Moving an analog signal to a shaky ground

[kender]

Colleagues,

Could you provide some ideas, insights or typical solutions for the following problem. I have 2 ground “domains” in my system. They are ohmically connected, but due to currents and resistances in my system, a potential up to 150mV can appear between the grounds. I would like to bring an analog DC signal which is generated on one ground to the other ground. E.g. 1V with respect to GND1 it will always be 1V with respect to GND2, provided that GND1 and GND2 are no further than 150mV apart.

At the moment I can see 2 ways of doing this: a differential amplifier or a current mirror. What could the other options be? How is it typically done?

Cheers,
- Nick

 

[FvM]

The second variant can be phrased more generally: using a current source on output. It's not limited to current mirror, also a differential pair with internal feedback resistors is an option, most likely with higher linearity.

To choose a suitable structure, the bandwidth and linearity requirements should be known.

 

[kender]

To choose a suitable structure, the bandwidth and linearity requirements should be known.

Bandwidth is very small (DC, as I wrote above). Linearity, let's say, 12-bit.

 

[pauloynski]

An isolation amplifier like ISO124 could also work.

 

[flatulent]

Another method is to use transformers on each end. The signal at each end is ground referenced and between the two it is differential. This has the additional advantage that the grounds are not connected and so there is no common mode current between them. There is little danger of damaging the electronics on either domain.

The above assumes that the signal is AC and is in a frequency range that a transformer can handle.

If the signal is near DC, chop the signal to make it AC. Then transport the signal and the chopping signal over two transformer coupled lines. Rectify the signal with the chopping signal at the receiving end.

Alternately, use a VCO to change the LF signal to an AC signal, then use a frequency to voltage converter at the receiving end to get the original signal.

A further idea is to use RF to transmit the signal as a modulation on a carrier or to use an ADC to convert the signal to digital and then a DAC at the other end to convert it to analog. The transmission could be by RF which totally eliminates any safety problems of standoff voltages or inconvenience of stringing wires between the two locations.

I hope this helps.

 

[dick_freebird]

I have been using instrumentation amplifiers to buck
moderate and variable ground offsets, with pretty
good results, for my wideband O2 meter tuning cars.
The AD620 and a couple of 9V batteries let me float
the signal across domains and the no-external-parts
for A=1 is nice. You might be able to get what you
want without the isolated supply, but you will need
voltages above and below your max signal swings.

 

[kender]

At the moment, I'm leaning towards a diff OpAmp circuit like the one below. The constant current driver on the right side is the receiver of the analog signal.

 

[tdy]

have you tried a optocoupler?

 

[kender]

have you tried a optocoupler?

No. I don't want to try it. I'd like to have something simpler.

 

[pauloynski]

Modern optocouplers can achieve very good linearity (0.01%).
**broken link removed**

 

[mandar_mahajan]

Grounding is always critical issue when it comes with analog & digital both.
U can reduce ur 150mv bet'n these two grounds.
1) Avoid looping of ur ground tracks.
2) See which component for ex LCD consumes max current, cut the ground track to it which certainly comes from other components, & give its direct connection to ur main ground.

3) All depends upon ur schematic, layout & tracking on ur PCB.
4) what is the source on ur pcb, means IC 7805,lm317.......
5) If speed(switching) is not required u can use OPTO-ISOLATOR PC817 which is also very cheap for ur analog portion(ADC/DAC/or whatever).

Regards
Mandar

 

[FvM]

For your specified requirements, the simple 1-OP differential amplifier as shown should be sufficient, apart from the interesting suggestions that have been made for opto-isolators and other more complex solutions.

The most important performance restriction for DC and low frequencies is the finite resistors matching. You can calculate, that 1% mismatch possibly jeopardizes 12 Bit accuracy. You get much better results with matched resistors arrays or differential/instrumentation amplifiers with internal (trimmed) resistors.

The problem could be also solved by moving the current driver to SGND, by the way. But it may be unwanted for other reasons.