Buffer for Kelvin Varley Divider?

[VintageNut]

I have a General Resistance KV divider of 7 decades. I would rather take readings directly than use a bridged configuration.

The ideal output buffer would be an electrometer grade op amp configured as a non-inverting buffer. The problem is that these buffers seem to have 10 microvolt per degree temperature change.

Has anyone implemented such a buffer successfully for use with a KVD?

[Vgkid]

Why, most meters have a 10Gohm input impedance. Wont matter much.

[VintageNut]

The output impedance of the KVD is measured at about 25k ohms max. Using a 10G input impedance this creates at 2.5 ppm error. The error is dependent on the KVD setting.

The KVD has 0.1 ppm resolution. Using a DMM without a buffer wipes out the 7th decade and two counts of the 6th decade.

The error can be observed using Fluke 332A.

[Marco]

AD8638? Auto-zero, so negligible tempco for offset, with below pA typical bias current.

[VintageNut]

AD8638? Auto-zero, so negligible tempco for offset, with below pA typical bias current.

Thank you for the suggestion. My reading of the datasheet reveals

Typical input bias current 1.5 pA, maximum 40 pA.  Not low enough
Input voltage range -0.1V to +3V  Not practical as a buffer

Please let me know if I missed something.

[edavid]

Jim Williams used an LTC1152 and LT1010 to achieve 0.4ppm error: http://cds.linear.com/docs/en/design-note/dsol11.pdf

[VintageNut]

Jim Williams used an LTC1152 and LT1010 to achieve 0.4ppm error: http://cds.linear.com/docs/en/design-note/dsol11.pdf

Perfect! Thanks.

[VintageNut]

Jim Williams used an LTC1152 and LT1010 to achieve 0.4ppm error: http://cds.linear.com/docs/en/design-note/dsol11.pdf

The datasheet for the LTC1152says that input bias current is +/- 10 pA typical and +/- 100pA max. I wonder if he ever evaluated the effect of the LTC1150 input bias on his KVD?

[edavid]

The datasheet for the LTC1152says that input bias current is +/- 10 pA typical and +/- 100pA max. I wonder if he ever evaluated the effect of the LTC1150 input bias on his KVD?

It's right there in his error budget.

[VintageNut]

The datasheet for the LTC1152says that input bias current is +/- 10 pA typical and +/- 100pA max. I wonder if he ever evaluated the effect of the LTC1150 input bias on his KVD?

It's right there in his error budget.

Ok. I see it. Thanks for pointing it out. His KVD must be 10K ohms output to have 0.1 ppm error at 10 pA bias. That would be 0.25 ppm for my KVD. I guess obtaining some samples and measuring will be in order. It will be a good use of the 236 SMU.

[Marco]

Typical input bias current 1.5 pA, maximum 40 pA.  Not low enough

You're not making electronics for an automotive environment here ... specs over temperature range are irrelevant.

Input voltage range -0.1V to +3V  Not practical as a buffer

You read the specs for 5V single ended supply, not the 16V supply (also you should feed it ~-2 to +14 to get better performance around 0).

The relevant bias current information for you is figure 14 page 7 (0.1 pA typical at 25 degrees Celsius). Worst case is irrelevant, test the IC and if it doesn't perform well enough try another till you get a typical one.

[splin]

Jim Williams used an LTC1152 and LT1010 to achieve 0.4ppm error: http://cds.linear.com/docs/en/design-note/dsol11.pdf

Wonder why he chose the LTC1152 rather than the LTC1052C? Also interesting that the current datasheet specs are considerably worse than listed in that app note:

LTC1052C

LTC1152

LTC1152 from Design Solution 11

Bias current (typ)   @ 25C

1pA

10pA

Bias current (max)  @ 25C

30pA

100pA

50pA

Bias current (max) over Temp

175pA

1000pA

Vos (typ)

0.5uV

1uV

Vos (max)

5uV

10uV

5uV

Vos drift (typ)

10nV/C

10nV/C

Vos drift (max)

50nV/C

100nV/C

50nV

0-10Hz Vnoise (typ)

1.5uV p-p

2uV p-p (0.1-10Hz)

CMRR  (min)

120dB

115dB

110dB

CMRR  (typ)

140dB

130dB

Avol (min)

120dB

110dB

140dB

Avol (typ)

150dB

130dB

[EDIT] Moved misplaced 50nV entry

[VintageNut]

Typical input bias current 1.5 pA, maximum 40 pA.  Not low enough

You're not making electronics for an automotive environment here ... specs over temperature range are irrelevant.

Input voltage range -0.1V to +3V  Not practical as a buffer

You read the specs for 5V single ended supply, not the 16V supply (also you should feed it ~-2 to +14 to get better performance around 0).

The relevant bias current information for you is figure 14 page 7 (0.1 pA typical at 25 degrees Celsius). Worst case is irrelevant, test the IC and if it doesn't perform well enough try another till you get a typical one.

Which datasheet has input bias of 0.1 pA typical?

[Marco]

http://www.analog.com/media/en/technical-documentation/data-sheets/AD8638_8639.pdf

page 7 figure 14