I've compiled an alpha version of the upcoming r24x release. There's quite some new functionality in the pipeline with this release. Getting feedback on the fix for the non-integer fractional meter constants bug would be very helpful right now. Leave a comment if you'd like to be at the cutting edge and migrate to r241. I'll then upgrade your FLM remotely. Please do not sign up if you lack basic ssh and linux command line skills as this alpha version might still have some rough edges.
Cheers
/Bart
Well, someone has to volunteer first. Count me in, if only as a spare ... - Jos
Sign me up please!
Michi.
+1
@jgysenbergs @michi @frumper Your FLM's have been successfully upgraded to r241.
@michi Let me know how the 0.4 meterconstant works out with this new firmware.
Count me in too please :) would love to try out the fix for non-integer fractional meter constants
Cheers
Adam
done
Hi Bart, looking really good so far, thank you!
The minute display is no rock-steady. I see very (very) small variations. I suspect these are caused by my new fridge, which has a linear compressor, and so adjusts consumption on a continuous basis. But certainly, the wild fluctuations I had previously have completely disappeared.
I'll take a reading on my DIN-rail meter today and compare with the value the Flukso reports 24 hours later. That should server as a sanity check as to the accuracy of the calculations.
Thanks again for this fix, this really makes a huge difference!
Michi.
Thanks Bart :)
I've only had the new version for a short time, but initial testing seems to show that the massive fluctuations in usage appear to have stopped - there seems to be some small fluctuations still, however i'm unsure if this is just normal variations in usage with different devices switching on/off and adjusting their usage on a continuous basis.
i'll keep an eye on it during the day and do some more testing, i'll get back to you towards the end of the day and let you know.
This is a link to my PVoutput graph for today, you can see some small fluctuations over night. The usage should be fairly stable during the night, whilst were asleep.
http://pvoutput.org/intraday.jsp?id=28541&sid=26142&dt=20140629&gs=0&m=1
This is a link to my PVoutput graph for a few days ago, where you can see the fluctuations over night from the integer bug
http://pvoutput.org/intraday.jsp?id=28541&sid=26142&dt=20140625&gs=0&m=0
i've also got a current cost running with 2 x iams on the fridge. I'll line up the usage of the fridge overnight with the flukso graph over night and see if the variations are indeed the fridge adjusting its usage etc.
I also just want to add a big THANK YOU to Bart! for this fix. From the initial testing I've done so far, it has made a massive difference.
I can now see the difference in usage when turning devices on and off, with the variation appearing to be in the neighborhood of a few watts now - previously i'd have a variation of anything from 100w to over 1.5kWh. It got worse as the consumption increased.
I think the remaining beat effects might be due to the fact that the Flukso can measure only when a pulse actually arrives. So, if a pulse just misses then end of a 5-minute sampling interval on PVOutput, you get a lower-than-actual reading, and if a pulse arrives just after the start of a 5-minute sampling interval, you get a higher-than-actual reading. As a percentage error, this should be more noticeable a low consumption levels than at high ones (because the pulse rate of the meter is a fixed number of Wh per pulse).
I don't think there is anything that can be done about this. The meter sends it's pulses whenever consumption has crossed the next Wh boundary for that meter (2.5 Wh with a 400-pulse per kWH meter).
It may also be difficult to truly hold consumption constant in a house. Certainly, anything with variable consumption, such as fridges, would have to be removed. And even things such as backup disk drives and the like will draw different amounts of power when they are active and asleep.
To be really sure, you'd have to disconnect absolutely everything that could have varying consumption and only leave purely constant loads on, such as lights.
Michi.
Thanks for the feedback!
The real-time readings are calculated based on the time difference between two consecutive (fractional) pulses. Time stamping is done in millisec resolution in the microcontroller. There might indeed still be some artefacts due to the discrete nature of the pulse interface. But this is the best we can do for calculating real-time readings on this type of interface.
Thanks for the comments, Bart!
I agree, perfection is impossible with a pulse output. And it's not needed either. I now get accurate consumption figures with my pulse meter. I'll try and see whether I can work out what the percentage error is when I get a quiet weekend with no-one else at home, so I can tinker with the house consumption by turning most things off at the circuit breakers and having only guaranteed constant loads running.
I would expect that if the actual consumption is low, the percentage error would be a bit larger, depending where exactly each pulse falls whereas, at high actual consumption, accounting for a 2.5 Wh pulse in the current or the next interval is neither here nor there, so the percentage error would be really small.
But, as far as I am concerned, the new 241 version is right on the ball. It's a massive improvement over the previous one!
Cheers,
Michi.
OK, I've taken two readings of my DIN rail meter exactly 24 hours apart. The meter reports 31.66 kWh consumption, and the log on PVOutput reports 31.741 kWh for the same period. That's a discrepancy of 0.081 kWh, or about ¼ of a percent. In other words, that's bloody accurate, well within what one could reasonably expect.
So, it appears that the consumption calculation is still accurate for my 400 pulse per kWh meter.
Cheers,
Michi.
Bart,
Can we take a small peek at the changelog of the upcoming release ?
Can I please be signed up for the alpha? Haven't got it connected yet, but will have 0.625 pulse meters plugged in soon...
Your FLM needs to be connected to the internet in order to be able to push an alpha firmware to the FLM.
Sorry, should have elaborated - no sensors are connected but it's connected to the net. Last heartbeat was a minute ago...
@Greenhouse Your FLM should now be running the r241 firmware.