Could you explain the graph of the cloud model? T67, corrected t, y-axis and abscissa, k1… k7. Is there any way to know the coefficients? I have been looking at the manual but it is not clear to me.
I have also noticed that if you send the cloud model to sqm in admin tab 4, the cloud flag shows 0 when it should be 30.
Thanks a lot
Last edit: Fernando 2021-07-31
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Fernando
My aplogies for the dealy in reply to your message.
Right now I am in the midst of the next big update for this project, so it is very time consuming.
If anyone out there has any tips that would be immensely helpful. The required readings are listed at the end of this reply.
First off, I own up and say that I have not played with this as much as others have. And I am sure they know I darn sight more than me about this.
The cloud model was/is part of the meteostation so all the work was done by others, not me. I can only talk about this from a personal perspective,
The cloud model.
1. Only valid during the dark hours - not valid during the day or twilight or sunrise
2. Requires the MLX90614 sensor
3. Sensor needs to point direct up to the sky and not be affected by stray lights or street lights
Warning: This is gonna take some time and effort by the user. it is not a 5 minute fix.
We start by using the values of IR ambient and IR object from the MLX90614 sensor.
These values are manipulated with a Model (called the cloud model) to generate a "Corrected" value for the sky temperature.
The MLX sensor, will give a different value for IR when there are clouds - compared to the same when there are no clouds.
By using a set of values we can extrapolate a model - we call it the cloud model - which we then feed in our mlx 90614 sensor readings (IR object/IR ambient) to give us a cloud percentage value 0-100 or Skystate (Clear, Cloudy, Very Cloudy)
This is covered near the end of the PDF
Paul is the one who enhanced this model (from Lunatico, and Indiduino MeteoStation).
Paul suggests you use the default settings. It will take you some time to tune the model.
Use the default settings.
Start when there is a partly cloudy night, not completely overcast, patchy clouds not dark thunderstorm clouds.
What you are trying to do is adjust a few things so that the cloud cover percent reflects what your observed clouds are,
What I like to do is use the profile setting. So I set the profile to say profile 2 and then save it.
So I can always go back to this profile2 if needed
We need to do this first
On a clear night I can see that my irobject is around -16. (anywhere between -14 and -18)
So I would set the tempclear to say -12.
On an overcast night with about 50% clouds or more I get an irobject reading of anywhere from -4 to +3. So I set tempovercast to -5.
Or if you logging data values you can go back over the logs for previous days and view the temperature readings for irambient and irobject. Then use the log to help you determine values for temp clear and temp overcast
having set the values of tempclear and tempovercast,
Then I select profile1 and load it.
you get a model by trial and error
So start with K1.
When changing a value, try adding or subtracting a value of like 16 at a time
If you change a value of K1 and it does not seem to make any difference then reset K1 to the default setting and move onto K2.
If say I subtracted 16 from K1 and the cloud percentage value got worse (compared to my clouds) then I would try adding 16. So I add 16 it gets a little better, I add another 16, no change.
So I subtract 16 for K1. save the profile and move onto the next K2
And so on. Hopefully I end up with something that is better than when I started
The values of Temp Clear and Temp Overcast are setpoints. That is why you needed to start with those set to meaningful value
They help determine the sky state.
This is handled in the code by getskystate().
This code takes the measured values from the MLX90614 sensor, applies the cloud model to it to generate a corrected temperature value,
The corrected temperature value is then compared to our setpoints.
If the corrected temp is < TempClear value then the sky is clear
If the corrected temp > TempCloudy value then the sky is cloudy.
and then we generate a cloud coverage percentage value from our results.
There is really no way to tell you to use "these magic numbers" because it is going to differ for every user. And to make matters worse, you may need to adjust the values for changes in seasons as well (esp TempClear and TempOvercast)
So tempclear on a clear night has to be a few degrees warmer than IR object (I put it as IR object) and temp overcast a little cooler than IR object on a partially cloudy night.
A few nights ago IR object was -15ºC and the next day it was -8ºC with the same room temperature in both nights, that's why I asked.
Do you know what temperatures are the ordinate and abscissa axes?
Thank you so much for everything.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
On a clear night I can see that my irobject is around -16. (anywhere between -14 and -18)
So I would set the tempclear to say -12.
On an overcast night with about 50% clouds or more I get an irobject reading of anywhere from -4 to +3. So I set tempovercast to -5.
I would have put tempclear -16 and temp overcast -4, but you put tempclear -12 and temp overcast -5.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I would have put tempclear -16 and temp overcast -4, but you put tempclear -12 and temp overcast -5.
But if you do that then an irobject value of -14 would not register as tempclear?
For tempclear
On a clear night I can see that my irobject is around -16. (anywhere between -14 and -18)
So I would set the tempclear to say -12.
mycase (-14 to -18 with tc = -12)
If the corrected temp < Tempclear value then the sky is cloudy.
if -14 < -12 then sky is clear // OK
if -18 < -12 then sky is clear // OK
so this works
your case (-14 to -18 with tc = -16)
If the corrected temp < tempCleary value then the sky is cloudy.
if -14 < -16 then sky is clear // FAIL
if -18 < -16 then sky is clear // OK
I would have put temp overcast -4,
For tempovercast
On an overcast night with about 50% clouds or more I get an irobject reading of anywhere from -4 to +3. So I set tempovercast to -5.
mycase (-4 to +3 with toc = -5)
If the corrected temp > TempCloudy value then the sky is cloudy.
if -4 > -5 then sky is clear // OK
if +3 > -t then sky is clear // OK
so this works
your case (-4 to +3 with toc = -4)
If the corrected temp > TempCloudy value then the sky is cloudy.
if -4 > -4 then sky is clear // FAIL
if +3 > -4 then sky is clear // OK
Last edit: brownrb 2021-08-04
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi Miguel
q1
As stated and credited in code/pdf, the mysqm+ cloud model is based on that code. the same forumlas etc with a little change here and there. The same sensor is used. So it would be fair to say its close enough.
q2
What is suggested in the pdf is essentially the same as the link u posted. The reference on how to set up towards the end of the pdf was provided by Paul from his real world experience on his gear. There should be enough guidence there. Short answer = yes
As pointed out in the pdf
1- dont play with that if it is not nightime
2-values will change often, as well as season to season (which is why the windows app provides the ability to save these into presets files.
3-it is done by trial and error, no shortcuts or magic wands are available
4-relying on some-one elses values is pointless (unless they are just over the neighbors fence
5-it takes time and time and time
Regards
Robert
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Robert, thanks for the info. However I would highlight something in the links provided:
"To tune this model, one should observe the Cloud Conditions graph from sunup to sunset for a clear day."
I know you don't have a cloud graph implemented, but we can log the values and come up with a graph for cloud % and assess the trend from there, all of this during daytime.
I'll try myself this method on a couple of cloudless days and will leave feedback in here for future reference.
Thanks again
Miguel
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Robert, thanks for the info. However I would highlight something in the links provided:
"To tune this model, one should observe the Cloud Conditions graph from sunup to sunset for a clear day."
I know you don't have a cloud graph implemented, but we can log the values and come up with a graph for cloud % and assess the trend from there, all of this during daytime.
I'll try myself this method on a couple of cloudless days and will leave feedback in here for future reference.
Hello Robert.
Could you explain the graph of the cloud model? T67, corrected t, y-axis and abscissa, k1… k7. Is there any way to know the coefficients? I have been looking at the manual but it is not clear to me.
I have also noticed that if you send the cloud model to sqm in admin tab 4, the cloud flag shows 0 when it should be 30.
Thanks a lot
Last edit: Fernando 2021-07-31
Hi Fernando
My aplogies for the dealy in reply to your message.
Right now I am in the midst of the next big update for this project, so it is very time consuming.
If anyone out there has any tips that would be immensely helpful. The required readings are listed at the end of this reply.
First off, I own up and say that I have not played with this as much as others have. And I am sure they know I darn sight more than me about this.
The cloud model was/is part of the meteostation so all the work was done by others, not me. I can only talk about this from a personal perspective,
The cloud model.
1. Only valid during the dark hours - not valid during the day or twilight or sunrise
2. Requires the MLX90614 sensor
3. Sensor needs to point direct up to the sky and not be affected by stray lights or street lights
Warning: This is gonna take some time and effort by the user. it is not a 5 minute fix.
We start by using the values of IR ambient and IR object from the MLX90614 sensor.
These values are manipulated with a Model (called the cloud model) to generate a "Corrected" value for the sky temperature.
The MLX sensor, will give a different value for IR when there are clouds - compared to the same when there are no clouds.
By using a set of values we can extrapolate a model - we call it the cloud model - which we then feed in our mlx 90614 sensor readings (IR object/IR ambient) to give us a cloud percentage value 0-100 or Skystate (Clear, Cloudy, Very Cloudy)
This is covered near the end of the PDF
Paul is the one who enhanced this model (from Lunatico, and Indiduino MeteoStation).
Paul suggests you use the default settings. It will take you some time to tune the model.
Use the default settings.
Start when there is a partly cloudy night, not completely overcast, patchy clouds not dark thunderstorm clouds.
What you are trying to do is adjust a few things so that the cloud cover percent reflects what your observed clouds are,
What I like to do is use the profile setting. So I set the profile to say profile 2 and then save it.
So I can always go back to this profile2 if needed
We need to do this first
On a clear night I can see that my irobject is around -16. (anywhere between -14 and -18)
So I would set the tempclear to say -12.
On an overcast night with about 50% clouds or more I get an irobject reading of anywhere from -4 to +3. So I set tempovercast to -5.
Or if you logging data values you can go back over the logs for previous days and view the temperature readings for irambient and irobject. Then use the log to help you determine values for temp clear and temp overcast
having set the values of tempclear and tempovercast,
Then I select profile1 and load it.
you get a model by trial and error
So start with K1.
When changing a value, try adding or subtracting a value of like 16 at a time
If you change a value of K1 and it does not seem to make any difference then reset K1 to the default setting and move onto K2.
If say I subtracted 16 from K1 and the cloud percentage value got worse (compared to my clouds) then I would try adding 16. So I add 16 it gets a little better, I add another 16, no change.
So I subtract 16 for K1. save the profile and move onto the next K2
And so on. Hopefully I end up with something that is better than when I started
The values of Temp Clear and Temp Overcast are setpoints. That is why you needed to start with those set to meaningful value
They help determine the sky state.
This is handled in the code by getskystate().
This code takes the measured values from the MLX90614 sensor, applies the cloud model to it to generate a corrected temperature value,
The corrected temperature value is then compared to our setpoints.
If the corrected temp is < TempClear value then the sky is clear
If the corrected temp > TempCloudy value then the sky is cloudy.
and then we generate a cloud coverage percentage value from our results.
There is really no way to tell you to use "these magic numbers" because it is going to differ for every user. And to make matters worse, you may need to adjust the values for changes in seasons as well (esp TempClear and TempOvercast)
For added reading see
READING MATERIAL
https://indiduino.wordpress.com/2013/02/02/meteostation/
http://lunatico.es/aagcw/TechInfo/SkyTemperatureModel.pdf
http://lunatico.es/aagcw/enhelp/
_
Thanks so much for the explanation.
So tempclear on a clear night has to be a few degrees warmer than IR object (I put it as IR object) and temp overcast a little cooler than IR object on a partially cloudy night.
A few nights ago IR object was -15ºC and the next day it was -8ºC with the same room temperature in both nights, that's why I asked.
Do you know what temperatures are the ordinate and abscissa axes?
Thank you so much for everything.
This is what I did not understand.
I would have put tempclear -16 and temp overcast -4, but you put tempclear -12 and temp overcast -5.
I would have put tempclear -16 and temp overcast -4, but you put tempclear -12 and temp overcast -5.
But if you do that then an irobject value of -14 would not register as tempclear?
For tempclear
On a clear night I can see that my irobject is around -16. (anywhere between -14 and -18)
So I would set the tempclear to say -12.
mycase (-14 to -18 with tc = -12)
If the corrected temp < Tempclear value then the sky is cloudy.
if -14 < -12 then sky is clear // OK
if -18 < -12 then sky is clear // OK
so this works
your case (-14 to -18 with tc = -16)
If the corrected temp < tempCleary value then the sky is cloudy.
if -14 < -16 then sky is clear // FAIL
if -18 < -16 then sky is clear // OK
I would have put temp overcast -4,
For tempovercast
On an overcast night with about 50% clouds or more I get an irobject reading of anywhere from -4 to +3. So I set tempovercast to -5.
mycase (-4 to +3 with toc = -5)
If the corrected temp > TempCloudy value then the sky is cloudy.
if -4 > -5 then sky is clear // OK
if +3 > -t then sky is clear // OK
so this works
your case (-4 to +3 with toc = -4)
If the corrected temp > TempCloudy value then the sky is cloudy.
if -4 > -4 then sky is clear // FAIL
if +3 > -4 then sky is clear // OK
Last edit: brownrb 2021-08-04
Hi Robert,
I wonder if the cloud model implemented is this one: https://www.lunaticoastro.com/aagcw/enhelp/topics/idh-topic180.htm
And I still didn't follow the calibration procedure in this topic, but would you consider the method in the following link a valid one for cloud model calibration?
https://www.lunaticoastro.com/aagcw/enhelp/topics/idh-topic350.htm
Thanks
Miguel
Hi Miguel
q1
As stated and credited in code/pdf, the mysqm+ cloud model is based on that code. the same forumlas etc with a little change here and there. The same sensor is used. So it would be fair to say its close enough.
q2
What is suggested in the pdf is essentially the same as the link u posted. The reference on how to set up towards the end of the pdf was provided by Paul from his real world experience on his gear. There should be enough guidence there. Short answer = yes
As pointed out in the pdf
1- dont play with that if it is not nightime
2-values will change often, as well as season to season (which is why the windows app provides the ability to save these into presets files.
3-it is done by trial and error, no shortcuts or magic wands are available
4-relying on some-one elses values is pointless (unless they are just over the neighbors fence
5-it takes time and time and time
Regards
Robert
Robert, thanks for the info. However I would highlight something in the links provided:
"To tune this model, one should observe the Cloud Conditions graph from sunup to sunset for a clear day."
I know you don't have a cloud graph implemented, but we can log the values and come up with a graph for cloud % and assess the trend from there, all of this during daytime.
I'll try myself this method on a couple of cloudless days and will leave feedback in here for future reference.
Thanks again
Miguel
Awesome, thanks, if you can sure your findings that will help others too, much appreciated.
Regards
Robert
On Tuesday, October 25, 2022, 04:55:32 AM GMT+13, molino molino26@users.sourceforge.net wrote:
Robert, thanks for the info. However I would highlight something in the links provided:
"To tune this model, one should observe the Cloud Conditions graph from sunup to sunset for a clear day."
I know you don't have a cloud graph implemented, but we can log the values and come up with a graph for cloud % and assess the trend from there, all of this during daytime.
I'll try myself this method on a couple of cloudless days and will leave feedback in here for future reference.
Thanks again
Miguel
graphic cloud model
Sent from sourceforge.net because you indicated interest in https://sourceforge.net/p/mysqmproesp32/discussion/general/
To unsubscribe from further messages, please visit https://sourceforge.net/auth/subscriptions/