Author Topic: Atlas Has Arrived  (Read 20328 times)

0 Members and 1 Guest are viewing this topic.

Offline CW2274

  • Forecaster
  • *****
  • Posts: 6731
    • Conditions @ CW2274 West Tucson-Painted Hills Ranch
Re: Atlas Has Arrived
« Reply #250 on: October 09, 2018, 10:15:23 PM »
Isn’t this the classic precision vs accuracy paradox?


If Davis wanted though they could still recalculate every 20 seconds but instead report on the previous 60 minute's totals. But I'm glad they don't do it that way.
Not sure what you're saying, however, it is recalculated every 20 seconds. Davis does nothing with the previous 60 minutes or any other time frame for that matter, other than that the rain rate is literally recalculated every 20 seconds, no more, no less. Elapsed time other than the packet pull for rain rate is non-existent.

Offline nincehelser

  • Forecaster
  • *****
  • Posts: 3337
Re: Atlas Has Arrived
« Reply #251 on: October 09, 2018, 10:28:26 PM »
Yes this is accuracy versus precision. By playing by the official WU rules then we loose precision by reducing the sampling rate to once per hour. We don't know at any given moment how fast it is raining by WU rules. But we do know that it is a certain amount within the last hour. That is still precise to the hour but not precise to the minute or seconds.

"rainin" is a value that changes quickly.  Saying the sample rate is only once per hour is very misleading.

In the case of a 5n1, "rainin" can change as often as every 36 seconds.

Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
Re: Atlas Has Arrived
« Reply #252 on: October 09, 2018, 11:12:45 PM »
Isn’t this the classic precision vs accuracy paradox?


If Davis wanted though they could still recalculate every 20 seconds but instead report on the previous 60 minute's totals. But I'm glad they don't do it that way.
Not sure what you're saying, however, it is recalculated every 20 seconds. Davis does nothing with the previous 60 minutes or any other time frame for that matter, other than that the rain rate is literally recalculated every 20 seconds, no more, no less. Elapsed time other than the packet pull for rain rate is non-existent.

I was trying to say that the recalculation frequency is inconsequential to timing interval period for total rain amount. They are two separate things.

If in the past hour it rained 0.5 inches then that is the rate for that period (the last hour). The period is 1 hour regardless as to the frequency of the calculation. It is like a permanently shifting sampling window that just adds the total rain amount for the previous hour. If that is done every 20 seconds or every 5 minutes it still doesn't change the fact that there is only one sample per given hour window.

You can only increase the precision if you reduce the size of the sampling window. So there is still a shifting window, and it is still every 20 seconds but now the sampling window time width is maybe the total rainfall for the past minute (much shorter). Then you multiply that the 1 minute rate result by 60 and you now have a more precise rate for that moment in time "per hour." All the while the recalculation keeps happening every 20 seconds.

If your shifting sampling window is 1 hour long then you don't need to multiply to get the per hour rate. Well technically you do still multiply but you multiply by 1.

Yes this is accuracy versus precision. By playing by the official WU rules then we lose precision by reducing the sampling rate to once per hour. We don't know at any given moment how fast it is raining by WU rules. But we do know that it is a certain amount within the last hour. That is still precise to the hour but not precise to the minute or seconds.

"rainin" is a value that changes quickly.  Saying the sample rate is only once per hour is very misleading.

In the case of a 5n1, "rainin" can change as often as every 36 seconds.

I don't think it is misleading. Yes it can change quickly. It can change like you said every 36 seconds. But the sample rate as for what is entered into the calculation is once per given hour frame (the shifting window width of time) . That is the total for the past hour times 1. That is what the definition means from WU when they say total rate for the past hour.
« Last Edit: December 06, 2019, 01:43:55 PM 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 nincehelser

  • Forecaster
  • *****
  • Posts: 3337
Re: Atlas Has Arrived
« Reply #253 on: October 10, 2018, 12:30:27 AM »
You can only increase the precision if you reduce the size of the sampling window. So there is still a shifting window, and it is still every 20 seconds but now the sampling window time width is maybe the total rainfall for the past minute (much shorter). Then you multiply that the 1 minute rate result by 60 and you now have a more precise rate for that moment in time "per hour." All the while the recalculation keeps happening every 20 seconds.

So if it is drizzling outside, and after 5 hours your 0.01" tipping bucket finally tips, you're going to chart a rain-rate value of 0.60" inches per hour for that minute?

Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
Re: Atlas Has Arrived
« Reply #254 on: October 10, 2018, 01:34:36 AM »
You can only increase the precision if you reduce the size of the sampling window. So there is still a shifting window, and it is still every 20 seconds but now the sampling window time width is maybe the total rainfall for the past minute (much shorter). Then you multiply that the 1 minute rate result by 60 and you now have a more precise rate for that moment in time "per hour." All the while the recalculation keeps happening every 20 seconds.

So if it is drizzling outside, and after 5 hours your 0.01" tipping bucket finally tips, you're going to chart a rain-rate value of 0.60" inches per hour for that minute?

Not sure happens with one tip. But I don't think it would report 0.60 per hour.
Ecowitt GW1000 | Meteobridge on Raspberry Pi
WU: KFLWINTE111  |  PWSweather: KFLWINTE111
CWOP: FW3708  |  AWEKAS: 14814
Windy: pws-f075acbe
Weather Underground Issue Tracking
Tele-Pole

Offline nincehelser

  • Forecaster
  • *****
  • Posts: 3337
Re: Atlas Has Arrived
« Reply #255 on: October 10, 2018, 01:39:38 AM »
You can only increase the precision if you reduce the size of the sampling window. So there is still a shifting window, and it is still every 20 seconds but now the sampling window time width is maybe the total rainfall for the past minute (much shorter). Then you multiply that the 1 minute rate result by 60 and you now have a more precise rate for that moment in time "per hour." All the while the recalculation keeps happening every 20 seconds.

So if it is drizzling outside, and after 5 hours your 0.01" tipping bucket finally tips, you're going to chart a rain-rate value of 0.60" inches per hour for that minute?

Not sure happens with one tip. But I don't think it would report 0.60 per hour.

It's what your explanation seems to predict.

Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
Re: Atlas Has Arrived
« Reply #256 on: October 10, 2018, 01:47:16 AM »
My explanation was just to simplify the math to explain the difference between the metbods. I'm sure there is more complexity. Like it probably knows it didn't rain for the full minute since it samples every 16 seconds. Which means it samples 3.75 times in 1 minute. So perhaps you need to divide by 3.75 before multiplying by 60. So maybe the rate is 0.16 on first tip. But who knows it may be smarter than that. I don't know the real formula.
Ecowitt GW1000 | Meteobridge on Raspberry Pi
WU: KFLWINTE111  |  PWSweather: KFLWINTE111
CWOP: FW3708  |  AWEKAS: 14814
Windy: pws-f075acbe
Weather Underground Issue Tracking
Tele-Pole

Offline nincehelser

  • Forecaster
  • *****
  • Posts: 3337
Re: Atlas Has Arrived
« Reply #257 on: October 10, 2018, 01:51:15 AM »
My explanation was just to simplify the math to explain the difference between the metbods. I'm sure there is more complexity. Like it probably knows it didn't rain for the full minute since it samples every 16 seconds. Which means it samples 3.75 times in 1 minute. So perhaps you need to divide by 3.75 before multiplying by 60. So maybe the rate is 0.16 on first tip. But who knows it may be smarter than that. I don't know the real formula.

Yeah... there's more to this than you realize.

Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
Re: Atlas Has Arrived
« Reply #258 on: October 10, 2018, 01:59:57 AM »
Found a 0.01 in my WU. Take a look at October 8th at 9:39AM. That was the first lonely 0.01 for the day. The rate is 0.05 inches per hour.

 [ You are not allowed to view attachments ]
« Last Edit: October 10, 2018, 02:16:19 PM 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 CW2274

  • Forecaster
  • *****
  • Posts: 6731
    • Conditions @ CW2274 West Tucson-Painted Hills Ranch
Re: Atlas Has Arrived
« Reply #259 on: October 10, 2018, 02:06:41 AM »
Found a 0.01 in my WU. Take a look at October 8th at 9:39AM. That was the first lonely 0.01 for the day. The rate is 0.05 inches per hour.
How do you get a rate from one tip? Maybe the spoon fell faster than normal... :-k

Offline nincehelser

  • Forecaster
  • *****
  • Posts: 3337
Re: Atlas Has Arrived
« Reply #260 on: October 10, 2018, 03:04:44 AM »
Found a 0.01 in my WU. Take a look at October 8th at 9:39AM. That was the first lonely 0.01 for the day. The rate is 0.05 inches per hour.

I take it you are not reporting as per the protocol, correct?

Otherwise, you should get a rate of 0.01" for a single tip.


Offline radioman61

  • Senior Member
  • **
  • Posts: 93
Re: Atlas Has Arrived
« Reply #261 on: October 10, 2018, 09:28:18 AM »
Since a rainfall event is cumulative, I don't think a snapshot in time is a valid measurement. 

Offline nincehelser

  • Forecaster
  • *****
  • Posts: 3337
Re: Atlas Has Arrived
« Reply #262 on: October 10, 2018, 09:32:35 AM »
Since a rainfall event is cumulative, I don't think a snapshot in time is a valid measurement.

There is no "snapshot" in time.  Like I said, saying the sample rate is an hour is very misleading.

"rainin" is updated every 36 seconds for a 5n1.

Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
Re: Atlas Has Arrived
« Reply #263 on: October 10, 2018, 11:12:00 AM »
Even though it tips just once it isn't a single tip being used in the calculation of rain rate. The absence of a tip also counts (it counts as 0 rain for that moment that it could have tipped). Every tipping mechanism has a maximum threshold for maximum tips per second. For ease of calculation and just as an example lets assume the tipping threshold is once per second. It could very well be higher than once per second, but lets keep it simple to grasp the concept. Please realize every tipping mechanism does have a maximum limit and this is why in really heavy rains with some tipping hardware is more prone to errors as there is then spillage if you exceed the tipping mechanism threshold.

Okay good, got that out of the way. Now back to my example of a tipping mechanism with a threshold of once per second. In this case having a single tip is not really a single tip because the system holds a memory of previous tips or possible tips that didn't occur and this is called data aging. This data age limit is hard coded in the software/firmware of the system. In my case since the reported rate for one tip was 0.05 inches per hour, I'm also going to assume my tipping mechanism is once per second (it could be more or less)...that part doesn't matter...it is what it is and then the calculation is done accordingly. So for 1 tip and with a tipping mechanism maximum of once per second for the result to be 0.05 inches per second means that the data aging is hard coded to about 10 minutes...or about 600 potential tips within that time frame. So it isn't one tip that gives you the rate. It becomes 1 tip out of 600 possible tips. Now you have a rate. 1 tip within the last 10 minutes. If you then multiply by 6 to get the rate per hour (10 minutes x 6 = 1 hour) you end up with 0.06 per hour. I didn't get 0.05 for some reason and it probably has to do with the fact that I don't really know the formula nor do I know the real data aging that Ambient is using nor do I know the threshold of the tipping mechanism (I'm also not using the Meteobridge to upload to WU), but I still think I'm close, and the bigger picture should now be clear. If you want to increase the accuracy then you increase the data age from 10 minutes to 1 hour. If you have more than 2 tips within 10 minutes then you now have also achieved greater accuracy. So for low drizzles of 1 tip per 10 minutes the accuracy is low. Having low accuracy for low numbers in a measurement is not the end of the world, our different hardware is constrained by many accuracy design limitations (temperature, humidity, wind...etc). All hardware is different. It doesn't stop us from comparing temperatures or wind or whatever.

The bigger point here it to learn that one tip alone doesn't give a rate. There really is more there than just one tip. The potential tips for the given time frame of the data aging also count.

Meteobridge data age setting change (default 10 minutes):
 [ You are not allowed to view attachments ]
« Last Edit: October 10, 2018, 11:40:29 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 DadCooks

  • Member
  • *
  • Posts: 17
Re: Atlas Has Arrived
« Reply #264 on: October 10, 2018, 01:02:09 PM »
Excellent representation of precision vs accuracy!  Probably explains why my chosen home defense weapon is a shotgun...

In reality, it’s only software we’re dealing with.  Would it be that difficult for them to give us an option to select pinpoint or averaging?

Simply Two Thumbs Up

Offline nincehelser

  • Forecaster
  • *****
  • Posts: 3337
Re: Atlas Has Arrived
« Reply #265 on: October 10, 2018, 02:42:15 PM »
Even though it tips just once it isn't a single tip being used in the calculation of rain rate. The absence of a tip also counts (it counts as 0 rain for that moment that it could have tipped). Every tipping mechanism has a maximum threshold for maximum tips per second. For ease of calculation and just as an example lets assume the tipping threshold is once per second. It could very well be higher than once per second, but lets keep it simple to grasp the concept. Please realize every tipping mechanism does have a maximum limit and this is why in really heavy rains with some tipping hardware is more prone to errors as there is then spillage if you exceed the tipping mechanism threshold.

Okay good, got that out of the way. Now back to my example of a tipping mechanism with a threshold of once per second. In this case having a single tip is not really a single tip because the system holds a memory of previous tips or possible tips that didn't occur and this is called data aging. This data age limit is hard coded in the software/firmware of the system. In my case since the reported rate for one tip was 0.05 inches per hour, I'm also going to assume my tipping mechanism is once per second (it could be more or less)...that part doesn't matter...it is what it is and then the calculation is done accordingly. So for 1 tip and with a tipping mechanism maximum of once per second for the result to be 0.05 inches per second means that the data aging is hard coded to about 10 minutes...or about 600 potential tips within that time frame. So it isn't one tip that gives you the rate. It becomes 1 tip out of 600 possible tips. Now you have a rate. 1 tip within the last 10 minutes. If you then multiply by 6 to get the rate per hour (10 minutes x 6 = 1 hour) you end up with 0.06 per hour. I didn't get 0.05 for some reason and it probably has to do with the fact that I don't really know the formula nor do I know the real data aging that Ambient is using nor do I know the threshold of the tipping mechanism (I'm also not using the Meteobridge to upload to WU), but I still think I'm close, and the bigger picture should now be clear. If you want to increase the accuracy then you increase the data age from 10 minutes to 1 hour. If you have more than 2 tips within 10 minutes then you now have also achieved greater accuracy. So for low drizzles of 1 tip per 10 minutes the accuracy is low. Having low accuracy for low numbers in a measurement is not the end of the world, our different hardware is constrained by many accuracy design limitations (temperature, humidity, wind...etc). All hardware is different. It doesn't stop us from comparing temperatures or wind or whatever.

The bigger point here it to learn that one tip alone doesn't give a rate. There really is more there than just one tip. The potential tips for the given time frame of the data aging also count.

Meteobridge data age setting change (default 10 minutes):
 [ You are not allowed to view attachments ]

 I. Just. Can't. Even.  #-o

Let's get this thread back on topic.  It's about the Atlas, not about the way wunderground handles rain-rate.




Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
Re: Atlas Has Arrived
« Reply #266 on: October 10, 2018, 03:19:06 PM »
nincehelser,
Did you not get anything out of all this???? I only had to get more and more technical because you just aren't accepting that things can be different than what you know or than what you are used to and maybe there is more than one right way. You are right in many ways for various topics. You are very helpful in this forum. I've learned from you. Or is it is true that you can't teach an old dog new tricks? (no offense intended). Come on man!

I gave it a really really good shot though.

By the way you are right about wind speed and gusts. The premise is almost the same as what we've been discussing with rain rate. Unless everyone agrees and standardizes the hardware design capabilities, the sampling rate, data age limit, and what the reporting period is, then we will always have slight differences in wind reporting.

It would be nice if someone else weighed in on this. Because maybe I made a mistake somewhere. Or maybe I didn't but there may be a better way to convey these ideas and methods across and someone else may be a better teacher than me.

Alright well whatever it is hopefully this helped someone out there. I have no problems dropping this.  Back to discussing the Atlas...
« Last Edit: October 10, 2018, 03:23:41 PM 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 nincehelser

  • Forecaster
  • *****
  • Posts: 3337
Re: Atlas Has Arrived
« Reply #267 on: October 10, 2018, 03:24:23 PM »
nincehelser,
Did you not get anything out of all this???? I only had to get more and more technical because you just aren't accepting that things can be different than what you know or than what you are used to and maybe there is more than one right way. You are right in many ways for various topics. You are very helpful in this forum. I've learned from you. Or is it is true that you can't teach an old dog new tricks? (no offense intended). Come on man!

No, I did not get anything out of it.    Your lack of knowledge and experience really shows (no offense intended).

Offline galfert

  • Global Moderator
  • Forecaster
  • *****
  • Posts: 6822
Re: Atlas Has Arrived
« Reply #268 on: October 10, 2018, 03:31:56 PM »
nincehelser,
Did you not get anything out of all this???? I only had to get more and more technical because you just aren't accepting that things can be different than what you know or than what you are used to and maybe there is more than one right way. You are right in many ways for various topics. You are very helpful in this forum. I've learned from you. Or is it is true that you can't teach an old dog new tricks? (no offense intended). Come on man!

No, I did not get anything out of it.    Your lack of knowledge and experience really shows (no offense intended).

Okay.  :shock:  Not the response I was expecting.  Hopefully we wont be stepping on each other's toes going forward.
Ecowitt GW1000 | Meteobridge on Raspberry Pi
WU: KFLWINTE111  |  PWSweather: KFLWINTE111
CWOP: FW3708  |  AWEKAS: 14814
Windy: pws-f075acbe
Weather Underground Issue Tracking
Tele-Pole

Offline worachj

  • Forecaster
  • *****
  • Posts: 475
  • Twin Cities, Minnesota
Access during power blackout
« Reply #269 on: October 10, 2018, 05:28:11 PM »
I’ve had my Access for little over two weeks now. As an experiment I turned off all power to the Access and my modem/router to simulate a blackout. After 30-minutes I turned everything back on. All data sent by my Atlas during the 30-minute backout was updated to MyAcuRite and worked as expected, but the daily rain total was lost and reset to zero, not expected.

Everything seemed to work as expected except the rain total being reset to zero. Rain total on WU was also reset to zero. Seems to be an Access bug to me.


AcuRite Atlas
KMNSAINT173

Offline DoctorKnow

  • Forecaster
  • *****
  • Posts: 1982
Re: Atlas Has Arrived
« Reply #270 on: October 11, 2018, 09:03:46 AM »
I wonder if any "bird spikes" have been discussed for the Atlas ?

Offline markmanxp

  • Member
  • *
  • Posts: 29
Re: Atlas Has Arrived
« Reply #271 on: December 06, 2018, 02:03:48 PM »
I've looked and searched here also but I unable to find my answer.  My apologies if I am just not looking in the right place.   Looking to purchase an Atlas and looking on Amazon I have seen 01007m, 01008m, and 01009m.  What's the difference between these three models?  Thanks

Offline Sir_MAK

  • Senior Member
  • **
  • Posts: 95
Re: Atlas Has Arrived
« Reply #272 on: December 06, 2018, 02:13:59 PM »
01007m comes with the HD display, Access Hub (for remote monitoring), and comes with the lighting detector.
01008m comes with the Access Hub for remote monitoring - no HD display, no lighting detector.
01009m comes with the HD display only.
Ambient WS-2000, Central Ohio - WU WC
Ambient
WS-2902, Lake Erie   -    WU WC

Offline markmanxp

  • Member
  • *
  • Posts: 29
Re: Atlas Has Arrived
« Reply #273 on: December 15, 2018, 02:43:18 PM »
Thank you Sir Mak!  Your information was very helpful.  I appreciate it. I went with the 01007m.

Offline ambientweather

  • Contributor
  • ***
  • Posts: 140
    • Ambient Weather
Re: Atlas Has Arrived
« Reply #274 on: August 04, 2020, 11:12:34 AM »
What is the insane deal with weather station makers insisting that you can only access YOUR DATA from THEIR CHOICE OF WEBSITES? As far as I am concerned this removes Atlas and virtually everything from Ambient Weather from consideration. It is absolutely CHEAPER to just include a USB port and let the OWNER decide what they want to do with THEIR data.

I'm sure that fanboys will screech at me, but AcuRite and Ambient have taken VERY SPECIFIC STEPS TO CONTROL YOUR DATA!

Or you could just pull the data directly from the airwaves with RTL_433. 

I expect it won't take long for someone to decode the Atlas transmissions.

If anyone Thinks that the ultimate goal is not to strongly encrypt the data so that you can consume it only when and where they want, then you know nothing of software and hardware development. They did NOT invest the money to develop the system this way for no reason - they did so under a calculated plan to keep charging you. Eventually they will charge you monthly just to see the data from the station that you PAID FOR. Netatmo encrypted their packets and were pretty honest that they simply don't care what YOU want to do with YOUR purchased product.

WU takes FREE DATA FROM YOU and now wants to charge you hundreds of dollars per MONTH for the PRIVILEGE of using their API to access the FREE DATA that YOU sent to them. Their next step is to make ALL ACCESS to their website subscription only.

I'm an extra class ham radio operator with a PhD in Electrical Engineering and a PhD in Mechanical Engineering - I can build my own system in my sleep that will allow me to send my data wherever I want, so this does not really affect ME, but I do realize that this will affect the vast majority of weather enthusiasts that wish to provide access to THEIR data on THEIR chosen venue, and THAT is patently, sickeningly WRONG.

And how exactly did you come to the conclusion that we are going to start charging for data? Did you receive inside information? Because I developed the service and it never crossed my mind.

Please elaborate.

Regards,

Ed