Turns out one of the 4 batteries was bad.

Many thanks to @SamGibson for his insight in solving this problem.

The well-deserved Accept is on his answer.

Here is what the charge plot looks like now:

Solar power - 6149 Watt-hour battery bank but only 1 hour of inverter time at 300W after dark

The system now provides near-real-time montage plots here: https://SDsolarBlog.com/montage

This is a 24 Volt system comprised of 6 panels and 4 batteries connected by about 25 feet of 8 Gauge cable.

  • 3 pairs of 100 Watt panels on single-axis hinges, with the mounts tilted south to catch the most sun. Each pair has positioning actuators - the three sets are run by a single Nano-based sun tracker which works well.
    It sets the panels flat when it loses track of the sun to protect them from wind during night.

  • The PV panels connected by about 25 feet of 8 Gauge cable. All the PV side is connected with standard MC4-style connectors. 8 Gauge is the largest that will fit.

  • Renogy Commander MPPT Charge Controller - installed and working. According to its remote monitor screen it is doing its job just fine.

[Just did a daytime spot-check and with 95.2V x 1.9A = 180.88 Watts in, it is putting out 29.2V at 6.0A = 175 Watts for an efficiency of 96.7% to go to the batteries and inverter]

enter image description here

  • Battery bank is comprised of 4x 29DC 122Ah deep-cycle marine batteries in series/parallel to present as a nominal 25.2 Volts. They are all about a year old.

They are wired together like this:

enter image description here

  • Inverter is a new 1000W Pure Sine Wave unit connected directly to the batteries with 10 Gauge cable. [During the same spot check it is producing 165W AC with the input of 175W DC, for an efficiency of 94%)

I set up a voltage monitoring and inverter control system using a Raspberry Pi and Arduino Nano. The Nano takes the readings and can send a shutoff signal to the inverter when the voltage is at 21 or lower for 5 readings in a row (1 minute intervals).

It also shuts it off at 4pm to give it time to top off the batteries each evening.

The Pi part is the data logger & RTC, and is used to plot the voltages which it displays continuously in real time by virtue of the gnuplot reread. I keep it up on the screen all the time via RDP.

enter image description here

I also set up an automatic power-transfer relay that is energized by the inverter, switching the entire load from house power to Inverter power. The automatic transfer relay feeds power out to two UPSs, so there is no interruption of AC power during the switchovers. Turns out you can't buy this little gem. You have to build it.

Total load is typically about 300 Watts to run the computers and the 2-tube 37 Watt LED room light. it is quite sufficient while the sun is up to run nearly everything in my little place while the sun is up. (No A/C or kitchen appliances, obviously).

I have a 24-to-12 converter to run the ham radios, and thank goodness for getting rid of the PWM charge controller. Much less radio noise (QRN).

enter image description here

The battery monitor shuts down the inverter at 4pm - time for the batteries to get topped off before the sun goes down.

The charge controller flashes the battery light at me on the panel to indicate that the batteries are fully charged. Almost any time the inverter is off while there is sun it indicates that it is full. The batteries can't hold all the power this system produces.

The controller indicates that the battery bank is full. The voltages when they settle do not, though.

The batteries are new, made by Johnson Marine. The water is fine in them, and they are barely 1 year old.

After the inverter is commanded off at 4pm, with 2 hours of sun yet to go, it is clearly shown on the MPPT's remote screen that the amount of current flow into the batteries decreases down to about 0.2A

When the inverter is turned on after the sun is down, it really drags the voltage down quickly. It is providing about less than 300 Watts of AC power according to the Kill-a-Watt meter before the inverter quits due to input DC low voltage.

The DC side is still good at that point for a lot longer. I have full confidence that if all I was running was a 12-V lamp and the ham radios I could run all night on DC.

The problem is when I try to use a reasonable amount of AC power. It is the inverter that decides when the voltage is too low.

With the MPPT charge controller screen showing a happy face and the icon of a full battery bank, it seems there should be more usable AC power in there.

With four 12V 122Ah batteries in series/parallel the system stores a total of 244Ah @ 25.2V when fully charged. This is 6,149 Watt-hours.

================== LOAD TEST =================

In these plots you will see that the voltage started to rise as the sun started to come up, and then I started the Inverter. But then thought better of it and used the breaker to disconnect the PV panels. Gave it time to settle down then at 7:30 began a 260W AC load test with the Inverter, sampling every 30 seconds.

It lasted less than a half hour before the inverter cut it off.

Then let the battery bank settle for a second run, again at 260 Watts - with sampling every 15 seconds. This time it only lasted a few minutes.

enter image description here

You can see it go off the chart when the PV Panels were turned back on.

After the test the PV panels went to work, feeding 10 Amps of 27-28V into the batteries for a while.

Then the batteries filled up and the amperage dropped to about 2 Amps which allowed the MPPT charge controller to go into boost at 29.8V

enter image description here

The charge controller is showing a happy face and a 'full' battery icon as it thinks the batteries are getting back up to full charge. It then drops down to the float voltage for the battery bank.

The green light on the charge controller is now blinking green to indicate a full battery bank. The amperage has now dropped to 200 mA.

Here is the overall voltage plot of both the input and output of the charge controller.

enter image description here

Obviously there was no input during the load tests.

There were some clouds interfering with the PV in the morning:

enter image description here

The charge controller handles partly cloudy pretty well, generally.

SO.. I can't figure out what I am missing here.

  • It all looks like it is working fine except for the drawing out of power from the battery bank.

  • During the day it is no problem to run 300 Watts AC while the batteries charge at the same time..

  • All indications are that the battery bank is full.

  • Then this. It simply does not add up.

Need someone with a practiced eye to look at all this and help, please.

So the question is: Why is it providing such short runtime after dark?

--EDIT: Thanks to @SamGibson's answer, it was determined that one of the batteries was defective. When fully charged then separated and left to settle for two days, three of them measured roughly 12.7 (100%) but one of them measured 10.02V. Also it was determined that by using 8 Gauge cable in yellow crimp connectors designed for 10 Gauge that extra resistance was presented there. So the solution so far has been to obtain a replacement battery and set up a new screw-terminal barrier strip where the inverter, battery bank and charge controller can all come together, and to replace the 8 Gauge cable with 10 Gauge.

After the new battery and harness were installed, here is the new charge plot:

enter image description here

Amazing the difference it makes to replace one bad battery.

Thank you, @SamGibson. You have the Accepted answer on this one. I really appreciate your insight.

  • \$\begingroup\$ Now it is September 2017 and I am on the next generation of the system. Microcontrollers and Linux machines all around. Everything is pretty much automatic now, including archiving the daily data collection to a hard drive for future plotting of seasons and stuff like that. Yet even with all the improvements (2 more panels and an MPPT charge controller) it still exhibits the original problem: Very little runtime when it is dark. It collects much more power than the batteries can hold. That doesn't add up. \$\endgroup\$
    – SDsolar
    Sep 2, 2017 at 1:26
  • 2
    \$\begingroup\$ You have all the logging equipment to do a proper load test of the battery. Why not charge it carefully for a few days, and then drain it with a load. Check the curves you get against the battery datasheet. Perhaps you're not charging it properly. \$\endgroup\$
    – tomnexus
    Sep 7, 2017 at 2:01
  • \$\begingroup\$ TNS, @tomnexus. The load test plots are posted now. \$\endgroup\$
    – SDsolar
    Sep 7, 2017 at 20:52
  • 1
    \$\begingroup\$ Load tests look terrible, like the batteries are almost empty. I'd expect several hours of steady >23 V, then dropping at end of life. Are your wires thick enough? What's the voltage drop on a single cable, from charge controller to battery terminal, when charging at full speed? Same for discharge. Perhaps add a proper circuit diagram of the whole system, showing all significant wires and their resistances. \$\endgroup\$
    – tomnexus
    Sep 8, 2017 at 6:50
  • \$\begingroup\$ Excellent questions. The cables are 8 Gauge from the battery to the charge controller, then there is a 6-inch 10 Gauge jumper from there straight to the inverter. The batteries are just below the table so the lengths are not very long at all. However, as I think about it, my yellow crimp connectors didn't like the 8 Gauge wire and I suppose I could have a bad crimp. So my first step will be to take it apart and measure each battery's voltage. Then measure resistance on each cable. And just for drill, will change out the 8 Gauge for 10, which fits better into the yellow crimp connectors. \$\endgroup\$
    – SDsolar
    Sep 8, 2017 at 7:01

4 Answers 4


I believe that if all of the data presented was correct, then the observed behaviour wouldn't be happening. Therefore something is wrong in the data given - we just don't know which part(s) to disbelieve.

Obviously there is lots of data here. There are a few points which are unclear to me within that data, but I formed a hypothesis and I can't see any data which disproves it (there just seems to be a belief that my suspect component(s) are OK).

All indications are that the battery bank is full.

Yes, but full to what actual capacity?

My hypothesis is a battery-related problem - one or more of the batteries either:

  • Don't have the capacity claimed; and/or
  • Have (at least now) a higher internal resistance, which prevents the capacity being used to power a higher-current load (even though they might have a higher usable capacity with a lower discharge current) [† see below]; and/or
  • Those batteries (despite any claims by the vendor to the contrary) may not have been designed to handle the specific load being applied.

Some of those possibilities overlap e.g. battery design could result in a higher internal resistance than required for supplying that load.

[ † See the update below for another possibility which would produce the same symptoms, and the test necessary to confirm or eliminate it. ]

Two examples of the behaviour which fits with that hypothesis includes:

  • The batteries seem to become fully charged during the day, due to the low eventual charge current in daylight of 200 mA;


  • Under the 260W (or 300W) load from the inverter, the overall battery voltage drops much more quickly than would be expected for the claimed battery capacity of 6149 Wh.

If one or more of the batteries has a lower-than-claimed capacity (or higher-than-acceptable internal resistance) now, then this is exactly the behaviour I would expect.

If I have missed some data which disproves this hypothesis then great - eliminating a hypothesis is a step towards finding the solution. However just believing the battery vendor's claims of the battery capacity, does not disprove this hypothesis.

I don't know the company "Johnson Marine" but are they really the battery manufacturer, as mentioned - or just the vendor (retailer)? Where is the datasheet for these "29DC" batteries? I couldn't find one online. But even if there was a datasheet, it could only be used to influence further testing; again, it would not disprove the hypothesis, since the batteries may not meet the specification in the datasheet.

You might find something useful from a quick check of the battery voltages even before disconnecting them. Obviously the total voltage of each series "pair" must be identical (since the two series pairs are connected in parallel). But what about the voltages of each battery within each series pair? Is there an indication of one battery with a significantly higher voltage that the other one, within a series pair?

Assuming that each battery has a very similar voltage when (they appear to be) fully charged, I would look to design a test similar to the following:

  • Start with (what the existing charge controller believes is) a fully-charged set of batteries.
  • Disconnect the series/parallel wiring from the batteries.
  • Discharge the batteries individually, while measuring their individual capacity, using a load which simulates the load on each battery of the 300W AC load on the inverter; my back-of-an-envelope calculation suggests that may be around a 12 A load on each battery, but do check this.
  • Charge them individually from a controlled, different (i.e. non-solar) source (thereby eliminating the "opinion" of the existing charge controller etc. as to when the overall "battery pack" is charged), while watching the voltage curve and measuring the charge current.
  • Review the results.

My hypothesis is that the data collected during the discharge and/or recharge tests, will not show the expected behaviour from one or more batteries.

Note: I'm assuming that we can believe the voltage readings provided (I guess they are from the charge controller's own ADC). I'm also assuming that there isn't an additional (e.g. unintended) current drain on the batteries, in addition to the inverter. I would probably use a DC current clamp meter on relevant cables during charging & discharging, to make sure that the currents shown were in-line with expectations.

One concern is that although the kill-a-watt claims 300 W AC power from the inverter during the testing, I don't see anywhere that the DC current from the batteries has been measured at that point, to confirm that it fits with that. Is it possible that relying on one piece of equipment (the kill-a-watt) and its reading might be misleading the investigation, if that reported reading is incorrect? Again, appropriate use of a DC clamp meter, would help give some confidence.


† Another possibility which would have the same symptoms as those described, would be if there is an unexpected high-resistance path in the wiring between the batteries and the inverter. That might be the short wires between the batteries, or the longer wiring between the batteries and the inverter.

This would have the same effects as one or more batteries having a high internal resistance as described in my hypothesis above, and would cause the observed increased voltage drop at the inverter under load only.

The effect would be that the batteries were being discharged only from 100% to, say, 90% of their capacity before the inverter correctly detected an excessive voltage drop. However the voltage drop wasn't at the batteries (the inverter cannot measure that) but instead the voltage drop was at the terminals of the inverter at the end of the wiring.

This would also fit with the apparently quick battery recharge times since, indeed, they wouldn't need much recharging.

The test for this would be to measure the voltage drop between the ends of each high-current cable, when under load. Although it might seem that you could measure the resistance of each high-current cable instead, there are disadvantages to that method. Measure the voltage drop across the cable instead.

In summary, I believe that somewhere you've got some unexpected resistance. It may be inside the batteries (e.g. due to their design, or their life so far) or in the wiring between the batteries and the inverter. It may be helpful to review the calculation of the expected voltage drop, for the wiring you have used, at the likely currents.

  • 1
    \$\begingroup\$ I think you are really on to something here. Resistance. Either in the cables to the battery bank or within the batteries themselves. So, first thing to do is redo all the wiring except the PV input. It is possible I did a crimp job that is sub-par on something. Perhaps the circuit breaker connections, even. Let's see how that affects things. Here is the info on the batteries: google.com/search?q=everstart+29dc+marine+battery+specs - Walmart is the vendor, Johnson is the manufacturer - same as from Costco. \$\endgroup\$
    – SDsolar
    Sep 7, 2017 at 23:38
  • 1
    \$\begingroup\$ I think you have hit the nail on the head. You also have figured out that I do not have a way to measure battery discharge current. The only current meter is the charge controller. As for voltage measurements, they are coming from separate ADCs that have been tuned to match what the charge controller's screens show. TO YOUR MAIN POINT: The inverter and charge controller sit side by side, and the inverter is actually connected directly to the charge controller's battery terminals. All 8 Gauge except the jumper to the inverter which is 10. \$\endgroup\$
    – SDsolar
    Sep 7, 2017 at 23:45
  • 1
    \$\begingroup\$ @SDsolar - (a) Personally, I wouldn't start by redoing "all the wiring". After all, if you have one high-resistance crimp, by redoing everything you risk introducing a new problem while correcting one old problem (or other variations of the hazards of "shotgun troubleshooting"). Instead I would be measuring voltage drops, cable by cable, as I mentioned (using the actual "end device" terminals as the points for meter connections, not the terminals on the cable, so that the crimped connections are included in the measurements). (b) If ADCs have been tuned to match the change controller... \$\endgroup\$
    – SamGibson
    Sep 8, 2017 at 0:18
  • 1
    \$\begingroup\$ ... then that relies on the accuracy of that one device, which risks misleading you with incorrect info. (c) Measuring DC current - I was considering a UNI-T UT210E recently, which claims it can measure up to 100A using its clamp. It's not up to the standards of a Fluke, but for < $40 (or equivalent here in the UK) I might take a chance and buy one. I read good reviews but make your own choice - I don't have it yet, so I'm not able to actually recommend. Knowing the DC currents will allow sanity checking of inverter measurements etc. and it can also measure the voltage drop across cables etc. \$\endgroup\$
    – SamGibson
    Sep 8, 2017 at 0:18
  • 1
    \$\begingroup\$ OK. I don't have any experience with DC clamp-ons; just AC. And no longer have the budget for the Fluke I found. Will look into that unit. As I began undoing the wiring I see a few places I can improve things. But then I went back and re-read that post here and realized that before "redoing" everything I can take the time to do end-to-end resistance checks and individual battery voltage testing. You are right about the ADCs. They have been a hassle but I have gotten to know them pretty well. "Shotgun troubleshooting" is a phrase I'll put in the toolbox. Don't want another bad crimp. \$\endgroup\$
    – SDsolar
    Sep 8, 2017 at 0:23

It's quite simple, really.

  1. Connect a dummy load that will draw more than the total power of your panels.
  2. Measure voltage at the panel connector. Measure current in series with the ground wire. Now you can calculate the input power using Ohm's law: Pin = Vpanel x Iinput
  3. Measure the voltage at the load. You can now calculate the power your load receives: Pload = Vload x Iinput

The losses will be: Ploss = Pin-Pload

If your system has a lot of losses due to cabling, then you want to improve the cabling.

If your Pin is too low, you simply aren't getting enough sun, or you need DCDC conversion to allow the panel to operate at a higher voltage than your load.

You can replace the load with an MPPT charger, and connect the load to the MPPT charger. Measure again. Your Pload should be higher, because of higher efficiency.


Turns out that when the battery bank was taken apart, after a long day of charging, then allowed to sit for 2 days, 3 of them measured 12.7 volts but one of them measured 10.02

That battery has now been replaced. Fortunately, the bad battery was the last installed so was still under warranty.

Here is what the voltage plot looks like now:

enter image description here

No more unexplained voltage drop.

There should be no more trouble getting the runtime I expected.


Load test started at 4:30 AM

enter image description here

With 4 good batteries in the bank, it handled it just fine.


Two 122Ah strings of batteries gives you a total of 24V 244Ah. But you're only charging them with 6A. That means around 40h to fully charge the batteries. Or it would still be about 20h if you want to avoid running the batteries below 50% charge.

Is the problem that there simply aren't enough hours of daylight in one day to recharge the batteries?

  • \$\begingroup\$ In this situation the MPPT charge controller drops out thinking they are full - I can watch the amps drop down to 0.2 about an hour after the inverter shuts off, then it blinks the green "full" light. It appears the panels are providing more than the batteries will accept. I do think you are right that the inverter drops out at the 50% level of about 24.2 while a full charge should be 25.2 \$\endgroup\$
    – SDsolar
    Sep 3, 2017 at 16:04

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.