When using SDR, you always don't leave your computer. But if you receive it, it is always inconvenient with a computer. Although sometimes it is still inseparable from the computer, it still wants to use a portable receiver. So I think I have one DIY. This time I use STM32F4 to do and test the ability of STM32F4 to provide a thinking for everyone DIY.
At the same time, use SDRPALY. It is found that MSI001 is simply artifact. The MSI001 uses zero multi-frequency mode, with the lowest bandwidth to 200K. The STM32F4 can be used at all. This allows you to receive some basic signals, although 200K is limited. Finally, I got the MSI001 data sheet. After the winter vacation, I played a MSI001 board, followed by the manual to use the single-chip drive to see if I could have a signal. But no matter how, you don't have a signal. Because I have an SDRPLAY here. I took a logic analyzer, and I wrote it in all the things that caught it. However, I can't drive.
However, in order to verify that STM32F4 can complete this work, I will first refer to the structure of MCHF and other machines, making a short-wave receiver. Test STM32F4.
Show the screen using 800x480, draws a simple UI, the effect is as follows.
SDR receives video
1. Use the zero multi-frequency structure, STM32F4 master, and Samples comes with STM32.
2. The program is still perfect, and the spectrum is stopped when listening to broadcast. I have time to improve the program and run.
3. The algorithm decoded is also perfect, the sound is very strange. And didn't write AGC, and the sound is ups and downs. Subsequent, if you add AGC, you should listen to the broadcast. But still can't hold excited mood, listen to broadcast XD with the original program
4. Thank you for your voice of China
5. When using the mobile phone record, there is a relatively large environment sound, and the broadcast sound is large, the volume is noted! I recorded video is really mudslide. Next, make a good equipment orz.
Waterfall
In order to observe the time of each function consumption, use the idle IO output state, observe the logic analyzer.
The time required is as follows
(1) Two ADCs set the sample rate of 200K, and 5.120ms required for calculation, the test results are the same.
(2) 0.625 ms is required for 1024 points of complement FFT operation.
(3) FFT results show that it is time consuming, generally between 4 ms to 7 ms.
(4) The waterfall map shows 3.699ms.
(5) 125th-order FIR filtering requires 3.058ms.
(6) Filter plus extraction and AM demodulation require 3.712 ms.
in conclusion
After testing, the STM32F4 operates indeed stronger, in the case of opening FPU, FFT only requires 0.625 ms. In the mode of only the FFT result, the FFT operation plus other handles require 2.8ms, but plus the 4ms to 7 ms of the display program far exceeds 5.120ms. At this time, you have to drop some buffering content. In a separate demodulated mode, it takes 3.712 ms, less than 5.120 ms, and separate demodulation is feasible. However, it is very bad that the spectral demultiplex is very bad, so the demodulation cannot be displayed. At the same time, if you want the waterfall map to widen, the demand for memory has also risen.
These situations are the results at 200k sample rate. If you use the MCHF, use 48K sample rate, time is very bad. However, there is no way to use MSI001. Of course, the program can also be further optimized, such as the drawing of the FFT results displayed to the non-filled can save some time, and the step of reducing the step of the FIR filter is replaced. Or reduce the screen resolution, saving time. Time issues and drive problems, the STM32F4 + MSI001 is still feasible.
The FM demodulation has not been tried, remember that the KCDemo project uses F4 to demodulate FM.
After this test. In the platform of STM32, this kind of thing is made, there is a feeling that is not good, and the application is also limited. So I think that I can use the card computer to add SDRPALY. For example, use the core board of the full-chromatic chip, plus the circuit and display circuits of the SDRPLAY, and Ubuntu. Resources should be much better. And there is no need to repeat the wheels. For example, use the following. I have time to try if I can implement it.
Our other product:
