current sense singularity

[itdaniher]

The MAX9919 has a mathematical singularity at VCM = 2V, causing a nonlinear offset in current measurement.

Most of this should be solvable in software with an if.. tree for common mode voltage dependent offsets.

This problem only presents itself as a functional impairment sourcing current with a common mode voltage of almost exactly 2V.

This problem will be resolved in the next hardware iteration of the CEE.

[itdaniher]

This issue stems from an unclean transition from the low-range level shifter to the high-range level shifter in the MAX9919 family of parts.

For future reference, possible replacements must...

• offer a common mode voltage range extending at least from ground to VCC
• offer bidirectional measurement
• take an external reference voltage to set zeropoint
• provide a high common-mode range rejection and a stable gain at frequencies up to 50KHz

Based on these guidelines, it looks like the AD8210 is one of the more suitable replacements, but comes in at 4x the price of the MAX9919.

[itdaniher]

Changes implemented in the AD8210 branch.

It looks like a lookup table spanning at least 1v..3v might be the best solution for CEE V1.0.

[darchons]

Maybe you can use a differential amplifier like this? http://en.wikipedia.org/wiki/Opamp_applications#Differential_amplifier

[itdaniher]

The 9919/8210 are specialized differential amplifiers offering a high common mode voltage range with a setpoint for zero.

Current sense is just a subset of differential amps.

[darchons]

That's true, but IMO for Vcm from 0V to 5V, a regular opamp will be cheaper and can potentially perform better. To set the zero point you can just connect Rg to your reference instead of ground.

[itdaniher]

After running some numbers, a difference amp built from a decent opamp and resistor array would be cheaper than the AD8210, but not the MAX9919. Thanks for the suggestion, I figure I'll give it a shot.

[itdaniher]

This sweep highlights the issue. Channel A is unloaded.

[itdaniher]

Solution implemented in V1.1 branch at 9963531.

This remains an issue for V1.0, but is not noticeable during normal use.

[itdaniher]

The discrete CSA in V1.1 has a common mode error of up to 10% of full scale voltage when made with 1% parts. Higher precision parts are considerably more expensive.

The LT1999-50 was explored, but has a large DC offset and nontrivial common mode gain error.