I have two circuits in two different ground domains GND1 and GND2. These grounds are shorted together at the power supplies which are far away. Domain 2 is a high power system so the voltage drop across the two grounds can be as high as 100 mV and depends on the absorbed current, which is variable. The signal from domain 1 has to be connected to domain 2. (see schematic. OPAMPs are not TL081s though) The bandwidth of the signal is about 20 MHz and connection must have no latency, so optocouplers are excluded and INAs are too slow. I thought a solution like the one in the schematic, a simple difference amplifier. The current from domain 1 to domain 2 will be in the order of 100 uA so DC current is not an issue, but I am concerned about the medium-high speed current, in the range of 100 kHz - up. To go back from GND2 to GND1 it has to travel back to the power supply which will be an antenna, and a probably big inductance. The only thing it came to my mind is to connect locally the two grounds with a cap, 1 uF, to short them at medium-high frequency. Is this a good idea or is there another solution I don't know?



simulate this circuit – Schematic created using CircuitLab

**My big concern is having net current from system 1 going into system 2, either via Vdd1 -> GND2 or GND1 -> GND2. An INA architecture would probably be the best solution, with GOhm input impedance I'll have neglecting current going through. I can surely make an INA out of three opamps but I was looking for an already integrated solution with only one IC and couldn't find high speed INA

  • 1
    \$\begingroup\$ bad idea. ** You cannot expect clean signals with 20MHz BW and 10k loads** It also has poor CM gain and there is DM gain on GND noise. Better use shielded twisted pair and use controlled impedance differential amp that is stable with x1 or x2 gain and use ferrite Balun around cable for improved CMRR or SMT 20MHz balanced twin coil choke. Even using an Ethernet (Balun) port with controlled impedance Analog driver & receiver loads would be better using CAT5 cable \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Oct 12 '16 at 14:51
  • \$\begingroup\$ I forgot to mention the two circuits are really close together, 2 cm max. \$\endgroup\$ – user126418 Oct 12 '16 at 15:00
  • \$\begingroup\$ then why 2 gnd domains so far apart? \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Oct 12 '16 at 15:15
  • \$\begingroup\$ Don't have an answer to your original question, but I would suggest that you use LVDS signalling. I also suspect that this is one of those cases where if we knew the full story, there would be some better way to do whatever you are doing. But LVDS should solve your immediate problem, and allow your two circuits to operate with 100mV GND difference. \$\endgroup\$ – mkeith Oct 12 '16 at 15:17
  • Gain of noise on GND1 = -1/(1+10) = -1/11
  • Gain of noise on GND2 = 10/11 * ( 1+ 1/(1+10))= 120/121

  • Thus GND differential gain is mismatched

| improve this answer | |
  • \$\begingroup\$ But the CM noise on GND1 is also on the actual signal which has the same gain. \$\endgroup\$ – user126418 Oct 12 '16 at 16:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.