New rain rate calculation

[dane]

That's interesting to know. I assume therefore that the actual figures update every three reads?

No: as far as I can tell, the EasyWeather program's display updates when the station has stored new info into a memory slot, and EasyWeather picks up the data on its next read.

I looked at a debug.log from Cumulus (attached).
Cumulus seems to read (up to) 8 records from the station to get all data from a memory slot. Right?
The "real" information seems to be in the last 4 records. Right?

Some times, though, Cumulus only reads the first 4 records.
This seems to be the case every time there is a change to a new memory slot (last 2 bytes of record 4). Consequently Cumulus seems to miss all data at that read, and only picks it up 48 secs later (latest build). See 20:44:42 in attached debug.log.

At 20:25:02 Cumulus also reads only 4 records. Reason: I don't know. But I happened to notice that the "last data read" time was not updated on the display between 20:24:14 and 20:25:50 - leaving me with the impression that Cumulus missed a "heart beat" at 20:25:02.

I'll leave it to you to make sense of all this....

[steve]

No: as far as I can tell, the EasyWeather program's display updates when the station has stored new info into a memory slot, and EasyWeather picks up the data on its next read.

You mean that some of the data updates more frequently than every 48 seconds?

Cumulus seems to read (up to) 8 records from the station to get all data from a memory slot. Right?
The "real" information seems to be in the last 4 records. Right?

Yes - the block of data at address 0000 contains the address of the block containing the current data. So it reads the first block, then it reads the data block.

Some times, though, Cumulus only reads the first 4 records.
This seems to be the case every time there is a change to a new memory slot (last 2 bytes of record 4). Consequently Cumulus seems to miss all data at that read, and only picks it up 48 secs later (latest build). See 20:44:42 in attached debug.log.

Yes, this is intentional. When the station moves to a new memory location, it updates the pointer in the address 0000 block before it has written new data to the next location. This is a bug, as far as I can see. It means that there is a window where the 'current' memory location actually contains old data. I changed Cumulus to skip a read when the location changes, to work around this problem. I had forgotten about this. I guess this explains the apparent missing timer event, and why Cumulus can miss data, particularly on the longer update intervals.

Hopefully changing the interval to 10 seconds will avoid missing any data, but it does depend on the answer to my first question.

Steve

[dane]

You mean that some of the data updates more frequently than every 48 seconds?

No, the station updates every 48 secs, and EasyWeather picks it up next time its 16-secs timer pops.
Sorry if I wasn't making myself clear.

Thanks amillion for taking the time to explain all this detail about Cumulus.
(I think I'll shut up for a while to give you time to make friends with the woman who appears from time to time... Say hello to her from me

[steve]

No, the station updates every 48 secs, and EasyWeather picks it up next time its 16-secs timer pops.
Sorry if I wasn't making myself clear.

Right, so rather like Cumulus with its 15 second interval.

Thanks amillion for taking the time to explain all this detail about Cumulus.

No problem - I'll get around to writing a complete 'how it works' eventually.

(I think I'll shut up for a while to give you time to make friends with the woman who appears from time to time... Say hello to her from me

We've just spent a couple of hours of quality time together - doing a crossword

Steve

[Super-T]

Steve
On a previous Beta, when you had the timer set down to a smaller amount, I had major fluctuations with the rain rate of 72 or 74 mm/hour whereas there might have been the odd rain shower around. Beta 556 fixed that problem. If you change the updates down to 15 seconds this problem may re-occur?

[steve]

Yes, as I said above, the change to 48s was a quick fix for the new rate calculation; if I go back to 15s (or 10s) I'll have to change the rate calculation some other way.

Steve

[steve]

I've now uploaded build 563 which changes the interval to 10s, and attempts to make the new rain rate calculation behave more sensibly.

Steve

[steve]

Anyone had any rain since installing this version? How does the rain rate look?

Steve

[Frank02]

Hi Steve,

Had a little rain but not much, can't say if the figures were accurate but the initial reading showed 108 mm/h which is much more than what was falling at the time.

Have a great night.

[steve]

Drat, that doesn't sound good. I've forgotten, do you have a web site with a graph I can look at?

Steve

[Frank02]

Drat, that doesn't sound good. I've forgotten, do you have a web site with a graph I can look at?

Steve

No, sorry. I am just using it locally on my pc.

[Super-T]

Steve
That 108mm/h sounds like the problem I had when you last had a shortened time in there. Problem went away once you increased the time.

No rain here at presetn so I can't test.

I'll send my lady out to water the garden and wash thcar....that normally makes it rain!

[dane]

Drat, that doesn't sound good.

Well - it's raining outside, and that's what I see.
And wasn't that to be expected? 1 tip of the bucket: 0.3 mm in 1 interval of 10 secs = 0.3 x 360 = 108 mm/hr.
So that's the number I see on the "All time His/Los" display.
But on the normal Cumulus display the rainfall rate is 3.3 mm/hr.
(see attached screen capture).

Steve, I think you're trying to do something impossible: predict the rainfall per hour (based on 10 sec. observations). Wouldn't it be so much simpler - and more correct - to simply report the amount that actually fell during the last hour? i.e. not report anything until you had one hour's worth of data?

I could certainly live with that.

[steve]

Well - it's raining outside, and that's what I see.
And wasn't that to be expected? 1 tip of the bucket: 0.3 mm in 1 interval of 10 secs = 0.3 x 360 = 108 mm/hr.
So that's the number I see on the "All time His/Los" display.

I've allowed for the 10s interval for the first tip, and assumed it fell sometime in the last minute. I've now written a rain simulator and found the bug, it's what happens afterwards that causes the problem - but as you can tell from the numbers, it's a similar problem to the one you're suggesting (0.3mm over 10s). I'm not currently sure how to fix it.

But on the normal Cumulus display the rainfall rate is 3.3 mm/hr.
(see attached screen capture).

It goes down again quite quickly, so maybe you didn't see it on the main display at its peak?

Steve, I think you're trying to do something impossible: predict the rainfall per hour (based on 10 sec. observations).

I'm trying to give an instantaneous representation of the current rain rate. The Davis and other stations do it; it's simply an indication of the time between tips of the rain gauge. The problem I have is that each Cumulus reading doesn't correspond to a real reading of the gauge (10s vs 48s) and I don't know exactly when the gauge tipped. I'm sure I'll work something out.

Wouldn't it be so much simpler - and more correct - to simply report the amount that actually fell during the last hour? i.e. not report anything until you had one hour's worth of data?

I could certainly live with that.

That's pretty much what I did before for the quick-and-dirty solution. But you already have an indication of 'rain in last hour', this is something else; the value is supposed to change more rapidly than that to indicate how heavily the rain has been falling over the last minute or two.

Steve

[dane]

I think I understand what you are trying to do.
And I think my problem is basically the "mm/hour" annotation when what we are really talking about is "tips per station interval". (48 secs)
So the "rainfall rate" on the main display is your sliding average of no. of tips over some (?) period?
And the "All time high rain rate" is basically "highest no. of tips in one station interval"?
If so, I'd very much prefer to see something like: "All time high rain rate: 0.3 mm in 48 seconds" rather than "22.5 mm per hour" - giving the impression that it rained heavily for a whole hour!

You know when the station updates (new buffer address), so picking up the no. of tips per interval should be no problem?