analogguru said:
JohnRoberts said:
That differential does not present an equal impedance to ground in both inputs.
It will probably work adequately for modest length, properly shielded, low impedance feeds, but the only common mode it it rejecting is differences between the local and distant signal grounds, not any common mode noise picked up in between.
JR
That´s interesting, can you please explain this ?
I always thought at the inverting input is "virtual ground". Therefore the input impedance at the inverting input would be 15k.
At the non-inverting input we see 15k and 10k resulting in 25k input impedance to ground. In parallel to this 25k is a 39k resistor resulting in a total value of 15,26 k. Where am I misleaded ?
analogguru
Because of the negative feedback, the opamp's - input will look like a virtual earth, only if the + input is connected to real earth. In fact the opamps - input is clamped by NF to the opamps + input voltage. This means that the - input is not 15k to ground, but 15k connected to whatever voltage is present at the + input.
To analyze the circuit for CMRR think of the CM noise as a current injected into both wires. The current coming into the + input will develop a voltage in that line that is defined by the parallel impedance of sending output termination resistor and total input termination. Roughly 15k in parallel with whatever source buildout R is in the output line. The resulting noise voltage at the opamp + input will be 10/25 or 40% of the noise developed in + input line voltage. This means instead of the - input looking like a 15k to ground it looks like a 15k to 40% of the input voltage or effectively 25k to ground. Since the CM noise current sees 15k on one line and 25k on the other, the CM noise voltage is different so won't cancel completely. Also since those lines are cap coupled the LF pole wrt CM noise will be different, but since the 100uf is tuned pretty low the error because of that won't be huge at mains frequency and harmonics. If these poles were higher, this could be another error source.
For CM balance, removing your 39k to ground at the + input (and adding the same value buildout R in series with the sending unit ground as is in the hot side) would make the the input impedance similar for the CM noise. However because of the opamps bootstrapping the - input will look different impedance wise if + input has no input voltage (15k), or will look like a lower than 15k impedance (10k or so) with opposite polarity signal swing at + and - lines.
For short, low impedance, shielded runs this is not a big deal. For longer high performance runs, it is desirable to have the interface impedance balanced for all modes, normal and common mode, to generate less stray ground currents that need to pass between chassis and can contaminate grounds. In a proper balanced input with differential audio input, signal ground currents will also cancel out.
This is pretty esoteric stuff and only important for long, high quality interfaces. For short, shielded, single legged, interfaces like TRS insert points, a simple differential is more than adequate, while I would lose that 39k that is actually imbalancing it CM and not doing anything useful for noise rejection.
JR
