Op-amp output oscillating at ~200 kHz — only happens when I add the anti-aliasing cap

opamp_oracle has the diagnosis right. One more thing to rule out while you're revising the circuit: check whether you have any capacitance from the inverting input node to ground.

A very common variant of this problem — especially if you've been following simulation tutorials — is adding a small bypass cap across the lower feedback resistor to "filter noise on the feedback divider." In simulation it often looks fine. On hardware it places a pole in the closed-loop response that eats phase margin in a way that looks almost identical to the capacitive output load problem.

If you have any cap there, remove it first. The correct way to add HF rolloff inside the loop is a small capacitor in parallel with the upper feedback resistor (51 kΩ). This forms a pole that rolls off closed-loop gain gracefully at frequency. For an 8 MHz GBW part at gain 6, something in the range of 5–10 pF across the 51 kΩ is typically enough to gain a useful margin improvement without affecting the passband.

The underlying framework here is the Bode gain-phase relationship: every pole in the loop adds –90° of phase asymptotically, and every zero recovers some. The op-amp fundamentals page covers GBW and slew rate — working through those numbers for your specific device before committing to a topology saves the kind of late discovery you've just hit.

The Riso fix will sort out the oscillation. Worth flagging one thing for the longer-term design though: whether a first-order passive RC is actually enough anti-aliasing for your application.

A first-order filter at 1.6 kHz only rolls off at –20 dB/decade. If you're sampling at, say, 10 kHz, you're getting roughly 14 dB of attenuation at Nyquist (5 kHz). Whether that's sufficient depends on how much signal energy exists above Nyquist and what SNR you need at the ADC — but for precision sensor work it often isn't.

If you find you need more rolloff, a Sallen-Key second-order low-pass is the cleaner route rather than cascading more passive RC sections. The op-amp is designed into the filter topology from the start, so stability is part of the design analysis rather than something you patch afterward. You still need to watch GBW vs Q-factor (high-Q Sallen-Key stages on a bandwidth-limited op-amp will peak), but the trade-offs are explicit.

It depends entirely on what your sampling rate actually is and how much alias rejection you need — sometimes a first-order RC is fine and sometimes it isn't. Either way, knowing which situation you're in before the board spins is worth the half-hour.