What are the advantages of such an accurate clock for a data mode with such a slow symbol rate. How can we evaluate the impact of oscillator accuracy on the overall performance of M17?
Most dev boards have 20-50ppm oscillators on them. Most Nucleo64 boards have a 50ppm crystal on the ST/Link (from MCO) that can be routed to HSE on the MCU. Most of these boards are not designed to take a TCXO. Requiring a high-precision oscillator will make experimentation more difficult.
If you think about an M17 radio, you will need to ensure that your frequency accuracy and stability are at least as good as other radios on the air. 1ppm on 70cms is still an error of 435Hz which is only just tolerable, and if you have a good reference oscillator for that, then you can tap that for your MCU and the ADC/DAC.
The MMDVM was able to accommodate the more wayward clock in the TNC, but it may have meant that the BER was not optimum at times. If you are serious about the M17 being at least as good as existing DV systems, you need it to be demonstrably better in all respects, not just the openness.
The oscillator used it the Kenwood TM-V71A is 5ppm. The oscillator used in a TK-790 is 2ppm. In both cases, the error is calibrated out via a DAC. At least for the TK-790, the oscillator used for TX cannot be tapped for an MCU because the oscillator itself is modulated. It uses two-point modulation. The same is true for the DMR radios OpenRTX are currently hacking on.
I’m not suggesting that you are wrong, or that a better clock does not help. I am simply asking “[color=#333333][size=small][font=Tahoma, Verdana, Arial, sans-serif]How can we evaluate the impact of symbol timing accuracy on the overall performance of M17?”[/font][/size][/color]
[color=#333333][size=small][font=Tahoma, Verdana, Arial, sans-serif]Are we talking a difference of 0.3dB between 2.5ppm and 10ppm? Or is it closer to 3dB?[/font][/size][/color]