2
\$\begingroup\$

So I want to charge a stack of six series-connected super capacitors from a small 10W solar panel.

The part number of the capacitors is TPLH-2R7/12WR10X30.
https://www.tecategroup.com/products/data_sheet.php?i=TPLH-2R7/12WR10X30

The part number of the solar panel is TPS-12-10W.
https://tyconsystems.com/documentation/Spec%20Sheets/TPS-W%20Solar%20Panels%20Spec%20Sheet.pdf

Like most super-capacitors, the cells are rated at 2.7V each. I want to limit the voltage one each cell to be in the range of 2.0V to 2.5V max. At first glance the cells could be mismatched in value by at least 60%. The solar panel looks like it could supply 500mA to 600mA in full sunlight. Unless properly regulated, the open-circuit voltage would rise to 21.95V.

Given the above, what is the lowest part count and lowest cost way to ensure that the voltage on each cell doesn't exceed 2.5V when the stack is charged at 700mA peak.

What I want to do is stick some sort of device in parallel with each cell to regulate their voltages. When all cells are charged I want the total voltage to be between 12V and 13.5V. The voltage on each cell should not exceed 2.5V.

I have looked into various options.

  • Advanced Linear Devices Supercapacitor auto-balancing (SAB) (ALD8100xx / ALD9100xx). These are just precision threshold MOSFETs. It seems that a lot of people are using and recommending these.
    1. They are limited to 80mA max drain current.
    2. The datasheet shows that the drain-source voltage typically crosses 2.7V at around 9mA.
    3. This would seem to require that the capacitor cells are either well matched, or that the charging current is very low. None of which are true in my case.
    4. They have a 500mW power limit per part so I would need a separate regulator to take any excess power from the solar panel to keep the load voltage in regulation.
  • Zener-diodes.
    1. I don't seem to be able to find 2.25V zener diodes with the 1 ohm or less output impedance required in this case.
    2. Zener didoes tend to have a lot of leakage as you approach the threshold. In my case I want the leakage to be less than 100uA worst case at a cell voltage of 2.0V.
  • TL431 or similar.
    1. Very low leakage below threshold.
    2. Precise, adjustable threshold.
    3. Limited to 100mA. In my case I probably need to be able to shunt over 800mA to be safe.
  • TL431 + power BJT.
    1. Very low leakage below threshold.
    2. Precise, adjustable threshold.
    3. Can sink multiple amps of current.
    4. Uses at least four components per cell (so 24 total for the design). I was hoping for something simpler.

Ideally there would be some sort of two terminal device that had < 50uA of leakage below 2.25V but would sink at least 1A in the 2.25V to 2.5V range. I have looked quite a bit but have not found such a device. This actually surprised me because the vast majority of super-capacitor cells are rated at 2.7V. The cells usually have very high tolerances, and it is not uncommon to want to charge them at currents beyond 1A. The only viable solution I see right now is the "TL431 + power BJT" option, but it uses 24 parts when I would rather be using six.

\$\endgroup\$

3 Answers 3

2
\$\begingroup\$

what is the lowest part count and lowest cost way to ensure that the voltage on each cell doesn't exceed 2.5V when the stack is charged at 700mA peak.

The BW6101 is a super capacitor charging protection chip with up to 700 mA bleed current that can be set to 2.45 V (alarm output at 2.55 V). Typical working current is 20 uA. It requires a minimum of 2 parts per cell (itself plus a bleed resistor).

It is used in some protection modules sold on eBay. Depending on what facilities you have, it might be cheaper to buy one of these boards and modify it for 2.5V operation, rather than making your own.

\$\endgroup\$
3
  • \$\begingroup\$ neat, so it dumps the power in an external resistor. \$\endgroup\$
    – Neil_UK
    Commented Mar 15, 2020 at 6:55
  • \$\begingroup\$ Well I can't read kanji characters... does an English version of the datasheet exist? \$\endgroup\$
    – user4574
    Commented Mar 18, 2020 at 1:21
  • \$\begingroup\$ Unfortunately no. But you can upload the pdf to Google Translate (click on 'Google Apps' on top right of Google home page, then 'translate', 'documents', 'browse your computer'). \$\endgroup\$ Commented Mar 18, 2020 at 2:04
1
\$\begingroup\$
  1. Don't worry about it. Do you know you need it? (Are you using mismatched units/types? Drawing unbalanced current from some taps?) Leakage goes up aggressively with bias; they tend to be self-balancing on their own. A clamp for the total supply will do to manage charging. (Or just use enough total voltage rating so that the solar panel can't overcharge even at OCV. Probably an expensive and bulky proposition, but hey, still worth noting.)
  2. Be careful with discrete circuits. They tend to have very soft thresholds (e.g. Neil's example), high leakage, or both. Note that most shunt references have a subthreshold conduction region: TLV431 for example might achieve rated 1.24V at 100µA, but it draws say 5-50µA for voltages 1-1.2V say. (The leakage does tend to be low, well below the threshold, which can be useful even for TL431.)
  3. A continuous solution isn't necessary; at least, unless you need very quiet power output, perhaps? Case in point, Bruce's example: simply a comparator with hysteresis, voltage reference, and a reasonably powerful switch to dump into a resistor.

As a worked example, I built a device like #2 myself, recently; it's for a somewhat different purpose (dumping excess power from a bidirectional power supply/load), but works just the same. See the link below to my website. I chose a discrete design, as leakage isn't a problem in my case. This serves to illustrate how to use a shunt reference as a hysteretic comparator.

https://www.seventransistorlabs.com/Images/ShuntBallast.pdf

(This is obviously nowhere near a complete solution for the present question; just for ideas, or a jumping-off point.)

In any case, definitely look up manufacturer recommendations for using their parts in series. Perhaps my claims in #0 aren't applicable to everything, for example!

\$\endgroup\$
2
  • \$\begingroup\$ With respect to point #0: I believe you that leakage goes up with bias, but if the charging current is high enough, I could see the maximum ratings of individual cells being momentarily violated unless this was validated through experiment. I do see that several manufacturers of super capacitors do sell stacks of multiple capacitors in one package (which presumably are well enough matched, or have balancing integrated into the package), so that could be an option. \$\endgroup\$
    – user4574
    Commented Jul 31, 2022 at 15:59
  • \$\begingroup\$ @user4574 I should probably clarify that point, that by using matched stacks (equal values), the voltage distributes evenly. \$\endgroup\$ Commented Aug 1, 2022 at 0:34
0
\$\begingroup\$

TLV431+powerBJT would be my favourite as well (note the V in TLV).

However, you might find a colour of LED for which LED+powerBJT gives a voltage in your desired window with fewer components, albeit with a less sharp current knee and wider tolerance.

Another common way to get a high power low voltage shunt reference is this, again the knee is not as sharp as when using a TLV431, but your specifications on voltage tolerance seem to allow for that. It doesn't meet leakage at low voltage, but replacing R2 with series silicon diodes might work. It's still a higher component count than you want.

schematic

simulate this circuit – Schematic created using CircuitLab

\$\endgroup\$
1
  • \$\begingroup\$ With no biasing current for Q2 (add a R from Q2's emitter to GND), and a wide variation over temperature and transistor types (VBE varies from 0.4 V to 0.9), this circuit cannot be recommended. \$\endgroup\$
    – jp314
    Commented Jul 31, 2022 at 15:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.