### Mini Design Review There are currently two stages, BOOST, and PUSLE. So if 20mW is insufficient to charge the LiPo, then I probably need to do the pulsing after all. Or just get WAY more power out. I can try two things right now: - Use the DC Power Supply to push 5V CV, and slowly ramp the power rating until I can confirm the battery is charging. This will represent the steady-state charging minimum - Use the Function Generator and a MOSFET to switch the DC Power Supply intermittently and deliver tuned-energy pulses into the Battery I know that the AA battery can push at least 100mW, so a steady DC 5V 100mW is worth testing. If that works, then all I would need to do is increase the boost stage's efficiency. I've seen the battery decay down to 50-70mW, so it's worth finding the floor. I also need to find the highest output current I can get from a battery by shorting it. And then I need to find the highest SAFE output I can get from a battery, which I assume will be through impedance-matching. I could just get any old wire and short the battery, but I might burn my hands, explode, start a fire, etc. A terminal block might be safer. I also have some nichrome wire I might be able to burn with it? GPT says safe and practical current draw is 10-500mA, and that bursts can reach 3-5A. I suppose a handheld multimeter in current sense mode could BE the short. That's probably fastest, so let's try that first. ### Shorting the AA Short Current is about 1.6A at peak. No obvious heating. Definitely no smoke or explosions. Start: 1.65V, 87mΩ, 100% After Shorting for ~2 minutes: 1.49V, 26mΩ, 95% Assuming a shunt/sense impedance: 0.100Ω Measuring the resistance of my leads: 0.03Ω x 2 Measuring the resistance of the battery holder's leads: 0.62 Ω x 2 So I've got around 1.0-1.5Ω Which explains the modest 1.0-1.5A current I'm seeing at 1.6V. And I can cut out ~1.2Ω of that (?) Even if I can form a 0Ω connection, I still need to sense the current. But a 100mΩ load (the meter) would probably be approximately max power theorem, since ESR is ~100mΩ anyway. So maybe I just touch the battery with the leads directly. Battery before meter-shorting: 1.59V, 36mΩ, **100%(!)** (lol.) Seeing around 8.0-8.5A with the direct short. That's more like it. Okay, building something safer, just in case. Got a jig set up. ![[Pasted image 20250210045557.png]] Battery never exploded, only got warm. Around 8A slowly and steadily dropping all the way down to 0.5A before my hand got tired. I want to see how fast I can discharge a battery from start to finish. Shorting the battery indefinitely, it really levels out around 120mA and stays there for a very long time. Measuring the Battery immediately after disconnecting: 1.32V, 97mΩ, 70% 110mΩ measured a few moments later. What I'm learning from this is I'm going to have to use the DC Electronic Load to figure out exactly what conditions are necessary to get the highest average max power out of the AA battery. ### Breaking out the PSU and LOAD Time for two more tests: 1. Using a Pulsed Power Supply to feed the BMS 2. Using the Electronic Load to drain the battery The electronic load has a Battery mode. I can set it to drain CV/CC/CP from the battery with specific stop conditions. It also keeps a total charge transferred counter, which is pretty nice. Turns out I already had a battery torture device right in front of me! Running a CC of 1A for 120s drained 33mAh, which amounts to about 1% of the battery's expected total capacity. It probably can't maintain that all the way to the bottom, but that would mean 200 minutes to drain. I want to target 1-2 hours total drain time if possible, so I think the electronic load is going to be key to achieving that. About 120mAh in 5 minutes. Seeing like 30mV by 1.45A Setting it to CR at 50mΩ gets me 120mW. That's better. 25mAh in 60s. It's interesting that setting it to 5W CP yields 60mW, but asking for 100mW gets 100mW. There's a lot I don't understand here. It's using a 19Ω load to drain it. It's like the ESR increases when the battery becomes loaded. I can ask for 250mW and it gives it. Finally making progress >1.251V, 0.197A, 0.25W, 6.332Ω 0.5W works as well >0.910V, 0.550A, 0.50W, 1.600Ω At 0.75W, it crashes (0.03V) and drops back down to 50mW. I think there might be some sort of protection inside the AA that throttles it to 50mW or something. When it crashes to 50mW, the voltage drops to ~20mV, and the current around 1-1.5A. ![[IMG_3567.webp]] Slowly tuning down the CR from 1.2, 1.1, 1.0... 0.5Ω lowers the power to 0.048W, so the sweet spot is around 1Ω Right around 550mW it's short circuit thing happens. Well, working with a 500mW target at around 1.2Ω seems to be the best I can do right now. ### AA Drainmaxxing Trying more random stuff to try and get more power out of the AA cell. Next is creating a pulsed load on the AA cell to see if I can get a higher average power drain. DC Load can measure DCR. **Which is different from ESR!** - ESR is the sum of all resistive losses including reactive elements - DCR is the DC Resistance measured by analyzing how the battery responds to current load. I'm getting a DCR of around 1.0-1.2Ω on the battery, which matches with the results I was getting before. A load resistance of 1.0-1.2Ω was producing the max power output. That's with 1A/1s, 0.1A/1s. With 2A/0.1s, 0.05A/1s, I get 0.4Ω. And 0.4Ω is the reading I was getting after the battery crashes out. The shortest time I can run a high current pulse is 20ms (machine limit). Steadily increasing the current from 1A causes a steady decrease in DCR from 1.2 down to 0.8 so far. By running an OCP test, I can set a dropout voltage and have it step current periodically until the voltage source crashes. Using this I was able to map out the parabolic power curves for the batteries and find that it doesn't really crash out every time per se, but slowly declines. There's still some sort of crashout mechanism going on if it's suddenly overburdened though. ![[IMG_3568.webp]] The peak power is around 0.4-0.5W, as expected. With ~0.6A and ~1.0-1.2Ω load. I don't believe the battery can sustain 0.5W indefinitely. But I think giving it a duty cycle might be able to prevent it from crashing out. ![[Pasted image 20250210050130.png]] ![[Pasted image 20250210050141.png]] ![[Pasted image 20250210050215.png]] Even with the pulsed load/duty cycle, it eventually tired out, and the average power begins to decline until the bubbles stop. Stepping the current even more slowly from 0.3A Peak power is typically around 0.45-0.55A it seems like. Testing cells for crashout at CC 0.5A now. Seems safe and reasonable, for now. #### Max Steady State Power: 400-500mW @ CC 0.5A Trying a pulse config: 0.5A, 5s period, 90% duty -> 360mW steady. Now, using that, let's calculate LTC3105 supply power based on datasheet efficiency. Loaded Cell Stats: - 0.725V - 0.495A - 0.36W - 1.467Ω ### LTSpice Remodel with new findings ![[Pasted image 20250208071550.png]] Now I can remodel the battery source as it behaves under loaded conditions (0.725V, 0.5A max). And now I'm seeing a steady 4.2V signal as intended, but with power delivery of around 160mW, which is MUCH better than what I was working with before. The Schmitt Trigger is still in place to keep the output voltage regulated within the LiPo's tolerance, but it's not oscillating because the trickle current is strong enough to prevent it from decaying to the lower threshold. With something like this, I can now try two things: 1. Switch the output voltage of the LTC3105 to a 5V USB Rail and confirm power there. 2. Use the DC Power Supply to mock up that input voltage and see how the charging looks. Apparently USB Rail Tolerance is ±5% I'm having trouble getting a solid 5V out of the LTC3105. Considered adding an LM7805 but it needs more than 5V input to regulate, and the LTC3105 goes up to 5.25V only, so that won't work. Considering the MIC2940A as well ``` **MIC2940A**: - **Features**: Low dropout voltage (~350mV at 1A), thermal and overload protection, adjustable output version available. - **Output Current**: Up to 1.25A - **Efficiency Considerations**: Best suited when the input voltage is close to 5V to minimize power dissipation. ``` But there's no LTSpice model for that. To keep things moving, I'll just use a 5.1V Zener for now. After adjusting the R2/R1 ratio to set the LTC3105's Vout to 5V, and adding a the Zener (1N4733A), I can put a 150Ω dummy load to get the 160mW power transfer I'm looking for. 30mA into the BMS. Now let's set that up in real life with the power supply. 100mW to 150mW functionality is the goal here. ### BMS Supply Spoofing Giving the BMS 5V @ 0.5A (2.5W) It reads 4.754/0.499A, and goes into CC mode. Giving it 5V/1A, it will swing between CC and CV, pulling the full 5W. The timer function on my PSU is for timeout, not pulsing. So I think I have to set up my little switcher circuit after all. Let the BMS charge for a while off of the PSU. I had the signal generator running a 50% square wave at 50mHz (20s) so 10s/10s for about 2.5 hours, and it went up ~10%. The current going through the switched FET was about 80mA. So 40mA avg. over time. 3.64V / 45% Now for some DC steady state charging at low power: After letting it charge for a while at 160mW (33mA), the cell rose to 3.66V / 50% After letting it charge for most of a day at 160mW 3.74V / 60% It's clearly going up, however slowly. But I can't help but think that I'm losing a lot of what little energy I have to powering blinking LEDs, and keeping small bits of compute and power energized that I don't need. I'm starting to think that I may need to learn how to roll my own BMS after all. I think that's the only way to prevent the storage efficiency from being slow and pathetic. Even if I can drain a battery at a (currently unrealistic) 500mW, the BMS needs to eat up as little of that power as possible. I think for similar power-constraint reasons I probably just shouldn't include any indicator LEDs on the main design. I wouldn't mind adding them as frills for a demo, but for the core design it only degrades my performance criteria.