Measuring voltages in -48VDC systems.

[cerickson]

I am hoping to have a ZX-1280n development board doing data logging and control stuff in a remote communications site and in most cases, it will be interacting with +12VDC powered and controlled negative ground systems. However I will also need to measure the -48VDC radio battery plant (positive ground) voltage as well as bidirectional current flow via a shunt and doing calculations for the radio battery plant depth of charge.

1. What would the best way be to measure the -48VDC battery plant (up to 65VDC when charging) while avoiding ground loops? I need 1/10th volt precision at minimum.

2. What other problems should I look out for?

-Christopher Erickson

[dkinzer]

What would the best way be to measure the -48VDC battery plant.

I've never attempted this before but my first thought would be to use a voltage divider to drop the maximum voltage down to something in the -5 to -10 volt range and then feed that to an op amp inverter (running on +/- supplies) with the gain set to yield a maximum voltage of +5 volts. If you then sampled the output of the inverter using a ZX ADC input, each step would represent about 0.064 volts (65 volts / 1024 steps).

If you needed better precision, you could add an external 12-bit or 16-bit ADC.

[cerickson]

A friend just suggested a single-chip voltage to frequency converter driving an opto-isolator to get around the ground loop problem.

I have no experience with such devices and wonder if anyone else does that might be willing to share their thoughts.

-Christopher Erickson

[pjc30943]

Voltage to frequency converters work well if the support components are stable (i.e. precision resistors, etc). By chance I have a similar setup where a V-to-f converter is optoisolated to allow a second system (at a different potential) to measure voltages of the first.

One way to measure current in both directions is via a small hall-effect type toroidal current detector, summed into an op-amp centered above ground (ex. 2.5V). Since they're hall effect (and thus not connected to the output) the measured system can be at any potential, for example your -48V. Some also already have outputs already centered at 2.5V, and swing both directions depending ont he current flow.

Don's suggestion is what I'd probably go with as well to measure the voltage: a divider, then op amp to invert, then center or scale as desired. There are probably more elegant solutions...

[cerickson]

I have been pointed to the Analog Devices AD7740, AD7741 and AD7742 chips and I would be interested in getting opinions from "the masters" here.

In my case it appears that they can be powered from the measured voltage via a regulator and have a very simple implementation. Internally it looks like they are implementing Doug's circuit design suggestion.

I really like the bidirectional current sensing solution using the toroid and Hall effect sensor too!

You guys are a great help.

-Christopher Erickson

[dkinzer]

I have been pointed to the Analog Devices AD7740, AD7741 and AD7742 chips

I haven't used any of those chips but my first impression after reviewing the datasheets is that only the AD7742 would be useful in your particular situation. It is capable of operating from a single 5 volt supply but it can handle input voltages in the range of +/-Vref. If you used the built-in 2.5V reference, that would allow you to use the -2.5V to 0V portion for measuring your supply voltage by employing a divider.

The question yet to be answered is whether you can get the 0.1V precision that you desire. That would require the ability to distinguish the output frequency change due to a 3.8mV change at Vin (0.1V * 2.5V / 65V). It appears that the maximum frequency change is 0.45 times the input clock so that would mean that your entire 0 to -65V range is mapped to 0.45F. If the input frequency were 100Khz, that would mean that the V-F transfer curve would be 692Hz/Volt using a 26:1 divider. Achieving a 0.1V precision would require the ability to accurately distinguish a frequency change of about 70Hz. Note that this simplified analysis ignores all of the error factors (divider accuracy, offset error, gain error, V-F transfer non-linearity, input frequency accuracy, frequency measurement error, etc.).

None of this should be taken as advice against using this technique or this particular chip. It may well be the best solution for you. You'll have to do some further analysis and test a prototype.

[cerickson]

Great advice Don!

I will be researching further.

-Christopher Erickson