Author Topic: ObserverIP and Meteobridge - Terrible for WU uploads (WS-2902A console better)  (Read 1244 times)

0 Members and 1 Guest are viewing this topic.

Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
For those of you with a WS-2902A, if you have upgraded to also have an ObserverIP (and maybe even a Meteobridge), you may be wondering which device should upload to Weather Underground. Since all 3 devices are capable of publishing data to WU, which is best? Well I did some testing. Here are the results...

I thought the Meteobridge would be the best to upload data to WU. It offers the fastest upload to WU at every 5 seconds rapid fire. The trouble is that the Meteobridge is dependent on how quickly it can read new data from the ObserverIP. Turns out that the ObserverIP, running the lastest firmware 4.4.7 (I also tried 4.4.5), only refreshes the Live Data page every 30 seconds. The WS-2902A display console refreshes every 14 or 16 seconds and also publishes to WU every 14 or 16 seconds, but even still the display console also responds to sensor changes quicker without doing data averaging.

To make matters worse I've closely watched the ObserverIP Live Data page and even though it is receiving the same information from the same outdoor sensor array as the display console, the ObserverIP seems to do some live data averaging over the short term. What this means is that even if you wait 30 seconds on the ObserverIP the temperature is not quick to react to quick temperature changes. For example if a storm cloud quickly rolls in or if day break just happens as the sun comes over the horizon, the temperature in both of these cases will drop or increase respectively in real life. What I noticed happens on the ObserverIP is that even after waiting the 30 seconds for a Live Data refresh the temperature didn't change much to match the quick change seen on the display console. Even after waiting yet another 30 seconds (or even a couple minutes) the temperature still sort of just slowly catches up to actual real temperature. The display console is better at really showing (and uploading) immediate live data.

The weird part is that the ObserverIP isn't the device with the temperature sensor. The temperature sensor is on the outdoor sensor array (pressure change is slow too from the WS-1000-BTH). The raw data being broadcast by the outdoor sensor array is the same for the console or for the ObserverIP to pick up at any given moment. Yet the ObserverIP seems like it is doing some sort of data averaging and that is what it then shows on the Live Data page. This is where the Meteobridge gets its data.

What this means is that if you want to publish data to Weather Underground if you publish with the ObserverIP or via the Meteobridge that is getting data from the ObserverIP, then your graphs will be very smoothed out and average slopes is what you will see. You will see the same amount of data points in the charts but it just has been averaged out by the ObserverIP.

If you have a Meteobridge and you are publishing to other services like CWOP, AWEKAS, PWS(WeatherForYou) etc it might not be a problem to have averaged data because to those services you only publish every 5 or 10 minutes anyway. In those cases having averaged data is just fine...or rather better I would say than actually grabbing a peak or low data point as luck would have it. But if the ObseverIP was faster at updating live data then I think the Meteobridge might still be smart enough to know that data being published every 5 seconds does not need to be averaged but data that is being published every 5 minutes to CWOP maybe could use some averaging. I don't know how the Meteobridge would behave with more instantaneous data.

But Weather Underground with its rapid fire updates is one of the few web services that I would argue is popular because you can see live data almost instantaneously. This means on WU you can see those wind speed and direction changes quicker, and you can see the temperature drop just as a storm cloud moves in. For these reasons I would say the WS-2902A display console is better for WU publishing despite being 14-16 seconds vs the Meteobridge 5 seconds (it is all the ObserverIP's fault).

I can see though that if you don't care much for the Weather Underground rapid updates and maybe you want to have another site to compare to the other sites that you publish with your Meteobridge and you want some greater degree of consistency between data at the various sites, then I could see perhaps choosing to have the Meteobridge be the one to upload to WU instead of the WS-2902A display console.

Attached is graph comparison I did for the same time period. This data was sent to WU using two different Station IDs, one using the WS-2902A display console and the other the Meteobridge. All the data was simultaneously coming from the very same outdoor sensor array. In case anyone is curious, yes, I also did this test using the ObserverIP to be the one to publish to WU and I basically saw the exact same graphing results of average soothing as you see in the Meteobridge graphs on WU.

 [ You are not allowed to view attachments ]

So what are those temperature tiny spikes and dips? I looked at my solar radiation charts and they correspond to clouds moving in and out during the day. The temperature really does drop and rise like the console indicates. The ObserverIP just averages it all out smoothing the curve.

I've also noticed the same issue but to a lesser degree in other metrics, wind, humidity and pressure.

* ATTENTION: This is not a Meteobridge problem. This is a limitation of having the Meteobridge get data from the ObserverIP. What we need is for Ambient Weather to release a newer better ObserverIP that is more up to the task of updaing quicker and perhaps even offer the Meteobridge a more direct way of getting data that does not involve screen scraping the data from the Live Data page.

« Last Edit: September 05, 2018, 09:15:50 AM by galfert »
Ecowitt GW1000 | Meteobridge on Raspberry Pi
WU: KFLWINTE111  |  PWSweather: KFLWINTE111
CWOP: FW3708  |  AWEKAS: 14814
Windy: pws-f075acbe
Weather Underground Issue Tracking
Tele-Pole

Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
Some people may look at that graph comparison I posted and think, "who cares, looks good enough." But please realize it is more than just the graphs. Notice in my graph that a 1PM the temperature suddenly drops. It is a quicker drop with the WS-2902A display console because it responds quicker and doesn't do data averaging. This was a storm rolling in at 1PM. With the Meteobridge (via the ObserverIP) on WU you would notice a few degree difference too warm just as the storm begins to hit, the graph is slow and less steep in change. Data via the ObserverIP has delayed response and is too slow for watching live weather. It is that bad. Not an issue if you are publishing to other services with 5 and 10 minutes uploads.
« Last Edit: September 05, 2018, 09:16:56 AM by galfert »
Ecowitt GW1000 | Meteobridge on Raspberry Pi
WU: KFLWINTE111  |  PWSweather: KFLWINTE111
CWOP: FW3708  |  AWEKAS: 14814
Windy: pws-f075acbe
Weather Underground Issue Tracking
Tele-Pole

Offline Kalibr

  • Member
  • *
  • Posts: 6
Thanks for this analysis,  I recently setup my WS-2902 and have the Meteobridge hardware on the way.  I had this very same question if I should leave my console to handle the WU uploading.

WU : KMIANNAR169
CWOP : FW3745

Offline Mattk

  • Forecaster
  • *****
  • Posts: 2135
The lesson being any system is only as good at it's weakest component

 

anything