kelvin varley divider schematic or theory info ?

Tony Williams wrote in news: snipped-for-privacy@ledelec.demon.co.uk:

You could salvage them from surplus Fluke Differential voltmeters. There's a lot of them in circulation.

--
Jim Yanik
jyanik
at
kua.net

Hmmm- it is basically a DAC. LTC has nice 50-page app note (AN-86 A Standards Lab Grade 20-Bit DAC with 0.1ppm/°C Drift) with some historical asides:

[snip]

See "Test-K-V-DAC.pdf" on the S.E.D/Schematics page of my website.

This is incomplete, having been published while the project was in progress, thus restrained by NDA.

Since the company has now been seized by the bean-counters and all technical capability destroyed, I'll now post a more complete version... sometime next week.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
 Anyone can be rude, but it takes a Democrat to be a real dirtbag.

The 10000 number is wrong. I argue as follows:

Imagine that the bottom most R1 is 0.1% too high.

When the switch first switch is rotated fully CCW, We have this:

Vin ! / \\ / 9R1 \\ ! +------+----- V2 ! ! \\ ! / R1 / 2R1 \\ \\ ! ! / ! \\ ! / KR1 ! \\ ! ! ! GND GND

V2 = Vin*((R1+KR1)*2R1/(R1+KR1+2R1)) / ((R1+KR1)*2R1/(R1+KR1+2R1) + 9R1)

To make the math easy lets use 1 ohm resistors and assume that K=1.001

V2 = Vin*((1+1.001)*2/(1+1.001+2)) / ((1+1.001)*2/(1+1.001+2) + 9)

V2 = Vin*((2.001)*2/(4.001)) / ((2.001)*2/(4.001) + 9)

V2 = Vin * (4.002/4.001) / (4.002/4.001 + 9) V2 = Vin * 0.1000225

When the switch is moved one place CW, we have:

Vin ! / \\ / 8R1 \\ ! +------ ! ! \\ ! / R1 / 2R1 \\ \\ ! ! / ! \\ ! / R1 ! \\ ! ! ! +------+--- V2 ! \\ / \\ KR1 / ! GND

V2 = Vin * KR1/(KR1 + 9R1)

V2 = Vin * 1.001/10.001

V2 = Vin * 0.10009

So we have an ABS error of:

Vin * (0.10009 - 0.1) = 0.00009

and a step error of:

0.00009 - 0.0000225

One of the really nice things about this circuit is that errors in the resistors tend to get deleveraged.

Nicer drawing than I could do left:

--
--
kensmith@rahul.net   forging knowledge

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

You can also use a many digit DVM to add high value trim pots in parallel with the right resistors to get 5 digits out of 1% resistors. It just takes a few hours of soldering and fiddling.

You can use a 10 turn POT at the last section.

--
--
kensmith@rahul.net   forging knowledge

Somebody, National maybe, made a dac that had a main resistor string, with a bunch of taps. Mux switches fed (I think) two followers, which in turn drove the ends of a second tapped string, so the second resistor string interpolated between the big jumps of the first string. It was sort of a K-V, except for the buffers and the fact that they "leapfrogged" when climbing the string. It was inherently monotonic.

John

Yep. That's what I have, except no leap-frogging, I think... it's been awhile ;-)

I do remember such a paper when I was researching this but, IIRC, this case only required resolution.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
 Anyone can be rude, but it takes a Democrat to be a real dirtbag.

[snip]

I think you have accidentally gone up a blind alley.

Take the case when the divider is set to 5000, and assume that the bottom three decades are perfect, so that the load on the MSD is exactly 2R. The top decade circuit then reduces to this......

Ra Rb Vin+---/\\/\\---+---/\\/\\---+0v \\|/ Vout (= nominally 0.5Vin)

Ra and Rb are nominally equal in value, but suppose you accidentally put all the high-tolerance resistor selections into Rb and all the low tolerance into Ra. For a 1 in 10000 matching tolerance, Vout will run at 0.5 in 10000 high...... that's half the divider's overall tolerance consumed in the top decade.

--
Tony Williams.

It's there, just slow as molasses. Looks like a piece of the east-west network is hosed... generally that means Cox :-(

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
 Anyone can be rude, but it takes a Democrat to be a real dirtbag.

Thanks. One clever "feature" of this architecture is that it's monotonic to any number of bits even if its absolute and differential linearity are ghastly. So its transfer function can be, well, interesting.

John

In a lot of applications it doesn't matter if its transfer function is "ghastly".

My approach used a very simplified version of K-V, but was contained within a comparison loop, so that the gm-C filters were tuned quite accurately... the DAC just afforded "memory" for the filter set point.

I don't remember the interval now, but the filters were re-tuned often enough to compensate for temperature drift.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
 Anyone can be rude, but it takes a Democrat to be a real dirtbag.

AD569, yes, but AD7569, no. :-)

--
 Thanks,
    - Win

Your website is down....

Hey- there was an Design Idea not too long ago that did that exact same thing with the digital pots.

[...]

I think not but lets see.

For a two stage circuit it will really be:

4AR1 BR1 CR1 5DR1 Vin -----/\\/\\/--+---/\\/\\/---\\/\\/\\--+----/\\/\\----0V ! ! ! ER1 FR1 ! ---/\\/\\-+-/\\/\\----

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

I think it was an early 16 bit DAC that did this. As I recall, it drew a lot of current off its reference input pin. The unit to unit spread on the reference current was huge. I don't remember it as being National. Analog Devices and Harris come to mind.

--
--
kensmith@rahul.net   forging knowledge

ADI. Patented by Peter Holloway. The first part using this trick was the AD569 IIRC, it's also been used in some others since that time (AD7569, perhaps? the numbers escape me).

Steve ADI employee by day...

Hi Pjotr,

I only find group "alt.binaries.e-book.technical"

where "e-book" is singular.

Is that where you posted it?

...Jim Thompson

-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | Phoenix, Arizona Voice:(480)460-2350 | | | E-mail Address at Website Fax:(480)460-2142 | Brass Rat | |

| 1962 | Anyone can be rude, but it takes a Democrat to be a real dirtbag.

In "alt.binaries.e-book.technical"

"book" singular

13.1MB for 14 pages :-(

Downloaded it, but it has that "dirty" look of being scanned multi-bit/grey scale.

Go back and re-scan, choose 2-bit, or line drawing, or text... whatever causes your scanner to do only black/white.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
 Anyone can be rude, but it takes a Democrat to be a real dirtbag.