Current-sense amplifier to voltage output

[jolshefsky]

At work, we use panel meters in our product. I end up getting a fair number of "bad" meters with things like a flickering decimal point or weak segments that I figure I can use. They are all 3.5 digit, 5V driven, and accept ±2V full-scale. I thought it would be neat to use them as volt meters (which is trivial) and current meters (not).

I figured I could use a current sense amplifier and picked up a few MAX9928 high-side bidirectional current-sense amplifiers. They are designed to interface to a microcontroller like in the typical circuit below, but I wanted to get a positive and negative output, so I used a digital switch to listen to the SIGN output and toggle between an inverting and non-inverting amplifier (I devised it without assistance and think it's pretty clever, although I'm sure it's been done before).

Unfortunately the output drifts all over the place, varying by ±0.1V at <2Hz seemingly randomly (next step is to hook up the 'scope). I am making a simple virtual ground and trying to sink the output current (from a current mirror, see the datasheet internal circuitry, below) to that virtual ground rather than to the 0V of the main supply. Could using a virtual ground be causing my problem? Is the virtual ground unstable?

The datasheet doesn't offer an answer, only to say that the output voltage can swing as high as VCC-0.1V, so I figured I was safe to use a virtual ground at approximately VCC/2. There aren't application notes that talk about using this chip. One option, I know, would be either to design the circuit as in the typical circuit and level-shift the resulting voltage, or to run the MAX9928 from the the virtual ground to VCC, but the chip is only specified to 2.5V, and that's precisely where +5V to +2.5V lay. But I'd like to be sure the design is the problem first.

[HAL-42b]

The bad panel meters are usually based on ICL7107. They need a negative supply in order to be able to measure down to 0.00V.

In the cheap Chinese ones that negative voltage is derived by abusing a shift register by making it operate as an oscillator. This is then fed to a diode and a leaky cap. Very noisy. This may be the cause of your drift.

If you can supply a clean negative rail it probably won't drift as much and you can use an ordinary current shunt.

[jeroen74]

Take a look at the LT1999 and the LTC6104.  These allow you to set the zero current voltage output at (almost) any voltage you like: negative current will give an output lower than this reference voltage and vice versa.

Your panel meter would be supplied with this reference voltage (1V sounds reasonable) on one input and output of the current sense amplifier on the other.

[jolshefsky]

Take a look at the LT1999 and the LTC6104.  These allow you to set the zero current voltage output at (almost) any voltage you like: negative current will give an output lower than this reference voltage and vice versa.

Your panel meter would be supplied with this reference voltage (1V sounds reasonable) on one input and output of the current sense amplifier on the other.

Thanks. I'm working with the LTC6104 and it's greatly simplified and improved the design. Curiously, some key parameters change from the Maxim chip. For one, the output current on the LTC6104 is designed to peak at 1mA rather than 250µA like the MAX9928. For another, the LTC6104 derives its power from the current source while the MAX9928 uses a separate supply. Finally, the application note design for the LTC6104 suggests a fixed gain of 100V[out]/V[sense resistor] while the MAX9928 datasheet suggest 50V/V—not a big deal in either case to get 100V/V or 50V/V, but at least it made me understand the LTC6104 design better!

[dannyf]

'bidirectional current-sense amplifiers. '

You can use two uni-firection amps wired back to back. Only one of them will provide a non zero output at any given time.

Thus, you can process the sign in software.

' (I devised it without assistance and think it's pretty clever, although I'm sure it's been done before).'

Without seeing it, it is hard to see what is wrong.