20:22, 15 January 2016 (UTC)
I redid both images using svg elements from File:SI base unit.svg. --IngenieroLoco (talk) 20:45, 28 June 2016 (UTC)
@IngenieroLoco: I should have said this much earlier, but thank you! (talk) 09:15, 3 January 2017 (UTC)

Date of redefinition[edit]

@Quondum: I put the word back to "likely", because "likely" is actually putting the case very mildly. It's basically certain, barring something very unexpected like the appearance of multiple inconsistent measurements of some of the constants. Perhaps the footnotes on that particular statement need improving, but the statement is WP:Verifiably true.

  • As the first footnote says, the BIPM director called it a "foregone conclusion" in 2014 already.
  • The 26th CGPM has already been scheduled for 13–16 November 2018.[1]. If you want to be specific, the redefinition is scheduled for the morning of Friday the 16th, and the press have already been invited. (Same source, right-hand side of p. 10.)
  • The proposal has been around, and accepted in principle, by two previous CGPMs (2011[2] and 2014[3]). At those meetings, they agreed to the redefinitions as soon as the measurements met a defined quality level.[4] That level has now been achieved.
  • The CODATA § Task Group on Fundamental Constants have already announced the deadline[5][6] for data to be incorporated into the redefinition: 2017-07-01 for the final fixed SI values.

The entire metrology world is assuming that it will occur. I can find half a dozen more sources (e.g. [7]) which state it in passing or obviously assume it. Can you find any fairly recent source which expresses the slightest doubt?

In fact, I've been thinking seriously about removing the word "likely" entirely, and either saying something like "planned" or "scheduled" or, even bolder, stating flatly that it will occur. Do you have any comments on that idea? (talk) 09:21, 13 January 2017 (UTC)

There is little doubt about it occurring then, but such an assertion remains a judgement. Using "likely" and "will" in the voice of Wikipedia is editorialization. WP should not be drawing conclusions (no matter how "obvious"), but should be citing sources, meaning stating what they say, or better, stating that they say something. To say that it is "expected" or "scheduled" would be fine, because that is what one or two of the sources say (though "expected" should be qualified by whom). My preference would be "scheduled", because that is a statement of fact (about intent), but does not attempt to predict the future. A statement in the future tense in WP is always suspect. —Quondum 16:00, 13 January 2017 (UTC)
@Quondum: I understand your point about future tense being suspect, but not the first point about editorialization. Expressing a value judgement (either overtly or via choice of wording) is editorialization. I don't see how "likely" does that.
The word is just a summary of the sources. (Remember, SYNTH is not just any synthesis.) It's certainly possible to go into great detail about who said what, but when everyone who knows anything about the subject agrees, that's not good sourcing, that's bad writing.
I haven't figured out the right wording for a more specific statement. Formally, only the vote is scheduled, and while the result of that vote is a "foregone conclusion" and everyone is making plans based on the outcome, they're also being polite about it in public. Fitting all of that into something brief is difficult. (talk) 19:34, 13 January 2017 (UTC)
It is a very fine point, but for the lead to say that it "likely" is WP drawing a conclusion about probability that does not appear to be stated by the sources even if that conclusion seems inevitable (let's be guided by how the sources are stating it). I suppose my reaction is because that wording immediately tells me "this could be an editor who really wants this to happen, so maybe they're skewing the interpretation". It just doesn't have the right tone to be convincing. How about something like "... is on track for adoption at ..."? —Quondum 20:44, 13 January 2017 (UTC)
@Quondum: I'm beginning to see the point here. I think that the conclusion is stated in the sources. As I described above, it's implied all over the place, but also stated explicitly in the sources already in the article. (Unless you think that "likely" isn't a reasonable synonym/paraphrase of the sources' words "foregone conclusion" and "expected".)
But even if that's a reasonable word in this context, it raises red flags for you in general and you'd like something which doesn't. That makes sense.
The problem is, as you mentioned earlier, anything that sounds like a prediction raises those flags. We could change the word to "expected" to match a recent source verbatim, but would that address your concern?
I could also add a footnote with details. But if it's going to be discussed in a footnote, I'm tempted to move all the caveats to the footnote, remove all the WP:WEASEL words from the lead, and just say it will happen.
I mean, obviously any statement about the future is subject to very unexpected caveats. There's no need to explicitly state that a major terrorist attack in Paris could preempt the 26th CGPM. But Summer Olympic Games says that "Tokyo, Japan will host the 2020 Summer Olympics" (emphasis added) because there's no reasonable doubt. (Someone added caveats to the lead of 2020 Summer Olympics, but the rest of the article uses the unmodified "will".)
Back to the topic at hand, do we have any remaining disagreement about what the sources say, or are we just trying to find the right wording to say it? (talk) 21:45, 13 January 2017 (UTC)
I think you've characterized it pretty well. Yes, it is probably only wording; I don't think there is any real disagreement with what the sources say, or that adoption is imminent at that date (I'm not well-learned here: it was a pleasant surprise to learn now that there is a definite timeline this time around). It is more that a reader might do what I did: see a red-flag word, check the footnote/references (I am guilty here: notes [1] and [2] don't directly deal with it and I thus barely scanned [3], then jumped to a conclusion about editing; I suggest perhaps keep only [3]; at least remove [1] as too dated and reorder [2] and [3] since [2] only confirms a premise in [3], as I scan it).
I take your point about how one expresses things (one is not expected to list all the caveats). In this case, I think word choice should suggest quoting something rather than an opinion of an editor. "Expected" would have raised less of a red flag for me, since it suggest at least that one is referring to a specific group who expects it and is less suggestive of editorializing. Being able to find the word in the footnote gives one a quick point of purchase. I'm not going to obstruct whatever you choose; I've learned something new and am not disagreeing Face-smile.svg. —Quondum 22:49, 13 January 2017 (UTC)
The most apropos parts of the sources are actually included in the citations as quotes. [1] was the first overt statement I could find, and the speaker is someone particularly knowledgeable. (Think of the BIPM director as the CEO of a company, the CIPM as the board of directors, and the CGPM as a shareholders' meeting. While the CEO can't speak for the latter, a good CEO knows what they're thinking.) [3] is by the head of the CODATA task group on fundamental constants, who's in charge of producing the numerical values that the CGPM will enshrine. Another good source, but not quite as good as [1]. [2] is the same thing as [3], just in "lay summary" form. Of the three, I'd say it's the most disposable.
It was clear in 2011 that the redefinition was going to go through "soon";[8] the 25th CGPM was held only 3 years after the 24th in anticipation, which turned out to be a bit too optimistic. But everyone who's been following the matter has watched the measurement uncertainties decrease below the specified limits through 2015 and 2016, and could draw the same conclusions as [3] states. (In truth, the hardest thing to track down was the exact date (as opposed to "Q4 2018") of the CGPM. It's not secret, just only communicated to the attendees.)
Anyway, I'll amend that to "expected". (talk) 01:35, 14 January 2017 (UTC)

Specific values in the 9th edition draft[edit]

@Petr Karel and Quondum: Sorry for that last undo; I hadn't realized it was a consequence of User:Petr Karel's edit. That's the problematic one, but this will take more discussion than a simple undo. (Feel free to undo my undo pending discussion if you like; I did it in a hurry and then stopped when I realized the situation was more complex.)

The problem is that the draft 9th edition does give those specific values, but those aren't the final values. I think it was done to make the formatting closer to final, but I worry that importing them verbatim will cause confusion among readers.

That those aren't the final values is verifiably true, As says, they will be adjusted based on measurement results submitted up to a deadline of 1 July 2017, which is still open. (And the reason the deadline is more than a year before the redefinition is to allow time for criticism of measurements.)

So we need to add some explanatory text... okay, I think I know how to do that. Perhaps consider this just an apology for my too-hasty undo. (talk) 05:51, 14 January 2017 (UTC)

No, I appreciate the undo; it was appropriate. And I was just commenting that the provisional nature of the values in Proposed redefinition of SI base units § Proposal should be made clearer, explaining that they will be finally revised after 2017-07-01, but you beat me to the talk page. —Quondum 05:58, 14 January 2017 (UTC)
@Quondum: Well, I was apologizing for undoing your edit and then stopping to discuss before undoing the previous one that was the really important one. Kind of shooting the messenger. Anyway, edits made. I felt hurried to correct misleading information ASAP, so I'd definitely appreciate a review/copy-edit. (talk) 06:31, 14 January 2017 (UTC)
@Quondum: P.S. Going over your recent edits, the Avogadro constant article actually distinguishes between "the Avogadro constant" and "Avogadro's number". The former is 6.022140857×1023, the latter is the number of hydrogen atoms is 1 g of gas, about 5.97538324×1023 based on the atomic mass of protium at 1.00782504(7) u. Although the terms are understood as synonyms today, would it be better to avoid the ambiguity? (talk) 06:50, 14 January 2017 (UTC)
I think it is beautifully clear and up to date now, with little chance of misinterpretation. The only thing is that the proposed values in the table could do with the same footnote, though this might increase the column width even more with the Planck constant and the Boltzmann constant. Perhaps that could be remedied by a forced line break before the footnote?
I kept the distinction between the Avogadro constant and Avogadro's number. All I did was fix minor variants in the terms to match the terms in the article Avogadro constant, specifically only adding/removing "the" and "'s".
A side issue: the table uses the symbol without ever defining it, or linking to the article Rydberg constant. —Quondum 16:27, 14 January 2017 (UTC)
@Quondum: You're right about the table, sigh.
Regarding the Rydberg constant, the table doesn't define , or , either. (Although the last is defined in the text immediately after it.) (talk) 17:41, 14 January 2017 (UTC)

Rationale behind choice of electromagnetic constant[edit]

Is there some background on the (proposed) SI's choice of the elementary charge rather than either the electric constant or the magnetic constant to be assigned an exact value in the New SI? They are nominally equivalent ways to specify the same thing in a sense. The latter are arguably more fundamental (even in quantum mechanics), but ultimately the choice in this context would be the one which leads to the most accurate and useful system. Precision of mass constants particularly might be affected. Looking at the table, the dominant source of imprecision for every interesting constant listed is the fine-structure constant, which is in a sense is just the square of e, at least as far as precision is concerned. It seems to me that the history of this choice and comments on it would be appropriate (and definitely interesting) content for this article. —Quondum 22:34, 14 January 2017 (UTC)

I see that there is a recent paper that seems to speak to this: K.A. Bronnikov et al, On the Choice of Fixed Fundamental Constants for New Definitions of the SI Units, Measurement Techniques, November 2016, Volume 59, Issue 8, pp 803–809. —Quondum 04:05, 15 January 2017 (UTC)
It's a bit technical for a lay audience, but we can try. I'll try explaining it here, and see if we can eventually turn it into something suitable for mainspace.
Basically, it's all about shifting the uncertainties around to minimize covariances. If I have a bunch of measurements over here which are accurate relative to each other to parts per trillion, then it's annoying to be measuring them relative to a standard which is only measurable to parts per billion.
That's what was happening in 1960 when the metre was redefined from the bar to a spectroscopic standard. The spectroscopists had all of these wavelengths measured to ridiculous precision relative to each other, but measuring a macroscopic dimension was much harder. The Ångstrom was an agreed-upon standard which was nominally 10−10 m, but that was known much more loosely. When someone finally figured out how to count the number of wavelengths in a metre, the metre bar quickly became the limiting factor.
Another place this has happened is the astronomical unit. It's much easier to measure the relative distances of the planets' orbits than to relate this to a metre bar on Earth. For a long time, the orbits of the planets were known to 3+ decimal places in A.U., but the A.U. was only known to one or two decimal decimal places. (Read up on the transit of Venus for details.)
These sort of situations result in a lot of measurements which are very strongly correlated. A is measured with an uncertainty of 25 ppb, B is measured with an uncertainty of 25 ppb, but all of that uncertainty comes from the 25 ppb uncertainty in the comparison to external standard Z. The ratio of A and B is known with an error of < 1 ppb. This quickly becomes an error-prone pain in the ass.
It would be better to find a standard which could be compared to A and B at their sub-ppb level and move the 25 ppb uncertainty over to Z, since it can't be measured to better than 25 pbb uncertainty anyway.
The interest in Plancks's constant and the elementary charge arise because the field of electrical metrology has surpassed mass metrology, yet the current definitions of the volt, ampere, etc. are all based on the kilogram. Quantum Hall effect resistance standards and and Josephson voltage standards are absolutely fantastic because, as far as anyone has been able to tell, there are no corrections.
When building a real cesium clock, there are all sorts of corrections that must be applied to correct for the non-zero temperature, Doppler effect due to the moving atoms, applied magnetic field, etc. etc. which mean that the atoms inside my clock are not absorbing radiation at quite 9192631770 Hz.
But if you build a Josephson voltage standard, nobody has yet found any source of intrinsic error. (Tested down to 3×10−19!) The only uncertainty is due to sources like thermal EMF and wiring resistance outside the Josephson effect itself. Likewise for QHE resistance standards. This makes them startlingly accurate and practical measurement devices.
And the two effects are defined by the constants KJ = 2e/h and RK = e2/h. There's been 26 years of research which has been done using conventional electrical units based on assumed exact values of those constants, because the SI-referenced values are not good enough.
Anyway, the reason for defining e and h specifically is to give exact values to those constants, and thus the measurements made with those instruments. (talk) 23:32, 15 January 2017 (UTC)
That is illustrating the underlying process of reasoning for choosing reference standards, not the actual choices made. Someone reading this article (I assume) is interested in which standards are intrinsically more accurately comparable with the quantities of interest. The choice of e versus ε0 is the one that I am interested in, and I'll use it as an example. The uncertainties in the table that change are:
Constant Symbol Proposed definition (e defined) Alternate definition (ε0 defined)
Relation to directly measured and fixed constants Significant factor(s) in uncertainty Relation to directly measured and fixed constants Significant factor(s) in uncertainty
Josephson constant exact
Von Klitzing constant exact
Elementary charge defined exact
Magnetic constant exact
Vacuum permittivity defined exact
Impedance of free space exact
Faraday constant exact
The changes probably amount to just moving uncertainties around (despite the apparent halving of error), and metrologically this is probably inconsequential (in terms of what you say above). But from a practical perspective, an uncertainty of a measured quantity that is not pervasive in physics equations (, , ) is less bothersome or problematic than fuzziness in the definition of a pervasively used constant (, , ). Having Maxwell's equations exact (in terms of SI units) is probably better than knowing the exact charge of an individual electron, especially since the strength of the field that it generates is uncertain. That the constants chosen to define values for are actually those used at present in measurement is just "something under the hood"; it is the accuracy and usability of the final system that counts.
Anyhow, what I'm getting at is that this article might relevantly summarize references that give how the actual choices were made. —Quondum 04:35, 16 January 2017 (UTC)
"But from a practical perspective, an uncertainty of a measured quantity that is not pervasive in physics equations (, , ) is less bothersome or problematic than fuzziness in the definition of a pervasively used constant (, , )."
That's where I disagree with you. How often do people operate circuits in a hard vacuum? There's little point knowing more accurately than you know the applicable medium's dielectric constant. People normally assume that the relative permitting of air is 1, but it's about 1.00059, and varies with the weather! If that's not stable to better than 1 ppm, uncertainties less than 1 ppm are negligible.
On the other hand, as I said, , , and make up the quantum metrology triangle (no article, but there should be one!) which are direct inputs to a lot of high-precision experiments. (talk) 05:24, 16 January 2017 (UTC)
In high-precision experiments, I expect that one would choose the standards one wishes to measure against. For example, if the caesium-133 standard proves to have a higher relative bandwidth or similar impediment to precision than some other source, the experimenter might use the latter. In the case being discussed, these constants are already built into the system, and one can use these instead of the units m and s for the experiment, if they prove superior under the circumstances. This does not seem to me to be a metrological argument, but rather one of allowing sloppy thinking by an experimenter, since correctly analyzed, this instance makes no difference. It comes down to aesthetics (and general usage), really. The accuracy of alpha, for example, stays just as imperfect either way. —Quondum 12:27, 16 January 2017 (UTC)
@Quondum: "For example, if the caesium-133 standard proves to have a higher relative bandwidth or similar impediment to precision than some other source, the experimenter might use the latter."
Yes, but a good global standard is one that is usually the best possible. When a significant area of research builds up around an alternate standard because it can't be compared to the global one, the result is awkward. That's what happened with my historical Ångstrom and a.u. examples, and is the current situation with the Conventional electrical units and , and with the International Temperature Scale of 1990. (And starting to happen with optical frequency standards, but that's not quite as pressing yet.)
It's absolutely true that changing the definition doesn't affect the accuracy of the experiments the tiniest little bit. All it means is that more and more people simply aren't using the unit to describe the result of their experiment because it's too inconvenient. That's the situation that international agreements like the Treaty of the Metre, and International yard and pound are designed to avoid.
As you point out, the two possible definitions are linked by the fine-structure constant, and it's a question of which side of that link to place the definition on, and which side to put the uncertainty on.
The fine-structure constant is known to 0.32 ppb, so that's the uncertainty that the losing side has to deal with.
The question is, which side is more likely to make measurements more precise than 0.32 ppb?
I's sure it's gotten better since then, but The PTB's Josephson voltage standard was achieving 0.15 ppb uncertainty in 2000.
On the other hand, the most accurate possible electric-field-type measurements are done with a gadget called a calculable capacitor. This is a special shape of variable capacitor where the relationship between the position of the plates and the capacitance can be calculated exactly. For realistic dimensions, the change in capacitance is about 0.5 pF. By measuring that change with high accuracy, the electrical standards can be linked to (and thus the current SI definition of the ampere).
More information:
However, the current experiments are struggling to break the 10-ppb uncertainty barrier. This is the limit at which anything defined in terms of the permittivity of free space can be measured.
So, given two domains, one with 0.15 ppb uncertainty, and the other with 10 ppb uncertainty, with a link of 0.32 ppb uncertainty, which side should we place the definition on, and which should be derived?

If we were to keep the current fixed , which can only be linked to and to 0.32 ppb, then we'd need to continue using "conventional" values of the latter for sub-10−10 electrical measurements.
If I had my choice about how to define things, I'd define the metric system in terms of Planck units, which would keep fixed. But unfortunately, they all depend on the gravitational constant, which is only known to 47 ppm = 47000 ppb. (talk) 23:39, 16 January 2017 (UTC)

──────────────────────────────────────────────────────────────────────────────────────────────────── On measurements using defined as exact, it would be effectively determined from measurements of and : there is no reason to require direct a more direct realization of the constant.
I guess the adoption of and is indicative of the difficulty of tracking uncertainties, what I rather irreverently termed "sloppy thinking", and I guess this carries your argument. But I think the article giving a traceable, notable argument advanced by CIPM would still make sense; this is only the talk page. I don't have access to the reference I cited; I was hoping someone could paraphrase its argument.
I also like a definition in terms of physical constants used in Plank units, but excluding (it does not enjoy the theoretical status of aspects of the Standard Model aside from the accuracy issues). A nongravitational constant can replace . Is there not some other suitable physical constant that does not rely on gravitation or a system as complex as an atom for its definition? Examples include the cosmological constant and , though these have similar issues. —Quondum 01:32, 17 January 2017 (UTC)