We are using the TL431BQDBZ as a regulator on the secondary side of a DCDC-converter.
The anode of the reference is connected to 0V.
The cathode is connected through a 1.5k-resistor to the regulator-optocoupler (optocoupler-diode paralleled with 680Ohm to bias the TL431).
The reference is connected to a voltage divider to the output voltage (Rref to 0V=2.2kOhm, Rref to Vout= 18.96kOhm).
The reference is connected to the cathode with an RC (PI-regulator): 22nF in series with 2.2kOhm.
With this setup, I measure: 2.476V on the reference of the TL431 and 23.81V on the regulated output-voltage. So the measured output-voltage coincides with the calculation 2.476/2.2k*(2.2k+18.96k). Measured cathode-current is 1.27mA.
Question: The regulated output-voltage is too low, because of the low reference voltage. This behaviour is reproducable on many units.
Why is the reference out of tolerance (should be 2.495V +/-0.5% = 2.483Vmin)?
How can I improve the accuracy to achieve the expected 0.5% accuracy (with ideal resistors)?
