kelvin varley divider schematic or theory info ?

I uploaded a part of my Fluke/Philips training manual considering your question.

You can find it in alt.binaries.e-books.technical, named

"kelvin varley divider theory.pdf"

Succes and greetings from Peter in Holland

Make that _1-bit_ ;-)

...Jim Thompson

The K-V (in conjunction with a Western Cell) was usually used as a precision calibrator for the voltmeters of its time... the best of which was probably a slide-back voltmeter using a linear potentiometer. There would be virtually nothing available then to verify the linearity of a K-V over the whole of it's 4-decades.

It is therefore important to be able to state a guaranteed accuracy of a K-V, from nothing more than the worst case combinations of 11 resistors, matched to within specified tolerances.

For simplicity on the top decade I took the case where the total series resistance from Vin to Vout was 0.01% low and from Vout to 0v it was 0.01% high. This included the effect of the end-end resistance of the next decades on two resistors. This is a reasonable worst-case to take for quick calcs.

BTW: I do remember matching the R's for the top decade and then putting 5 in // as the Rref for the next decade, and so on. Matchings of the second decade (to Rref) was probably within 0.05%.

You've raised my eyebrows. Let us in on it.

In article , Tony Williams wrote: [...]

That statement is not true but point taken.

It's often true that dimensionless measurements can be done to arbitrary precision with just a lot of patience. Things like matching resistors and resistor ratios, matching sets of weights, lengths, etc.

John

If you go to the effort of getting 0.01% resistors for one, you then have the instrument needed to check the linearity of one with 0.1% resistors, but lets say you don't.

Assume things are battery powered. Consider this case.

Unit B Unit A --------------------- ! ! V+ -+- -+- ! ! / / \\\\ / / \\ \\ ! ! V- -+- -+- ! ! ---------------------

Unit A can be set to its mid span and Unit B can be adjusted until the meter nulls. If we swap the supply leads of Unit A, ideally the meter would still be nulled. The meter will indicate the offset of the center point of Unit A from the ideal center point.

If you don't want to use a meter because it is not accurate enough, you can add a unit C, with its own battery, floating on the wiper of Unit B.

This same unit C comes into play in getting the 0.25, (0.5-0.25), (0.75-0.5), (1.0-0.75) differences.

Now we can add unit D and really go wild.

Flatness, too. You just rub three things against each other.

John

What's the story there?

Ha! But seriously, it's an old technique. Term being "lapping". Add some grit and slide and swirl plate A on B, then B on C, then A on C and repeat. Any two surfaces can meet in any smooth curve (depending on how you've been grinding it), but three together can only meet flat. If, for example, A is convex and B is concave, only one can fit into C's curvature, whichever direction it may be. Statistically, by rotating and grinding the plates like this, the curvature drifts towards a perfect flat.

Once you have the flat (or any other shape for that matter), you can add ink or paint and press another surface onto it, and grind, scrape or machine off the indicated (high) spots where the ink transferred. By repeating many times while reducing the thickness removed after each spotting, you can get a near-perfect match in the curvatures.

Variations of 50 microinches over several feet area aren't unheard of for surface plates (just don't breathe on it, else you'll expand that area from the heat!).

Tim

-- Deep Fryer: a very philosophical monk. Website:

In article , John Larkin wrote: [...]

Even the old method for grinding lenses by hand falls into this class. The method ensures that the shape must be part of a sphere but not the radius of the sphere.

Go away you foolish troll.

The Kelvin-Varley's cool idea is that it does parallel the section, which is cool because switches to swap the section would be painful.

No. K-V voltage dividers are always resistive because they must work correctly at DC.

It does not "parallel" it swaps out the element with the next lower divider.

In article , Joseph2k wrote: [....]

Just because you said that, I'm going to make one out of capacitors to only work with AC. The advantage would be that the capacitors won't make much heat so the self heating tempco issue can be ducked.

In article , John Larkin wrote: [.... me ...]

Yes, that's just a sphere with a very large radius. You can also match lenses by alternating the grinding between the pairs.

In article , Tim Williams wrote: [...]

I believe the term "lapping" overlaps with the term "grinding" but not completely. The method called a "pitch lap" would not be refered to as grinding. The process of getting the gross shape correct with the large grit would not be refered to as lapping.

Have you seen it otherwise?

I have seen it otherwise many times, for example in the book "Grinding and Lapping" somewhere on one of my shelves, as well as several books on amateur telescope making, "Optical Production Technology" by Horne, and common usage in industry in the days when I was lapping mechanical seal parts as close as 1 microinch in 6 inches. Grinding is done with a solid abrasive wheel or other solid abrasive tool, and lapping is done with loose abrasive particles or loose abrasive embedded in the surface of a tool. Abrasive particle size is irrelevant as to what kind of process it is except when fine abrasives are used with a pitch tool the identical process is generally referred to as polishing.