02.10.2019 Views

UploadFile_6417

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

514 Chapter 9 SAMPLING RATE CONVERSION<br />

Significance of δ 1 and δ 2 The filter ripple parameters δ 1 and δ 2 have<br />

the following significance, which must be taken into consideration while<br />

specifying their values:<br />

• The passband ripple δ 1 measures the ripple in the passband and hence<br />

controls the distortion in the signal bandwidth ω p .<br />

• The stopband ripple δ 2 controls the amount of aliased energy (also<br />

called leakage) that gets into the band up to ω x,s .<br />

There are (D − 1) contributions due to k ≠0terms in (9.53). These<br />

are expected to add incoherently (i.e., have peaks at different locations),<br />

so the overall peak error should be about δ 2 . The actual error depends<br />

on how X(ω) varies over the rest of the band |ω| >ω x,p . Clearly, the<br />

filter stopband ripple δ 2 controls the aliasing error in the signal passband.<br />

Therefore, both δ 1 and δ 2 affect the decimated signal in its passband.<br />

Comment: Comparing the FIR decimator filter specifications (9.57) to<br />

those for the FIR interpolator in (9.52), we see a high degree of similarity.<br />

In fact, a filter designed to decimate by factor D can also be used to<br />

interpolate by the factor I = D, aswesee from the following example.<br />

This means that the function intfilt can also be used to design FIR<br />

filters for decimation.<br />

□ EXAMPLE 9.13 To design a decimate by D stage we need values for ω x,p and ω x,s (remember<br />

that these are signal parameters). Assume ω x,p = π/(2D), which satisfies the<br />

constraint ω x,p ≤ π/D and is exactly half the decimated bandwidth. Let ω x,s =<br />

ω x,p. Then the FIR lowpass filter must pass frequencies up to ω p = π/(2D) and<br />

stop frequencies above ω s =2π/D − π/(2D) =3π/(2D).<br />

Now consider the corresponding interpolation problem. We want to interpolate<br />

by I. Weagain choose ω x,s = ω x,p, but now the range is ω x,p

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!