deletions | additions       diff --git a/layout.md b/layout.md  index dae3e44..38cfda6 100644  --- a/layout.md  +++ b/layout.md 
...
PeriodLimir.tex  figures/MaximumDMvsP3/MaximumDMvsP3.png  considerations.tex  scattering part5.tex  figures/FFT_size_100ns_resolution2/FFT_size_100ns_resolution2.png  scatter part6.tex  diff --git a/scattering part5.tex b/scattering part5.tex  deleted file mode 100644  index f7234cb..0000000  --- a/scattering part5.tex  +++ /dev/null 
...
Given that the scattering time grows with $\nu^4$ while dispersion delay grows as $\nu^2$ the limiting DMs calculated above define the maximum delay that must ever be handled within the SKA1 pulsar timing instruments. Above the limiting DMs we suggest that the data be channelized such that the effective temporal resolution is always greater than the scattering time time. If rechannelization is applied, the limiting DM delay defines the maximum size of Fast Fourier Transform required to be supported by the instrument and thus the required FLOPS for the instrument (under the argument that decreasing the temporal resolution by a factor of 2 reduces the required FLOPS by a factor of $1-1/\log_2(n)$, i.e. the cost of two FFTs of length $n/2$ is less than the cost of one FFT of length $n$). Figure \ref{fig:fft_size} shows the maximum FFT size that is required to be supported by the SKA1 pulsar timing instruments as a function of observing frequency. For efficiency reasons, the FFT size derived from the dispersion delay must be rounded up to the next but one power of two (i.e. a suggested FFT size of $2^{22.4}$ becomes $2^{24}$).