We　study　the　impacts　of　non-ideal　filters　in　Modulated　wideband　converter　(MWC).　An　analog　filter　compensating　method　is　presented.　The　limitations　of　the　conventional　back-end　of　the　advanced　MWC　is　analyzed　and　an　improved　back-end　based　on　Fast　Fourier　transform　(FFT)　is　proposed.　By　setting　proper　length　of　the　digital　signal　processed　each　time,　the　proposed　system　increases　the　recovery　Signal-to-noise　ratio　(SNR)　and　significantly　reduces　the　computation　loads.　The　compensating　operation　is　accurate　to　each　frequency　point　and　the　expanding　operation　is　simplified　to　a　splitting　of　the　frequency　points.　Numerical　simulations　demonstrate　our　proposed　system　against　the　ideal　MWC　with　different　filters.　The　results　indicate　that　the　recovery　SNR　of　over　70　decibel　is　achieved　with　almost　only　one-tenth　the　computation　complexity　of　the　conventional　MWC　backend.