Am Freitag, den 16.09.2011, 18:44 +0200 schrieb John Paul Adrian Glaubitz: > Ok, changing the delay back to 1000 ms fixes the reliability problems > with my MZ-RH10. In that case, please commit with 1000ms. > I see that Michael claimed that 1000 ms delay (translating into 1 Hz > polling frequency) is actually too long and should be changed to > 200 ms (5 Hz). But why? Because 1Hz polling frequency is excessively low. For comparison, the python netmd code uses 100 ms (10Hz), see line 172 of libnetmd.py. Probably some timeout is tied to the number of iterations and needs to be increased by a factor of 5 when the delay length is decreased. In this case, it is NETMD_RECV_TRIES, being 30, resulting in a 30 second timeout. It appears reducing it to 6 seconds by decreasing the poll interval is not a good idea, so (after committing the patch with 1000) trying again with 200 in the sleep duration and 150 for NETMD_RECV_TRIES seems like a good idea. The problem with the 1 Hz polling frequency is that old NetMD units typically miss the first iteration (at least the Sharp IM-MT899 does), so there is a 1 second delay introduced for each command sent to that unit. The RH1 on the other hand manages to reply fast enough on most "simple" NetMD commands, so that the frequency of the polling loop does not matter.