Classic powers up, but dead short at input

[MadScientist267]

Steve, I have a question if you have the answer...  I'll also ask Jose in Warranty.
What was the high voltage reading of the input to your Classic ?  How high over
150V did it go ?

BoB -

The number given to me was 177.3V

Since being returned, the highest I've seen was 150 (and possibly some tenths, can't recall because the 150 and rate of rise was making me panic.

When it hit 150, it immediately went to rest, and the voltage began dropping, about as quickly as it was rising. Presumably this was hyperVoc engaging, as the relay let go at that 150V mark. After the voltage reduced to about 105V or so, it reentered float, but at that point I had disconnected the source, so it continued to fall until eventually it went back to resting.

I've witnessed it doing this three times now, personally. The first one was what started all this. Neither of the first two were captured. The first was also a "natural" mode change into float. The subsequent mode changes were forced, but the behavior is exactly the same.

Video of the third "hyper Voc" event:

https://www.youtube.com/watch?v=bWa_JXjgK7Q

I was able to get it to rise well over 100V time after time, like clockwork. One of them I caught on video as well, reaching 129.3V before topping out and then slowly falling back down again.

I did two resistor experiments to see if the voltage could be suppressed to help protect the unit.

My first was with a 150R 10W, from PV + to ground. When the test was run, the voltage climbed quickly and held at about 35V for around 20 seconds, then dropped back down to less than a volt, with the classic returning to resting mode. The resistor got quite hot during that time frame.

The second resistance test, I doubled the resistance to 300R/20W (2 x 150R/10W in series). During this test, the voltage rose to ~51V, again holding this level for 20 seconds or so. When I went to record the video of the 300R test, the voltage didn't go much higher than 42V or so, but did hold there with similar timing and the like.

Video with the 300R:

https://www.youtube.com/watch?v=PxH3A0gMDvI

As of right now, there's no "problem", meaning it hasn't resulted in damage. However I very strongly believe that it was an event similar to these that caused the damage to the original MOSFETs, as the software had been acting strange just before I did the factory reset and the big bang occurred. I can't help but wonder if one of the effects of that software glitch was that possibly it didn't catch hyperVoc and disconnect the way this is doing at this point, resulting in a voltage that just kept rising, silently killing them.

Other circumstances surrounding this:

While it appears that it will rise somewhat without heavy loading, this is when it rises the highest.

It seems it needs to be in bulk for a few minutes, then transition to absorb to be most prominent. I typically leave it in absorb for a couple of minutes to ensure that it isn't going to bounce back out to bulk. The effect can be seen at any point after having been in bulk followed by absorb... anywhere from the beginning of absorb, to the end, by forcing a float, or left to it's own where it naturally switches over (via timer or end amps).

Hopefully it's a little clearer now with the video of the events. I don't know what the best solution is, but at least the resistors help keep it in check by not allowing it to rise to potentially destructive levels.

[boB]

Steve, what are you using for an input source here ?

Maybe you could do a short video to show what you have connected and your complete setup there.
I think that may help explain this.

If the classic is holding the battery at Float voltage or Absorb voltage, you must have some kind of input voltage source.
Also, why would the input voltage be right above battery voltage ?

Try replacing your source with a real solar panel.

What version of software ?  I have not seen this input voltage rise happen, at least not in a very long time
and certainly not with a solar panel connected to the input rather than the grid.

Didn't you originally connect a rectified grid to the input before using a transformer ? That would explain the 177 volts maximum input reading.
The software will always disconnect when the input voltage goes above 150 volts.

[MadScientist267]

boB - on the solar panel... I'd absolutely *love* to... I simply don't have a practical means to do it, especially at the power levels where it seems to be the worst.

I *strongly* believe that the transformer has nothing to do with the issue we're seeing here. Inductive kick would be there, and gone... It wouldn't last *seconds* and continue to push a basically regulated current thru resistors like it does. It simply cannot be the problem. I'm involving you guys because I'm trying to work this out. That was a relatively expensive "mistake" that you can imagine I wish to avoid a recurrence like the plague.

Without a schematic of the grunt section of the classic, I can't do anything but speculate based on behavior. What I *think* is happening is the software is giving the converter some kind of arbitrary signals that cause it to run "in reverse" as a boost converter. As you know, unregulated, a boost converter can run away, and I truly believe that's what's happening here. The resistors provide somewhere for the power to go so that it doesn't just build up and go out of control.

One thing that puzzles me is the load vs voltage rise, as well as how hard it pushes into the resistors... There's *some* correlation, more load means more power/higher voltage, but if one were to plot it, I would hazard to say that it would not follow a curve, line, or anything else... very erratic.

The other aspect to this is that the software can "intercept" at 150V, is there a possibility that it could let the relay go sooner?

Another thought just crossed my mind, I'll have to run a few and see if I'm right: the /length/ of time it pushes against the resistors /may/ correlate with the batteries settling into the float set point. I'll run a few and report the findings.

The transformer is from a 900VA APC UPS unit, the Voc on it is ~22V, and I'm not sure where it's absolute maximum is, but at about 15V I am pulling ~300W through it. Using this transformer was based on the discussion about keeping the transformer as close to battery voltage as possible. As you pointed out basically, raw math would keep it from passing much more power than that, but both it and the classic are happy working with it, no unexpected heat, etc. It does just fine *except* for this transition to float issue.

[boB]

I you aren't using the Aux 1 or   2 terminals, you could set Aux 1 to go high when it goes over some PV
input voltage an drive Aux  2 from that.  Then set Aux 2 to turn the Classic off (Resting) when its input
goes high.  Set Aux 1 PV on High for whatever voltage you want to see it turn off at.

Doesn't really solve the problem except in an indirect way but still something you could do.

What firmware version do you have again ?

boB

[MadScientist267]

Firmware - MNGP=1933, Classic=1923

Sorry about that, got lost in trying to provide details and forgot one of the questions...

Aux 1 is controlling the fan for the transformer, using waste not low I believe... Still experimenting with that.

[EDIT] - Aux 2 has the WBJR on it

I'm working out how to do a temporary setup to do tests with actual PV. It's a logistical issue, but I'll get around it. There's always a way, right? Haha

The *real* solution to that aspect would be to actually do the real install, but some lengths of odd shaped extruded aluminum are standing in the way... gotta make some phone calls and get the metal guy back on the ball I guess now that the weather is starting to show signs of being favorable.

It might take a bit... This weekend is a possibility, have to see how it plays out.

I'd like to find a way to reproduce this for you in your shop so you can gather a better idea of what's happening. My tests will be with 4x 100W @ 12V nominal (72 cell) PV in 3 configurations... 4s, 2s2p, and 4p. The latter most will be the closest to this transformer.

There's a difference as well between the other transformer that was connected when the MOSFETs went... That was in full wave config, this is connected as center tapped half wave on either leg. Theoretically *shouldn't* be a difference, but throwing it out there just in case. I know you've got to be as curious as I am about what's happening here... I don't care about right or wrong *nearly* as much as finding the answer.

You mentioned video of how this is all connected... not sure what anyone would be able to really infer from it, but I suppose I can shoot some as well as draw some schematics... they need to be done anyway LOL

[MadScientist267]

Here are a couple pics along with pertinent schematics...

As you can see, pretty straight forward.

I'll do the real PV tests as soon as I'm supposed to be getting a clear day... other than that, this is about all I can offer up besides the descriptions and video in the other posts.

My theory about it pumping in reverse until the battery sinks to the float set point was indeed correct. It appears it'll push until the cows come home if something were to hold the battery above it. I'm throwing a link to the video I shot earlier where it pushed against those resistors for 3+ minutes, over 100V, until the battery settled to the set point. This was a natural mode change, triggered by the absorb timer. There doesn't however seem to be much rhyme or reason (yet) as to how hard it pushes.

https://www.youtube.com/watch?v=fFgClMw44K0

[MadScientist267]

And the pics of the implementation...

[boB]

Looking at the video, I wonder what happens if you do the same thing BUT unplug the transformer
right when it goes from Absorb to Float and don't hook up your resistor.

Instead of unplugging the transformer from the wall, disconnect the positive output of
the diodes (cathodes) from the PV input of the Classic right after the transition.


ALSO, when doing this, go to the RFR screen from status by holding the Left-arrow key
and pressing the ENTER key.  I would like to see those 4 + and - numbers in the middle
which will show all of the current sense numbers.  Video that too if you can.

[MadScientist267]

boB - As requested... I wasn't sure if you'd want resistors or not looking at the raw current readouts, so I did a video both ways. Hopefully these help a bit. I'm still looking for a correlation with the peak voltages... I tried a few more things before posting, looking for a connection, but still not lining up with anything that's visible.

The videos:


Bridge disconnected immediately after mode change, No resistors, Status display:

https://www.youtube.com/watch?v=SGje_sBXDCM


Bridge disconnected immediately after mode change, No resistors, RFR display:

https://www.youtube.com/watch?v=zbcmYDfA5I8


Bridge disconnected immediately after mode change, With resistors, RFR display:

https://www.youtube.com/watch?v=HPduZ-Hke38

[MadScientist267]

Ok, I know there's been a little curiosity going on, and here are my findings over the last couple of days.

I'm 100% positively sure, with no doubt in my mind, that it is coming from inside the classic, and no chance whatsoever that it's coming from anywhere else. At all. Period. To wit, the recent development:

The rise can be triggered simply by forcing bulk then forcing float while battery is above float voltage.

The rise will occur with *nothing* connected to the input at all whatsoever.

I was satisfied that real PV would not trigger it... as I was only seeing minimal rises, just a few volts. There was a blocking diode between the PV and the classic, so there could not have been any loading by the panels. I had more or less resigned to the idea that in some form or another the transformer was at fault, so the PV tests ended early.

But then...

After disconnecting the PV, I let the batteries drain, then charged them again via the transformer. I was trying to catch the peak with the daily "hi pv" register, when it became apparent that this was a product of "yesterday". So I set out to confirm it by letting the batteries discharge for a full 24+ hour window. Sure enough, all it caught was the transformer's Voc at the beginning of the discharge when it was disconnected from the grid: 22.5V

Tonight, I set out to catch any anomaly in disconnecting the transformer, and found the above instead - there was no rise, so I just forced it into float (with no input) so as to avoid the spike, and then *would* have turned the transformer back on. I didn't make it to reconnecting power, as I had then gone back to the status screen, to see it in excess of 65V!

As a sanity check, I went back to tweaks, hit force bulk, and returned to status, it was then sitting at 16.5V (the MPP set point), but of course not actually doing anything since there was still no input.

Back to tweaks, force float, then status, and it shot up again, this time only reaching 27.1V

So I disconnected everything completely from the input, and repeated the bulk/float tests. Same exact results!!! The behavior didn't stop until the float voltage set point was reached, at which time it returned to the 16.5V, and about a minute later, the relay kicked out and it went to rest, voltage falling to ~7.5V

Ross hinted that I should maybe see that with a lower float set point, would it also jump. The answer is yes. Float set to 12.0, battery sitting at 13.0, nothing connected to the input whatsoever, force bulk, force float, rose to 27.1V once again. Another run with the same settings produced 26.6... With the float voltage set at 10.5, input rose to 26.1

I'm STILL not sure what decides how high it goes... but it always jumps... the delta appears to be mostly related somehow to the amount of time spent in absorb... or now even simply bulk? It's very unclear, and temperatures still don't really appear to mean anything. The relationship with time, if there is one, is non-linear.

With the float point set at 10.5, I left it a bit while throwing the basics of all this out on IRC, which took several minutes. It held that voltage (26.1) in float mode, and seemed likely would stay that way until the battery reached the float set point. For what that's worth.

If that's all the further it ever shot up, who cares right? Unfortunately this is just not the case, and now I have no idea what to do. I can only put the resistors on it to help keep it from rocketing
 
Something is also way out of whack with the daily "hi pv" or *something*... it's calling the last high peak 118.4... and I know it didn't rise that high on the status screen - I was watching intently during the disconnect, and it was still rising for each of the subsequent observations... so either it's measuring it independently, or simply not pulling the data from the right place. All I know is, by my count, it should be holding a value of approx 65V, but it's much higher.

[boB]

With 22 volts Voc the Classic input WILL go higher and 26 volts is not high at all.

PV V is allowed to go higher than measured Voc because if Vbatt is above the set point,
the Voc may have changed since the last Resting period and so it has to be allowed to
go a bit higher at least.  Going from 22 Voc up to 150 V or even 118 Volts to reduce
the Vbatt would be right out of course.

I am not sure why you would have seen 118 Vpv high but were you watching the MNGP
all day ?  You would have to log that every couple of seconds or be staring at it to
notice usually.  That high Vpv voltage logging does require it to be there for a
while, at least a couple of seconds before it logs it IIRC.  I will take a look though
and make sure nothing has changed.

[MadScientist267]

With 22 volts Voc the Classic input WILL go higher and 26 volts is not high at all.

I'm in agreement there... if that's the furthest it ever shot up (or even double that), I wouldn't be concerned.

PV V is allowed to go higher than measured Voc because if Vbatt is above the set point,
the Voc may have changed since the last Resting period and so it has to be allowed to
go a bit higher at least.  Going from 22 Voc up to 150 V or even 118 Volts to reduce
the Vbatt would be right out of course.

I never put it into solar mode during the PV tests, because I was trying to keep from changing too many variables at one time. It stayed in manual hydro the entire time (and I adjusted the MPP accordingly so conditions could be met).

I'm not entirely sure what you mean by "Going from 22 Voc up to 150 V or even 118 Volts to reduce the Vbatt would be right out of course"... can you elaborate/clarify?

I am not sure why you would have seen 118 Vpv high but were you watching the MNGP all day ?

No, however it should have stayed in resting mode... I timed everything so that immediately after the midnight reset, I turned the transformer off, and was running from battery while I wasn't able to babysit it (specifically out of recognition that the logging wasn't frequent enough to catch it elsewhere)...

[boB]

I'm not entirely sure what you mean by "Going from 22 Voc up to 150 V or even 118 Volts to reduce the Vbatt would be right out of course"... can you elaborate/clarify?

The classic raises the input voltage to reduce charging current and/or battery voltage so if your Voc was 22V it wouldn't need
to go very high of V pv.

I am not sure why you would have seen 118 Vpv high but were you watching the MNGP all day ?

No, however it should have stayed in resting mode... I timed everything so that immediately after the midnight reset, I turned the transformer off, and was running from battery while I wasn't able to babysit it (specifically out of recognition that the logging wasn't frequent enough to catch it elsewhere)...

I wonder if that log was simply from the previous day ?

[MadScientist267]

Ok found the problem with the hi Vpv... for whatever reason, "Day 1" was reading data from mid February... Bumping it to Day 2 caused it to find itself and going back to Day 1, it's realigned. The voltage in question (~65) is logged as 69.2V.

So THAT part is at least cleared up.

On to the other parts...

[boB]

OK, good deal on the Logs...   Next...

Next time you catch the input voltage rising above, say 40.0 or 50.0 volts,
press the STATUS button twice and get to the screen that shows Voc and
note that number to see what the Classic thinks the Voc is.
Or, how about watching that screen for a little while or better yet, video that
third status screen if that number happens to be changing or moving around
while actively in Hydro mode and not resting.