THS4531: Single ended to diff circuit -> AC simulation shows resonance/spike?

morning Anber

That feedback C is taking the noise gain to 1 at LG xover, low phase margin results - adding that same value across the inputs shapes the noise gain to 3 looking a lot better, 

Here, the 10nF across the inputs fixes the issue - sometimes a small series output R is then required - 0ohm here but sometimes about 20ohm in each leg. 

This article touched on these issues referring back to earlier material as well, 

https://www.planetanalog.com/stability-issues-and-resolutions-for-high-speed-fully-differential-amplifiers-fdas-insight-8/

I am not up on Pspice for TI, working with TINA still,