It　is　a　challenging　task　to　detect　planetary　gearbox　fault　type　under　time-varying　speeds.　Firstly,　under　time-varying　speeds,　the　fault-induced　frequency　harmonics　of　the　planetary　gearbox　are　time-varying　and　close-spaced,　and　the　fault　feature　is　difficult　to　detect　from　the　spectrum　or　traditional　time-frequency　analysis　(TFA)　results.　Secondly,　the　interference　component　in　the　vibration　signal　resulting　from　the　complicated　gearbox　configuration　is　another　obstacle　for　fault　detection.　As　such,　the　scaling　operator　demodulation　spectrum　(SODS)　technique　with　high　readability　is　developed　for　planetary　gearbox　fault　diagnosis　under　variable　speed　conditions.　The　developed　technique　consists　of　two　terms:　scaling　demodulation　operator　term　for　transforming　all　time-varying　fault-related　frequency　harmonics;　and　spectral　selection　criterion　term　for　obtaining　fault-related　frequency　harmonics　and　removing　the　other　smeared　frequency　components.　With　the　proposed　technique,　the　time-varying　and　close-spaced　fault-induced　frequency　harmonics　can　be　quantitatively　characterized　by　the　spectral　peaks.　The　performance　of　the　SODS-based　planetary　gearbox　fault　diagnosis　technique　is　validated　by　a　simulated　signal　and　two　experimental　signals　collected　from　different　test　rigs.　Furthermore,　comparison　results　with　the　traditional　TFA　methods　show　that　the　proposed　technique　has　much　better　ability　to　identify　the　fault-related　frequency　harmonics　of　the　planetary　gearbox.