Where am I going wrong with GNU Radio Companion that I keep getting the overflow message?

Question

I am new to SDR's as well as GNU Radio Companion. I am trying to set up a HackRF One to receive a text beacon from another transceiver. They are attached via SMA's with attenuators in between.

The beacon transceiver is set up to have a transmission frequency of 435MHz with a GFSK Modulation at a modulation index of 0.5 at a rate of 9.6 symbols per second

When I try to run the flowgraph, I end up with the overflow warnings, "OOOOOOOO ..." until I shut down the program.

I know that the problems lies either with how I am setting up the blocks or with the computer I am using.

I have made sure that the HackRF can received something from the transceiver by just directly connecting the osmocon source to a waterfall plot and I saw the fluctuation as the transceiver transmitted.

I have verified that the "OOOOOOOO" comes after I implement a filter into the chain.

I have been trying to translate what I know about analog RF into what the blocks are asking for, but it is obvious that I am not understanding how I need to calculate the number for the block parameters.

Answer 1

You have a transition with of 1 Hz, at a sampling rate of 2 MS/s. That means your transition width is 1/2,000,000 of the sampling rate – so the filter must be in the order of 2 million taps long!

That's an extremely long filter. The same as for analog filters applies to digital filters: the steeper you make their flanks in the frequency domain, the harder they become to build and the longer their impulse response. In the digital world, a long impulse response means that every input sample needs to be multiplied with many filter coefficients, and that simply takes time to calculate.

Design a filter that fits your needs AND is possible to execute in real-time.

I don't know what your need in filter sharpness is, but it certainly isn't only 1 Hz between stopband and passband.

I would try to infer from the rest of your flow graph, but sadly, that doesn't make sense, either: You're first filtering to 75 kHz bandwidth (within the 2 MHz sampling rate), and then are trying to demodulate at 2 Samples per symbol – so at a symbol rate of 1 MSym/s. How are 1 million symbols per second going to "fit" into a bandwidth of 75 kHz?

So, either your filter bandwidth is wrong, or your GMSK demod's samples per symbol is wrong, or both!