My next step after https://muwlgr.dreamwidth.org/49507.html was to replace 3 separate AC "wall-warts" powering my network boxes with more centralized PSU, potentially wasting less heat, and backed-up by a separate battery (to reduce AC load on SmartUPS 1000).
To this end, I took an [Carspa?] SC-120-12 battery-backed PSU from my initial experiments described at https://muwlgr.dreamwidth.org/47435.html , with 12V 14A*h SLA battery connected to it.
To its output, I connected a [Tusotek?] TS-IPS-V02 buck-boost DC-DC convertor which I screwed to output 12V, and [Tusotek?] TS-IPS-V04 buck-only convertor, screwed to output 5V.
For 12V output I used boost-buck convertor because when working from battery, SC-120-12 would output less than 12V, down to 8V when the battery is deeply discharged.
To its output, I connected a cable with appropriate plug (found among broken PSUs I collected from my past), from which I powered an ONU1GEZ GPON Ether-to-optic convertor.
The great feature of TS-IPS-V04 is its built-in volt-meter which can show either output or input voltage, selectable by pressing a tiny button.
To its 5V output I connected a pair of cables with appropriate plugs, from which I powered a D-Link DGS 1100-08 GbE switch, and a Step4Net Ether-to-optic convertor.
AC-DC PSUs provided for these 3 boxes by their respective vendors, I just plugged out off my SmartUPS output.
After these moves, SmartUPS load reported by apcaccess went down to 4.6%
I was able to reduce it even more, down to 3.6%, by running powertop and switching all its "tunables" from "Bad" to "Good".
Then I found that powertop provides a startup script with --auto-tune option, so I just run systemctl enable powertop to have it started on next boots.
After the previous step (replacing an integrated AC ATX PSU with separate AC-DC rectifier/stabilizer and DC ATX PSU), battery run time for my main PC box had extended from about 2h45m up to a hefty 4h. [update: Recently I had a power outage of 5h47m, of which, the PC box lived on battery for 4h38m. ]
I still have to measure AC power consumption by some decent watt-meter in a number of places, at least, before the SmartUPS; between the SmartUPS and DC ATX; before the SC-120-12; and before the AC extender when all 3 network boxes are powered with their provided "wall-warts".
Know there are some LVC (low-voltage cutout) modules used to protect chemical batteries from deep discharge, but need to avoid cutting my consumers off the UPS battery earlier than getting "low battery" signal from the UPS.
To this end, I took an [Carspa?] SC-120-12 battery-backed PSU from my initial experiments described at https://muwlgr.dreamwidth.org/47435.html , with 12V 14A*h SLA battery connected to it.
To its output, I connected a [Tusotek?] TS-IPS-V02 buck-boost DC-DC convertor which I screwed to output 12V, and [Tusotek?] TS-IPS-V04 buck-only convertor, screwed to output 5V.
For 12V output I used boost-buck convertor because when working from battery, SC-120-12 would output less than 12V, down to 8V when the battery is deeply discharged.
To its output, I connected a cable with appropriate plug (found among broken PSUs I collected from my past), from which I powered an ONU1GEZ GPON Ether-to-optic convertor.
The great feature of TS-IPS-V04 is its built-in volt-meter which can show either output or input voltage, selectable by pressing a tiny button.
To its 5V output I connected a pair of cables with appropriate plugs, from which I powered a D-Link DGS 1100-08 GbE switch, and a Step4Net Ether-to-optic convertor.
AC-DC PSUs provided for these 3 boxes by their respective vendors, I just plugged out off my SmartUPS output.
After these moves, SmartUPS load reported by apcaccess went down to 4.6%
I was able to reduce it even more, down to 3.6%, by running powertop and switching all its "tunables" from "Bad" to "Good".
Then I found that powertop provides a startup script with --auto-tune option, so I just run systemctl enable powertop to have it started on next boots.
After the previous step (replacing an integrated AC ATX PSU with separate AC-DC rectifier/stabilizer and DC ATX PSU), battery run time for my main PC box had extended from about 2h45m up to a hefty 4h. [update: Recently I had a power outage of 5h47m, of which, the PC box lived on battery for 4h38m. ]
I still have to measure AC power consumption by some decent watt-meter in a number of places, at least, before the SmartUPS; between the SmartUPS and DC ATX; before the SC-120-12; and before the AC extender when all 3 network boxes are powered with their provided "wall-warts".
By the hand touch, SC-120-12 is getting noticeably warm during its continuous operation. I put SLA battery on its top, which is getting warm too. So when working from AC, there could be no power saving at all, but the added battery backup could be worth it. [update: on recent long power outage, the battery on SC-120-12 has been barely discharged; probably I should join it to DC feed for LR1106 in some way, to reduce SmartUPS load even further, I just did not measure the power draw for network boxes]
Next step would be to remove AC invertors and rectifiers between the SmartUPS and the PC but I don't have it thought out clearly enough yet.Know there are some LVC (low-voltage cutout) modules used to protect chemical batteries from deep discharge, but need to avoid cutting my consumers off the UPS battery earlier than getting "low battery" signal from the UPS.