Newbie
Joined: 25/10/2020
Location: United StatesPosts: 3
| Posted: 04:39pm 29 Oct 2020 |
I'm very confident that the OCP circuit is drawn correctly—it took me awhile to understand how it works. I too thought that the positive input on the LM358 should go to the current sense shunt, but it does not: It is clearly connected directly to the signal ground bus through the 2K resistor, which creates a resistor divider with the slight internal input pullup of the LM358, raising it slightly above ground. The current feedback actually comes from the inverting input, through the parallel 100 and 120ohm resistors, which explicitly avoid the nearby signal ground bus (that the 2k resistor is connected to!) and are connected to power ground by a through-hole via. Thus, the OCP circuit is measuring the input current (voltage dropped through the circuit board traces) instead of the output current. Hence my convoluted way of drawing those resistors up to power ground instead of staying in the visual envelope and drawing them down to the signal ground bus for a cleaner look.
Ironically, I have tried unsuccessfully to switch that op-amp to a positive-input based output current sense, because I don't need the OCP function (each unit is only connected to 1000w of panels), and I could really use a "voltage overregulation" function, where the output voltage is linearly boosted under higher current loads to compensate for wiring losses (output of that op-amp would be reconnected in some way to pin 2 of the TL494). But I haven't had success so far—unlike the TL494's op-amp, the LM358 seems to just get "stuck on" when dealing with voltages that close to zero on the positive input.
You're probably right about the buck converter chip. On my boost converter, its top has been completely ground off, obscuring both its part number and its orientation. Combined with a small bar running across the "top", I assumed its orientation was upright, just like the LM358. Judging by photos found online, it appears that whoever made the 1500w boost converters with the four-lead input block didn't grind that chip, but everyone else did. Sadly, none of the photos are clear enough for me to read their part number. It also looks like that unit uses pin 8 (or pin 4) of the chip, whereas it appears to be unconnected on my 1800w unit—so I don't know for sure if they're using the same chip or not.
I personally think that their using LEDs as "or diodes" while simultaneously functioning as indicators was brilliant—lower component count and slightly higher efficiency, since the current would be flowing anyway. They just needed to consider the blue LED voltage drop and the LM358 drive overhead! If they had omitted the OCP feedback capacitor and swapped their use of red and blue LEDs (so blue for power and red for UVP and OCP), it would've saved them money (red LEDs are cheaper), the 5-10v "fry zone" would be eliminated, the color coding would make more sense (red for warning), and UVP would regulate the input voltage properly. They could've also eliminated the 75k power LED resistor by putting the power LED in series with the 200k voltage adjustment knob, resulting in a constant current over the entire adjustable voltage range for that LED. They could've saved even more money and eliminated a TL431 plus a few resistors if they had tapped into the TL494's 5v reference and adjusted a couple resistors instead of creating a 10v reference just for the fan. Or saved even more money removed all those fan control components like I did, adding variable fan speed and OTP in the process!
The output shunt is too small for any of my meters to read. I plan to calculate its resistance by putting one of these boost converters under a controlled, high load and measuring the resulting voltage drop. After confirming a few things and performing a few more tests, I plan to update the schematic and post some visual material regarding the changes I've made to these converters. Thanks for the AP1509 part suggestion, that does look like the chip they're using—or at least a plausible replacement!
Edited 2020-10-30 03:38 by Techie007