• pic12f675/main.c - MCU source code with description and documentation;
• pic12f675/1.0_schematics.jpg - main schematics
• pic12f675/U_schematics_extension.jpg - extension to main schematics for UVLO feature ('U' topology)
• pic12f675/... - older schematics (Basic_Scheme - legacy file) and sample images of prototype #2 with extra comments
• pic12f675/compiled/... - firmware images (current version - 1.3.THR.OPTION).
• pic12f675/compiled/pic12f675_lp_mppt__1_1_198_B.hex - latest tested version.
• pic12f675/compiled/pic12f675_lp_mppt__1_3_202_F.hex - latest tested version.
• pic12f675/compiled/pic12f675_lp_mppt__1_3_202_U__uvlo138.hex - latest tested version.

Version format is major.minor.THR.OPTION where:

• THR is a threshold constant, which should be chosen according to used input diode or potential loss on input wires for optimal work:
  • compiled THR versions are 202, 198, 194; 202 is assumed for "no input diode" connection, 198 - for low drop diodes, 194 - for worse diodes.
  • THR choice affects efficiency and output pulsations and it's preferable to use lower THR (leave a little room until MPP) - see "PROJ_CONST_THR" notes and "Input diode Vf" notes in "main.c".
• OPTION is used pinout/topology, see "FW_VER_OPTION vaiants" in "main.c".
• version data is built into FW at the end addresses (before OSCCAL) so it can be easily checked, those bytes/commands aren't used by firmware.

• TBA: very high efficiency solar powerbank design;

• added 1_3_194_U, 1_3_198_U, 1_3_200_U; initially project was oriented for 18v and higher voltage panels as I thought it wouldn't be reasonable for low voltage, but luckily after deeper research it turned out that it's fine and reasonable even for single li-ion cell charging (3.7v output, 6v MPP rated solar, 3.3v voltage stabilizer for PIC) if everything is done right. The THR param is percentage of OCV voltage to be maintained on input (input voltage = OCV * THR/256), keeping in mind that measured OCV can be slighly smaller than real OCV and even without diodes there is some voltage loss on input wires, so even without input diodes there's some room of error which should be considered/corrected for small input voltages or for big currents and not-too-tough wires, so extra options were prebuilt and some of them will suit the setup best way and win some 1-2% over some other version (e.g. if expected maximum current is somewhat big and loss on wire is somewhat noticeable - it may be better to use *200 or *198 version instead of default *202);
• attempted to implement wider continuous mode range (to support lower duty-cycle ratios and yield less ripple current in certain cases - like for charging almost discharged li-ion cell), but it ended up being too poorly balanced and causing significant pulsations and coil wine so I discarded this idea (*attempt assumed keeping existing continuous mode part the same so maintaining high frequency on 1mhz MCU).

• fixed recovering cached last adc value for fastest case of measuring OCV - this was causing flipping of UVLO pin ON/OFF in certain cases

• added pic12f675/U_schematics_extension.jpg;
• added more documentation notes in top of main.c;
• FW has no other differencies with v1.1;
• assembled and partly tested new prototype for 'U' feature - very high efficiency Li-Ion charger for 6v-MPP-rated solar panel, no input/output diodes, no BMS (direct charging, no reverse current), up to 94% tested efficiency (typically 90+%), used 3.3v linear stabilizer for powering PIC and nice IRLML6244 SMD mosfets (except from 'FET3'), more info later.

• fixed minor pulsation in specific case of limiting (missed pwm-ON in one copypasted place);
• optimized OCV measuring in some cases (optimized multiplication with smarter approach);
• tested replacing 33uh indictor with 18uh - overall good, and when there's coil whine - it's less audible (with such change - the ripple current is higher, which implies higher requirements for diode and capacitors);
• tested real reverse current from battery to solar panel when it's dark/almost dark - nearly no reverse current (~2ma for 50w 18v panel with 12v battery connected), input diode doesn't seem reasonable anymore (personally I removed it from my setup), more details on overall approach will be sometime later;
• designed solution for reverse current protection without involving diodes and with nearly zero losses (for custom powerbank, current won't flow back to device when panel is removed/got dark), the solution is TBA and it will have a new pins topology (extended feature);
• rethinking usage for different kinds of output: when in non-continuous (discontinuous) mode - ripple current can be very high, which is important for high-current applications, (happens when output voltage is less than ~60% of rated MPP - e.g. relates to almost discharged li-ion batteries regardless of chosen solar panel voltage, solution/improvement is planned for implementation and is TBA).

• added TLDR_DIY.md
• added HOTWO_PROGRAM_PIC12F675_USING_ARDUINO_IN_5_MINUTES.md
• added pic12f675/1.0_schematics.jpg - extended schematics info
• added pic12f675/lp_mppt_pic12f675__prototype2_top_view.png
• added pic12f675/lp_mppt_pic12f675__prototype2_bottom_view_flipped.png
• added pic12f675/lp_mppt_pic12f675__prototype2_bottom_view_flipped_commented.png

• reworked mid and high PWM modes (now 72 modes in total);
• improved limiting logic and accuraccy;
• increased range of continuous mode (more cases/more possibility of falling into continuous mode);
• reduced ripple current;
• reduced coil whine is some cases;
• eased output capacitor requirements;
• optimized OCV measuring algorithm and used EEPROM for faster multiplication (smaller no-output-feeding gaps during OCV measurings : minimal possible gap is 52us);
• updated documentation (in main.c);
• ran wide testing using different equipment, efficiency results appeared slightly smaller (measuring errors +/-) but up to 95% conversion efficiency.

• revised and drastically reworked PWM modes, now 90 modes instead of old 26 modes;
• supported inductor's continuous mode (in part of cases / some power ranges it works in continuous mode);
• improved efficiency due to reworked PWM modes (mostly for higher-power ranges);
• significantly reduced ripple current;
• significantly improved pulsations (especially when in continuous mode, middle range is now well balanced too);
• significantly eased output capacitor and inductor requirements;
• significantly reduced coil whine (is some cases fully removed);
• improved output limiting accuraccy;
• revised/fixed adc mode/timings, revised and profiled OCV (Open Circuit Voltage) measurings;
• it's necessary to add an extra 47uf (any, e.g. electrolythic) capacitor (along with existing ceramic 1-2uf cap) to output of linear voltage regulator to make it work correctly*
  • *this fixes OCV measuring on higher power ranges as otherwise the used regulator (LP2950CZ) overshoots output voltage when input voltage rapidly increases, which significantly distorts ADC readings for OCV evaluation and has very long settling time (when only recommended small ceramic cap is used);
  • *adding this capacitor BEFORE updating the FW to 0.10 is NOT recommended due to [see 'startup voltage' note below];
  • *exact value of referred capacitor might be reconsidered but 47uf overall looks okay in current testing for both assembled prototypes.
• removed a feature of measuring startup voltage (which was used as minimum allowed MPP voltage) due to too many related problems:
  • apart of secondary problem PIC starts up with voltage around 2v, while ADC requires higher voltage to provide correct results, which makes it too complicated to attempt to measure minimal stable voltage;
  • so minimum voltage is now hardcoded to be 21.8% of measured range (actual voltage range depends on PIC input voltage and voltage divider on ADC line);
  • actual THR voltages will be from around 50% (to 80%) of divider range (solar panels would output somewhat power starting from around 50% of their nominal OCV);
  • *with older FW ADC provides completely invalid starting measured level if larger capacitor is added to PIC power supply.
• updated internal documentation/notes in main.c;
• tested conversion efficiency reached 97% @ 11.7v 6w output;
• still present issue: when input power is more than output load can consume (cut by limiting logic) - there's coil whine and some extra power loss (e.g. input is capable of providing 22watts, output can consume 19 watts, actual output is 18.5 watts), planned for further improvement of limiting logic;
• now when output can be (sometimes) as flat as a battery (not just voltage, but even current in 3 digits) it's very clear how much PCB layout/design matters as second prototype with smarter assembly works much more smoothly in same conditions, so at some point PCB layout example/recommendations will be provided, minimal recommendation would be to keep power circuit (fets+inductor+diode+output capacitor) as dense as possible and solder to capacitors as close as possible.

• added topology 'F' (used in second assembled prototype) with two PWM-driving pins (both not inverted, 100% synced), more PCB-frindly pinout;
• reworked balancing logic - will work well with PWM updates, still looks better as is (middle part of PWM modes is especially poorly balanced yet);
• new balancer assumes no significant noise on ADC line (not likely to ever be an issue in any circumstances);
• improved limiting logic/accuracy (still being improved, currently there are significant output spikes at higher power range);
• significantly reduced cycle delays (less response time - faster balancing);
• MPP voltage threshold is slightly lowered due to changed logic (by 1/256), might be compensated internally in future;
• *dual PWM pin in topology 'F' is rather for better temperature stability on chip for stronger mosfets, even strong (6 amps rated) external Low-Side mosfet driver TC4420CPA has made no impact on efficiency whatsoever, so no insufficient current for driving fets was observed in tested conditions (IRLZ34N fet, IRL2203 fet);
• *overall measured efficiency looks good and matches appropriate charts of buck converters with such topology (with diode), on 18v panel with output at 5v efficiency was from 78% (below 1 watt output) raising to 86% (at around 2 watts output) which is an expected characteristic (comparing to data of XL4015 in it's datasheet), with output at 12v characteristic is similar but higher and reached around 92.5% efficiency at 2 watts output (core components of used prototype here are ILR2203, IRLZ34N (pwm), SR840 diode, CDRH127/LDNP-330MC - 33uH inductor).

• added output limiting pin (GPIO4; pull to gnd when outputting power needs to be suspended; has internal pullup resistor);
• added inverted PWM output (GPIO5; strict phase matching with non-inverted PWM) - might be used in some cases (e.g. when some output mosfet driver is involved);
• compatible with 'A' versions if GPIO4 and GPIO5 weren't used (as it was recommended) - see 'FW_VER_OPTION vaiants' updated;
• this release has been more or less tested in various scenarious involving initial and second assembled prototype boards;
• plenty of testing has been done during last ~1 month, related info/assumptions are to be updated sometime later ('Releases' section will get a new tag as well).

• added input diode Vf consideration (since now several .THR options are built for optimal choice) - significantly reduces pulsations, improves efficiency (avoids solar power generation loss in MPP overshoot);
• excluded debug/configure code part - reduces pulsations;
• used constant multiplication (hardcoded autogen algorithm) - reduces pulsations.