THS4521: output is not equal to input

Your schematic shows an ADA4940 - indeed drop in to the THS4521, but is the issue with the 4940 or with the 4521. Keep in mind this is an FDA, so the output is a combination of the common mode voltage (Vcc/2 in your drawing) and plus and minus around that - the top input side goes negative from Vocm and the other side goes the same amount positive. 

Hello,

 As Michael mentioned, your VOCM is set to mid-supply in your circuit. This means the DC common mode at each of the FDA's output is 2.5V, and your DC value will add on one end and subtract on the other end, but your differential measurement will be the same as your input DC since the common-mode will subtract out. 

 Vout+ = (VN*(Rf+Rg)/(Rg)) – (Vin-*(Rf)/(Rg))

 Vout- = (VP*(Rf+Rg)/(Rg)) – (Vin+*(Rf)/(Rg))

 Vod = Vout+ - Vout-

This E2E thread has a PowerPoint with simulations to analyze this equations visually. 

 For example, at an input of 0.530V as shown in your picture,

Your +OUT pin will read 2.5V + 0.265V = 2.765V

Your -OUT pin will read 2.5V - 0.265V = 2.235V 

Differential Output will read 2.765V - 2.23V = 0.530V

The voltage directly at VIN+ and VIN- of the FDA is a value determined by VOCM pin divided down by the RF/RG resistors in combination of the input voltage also divided down. The voltage will be 1.38V at both inputs. The nodal analysis (0.530-1.38)/1000 = -0.00085A which matches your measurement. 

Even though I said 1.38V at both inputs since amplifier should keep the voltages at its input the same when operating in a linear range ideally, the FDA will not work properly or in a consistent manner when the common-mode voltages at the inputs are not the same. In your case, you are setting one DC at 0.530V and the other at 0V. The TI Precision Lab video series on FDAs will explain what happens in this situation: https://www.ti.com/video/series/precision-labs/ti-precision-labs-ti-pl-amps-fully-differential-amplifiers.html

Please click on the link, and watch the video titled Input, Output, common-mode and differential swing. I would suggest watching the full video, but the important information regarding difference in common-mode voltages at the input is discussed around the 6:15. In general, it shows KLC nodal analysis calculation with single-ended input as discussed in example above, and mentions that a common-mode different at input will cause an error since amplifier will amplify the difference between these two common-modes. 

Also, another error contribution could be coming from your DC power supply. These equipment have an internal impedance, this will cause a difference in impedance between the two paths of the amplifier, which will cause asymmetry in the paths, leading to an offset error. Adding a buffer like you did in your debugging process will resolve this error since now both paths of the FDA is symmetrical in impedance since now the input signal path will now see a relatively low to zero source impedance at DC. 

Thank you,
Sima