It's a year since this problem was first reported. It seems to affect many Voltronic Power models, particularly the PF1 models. However since it was
first reported on this forum on an Axpert King / PIP-5048MK, I'll post my thoughts here.
First, two brief definitions:
- A PV dip is when the PV current goes to zero, or a very low value like 1 A, for about 30-60 seconds, occasionally longer, such as half an hour. Afterwards, the PV charges normally, until the next dip. Dips seem to be less frequent in the middle of the day, but still happen then.
- A PV freeze is when the PV current goes to zero, and stays there until either the start of the next day, or when the user does something to "reset" the solar charge controller, e.g. by changing a charge related setting. Freezes are more serious, but fortunately less common, and may not be related to dips. So for now, I'll ignore the freezes, and hope that when the dips are fixed, freezes will be fixed as well. If not, freezes can be examined separately.
My latest thought on this is that the inverter-charger is in the float stage before the dips, and changes to bulk stage after the dip. PIP/Axpert chargers are unusual in that to go from float stage to bulk stage, they first have to go through a zero-charge stage for 15-30 seconds, then when bulk charging starts, it charges at just one amp for some 15 seconds, and then slowly ramps up the charge current. Despite the slow ramp, it may overshoot the target charge current for ten or twenty seconds or so. The effect is that the dips typically last one to a few minutes.
There are two ways of the exiting float charge stage; both of these end up in the zero-charge stage. The first, and presumably most common, is that the battery voltage has to be measured at 4.0 V below the float voltage setting (setting 27) for ten seconds, no exceptions, testing 50 times per second (500 measurements total). This undocumented 4.0 V figure is way too large for LFP batteries, so LFP flavours of fully patched firmware changes this figure to 1.2 V (some older patched firmware used 1.0 V). Even in patched firmware where KettleKomp™ is implemented, for this and the following measurements in float mode, uncompensated battery voltage is used. It seems unlikely that this path from float stage to bulk stage is involved with the PV dips.
The other way to exit float mode, which I'd forgotten about, works as follows. Regardless of the above measurement, another measurement of battery voltage is made, and it is tested against the float voltage setting (setting 27), less 1.0 V (for 48 V models). So if the float setting is 54.0 V, the battery voltage is tested against 53.0 V. If the measured battery voltage is lower than but not equal to this value, a counter is incremented by one. If the measured battery voltage is higher than or equal to this value, the counter is decremented by two, never allowing the counter to go below zero. I think of this as a leaky bucket. Every time the battery voltage is lower than it should be, one drip goes into the bucket. But if the battery voltage is near where it should be, two drops come out of the bucket. If the counter ever reaches 90 000, then the charger exits float stage, goes to zero charge stage, and a dip occurs. I note that it would take 90 000 / 30 / 50 = 30 minutes to fill the bucket if nothing ever dripped out. So these dips should be a minimum of 30 minutes apart for this theory to be true.
I note that in the middle of the day, the battery is more likely to be fully charged, so there will probably be more leakage from the bucket, predicting fewer dips in the middle of the day (as has been observed).
For this theory to be true, the inverter's charger has to be in float mode to start with. Some cases are late in the day, when this is expected. But some are early in the day, requiring one of the
premature float bugs to have put the charger into float stage too early. Dips have mainly been reported on PF1 models, which until recently don't have patched firmware.
Now to what users can do to prove or disprove my theory.
- The charger needs to be in float mode before the dip. One way is to watch the middle (charge) LED; it should be on solid during the float stage. I would expect the LED to start flashing, indicating bulk or absorb/CV stage, at least for a short time. The other way is to check data logs that record the charge stage. The premature float bug should kick in soon after the end of the dip.
- The dips should be at least 30 minutes apart, probably a fair bit more than that if the battery isn't yet fully charged, for the premature float bug to return the charger to float mode. This should be obvious from data logs.
- You could watch the battery voltage, comparing it with one volt less than the float voltage, to see if the leaky bucket idea holds water, so to speak Keep in mind that the inverter is measuring much faster than you can see data logged, so there may be more sags of battery voltage when fridges start or the like, which you may not see often in the logs.
It's possible that there are more than one mechanism triggering these dips, so if occasional data contradicts the theory, it may not be dead in the water.
Observations welcome!
As for a fix for this problem, that's another story. But first, we need a good theory of what is going on.