How High a Source Impedance can be to Opamp Input?

[Mechatrommer]

i dont think "current probe" is a good idea to be included in this study right now. they are... different "beast", it will make my life more miserable. yeah, it should be general purpose high side measurement offline/mains circuit. high accuracy is not critical, since as oscilloscope, its not high accuracy wave representation. but i'll try struggle for higher standard and to be able to measure things like power bjt/mosfet Vgs etc in the range of 0.3V differential, lets see if its achievable at hobby/noncertified/limited resources engineering level. those blue boxes in the pic above are trimmers, and dont ask about components used, they are 1% and worst. but from the testing, i believe the critical part lies (as you said) at the input attenuator (ie the voltage divider, and probably at 2nd stage, ie before differencing amp) impedance matching. both signal get phase/magnitude shifted at high freq (from observation), compensating them and match them under varying condition is.... difficult. asking such question in this forum will invite to "dissapointmentville". while searching for info time ago, even our own good Dave Jones admitted it on another forum.

yeah its kind of disturbing i have to attenuate all AC and DC in one go, but not as much disturbing as splitting them into separate channel. i tried to make my life less miserable, with addition that if there is such 100MHz at fullswing diff of mains voltage, then my ins-amp will be able to measure that easily! we may call it as... feature and ofcourse for smaller diff signal at high CM, we can implement less aggressive attenuator and AC coupler upfront, thats i believe... easy? when things i've asked are settled.

[ejeffrey]

and ofcourse for smaller diff signal at high CM, we can implement less aggressive attenuator and AC coupler upfront, thats i believe... easy? when things i've asked are settled.

Keeping in mind that my frame of mind is usually to go for the highest precision you can afford rather than 'good enough', my solution here would be a small floating power supply biased to the HV supply rail or to the common mode voltage of my input signal.  Then I would use a normal instrumentation amp with the appropriate gain.  Now I have a nice signal and I just have to get it down to ground.  I can do that easy by capacitively coupling it or if I need the DC component I can use a circuit like yours as a level translator.  Even though I am still using the instrumentation amp with the attenuator, I win on one, possibly two fronts.  First off, I have added some differential gain before the attenuator network.  This improves both the SNR and the CMRR by the gain of the floating amplifier.  Second, if I can choose the common node for my floating supply to be something that doesn't have any high frequency ripple relative to the "zero volts DC", my common mode signal for the second stage amplifier is only at DC/low frequency.  This makes it a lot less susceptible to phase shifts in the attenuator network.

Anyway, that is what I would do if I wanted to measure millivolt differential signals on top of a 330 volt supply.  If you are measuring a volt or so I think your circuit will work fine. Again, if what you really care about is the current and you are measuring a sense resistor, then a current probe will give you better performance.

[free_electron]

right, explain this time , COMPLETELY, what it is you are trying to do. it almost looks like you are after a PLC-modem kind of system ( power line carrier ). you want to pick up the datapackets off the power line


So , you don't need high dynamic range as you said before, you just need high CMRR since you are trying to pick up a small signal superimposed on a high common mode signal.

[Mechatrommer]

right, explain this time , COMPLETELY

beside the small diff signal ejjef explained, i also need it to be able to measure full swing oscillation of mains somewhere, such as smps transformer primary winding input, Vds of a mosfet etc etc, pretty much every unknown nodes in offline circuit unknown swing magnitude signal. it also should be able to measure large signal on floating or high side grounded circuit (offline or not offline) as i said, general purpose.

thanks ejjeff for the input, i will keep your advice in mind. infact i already have in mind that... if necessary, i will separate the attenuator from the ins-amp, just like agilent differential probe. so, ins amp is another part, attenuator head is another story. different attenuator for different signal measuring purpose. i'll start with this "all in one go" attenuator head first, if somehow its possible, i may not need the other specialized ac coupled head.

i've read about this floating psu, or in reference to high side of a circuit, but well... thats another problem to be solved. the psu for this project is already floating, but since the scope's ground (earth) is connected to this psu's ground, so effectively it is grounded i read in AoE demanding 2 separate psu for both ins-amp stage (one floating one not), well, thats extra psu and study is needed, i'll try it last when i get this grounded psu cmrr right.

[Mechatrommer]

as per free_electron's comments on input bias current. i've made some calculation for the effect of Voffset, Ioffset and Ibias. my finding is that this is not AC related error?, its in the form of voltage offset error, which i think can be "easily biased"? from the formula below, assuming opamp's input impedance infinity (ignoring opamp input impedance, it seem doesnt matter when looking at 1Kohm source from 1Mohm input opamp), and assuming circuit is working in nonreactive region of Zs (external resistors), the plot in second picture, shows this offset error. though there's 1200% error at low voltage, but its not as bad as it sounds i think :-|

summary from the formula:
-Voffset will be affected (exagerated) by opamp (closed loop) gain.
-error generated by Ioffset at noninverting input will be affected by opamp gain, and source impedance (including the divider)
-error generated by Ioffset at inverting input will be affected by feedback resistor.
-they all accumulated to generate DC voltage offset. which i believe can be corrected with voltage bias somewhere.
-since In error is negative term, it can be made to cancel the other 2 positive error terms, ie caused by Vio and Ip
-POM = point of measurement. loading effect is also studied, but... at nonreactive region only

FWIW. i stand to be corrected.