Hello guys,
my apologies for the late reply (and the $#|+ty pictures further down). Lots of stuff has been happening in my life. I have been very busy looking for an education (and I finally found one :]) and I also did a serious upgrade/overhaul on my PV system, but more about that later.
Replies
How fast does it need to operate? And how slow does the 100nF/100K filter slow it to? (Is it now too slow to adequately protect your sensitive stuff?)
Back in October I didn't have any fancy measuring equipment, so I have no idea of exactly how long it took the fuse to trip. I didn't do any fancy calculations either. I just guesstimated a few values and after a try or two 100n/100k turned out to be perfect for my system. Not enough delay to fry any delicate electronics, but not too little to trip from a bit of
inrush current.
In the event you have a catastrophic failure and the electronic fuse does what it's supposed to - ie, turns off... when you come and press the "On" switch, you're forcing the FET on without the protection circuit being able to help. In this case, are you going to blow stuff up by the mere fact of "resetting" the fuse? I'm wondering if it would be safer to pull the non-inverting input to ground via a small cap (say, another 100nF) that's discharged through the pushbutton.
Yes indeed, pressing the "On" button with the fault still present is a very bad idea and would cause the weakest part to go up in smoke (cool!) until the secondary fuse finally cuts the power.
The new stuff
The regulator described above is history! Sure, it was working allright with the old setup, but with the new panels and the heavy wiring I just didn't feel comfortable with it anymore. The P-channel FETs had high internal resistances and the current sensors were flimsy at best. Lastly I wasn't happy with the bad maintainability due to the cramped space inside the regulator, so in February I deciced to build a new charge controller, and here's what I came up with:
Most importantly, I bought a much bigger enclosure giving me plenty of room to work with. This allowed me to use heavy wires and proper power electronics all around. I got rid of the terminal strips, too, since those were awkward to work with when dealing with 16mm² cables. Instead, I am now using ring terminals and M8 nuts and bolts for clamping them down. Having this much space also allowed me to incorporate some bigass caps (2x 200mF@35V, rated for 23A ripple current each IIRC) for filtering out the PWM noise. Those capacitors have to swallow quite a bit of AC current despite the proximity to my batteries.
The new design is supposed to survive dead shorts across the output or anywhere else in the power subsystem (and it does, I tested it with poor Mr. Screwdriver!), hence the ACS714 current monitors had to go as those were the weakest link in the chain, burning up immediately when experiencing a short across the output (until I came up
with the electronic breaker that is). I picked up a bunch of LEM sensors (the three blue things) and adapted them a bit to make them fit my needs. The new current sensors are capable of measuring up to 57 amps nominal and will certainly survive a short as the current is basically just traveling through the LEM's magnetic core via a rather thick piece of wire.
In addition to the new current probes, I am now using three dedicated 12-bit A/D converters (lower right on the main board) with averaging for measuing the current, giving me 50mA resolution and good accuracy. I didn't bother changing the voltage ADCs as 10mV resolution (10 bits+1 bit of oversampling) worked out to be more than good enough.
In my plan of making the regulator short circuit proof and in order to reduce power losses I sacked the P-channel FETs. The new regulator got two IRF3205 N-channel MOSFETs in parallel per channel (input, dumpload, load), easily capable of switching 50+ amps and maxing out both, my PV system and my 32A bench PSU at the same time. They do get warm due to switching losses once the PWM kicks. The fan doesn't come on until 50 amps of combined current have been reached and it's not really necessary anyways as the heatsink doesn't ever get too hot in everyday use. I was very lucky with my circuit design as I was able to bolt the load and dumpload's FETs directly onto the heatsink - without any sort of electrical insulation. Not only does it improve heat transfer, it also saved me a bunch of cumbersome wires :)
All these changes to the power subsystem allowed me to revert to an old fashioned blade style fuse, further reducing the total impedance. I calculated the resistance from one battery terminal to the regulator output and back to the battery to be lower than 20 milliohms - a single P-channel MOSFET had this much resistance!
As the icing on the cake, I separated the user interface from the charge controller. This allowed me to place the charge controller underneath my table, right next to the battery while still granting me easy configuration access at all times.
In an effort to make the regulator even more reliable and to cut down quiescent current I cleaned up the controller's firmware, shrinking down code size and execution time considerably and making the unit nicer to work with. Average power, overcurrent alarm, real time clock, programmable outputs and datalogging, I sacked them all. I only kept the bare essentials, as immediate voltage, current and power as well as the energy meter. Current offsets and voltage tresholds are #define'd in the source code and cannot be changed during runtime anymore. This is also to ensure
nothing can mess up the values and cook my batteries or the load.
The new controller's standby current consumption averages to around 7mA at with everything in sleep mode, although I'm blaming the 5V regulator for drawing at least 5 of the 7mA. I was planning on using a low quiescent
current regulator at first, but those don't even come close to the 7805's line and load regulation. I didn't have any luck with the LF50 and LP2950 at all as their output voltage dipped quite a bit with the LCD on, causing unacceptable glitches on the gate drive signals.
What's to come?
Currently I'm just far too busy with my education to even think about any future plans but I'll make sure to let you guys know as soon as I come up with something fun and interesting.
So long,
Cheers!
