WS2300Sync=1 question

[anthros]

Hi,
I tuned my scanner to the wireless data frequency and listened to the transfers from the outside unit and noticed cumulus missing a few readings now and then, especially when the wind speed activated the 32 second data transfer interval. Ok, it happens once or twice per hour, not the end of the world as such, but being something of a stickler for accuracy i activated the WS2300Sync=1 in the cumulus.ini. After activating it, the amount of lost readings increased.

I noticed that cumulus read data from the station while the outside unit was still sending the data over wireless. On my station the transmission duration is on average about 2.1 seconds and every single time while the WS2300Sync=1 option is activated, cumulus reads data in from the main unit while it still is receiving the wireless data. This leads to the result that quite often, the data downloaded from the station is only partially up to date.

Is there any way to increase the wait time from what it is now ? i think something around 3.5 to 4 seconds would be ideal.

//Alex

[steve]

I'd have to change the code to change the value - I can make it configurable. The current value was one that someone with a La Crosse station came up with which worked. Well, it achieved the objective of not reading while the station is transmitting; it was an attempt to reduce the number of corrupt data readings, which it didn't do. That's all that setting was for. If the transmission takes 2.1 seconds, and the delay is 2.5 seconds why is that not long enough?

I don't understand how (ignoring the 'sync' setting) Cumulus can be missing data when it's reading every 6 to 8 seconds or so, and the update interval is 32 seconds?

[anthros]

Well, without the sync, i get a variable read time. At good moments it can be as soon as 10 seconds but on average it seems to be around 20- 40 seconds, sometimes even up to 50 seconds.

This is based on the Last data read time stamp in the bottom of the cumulus window. Data loss occurs if the wind speed activates the 32 second threshold, and cumulus happens to load in data from the station just before the update. This is not a big problem as such, as it only happens on windy days. A user adjustable read time would be nifty though.

I had a much higher update rate when i was running on usb, but that of course created huge amounts of corrupted data. When i moved over to serial port connection, I did, as an extra safety precaution, lower the baud rate on the serial port to 4800 bauds and dropped the fifo buffers to as low as they can be in the hopes to minimize the chance for communication errors.

Could the data transfer be tied to the baud rate ? I noticed no signs of the baud rate influencing the data transfer rate while the station was connected through the virtual com port that the serial=>usb adapter uses. Havent really dared to play around too much with the baud rate on serial in case i create more problems than i might solve with a higher transfer rate.

//Alex

[steve]

I suspect the long read times are down to retries because of bad data. The way the protocol, if you can call it that, works (or not), relies on retrying each read several times until a successful response is returned (and a 'successful' response can sometimes contain bad data). There are several steps involved in reading each data item, and each step may require several retries. If you turn on the debug log, that will have some tracing which may show what's going on.

It does look as though the 'sync' facility may be useful after all, as a way of hitting the station when it's most likely to respond. I'll look at making the delay configurable when I get back to working on Cumulus 1 after I've released the next Cumulus 2 alpha. I don't think I've bothered putting the 'sync' code in C2 as it didn't seem to serve any purpose, but from what you say it looks like it's worthwhile.

[anthros]

The debug log does contain lots of resets with retries. Looks like read errors are the cause of the slow update rate. Thank you for considering to add a user definable update interval time, i certainly would enjoy that feature.

//Alex