Inverters and batteries

[bob prohaska]

This is to follow up on Jim Wilkins' inquiry about
experiments with batteries and inverters for UPS
service. Originally the title was inverters vs wallwarts,
I think.

The first experiment was to buy the cheapest pure-sine
inverter on Amazon, an 800 watt Ampinvt, and pair it with
an 80 AH Delco Voyager deep cycle battery. It managed to
run my fridge overnight, but the battery was clearly too
small. In addition, the "green eye" charge endicator went
blank and wouldn't come back after prolonged recharge.

The battery nonetheless seemed to work fine. I returned
it for replacement (only a few months old) thinking the
eye was somehow defective. The second battery did exactly
the same thing. I've opted to keep that one to see how it
ages.

Next I bought a "Power Queen" 100 AH LiFePO4 battery, which
seemed to work much better. It'd run the fridge for close to
20 hours before the low voltage beeper want off. Charging was
slow at 15 amps but the battery voltage came to setpoint and
was stable. The cheapest inverter and the cheapest battery on
Amazon seemed like a double win.

So, I ordered a second inverter and battery for use in another
location. There are slight differences in the inverter, the
charge indicator LED and alert beeper are different, but in
terms of discharge performance the second pair seem fine.

The charging behavior is puzzling. Set to charge to 14.6 volts
per the battery maker's specs, the inverter runs up to setpoint,
stops charging for perhaps an hour and then resumes charging.
Careful battery monitoring show a voltage ramp to 14.9 v, then
a decline to about 14.4 with persistent charging attempts that
never get anywhere.

Initially I thought this was some fault of the battery or BMS,
but an isolated test of the holding voltage showed the battery
resting at 14.39 volts, which seems pretty good after 2 hours.
A capacity test delivered just over 1 kWh to the fridge before
the low voltage alarm went off, which seems reasonable given
the idle draw of the inverter and plausible efficiency.

I'm tinkering now with the battery charge voltage setting. One
attempt at charging to 14.5 volts resulted in a stable finish
voltage, but that result hasn't repeated so far. Now I'm just
slowly backing down the charge voltage looking for a stable
endpoint. The battery maker says 14.2 v is enough to balance.

So far, I think the experiment is a qualified success. The load
pickup by the inverter transfer switch is fast enough that a
Raspberry Pi 4 isn't disturbed by the transient. The fridge
seems happy with the power quality and makes no odd noises.

The inverter fan is variable speed but never stops entirely,
a small minus. The battery cables furnished with the inverter
are too thin for anything but initial testing if rated load
is to be supported. The integral AC outlet has hot and neutral
swapped but that's easily worked around by using the terminal
block. Fifteen amps of charge current is kinda low, but, it's
the cheapest inverter I could find!

Both Ampinvt and Power Queen have been prompt, courteous and
minimally informative in terms of tech support. The inverter
has an RS-485 interface, it would be really nice to monitor
via computer. I haven't yet tried to ask. It seems a long shot.

The real test will be of longevity. Five years will be a bit
dissapointing, ten a success. Any more is a windfall.

Thanks for reading,

bob prohaska

[Jim Wilkins]

"bob prohaska" wrote in message news:tbmjps$1bu9c$***@dont-email.me...

The charging behavior is puzzling. Set to charge to 14.6 volts
per the battery maker's specs, the inverter runs up to setpoint,
stops charging for perhaps an hour and then resumes charging.
Careful battery monitoring show a voltage ramp to 14.9 v, then
a decline to about 14.4 with persistent charging attempts that
never get anywhere.

-------------------

That sounds like a lead-acid charging profile. 14.9V to quickly top off and
equalize followed by 14.4V float is similar to the options I've seen on PWM
solar panel controllers. Personally I set the float voltage around 13.5 to
13.8V which is in range for flooded, AGM and LiFe and hopefully maximizes
their storage life as mainly backups. The higher voltage may be better for
daily cycling because the battery recharges faster.

The SIN9020S that I like to monitor battery charge and discharge current is
unavailable from Amazon, apparently I bought the last one. I haven't risked
buying directly from Chinese importers. The +/-50A version of this meter has
proven to be a good accurate alternative although it requires homebrew
packaging and a separate isolated power supply, I use a 2405 isolated DC-DC
converter, also now unavailable on Amazon but listed elsewhere.

https://www.amazon.com/DROK-Display-Voltmeter-Ammeter-Voltage/dp/B019R8LMDO

The Ampinvt looks like a good choice, if I didn't already have a true sine
UPS. My two smaller AC refrigerators have startup surges that measure around
12A which the Ampinvt should handle easily.

[bob prohaska]

Post by bob prohaska
The charging behavior is puzzling. Set to charge to 14.6 volts
per the battery maker's specs, the inverter runs up to setpoint,
stops charging for perhaps an hour and then resumes charging.
Careful battery monitoring show a voltage ramp to 14.9 v, then
a decline to about 14.4 with persistent charging attempts that
never get anywhere.
-------------------
That sounds like a lead-acid charging profile. 14.9V to quickly top off and
equalize followed by 14.4V float is similar to the options I've seen on PWM
solar panel controllers. Personally I set the float voltage around 13.5 to
13.8V which is in range for flooded, AGM and LiFe and hopefully maximizes
their storage life as mainly backups. The higher voltage may be better for
daily cycling because the battery recharges faster.

So far s I know the charger section is a simple constant-current design
with a hard voltage limit. There's no profile from what I can gather.
There's a "battery type selector" but I think that's just a preset
voltage limit. The "lithium-ion customize" type selection allows
setting of an arbitrary voltage limit up to 15 or so volts. The
charge current can be set from zero to 100% of 15 amps. For a
100AH battery fifteen amps does not seem excessive.

Initially I thought the trouble was some combination of too much
gain in the voltage error amplifier and perhaps too much gain
in the BMS controls.

FWIW, yesterday's discharge/recharge cycle ended stably at 14.3 volts.
I'm going to run the test again to see if it repeats. If it does, I'll
call it good for a while.

Post by bob prohaska
The SIN9020S that I like to monitor battery charge and discharge current is
unavailable from Amazon, apparently I bought the last one. I haven't risked
buying directly from Chinese importers. The +/-50A version of this meter has
proven to be a good accurate alternative although it requires homebrew
packaging and a separate isolated power supply, I use a 2405 isolated DC-DC
converter, also now unavailable on Amazon but listed elsewhere.
https://www.amazon.com/DROK-Display-Voltmeter-Ammeter-Voltage/dp/B019R8LMDO
The Ampinvt looks like a good choice, if I didn't already have a true sine
UPS. My two smaller AC refrigerators have startup surges that measure around
12A which the Ampinvt should handle easily.

For $250 it's a decent all-in-one package. I'm not entirely happy with it,
but it's been both instructive and useful. The ability to communicate with
the controller over the RS-486 interface might be highly valuable. The
contols seem fairly generic, but I haven't found any clues online.

bob prohaska

[bob prohaska]

Post by bob prohaska
FWIW, yesterday's discharge/recharge cycle ended stably at 14.3 volts.
I'm going to run the test again to see if it repeats. If it does, I'll
call it good for a while.

Alas, the test didn't repeat. After discharging overnight today's
recharge session ended at 14.3 volts but after a couple of hours
lapsed to 14.2 with the inverter's charge light flashing, which
continues as I write this about four hours later with no change
in battery voltage. That's exactly the same behavior observed with
charge voltages of 14.4, 14.5 and 14.6. The resting voltage of the
battery is 14.39, so it seemed possible 14.5 and 14.6 were simply
too high.

Not sure what to make of this. I've reported to my contact at
Ampinvt but have asked only what to try next. The behavior is
wrong, but it might not be harmful.

Thanks for reading,

bob prohaska

[Jim Wilkins]

Post by bob prohaska
FWIW, yesterday's discharge/recharge cycle ended stably at 14.3 volts.
I'm going to run the test again to see if it repeats. If it does, I'll
call it good for a while.

Alas, the test didn't repeat. After discharging overnight today's
recharge session ended at 14.3 volts but after a couple of hours
lapsed to 14.2 with the inverter's charge light flashing, which
continues as I write this about four hours later with no change
in battery voltage. That's exactly the same behavior observed with
charge voltages of 14.4, 14.5 and 14.6. The resting voltage of the
battery is 14.39, so it seemed possible 14.5 and 14.6 were simply
too high.

Not sure what to make of this. I've reported to my contact at
Ampinvt but have asked only what to try next. The behavior is
wrong, but it might not be harmful.

Thanks for reading,

bob prohaska

-----------------------------------

You won't know what is happening unless you measure the charging current,
it's much more informative than just the voltage since both voltage and the
battery's state of charge determine it. Once the LiFe battery is full you
may see very low current over a range of voltages. AGMs are similar, with
somewhat higher float current when full, around 1% of C.

For temporary use a 75mV at 75A shunt and a DVM on the millivolt scale
should be accurate enough, and will tolerate much more accidental
overcurrent than a DVM's current range. The shunt can be shielded inside a
pvc conduit tee. 75mV at 75A is 1 milliOhm, so the meter reads Amps directly
as milliVolts.
tinyurl.com/mt524n7y

The DVM leads connect to the smaller screws and the current-carrying leads
to the larger ones. They are meant for solderless crimp lugs and stranded
wire, which withstands flexing from handling far longer than solid wire
would. The notches are from calibration after assembly.

As a bonus the DVM measures both charging and discharging current, unlike
most of the panel meters I've mentioned, though they are safer for permanent
use. If the DVM connects to a computer you can record the charge and
discharge cycles. The TP4000 DVM I use writes a file that can be imported as
*.csv into a spreadsheet to adjust calibration (such as a different shunt
resistance) and calculate Amp and Watthours of battery capacity with the
actual load.

[Jim Wilkins]

"Jim Wilkins" wrote in message news:tbsfkp$2lav5$***@dont-email.me...

tinyurl.com/mt524n7y

----------------

It worked the first time. This is the title on Amazon:
"uxcell 75A 75mV DC Current Meter Shunt Resistor Resistance for DC Ammeter"

[Jim Wilkins]

"Jim Wilkins" wrote in message news:tbsfkp$2lav5$***@dont-email.me...

You won't know what is happening unless you measure the charging current,
--------------

Here's a nice meter that doesn't need an enclosure, just tape over the
terminal screws.
https://www.amazon.com/PEACEFAIR-PZEM-031-6-5-100V-Electric-Industry/dp/B07RQTXNRN

It records total Watt-Hours but not Amp-Hours, and its current capacity is
more than the 15A charger output. I use Anderson PP45 connectors to
temporarily insert the meter into battery leads, in either the charge or the
discharge direction.

This is the same idea, for more money.
https://www.amazon.com/Powerwerx-Meter-PP-Continuous-Powerpole-Connectors/dp/B088KTMQNB

I bought a less expensive version that isn't accurate enough at low
currents. The PZEM series is accurate enough and can be recalibrated, and
the external shunt versions can be wired for current above the limit of 12
AWG wire. My UPS can draw over 50A from the 24V battery.

[bob prohaska]

Post by Jim Wilkins
You won't know what is happening unless you measure the charging current,
it's much more informative than just the voltage since both voltage and the
battery's state of charge determine it. Once the LiFe battery is full you
may see very low current over a range of voltages. AGMs are similar, with
somewhat higher float current when full, around 1% of C.

In principle, I agree. In practice, I don't have a suitable shunt handy.

On a whim, I tried connecting a tail light bulb of about half an amp
across the battery. After sagging, the charger started working and
the voltage settled slightly above setpoint, with no oscillation.
That suggests to me the charge current is too high near float voltage.
Float voltage and maximum current are the only adjustments available.

For the moment, I've removed the light bulb, lowered the charge current
to 80% of max, or 12 amps, and am now running a discharge/recharge cycle.
It'll take a few iterations among voltages and charge currents to get a
picture of what's going on, but I think the problem is wrong PID values
for the charging regulator. The notes I take are at
http://www.zefox.net/~bp/power_queen [named for the battery as the culprit]
with the newest entry last.

Thanks for posting!

bob prohaska

[Jim Wilkins]

"bob prohaska" wrote in message news:tbvec1$3d2tb$***@dont-email.me...
...
The notes I take are at
http://www.zefox.net/~bp/power_queen [named for the battery as the culprit]
with the newest entry last.

----------------------

I've done a lot of measurement like that, initially recording the
Kill-A-Watt values by hand at irregular intervals as you appear to have
done, then I automated the measurements with a pair of TP4000ZC DVMs logging
battery voltage and current to a laptop, a much better arrangement that
graphs the readings and shows trends and events such as when the battery
reached full charge or discharge, and that my UPS inverter trips out on low
voltage a few seconds after the fridge turns on, so Ah capacity and run time
measurements are somewhat random. The minimum sampling interval is 1 second
and it doesn't accurately capture compressor starting surges, I use a
digital storage scope for that. The TP4000ZC can also log temperature with
an included sensor or a K type thermocouple.

Besides the DVMs I bought very flexible silicone-insulated wire to connect
the meters without knocking them over, screw-on banana plugs to make test
leads, and non-gendered Anderson PP45 connectors to join (or bypass)
everything in various configurations. The 30A pins can be soldered, the 45A
ones need a crimp tool. 12AWG Si wire fits the 30A pins and housings without
issues, 10AWG can be forced with difficulty. Gendered pin / socket
connectors are dangerous on batteries where both sides can be live.

Since the output files can be imported into a spreadsheet and manipulated
you can use a lower current, higher voltage shunt for more resolution and
rescale the mV reading to Amps. Each meter writes to its own log file which
I combine by aligning the time stamps, thus they can be different types. One
reading per minute seems a good compromise between resolution and file size.

To go further, the DPS5020 is a digital voltage and current regulator that
can also log its output readings to a computer. It needs a DC power source
such as an old 19V laptop power brick, battery charger or whatever else you
may have, up to 60V.
https://www.electronics-lab.com/ruideng-dps5020-50v-20a-power-supply-module-review/

The designer assured me that it isn't harmed by voltage from the battery
back-feeding into its output, it simply powers up, but the voltage does pass
through to the input so its input supply may need protection such as a large
series Schottky diode.

I built a rectifier supply for it from an old arc welder transformer that
takes it to 20A at up to 30V. A case is available if you don't want to make
one. I made a box for a DPS5015, bought one for the DPS5020. The purchased
case was a tight fit for an added diode on an insulated heatsink.

[Jim Wilkins]

"Jim Wilkins" wrote in message news:tbn2s0$1fp6m$***@dont-email.me...

The SIN9020S that I like to monitor battery charge and discharge current is
unavailable from Amazon, apparently I bought the last one. ..

---------------------

Drok introduced a 50A battery Amp-hour meter (fuel gauge) with a Hall sensor
instead of a shunt, so I ordered one, due 8/5/2022. The disadvantage of a
Hall sensor is less accuracy than a shunt. It's claimed to work down to 10V
which improves on the 9020 that becomes inaccurate as a lead-acid nears full
discharge under load, exactly the region of concern.

I was assigned to test and repair industrial battery packs that included
fuel gauges, as well as Lithium supervisory and cell balancing circuits, so
this is more than a casual hobby for me, I'm trying to keep up with tech I
think is important for the future.

Bob, you may not be equipped to design, prototype, test and package
experimental circuits as I did for a living but other readers with similar
interests might be.