If you are already able to generate the full scale signal, you will have
to scale down the full scale value by multiplying it with a factor of
0.17804. Because thats not the easiest task to build in hardware, its
more convenient to multiply it with a value eg. 2917 and divide it by
16384 = 2^14. Another pair of values with less error is a multiplication
with 5974013 divided by 33554432 = 2^25. But here you should have a
close look for the implementation because the AD value is 24 bits, plus
the downscale is 25 bits, therefore you will need a intermediate result
with 49 bits.

> How do u get the number of 2917 and 16384?
Take the 0.17804 and multiply it by 2, repeat this until the fractional
digits are 0. Count the multiplications. In case of 2917 it took 14
multiplications to get 2917,00736. So the resulting factor is 2917/16384
= 0.178039550781. This causes an error of 2.5e-6, which is 42 LSB of 24
bits.

So this isn't enought accuracy for a 24 bit DAC. You will have to
increase the accuracy at least by the factor 2^6 = 64, so the next
matching number will be 5974013/2^25 = 0.178039461374.

> Actually i want to generate a sine waveform with adjustable amplitude.
Multiply the output with a value of 0...65535 and divide the result by
2^16. this will give you a resolution of 16 bit in the output amplitude.