Oversampling implication for data rate to host?

Quick question about oversampling … in RxTSP, is the oversampling performed in the HW or the SW on the host ? ie. if trying to sample 48 MHz, and RxTSP clock is showing as 96 MHz, is the reduction to 48 MSPS done in the host or the Lime ?

I have a LimeMini which when used with GNURadio is very happy at 30.72 MHz and oversampling bypassed with no dropped samples. My LimeUSB refuses to run in bypass oversampling mode, and trying to capture 48 MHz I am battling dropped samples. Trying to figure out where to look.

It is done inside the Lime chip i.e. before the FPGA, so it should have no effect on dropped samples. Oversampling is disabled for LimeSDR-Mini as it is prone to overheating.

Where is your source that says that oversampling is disabled in the LimeSDR-Mini?

Sorry, i was not exactly clear. Oversampling is disabled by default if your bandwidth is larger than 20M. In general, Mini selects lower oversampling values e.g. for 20M bandwith Mini would set oversampling of 2 while the USB board would use 8.
This applies for the default logic, i guess you can override it if you use gr-lime blocks, not sure about this.

Source for the oversampling logic:

i am giving sample rate 500khz and oversampling is in default. so finally the sample rate i am seeing is (500khz * 32) 16Mhz. the oversampling factor is changing(32,16,8,4,2,1) depends on the sample rate which we given. why the oversampling is needed? for me 500khz sample rate is enough but the limesdrmini doing some inbuilt operation based on the input sample rate.can anyone please tell me about this oversampling operation in limesdrmini.

As far as I’m aware, oversampling the signal with subsequent decimation to your required rate improves signal to noise ratio because the noise error averages out between 32 samples.