More Details regarding Bit Resolution on Biquad Filter

I have attached slides with magnitude bode plots for a 2nd order low-pass filter with critical frequencies of 1 MHz and 10 Hz (and a sampling frequency of 10 MHz). The set of points with lower bits of resolution have been normalized so that they all share the same value (-3 dB) at the respective critical frequency. Slides 2 and 3 show details about the filter with a 1 MHz critical frequency. Behavior of lower resolution deviate as expected with differences becoming more noticeable for values above the critical frequency. The slope for values above 1 MHz is also generally well-behaved, with most being near -40dB/decade as expected for 2nd order filters. 4 bits seems to be a sufficient amount for preserving filter effectiveness. Slides 4 and 5 show details about the filter with a 10 Hz critical frequency. It is worth pointing out that the filter required at least 36 bits of resolution to function; lower values result in a nan error which I am still trying to investigate. It is also worth noting that all the slopes deviate from the expected -40dB/decade, even the float resolution. However, the upside is that the differences plot on Slide 5 shows very little divergence from the float values for lower resolutions, with differences decreasing with more bits.