input data rate calculation and max output bandwidth per channel

Hello,

Since your application will not require digital modulation, one will need to set Modulation Mode Selection to "None" via Reg 0x111, Bits [3:2].
In this setting, the "I" and "Q" data can be treated as independent data with "I" data being reconstructed by  "I" DAC and "Q" data being reconstructed by "Q" DAC.

As described in datasheet, the use of a digital filter results in reducing the "usable" bandwidth of the input signal by 80% vs bypassing (1x) the digital filter (i.e.   BW=FDATA/0.4  with interpolation vs BW=FDATA/0.5 without interpolation).  The benefit of interpolation is that it can simply any post filtering requirements in the system should it be required to filter out the "images" that inherently (due to sampling theory) result when reconstructing a digital signal.   The other restriction is the maximum DAC update rate (i.e. 2.25 GSPS) which limits the max FDATA rate (hence usable bandwidth) when operating with different interpolation rates as shown in Table 56...................while noting for the bypass case (1x), the max DAC update rate of 1.238 GSPS is limited by the JESD204B interface.

Most applications benefit from use of interpolation filters and would set the interpolation factor to highest possible level based on signal bandwidth requirements that need to be supported.   For example,  if the reconstructed signal must support a bandwidth >225 MHz and less than 450 MHz.......then 2x interpolation can be used (as shown in modified table below).

Hello PMH

Thank you for the very clear and comprehensive answer. I think I understand it now. Just one more question, when no interpolation is used, it gives .5 of usable bandwidth. Assuming the JESD link is limited to 1125MSPS, this will give 562.5MHz bandwidth while 2x interpolation only gives 450MHz. The benefit however with 2X interpolation the images are pushed away from the signal frequency range allowing us to use a simpler reconstruction filter? 

Thank you

Dan

Hello,

Your understanding is correct.   That said............while DAC can give you 0.5 x FDATA bandwidth when operating without (1x) interpolation, the analog anti-aliasing filter would be impossible to design to reject the closes DAC image falling between 0.5 x FDATA to 1.0 x FDATA since no transition region exists for the analog filter.   In practice, one would need to design a very aggressive analog filter with cut-off frequency starting at 0.4 x FDATA and rejection band starting at 0.6 x FDATA (i.e. transition band of only 0.2 x FDATA).....................thus one is better off using the 2x Interpolation filter since one will get the same usable BW in the end.