On Sat, Jul 25, 2020 at 12:23 AM Al Beard <[email protected]> wrote:
> Hi all, > > Am I right? > > Mode 2020 needs a lot of CPU power eg. AVX/AVX2 to work. > > All the other codecs and modems (modes) run well on the > STM32 series and almost any Pi or clone. > > So, can our experts explain the real usefulness of LPCnet? > It has radically better quality per bit than alternatives. It also can be run in a _fast_ embedded computer (like a modern generation rpi), just not underpowered ones. Computing power has become radically cheaper and will continue to do so for at least the immediate future. Make a codec that is too slow then wait until computers catch up is a time proved technique (e.g. it's what H264 did). One of the radical advantages of DSP driven radio is that it's possible to improve your designs by throwing more computing power at them. Maybe more engineering time would be better but more computing comes extremely cheaply. If anything, with the exception of off-grid battery powered usage, to me it seems like a distraction to spend too much attention on tiny microcontrollers that have 1/5000th the computing power of a $100 high power part especially when it comes to a codec rather than a modem (low power modems being more useful/interesting as unattended relays and the like). I think this is especially true when it comes to open source and community development. Squeezing into a tight performance envelope requires a lot of tedious optimization and the result is brittle and not easy to modify and experiment with because there isn't any room left. All that optimization may be unjustified if you're just going to throw out the design next year when the community comes up with something better.
_______________________________________________ Freetel-codec2 mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/freetel-codec2
