Earth Notes: On Website Technicals (2021-04)
Updated 2022-10-16.2021-04-29: Core Web Vitals All Zeroes?
For a little while the GSC core be vitals poor, need improvement and good values have all been zero. All other metrics including page experience look good. I don't know how to interpret all three being zero! 
2021-04-26: Max Green?
MBA and water boiling for tea off-grid, hitting ~160W pulled from panels, and primary panels still floating above Vmpp! Yes it's sunny out there, and close to solar noon, and cool.
2021/04/26T13:17:25Z AL 10903 B1 14393 B2 -1 P 27793 BV 14030 ST VH D V A1P 155063 B1T 12 UC 100 A1V 31351 2021/04/26T13:17:32Z AL 10792 B1 14393 B2 -1 P 25980 BV 14030 ST VH D V A1P 156980 B1T 12 UC 100 A1V 31284 2021/04/26T13:17:38Z AL 10978 B1 14393 B2 -1 P 28585 BV 13047 ST VH D V A1P 160725 B1T 12 UC 100 A1V 31278 2021/04/26T13:17:44Z AL 10896 B1 14393 B2 -1 P 27506 BV 13986 ST VH D V A1P 155863 B1T 12 UC 100 A1V 31330
Yes, I will have green green tea again, not coffee! 
2021-04-25: JXL Easter Egg
The mechanism and details may change, but as of today I have embedded one of JS' JXL art images to appear as the hero image at the top of this (full-fat) page for capable browsers.
<picture> <source type=image/jxl srcset=img/abstract/art-1-JS.jxl> <source media=(max-width:640px) srcset=img/a/h/tools-1280w.l354283.640x80.l.png> <img class=resp width=800 height=200 alt=tools title=tools style=background:#ccc decoding=async src=img/a/h/tools-1280w.l354283.800x200.png> </picture>
Now to inline it properly...
<picture> <source type=image/jxl srcset=data:image/jxl;base64,/wr4j/4PWAkIGBAAUAAAAEAABABLOEHZrRc0OTx/AevCGRBIkCQcAAMD> <source media=(max-width:640px) srcset=img/a/h/tools-1280w.l354283.640x80.l.png> <img class=resp width=800 height=200 alt=tools title=tools style=background:#ccc decoding=async src=img/a/h/tools-1280w.l354283.800x200.png> </picture>
WebPageTest instances that I'm trying don't seem to work running Canary at the moment (with or without JXL enabled) so this really is bleeding edge!
2021-04-24: JXL
Jon Sneyers of JXL fame has created me two tiny (~40 bytes each!) JXL abstract artworks to use as possible hero images (inlined, for browsers that understand JXL) for these site-technicals pages!
I created the appropriate image/jxl MIME type for .jxl files being served for EOU by Apache.
Chrome Canary (92.0.4487.0) with the JXL flag enabled (chrome://flags/#enable-jxl) now displays them! Woot!
Also for giggles I created the same MIME type for .jxlL and .jxlLL variants.
2021-04-23: MacBook Off-grid
I have temporarily hard-wired in the car adaptor previously used to power the MacBook. (Its plug had partially melted under heavy load.)
With my phone charging and MacBook running, off-grid load has risen to ~35W, but at solar noon still the primary panels are floating above Vmpp!
2021/04/23T12:00:06Z AL 3259 B1 14335 B2 -1 P 13862 BV 14030 ST VH D V A1P 46715 B1T 14 UC 100 A1V 32993 ... 2021/04/23T12:09:55Z AL 4255 B1 14332 B2 -1 P 34498 BV 13846 ST VH D V A1P 62131 B1T 14 UC 100 A1V 32569
Giving the MacBook a beefy task (make pages!) shows load rise:
2021/04/23T12:25:24Z AL 4226 B1 14332 B2 -1 P 37335 BV 13855 ST VH D V A1P 53419 B1T 14 UC 100 A1V 32554 2021/04/23T12:25:31Z AL 4286 B1 14332 B2 -1 P 40001 BV 13653 ST VH D V A1P 69680 B1T 14 UC 100 A1V 32224 2021/04/23T12:25:37Z AL 4810 B1 14332 B2 -1 P 49131 BV 13653 ST VH D V A1P 68805 B1T 14 UC 100 A1V 32233
Leaving the adaptor powered up all night with no load consumes ~1W. Not unbearable outside winter, but better to disconnect when not in use.
2021/04/24T00:00:06Z AL 0 B1 12549 B2 -1 P 2887 BV 12424 ST OK - m A1P 0 B1T 14 UC 56 A1V 95 2021/04/24T00:10:06Z AL 0 B1 12549 B2 -1 P 2736 BV 12407 ST OK - m A1P 0 B1T 14 UC 56 A1V 83 2021/04/24T00:20:06Z AL 0 B1 12549 B2 -1 P 3640 BV 12424 ST OK - m A1P 0 B1T 14 UC 57 A1V 92 2021/04/24T00:30:06Z AL 0 B1 12549 B2 -1 P 2824 BV 12407 ST OK - m A1P 0 B1T 14 UC 56 A1V 83 2021/04/24T00:40:06Z AL 0 B1 12549 B2 -1 P 2736 BV 12380 ST OK - m A1P 0 B1T 14 UC 56 A1V 74 2021/04/24T00:50:06Z AL 0 B1 12549 B2 -1 P 2799 BV 12380 ST OK - m A1P 0 B1T 14 UC 56 A1V 71
2021-04-21: Phone Dump
Plugging in my Fairphone 3 to recharge a few Wh via my Skross 12V cigar plug USB supply to shift a little more load off-grid, maybe 6W:
2021/04/21T09:02:25Z AL 1607 B1 13349 B2 -1 P 3845 BV 13223 ST OK - s A1P 21410 B1T 14 UC 100 A1V 28852 2021/04/21T09:02:31Z AL 1607 B1 13349 B2 -1 P 4366 BV 13188 ST OK - s A1P 21440 B1T 14 UC 100 A1V 28855 2021/04/21T09:02:37Z AL 1605 B1 13349 B2 -1 P 3712 BV 12293 ST OK - s A1P 21425 B1T 14 UC 100 A1V 28855 2021/04/21T09:02:43Z AL 1605 B1 13349 B2 -1 P 5554 BV 13135 ST OK - s A1P 21425 B1T 14 UC 100 A1V 28855 2021/04/21T09:02:49Z AL 1609 B1 13343 B2 -1 P 12583 BV 13056 ST OK - s A1P 21471 B1T 14 UC 100 A1V 28852 2021/04/21T09:02:55Z AL 1609 B1 13337 B2 -1 P 11870 BV 13074 ST OK - s A1P 21440 B1T 14 UC 100 A1V 28855 2021/04/21T09:03:01Z AL 1612 B1 13334 B2 -1 P 11868 BV 12161 ST OK - s A1P 21350 B1T 14 UC 100 A1V 28855
(I see evidence of draw of up to ~10W at below 90% phone charge.)
Testing a 12V car fan new today (though not what was ordered!), maybe 8W:
2021/04/21T12:05:26Z AL 2334 B1 14271 B2 -1 P 14657 BV 13969 ST VH D V A1P 33806 B1T 16 UC 100 A1V 32123 2021/04/21T12:05:33Z AL 2368 B1 14271 B2 -1 P 15142 BV 13021 ST VH D V A1P 33776 B1T 16 UC 100 A1V 31479 2021/04/21T12:05:39Z AL 2919 B1 14271 B2 -1 P 22620 BV 13890 ST VH D V A1P 41642 B1T 16 UC 100 A1V 30704 2021/04/21T12:05:45Z AL 2902 B1 14271 B2 -1 P 23077 BV 13969 ST VH D V A1P 33383 B1T 16 UC 100 A1V 31403
2021-04-18: Two Brews
Making myself an early off-grid coffee before the sun is fully on the large panels, so the ~100W load is pulling mainly from the batteries. At 12.7V they don't seem to be suffering too badly, and 20W is coming from the main panels. Maybe the same from the secondary panels that are in full sunlight.
2021/04/18T10:18:11Z AL 1621 B1 12717 B2 -1 P 15057 BV 12380 ST OK D r A1P 20550 B1T 12 UC 70 A1V 29841 2021/04/18T10:18:17Z AL 1629 B1 12714 B2 -1 P 14660 BV 12372 ST OK D r A1P 20640 B1T 12 UC 69 A1V 29841 2021/04/18T10:18:24Z AL 1631 B1 12711 B2 -1 P 14377 BV 12345 ST OK D r A1P 20701 B1T 12 UC 68 A1V 29841 2021/04/18T10:18:30Z AL 1638 B1 12708 B2 -1 P 15021 BV 12372 ST OK D r A1P 20791 B1T 12 UC 68 A1V 29841 2021/04/18T10:18:36Z AL 1641 B1 12708 B2 -1 P 15110 BV 12302 ST OK D r A1P 20640 B1T 12 UC 68 A1V 29841 2021/04/18T10:18:42Z AL 1643 B1 12705 B2 -1 P 15043 BV 12372 ST OK D r A1P 20912 B1T 12 UC 67 A1V 29841 2021/04/18T10:18:54Z AL 1650 B1 12705 B2 -1 P 14954 BV 12302 ST OK D r A1P 20912 B1T 12 UC 67 A1V 29841
And after, with the sun just having reached one of the large panels...
2021/04/18T10:35:41Z AL 6448 B1 13901 B2 -1 P 4101 BV 13855 ST H - t A1P 90305 B1T 12 UC 100 A1V 17038 2021/04/18T10:35:48Z AL 6441 B1 13923 B2 -1 P 4080 BV 13934 ST H - t A1P 90094 B1T 12 UC 100 A1V 17038 2021/04/18T10:35:54Z AL 6436 B1 13944 B2 -1 P 4086 BV 13916 ST H - t A1P 90350 B1T 12 UC 100 A1V 17038 2021/04/18T10:36:00Z AL 6429 B1 13965 B2 -1 P 4120 BV 13960 ST H - t A1P 90396 B1T 12 UC 100 A1V 17038
The panel is still partially shaded, thus the low A1V, I think. And then not!
2021/04/18T10:39:01Z AL 6347 B1 14377 B2 -1 P 4414 BV 14258 ST H - t A1P 91317 B1T 12 UC 100 A1V 16846 ... 2021/04/18T10:39:56Z AL 5924 B1 14386 B2 -1 P 3899 BV 14258 ST VH - t A1P 85005 B1T 12 UC 100 A1V 32126
And a cup of solar green tea later, with hazy sun on the panels:
2021/04/18T12:59:40Z AL 10171 B1 14261 B2 -1 P 14889 BV 13091 ST VH D V A1P 147649 B1T 14 UC 100 A1V 27805 2021/04/18T12:59:47Z AL 10258 B1 14277 B2 -1 P 14763 BV 14065 ST VH D V A1P 146925 B1T 14 UC 100 A1V 28650 2021/04/18T12:59:53Z AL 10219 B1 14289 B2 -1 P 14775 BV 14065 ST VH D V A1P 146442 B1T 14 UC 100 A1V 29346
(I am considering how to get my MacBook at least partially off-grid again to minimise grid load.)
Anticipating a hot summer like recent ones I have ordered myself a 12V fan as designed for cars and trucks. That may provide some simple comfort cooling exactly when needed when the sun is blasting away!
2021-04-13: 2%
Today with patchy sunshine, by , the grid-tie system has generated ~12.1kWh, the off-grid system ~220Wh. So off-grid with a very nominal peak capacity of ~10% of grid-tie, has taken ~2% of demand that would otherwise have been carried by the grid.
2021-04-10: Less Dumping
Dull day today: will maybe have pulled ~150Wh off-grid, and the big array has not floated over Vmpp so far. (The grid-tied system had generated ~4.6kWh by ~16:00Z, vs 14.9kWh yesterday, for reference.) The dump load has been operating maybe 50% of the time.
So it seems possible to absorb everything available with the current setup.
2021-04-07: Moar Dumping
Looking at all times in April up to ~6pm BST today where the panels have been floating over Vmpp, and dumping was not happening, shows:
cat /var/log/powermng/2021040?.log | awk '$23>=30000 && $14!="D"{print}'
2021/04/01T10:20:06Z AL 3528 B1 14271 B2 -1 P 2070 BV 14144 ST VH - s A1P 50324 B1T 16 UC 100 A1V 31577
2021/04/05T08:10:06Z AL 1380 B1 13007 B2 -1 P 14464 BV 12679 ST OK - s A1P 17847 B1T 11 UC 100 A1V 30323
2021/04/05T09:50:06Z AL 2624 B1 13227 B2 -1 P 15119 BV 12889 ST OK - s A1P 34712 B1T 11 UC 100 A1V 30619
2021/04/05T10:00:06Z AL 3863 B1 13727 B2 -1 P 2966 BV 13635 ST H - t A1P 52785 B1T 11 UC 100 A1V 30127
2021/04/05T10:10:06Z AL 4228 B1 14185 B2 -1 P 2100 BV 14030 ST H - t A1P 59700 B1T 11 UC 100 A1V 30509
2021/04/05T10:20:06Z AL 2556 B1 13846 B2 -1 P 4016 BV 13714 ST H - t A1P 35512 B1T 11 UC 100 A1V 30274
2021/04/05T16:40:06Z AL 1088 B1 14094 B2 -1 P 3214 BV 13995 ST F - t A1P 876 B1T 12 UC 100 A1V 33069
2021/04/05T16:50:06Z AL 431 B1 14090 B2 -1 P 2424 BV 13934 ST F - t A1P 6734 B1T 12 UC 100 A1V 31934
No missed opportunities since the last algorithm tweak. Today and yesterday have been mixed sunny/overcast.
A more general look at all times floating over Vmpp so far in April shows 215 samples or ~36 hours (out of ~162 hours) where available energy was not being fully used from the off-grid system (including the 12V-nominal part):
% cat /var/log/powermng/2021040?.log | awk '$23>=30000{print}'
2021/04/01T10:20:06Z AL 3528 B1 14271 B2 -1 P 2070 BV 14144 ST VH - s A1P 50324 B1T 16 UC 100 A1V 31577
2021/04/01T10:30:07Z AL 2969 B1 14271 B2 -1 P 2141 BV 14144 ST VH D e A1P 42367 B1T 16 UC 100 A1V 32676
2021/04/01T10:40:06Z AL 4433 B1 14298 B2 -1 P 16100 BV 14074 ST VH D T A1P 63430 B1T 15 UC 100 A1V 31809
2021/04/01T10:50:06Z AL 3871 B1 14301 B2 -1 P 17662 BV 14074 ST VH D h A1P 56786 B1T 15 UC 100 A1V 33054
2021/04/01T11:00:06Z AL 2759 B1 14267 B2 -1 P 13697 BV 14004 ST VH D h A1P 41703 B1T 16 UC 100 A1V 33671
...
2021/04/07T15:10:06Z AL 1658 B1 14496 B2 -1 P 15178 BV 14118 ST VH D A A1P 24687 B1T 9 UC 100 A1V 33219
2021/04/07T15:20:06Z AL 1745 B1 14496 B2 -1 P 14960 BV 14118 ST VH D A A1P 25517 B1T 9 UC 100 A1V 33967
2021/04/07T15:30:06Z AL 1665 B1 14499 B2 -1 P 15311 BV 14179 ST VH D A A1P 24158 B1T 9 UC 100 A1V 32871
2021/04/07T15:50:06Z AL 1701 B1 14493 B2 -1 P 15175 BV 14170 ST VH D A A1P 14903 B1T 9 UC 100 A1V 32554
2021/04/07T16:10:06Z AL 2109 B1 14493 B2 -1 P 15378 BV 14223 ST VH D A A1P 8818 B1T 9 UC 100 A1V 33149
The maximum that that could have represented (an unrealistic upper bound) is ~500W * 36h = 18kWh, or ~2kWh/d. Very likely the excess energy most of the time, except with bright sunshine direct on the big panels unshadowed near noon, is a small fraction of that. The fact that I have been able to run a ~100W water heating load without the system batting an eyelid suggests that significant money is definitely being left on the table though!
Indeed 205 out of the 215 samples are with battery state of HIGH, VHIGH or FULL, so likely far from marginal excess available.
In principle I could craft a hardware dump to bleed off all energy above Vmpp, ie keep the voltage down to ~30V, and measure what energy is dumped that route.
2021-04-05: Dumping Setup Is Tweaked
Spoke too soon! This short dropout stopped dumping at a key time (16:30Z):
2021/04/05T16:10:06Z AL 1404 B1 14094 B2 -1 P 14644 BV 13758 ST F D F A1P 19674 B1T 12 UC 100 A1V 33085 2021/04/05T16:20:06Z AL 1595 B1 14060 B2 -1 P 15930 BV 13776 ST F D F A1P 22316 B1T 12 UC 100 A1V 30271 2021/04/05T16:30:06Z AL 805 B1 13413 B2 -1 P 14688 BV 13109 ST OK - s A1P 8939 B1T 12 UC 100 A1V 28574 2021/04/05T16:40:06Z AL 1088 B1 14094 B2 -1 P 3214 BV 13995 ST F - t A1P 876 B1T 12 UC 100 A1V 33069 2021/04/05T16:50:06Z AL 431 B1 14090 B2 -1 P 2424 BV 13934 ST F - t A1P 6734 B1T 12 UC 100 A1V 31934 2021/04/05T17:00:06Z AL 764 B1 14090 B2 -1 P 3269 BV 13951 ST F - t A1P 10675 B1T 12 UC 100 A1V 29517 2021/04/05T17:10:06Z AL 677 B1 13965 B2 -1 P 3073 BV 13793 ST F D F A1P 9407 B1T 12 UC 100 A1V 27594 2021/04/05T17:20:06Z AL 595 B1 13199 B2 -1 P 12988 BV 12942 ST OK D T A1P 7837 B1T 12 UC 100 A1V 27591 2021/04/05T17:30:06Z AL 474 B1 13163 B2 -1 P 12282 BV 12881 ST OK D E A1P 6161 B
And in fact the day before at much the same time, FULL dips to OK once:
2021/04/04T16:20:06Z AL 1351 B1 13648 B2 -1 P 14290 BV 13425 ST F D F A1P 18421 B1T 12 UC 100 A1V 27732 2021/04/04T16:30:06Z AL 1409 B1 13852 B2 -1 P 15362 BV 13547 ST F D F A1P 20550 B1T 12 UC 100 A1V 29056 2021/04/04T16:40:06Z AL 1100 B1 13431 B2 -1 P 14869 BV 13109 ST OK - s A1P 14812 B1T 12 UC 100 A1V 28025 2021/04/04T16:50:06Z AL 810 B1 14090 B2 -1 P 3072 BV 13951 ST F - t A1P 10177 B1T 12 UC 100 A1V 28913 2021/04/04T17:00:07Z AL 687 B1 13877 B2 -1 P 3026 BV 13688 ST F - t A1P 9498 B1T 12 UC 100 A1V 26462 2021/04/04T17:10:06Z AL 602 B1 13736 B2 -1 P 3654 BV 13574 ST F - t A1P 8184 B1T 12 UC 100 A1V 26175 2021/04/04T17:20:06Z AL 588 B1 13617 B2 -1 P 2969 BV 13468 ST F D F A1P 7942 B1T 12 UC 100 A1V 26127 2021/04/04T17:30:06Z AL 474 B1 13172 B2 -1 P 14674 BV 12907 ST OK D T A1P 6191 B1T 12 UC 100 A1V 25733 2021/04/04T17:40:06Z AL 389 B1 13120 B2 -1 P 14590 BV 12863 ST OK - s A1P 5104 B
This may even be a bug/feature in the charge controller. 
So I have moved the 's' case directly below the 'r'ide-through case (they were the other way round) to try to avoid extended dropouts from the sun briefly going in (or a controller 'feature').
This would have been a good moment to continue grid support, and this possibly lost the opportunity to pass 40 minutes of sun to the load.
It is possible that the ride-through period should be reduced from the current 40 minutes.
2021-04-04: Dumping Setup Is Good
I'm reasonably happy from examining when the large PV panels are being allowed to float above Vmpp (maximum power point) that I'm dumping ('D's, taking load off-grid) whenever I reasonably might.
Vmpp seems to be just short of 30V (on colder mornings). Looking at times when the array voltage has been allowed to float above that, but the system was not dumping, shows only one 10-minute slot in the first three days of April with mixed/sunny weather:
% cat /var/log/powermng/2021040{1,2,3}.log | awk '$23>=30000 && $14!="D"{print}'
2021/04/01T10:20:06Z AL 3528 B1 14271 B2 -1 P 2070 BV 14144 ST VH - s A1P 50324 B1T 16 UC 100 A1V 31577
I will resist the urge to tinker with the power management code for now.
Energy harvested off-grid since restoring the router dump-load on the 28th has risen from ~150Wh/d to ~300Wh/d:
2021-03-24 0.1 0.168 2021-03-25 0.1 0.170833 2021-03-26 0.2 0.173913 2021-03-27 0.1 0.177273 2021-03-28 0.1 0.180952 2021-03-29 0.3 0.18 2021-03-30 0.4 0.184211 2021-03-31 0.2 0.188889 2021-04-01 0.3 0.194118 2021-04-02 0.3 0.2 2021-04-03 0.3 0.2
Auto-tweeting
Also today I set up a mechanism (via the existing intensity Twitter poster) to be able to send additional tweets ad hoc.
I'm using this to emulate what I currently do semi-manually with Hootsuite, ie send one article tweet during the day, and one down-under-skewed tweet in the small hours of the night (thus antipodean daytime).
The down-under tweak is to pick featured articles 6 months out of sync, to pick up more that are seasonal for there.
Articles selected are featured for the current month (or 6 months off), popular, or a few at random (that are not noindex).
My Java Twitter connector is old enough that it thinks tweets are limited to 140 characters: I should upgrade that to allow longer tweets including the page description.