One problem I have noticed with the PIC TNC I recently built is that it is less tolerant of different packet signals than any of my radios. It decodes my two Kenwood transceivers just fine but it will only decode the VX-8G at a specific audio level that is impossible to set when using the fixed output of many radios. And it won't decode my WX-1 weather station at all.
My Kenwood TH-D72 won't decode the weather station either. However it is the VX-8GR I am more concerned about. With the volume of the packet channel turned up, it's braaps sound a bit thin and weedy compared to those of the Kenwoods and other radios I hear over the air. I thought that I would try to analyze the signals to see if this would give me an idea of what was causing the problem.
I used Spectran, the only free software I know that will do audio spectrum analysis. The receiver was the old Kenwood TH-205E, which being over 25 years old had IF filtering wide enough not to cause any deviation limiting. Each capture was made at the same volume level so the signal levels shown should represent the relative signal deviation.
Because packet bursts are fleeting it took a few attempts to capture the screen at just the right moment. But eventually I obtained plots for each of four radios, including the weather station. Incidentally I am puzzled that the spectrograms show a comb of frequencies. I thought 1200 baud packet was FSK using two frequencies, 1200Hz and 2200Hz. I have seen this before when using sound card decoder software for packet but I have always been puzzled by it.
The top two plots are for the two Kenwood radios. They look pretty near identical. In the absence of any test equipment to actually measure the deviation levels I have to assume that these two radios were correctly set up at the factory and represent the ideal signal to aim for. It is interesting that the highest frequency which I would have assumed to be 2200Hz actually peaks at about 2235Hz. The peak closest to the lower frequency of 1200Hz is actually 1185Hz. But there are six peaks at intervals of about 150Hz between the two and some spaced the same distance going below the lower frequency. I'm sure there's a reason for it.
If you look at the plot for the VX-8G the top peak is at about 2230Hz and 5dB weaker than the corresponding peak of the Kenwood traces. The other peaks are lower still with the one at about 1180Hz around 8dB lower than that from the Kenwood. Some VX-8 users have complained about low packet deviation of the radio but have been told by Yaesu that it is within specification. As far as I know there is no adjustment to increase it. You would have thought from this that I would need to increase the audio level to get reliable decoding of the VX-8 compared to the Kenwoods. In fact, I have usually had to reduce it a little. As previously stated, the volume setting at which the PIC TNC will decode the VX-8G is quite critical, whereas the Kenwood signals would decode over quite a wide range of audio input levels.
When you look at the signal from my WX-1 weather station, which is modulating a Radiometrix VHF transmitter module, the peak signal levels are close to that of the Kenwoods. The lower frequency components are in fact a couple of dB stronger. However, it's clear that the frequencies are too high. The top peak, which should be 2200Hz, is about 2290Hz. And the one closest to 1200Hz is about 1230Hz. When setting up my FoxTrak APRS tracker I had to set the frequencies using the PIC calibration routine as low as they would go before my TH-D710 would decide it, so clearly it is the frequency offset that is responsible for the packets not being decoded. The WX-1 firmware unfortunately does not have a calibration procedure. Either the PIC clock crystal needs to be slowed down a bit or I need to make a change in the source code to shift the frequencies and recompile the firmware.
But it's the VX-8G that most bothers me most. I wish there was a way to boost the level of its packet modulation and make it more like the Kenwoods.