Solar Cells as DC Power Supplies

Printer-friendly versionPrinter-friendly version

Photovoltaic sources are simply direct-current power supplies. They have the advantage of being "free" to operate, but they have their own peculiarities in design. We'll start here with a general look at the nature of DC power supplies and then extend that into solar.  It requires a little bit of arithmetic to see how the pieces fit together.

Concepts from basic electricity.
A power supply is a power supply. It has a (+) and a (-) terminal. A normal, wholesome power supply puts out a constant voltage. It delivers current to a load in proportion to the size of the load.

The load determines the current -- up to the point where the power supply can't take it anymore and then not-very-good or maybe even bad things happen. This is called an overload condition.

Power is measured in watts. Watts = Volts x Amps. Since the supply voltage is assumed to be constant, the power in the system is related to the current. Current is measured in amps.

Takeaway: Voltage is constant. The load determines the current and hence the power.

In general, that's all there is to it regardless of the type of power source.

Low quality power:

A freshly charged battery produces a constant voltage, even as the load varies within reasonable limits. As a battery reaches the end of its life, its voltage will change if the load changes. Power engineers call voltage that sags under load (or has other artifacts) "low quality power". Flashlights get dim as their batteries discharge. An _ideal_ battery would deliver its rated voltage until it was completely empty, then suddenly stop.

Since real batteries don't behave this way, battery powered devices have to be able to deal with low quality power. Mostly this is only annoying, but there are exceptions. MCU's can suffer program memory corruption because of random instruction execution during brownout. Many MCU's have BOD systems to take corrective action before this occurs. (Brownout detection is configured in the application source code.)

Normal AA size batteries aren't a big conceptual deal in a battery powered MCU-based device. When the thing stops working you change (or recharge) the batteries.

Solar Cells:
The voltage of a solar cell is fixed by the chemistry of the semiconductor. Nominally, a solar cell in "full sun" produces around 0.45 volts at its rated current. Photons move the current in the cell. The more photons, the more current is available. Photons are collected across the area of the cell's surface. More area will capture more photons and make more current. So there are two, nearly independent ways to get more current: a bigger cell, or more light. Power ratings are typically optimistic and assume a particular rule of thumb: Solar flux is about 1kW per square meter at noon at the equator. In Chicago the average is 350W per square meter. Solar cells are 13-14% efficient in bright sun, so a 1m x 1m array would give an average of about 50 watts. The figure of 45mW per square inch in full sun is a rough figure for estimating the capacity of a solar cell.

Natural power sources produce low quality power. The voltage is highly variable and unpredictable. Solar cells are typical this way. They violate the first criterion of a wholesome power supply -- their voltage isn't constant. Not even remotely close.

There are a few ways of dealing with this.

Match the cell and the load closely, and keep a constant light on the cell when the device is required to operate. [solar toys]

Use an over-sized cell and keep some light on it all times, but never allow the light to go below an operational minimum. This works if the load is constant and also if the load is changing. [car ventilator fans]

Use a rechargeable battery to even out variations in the cell output as the load is powered at all times. The cell must be big enough to charge the battery AND run the load. [most commercial applications]

Same as above, but allow the load to shut down while the cell charges the battery. [solar flashlight]

Batteries deliver current to a load. Current is measured in milliamp. The total current used, either charging or discharging a battery is measured in milliamp-hours, mAh. (milliamps x hours). A load which uses 150 mA for 2 hours, uses 300mAh. An AA NiCad supplies about 700mAh, and an AA NiMH supplies about 2000mAh, both at 1.2 volts

For a load to consume 2000mAh, it seems fair that you would have to put 2000mAh into the battery. Actually the batteries lose almost half their power in the charging process, so the solar cell charging the NiMH would have to produce about 4000mAh. Under normal contritions, charge time is inversely proportional to charge current. Twice the mA going into a battery cuts the charge time by about half, but the NIMH will not tolerate charge current over 2500mA. Since small solar cells only put out a small fraction of this current, the concern would more likely be a long, tedious charging time, rather than overcurrent. It is also possible that at very low charging currents, the charging process looses some efficiency.

Electronic charge control:
Solar cell output voltage is low and variable. A solar cell plus a blocking diode will charge batteries directly, but the system is usually weak and unreliable. An improved method is to use a small switch-mode regulator circuit to take chronically under-voltaged output from a solar cell, and exchange some of the solar mA to get the voltage up to a nice level for charging the battery. This approach has become so cheap that LED lawn lights use this arrangement. The charge current is still below 50mA (I measured 25mA), so it's doubtful that the lawn lights are ever really charged very much. They do work though, because I see them all around my neighborhood. Lowes sells a 10-pack for ~$40.

Voltage considerations:
Assume you've got solar feeding a battery. It is a DC supply of around 1.2 volts. From now on we can treat that supply generically and use it like a plain battery.

Boost conversion
Most electronics runs at 3.3V or 5V. TO get from cell voltage from something other than an LED or similar device, the solar derived voltage has to be increased. This is done at the expense of supply current in a boost type switching regulator, or in a charge pump. Sparkfun sells one for $15 DEV-08466. It take a single AAA battery and makes 5V at 100mA. Also they sell the NCP1400 is a 5V DC-DC converter. The breakout board will accept voltage inputs between 1 and 4 Volts and output a constant, low ripple 5V output capable of sourcing up to 100 mA. This board is great for supplying power to 5V sensors on a 3.3V board, or providing 5V from a AA battery.

Switching regulators can be very efficient, but I'd count on losing about 60% of the battery power in the inverter.

An LED is about 10mA
A pager motor is around 50mA
An AVR chip running at 16MHz is around 25mA.

So, for example,
3 LED's (3 x 10mA) +
1 pager motors (50mA)+
1 MCU (25mA)

comes to 105mA at 3.3V or 5V.

This is a bit over the PRT-08999's 100mA max, but we won't run the pager motor
long anyway. Power is .105A x 5V = 525 mW total.

We'll get 40% of the energy stored in the battery, so to use 525mW, we'll need
to draw 525 / .4 = 1312 mW

From a 1.2V NiMH battery, we'll draw 1312mW / 1.2V = 1093mA
For a fully charged NiMH holding 2300mAH, the setup should run about 2300/1093 =
2.1 hours.


A sloar cell that can produce the same power 1312mW / (45mw/sq in) = 29 sq in. About 5.5 x 5.5 inches square.

A Caveat. All the numbers here a very, very rough. They're probably in the right order of magnitude. However. The relationships between supply and load characteristics in all the cases discussed here are highly non-linear. That is to say that the shapes of the curves that relate light, voltage, current, and load in a solar cell are not straight lines. Batteries don't charge or discharge on straight line curves. The efficiency of switching regulators varies with input and load conditions, and this affects the load on the upstream supply. And so on.

A few open-ended experimental setups for evaluating hardware under a few different conditions would yield informational vectors pointing toward real utility. A small number of useful device arrangements in the form of mixes of loads, cells, power electronics, and storage batteries should be built and tested under different conditions. Each device arrangement would be assesed using parameters typical for this type of design.


    Example parameters would be :

  • maximum and minimum availabe light intensity

  • maximum light duration (hours)

  • charge time

  • discharge time

  • total energy capacity (w-h)

  • peak load consumption (mA)

  • average load consumption (mA)

  • load duty cycle (short and long duration averages) (percent)

  • weight

  • volumetric size of circuitry (by LxWxH dimensions)

  • length and width of the solar panel(s)

  • cost

  • availability of materials

  • fabrication skill required for assembly

  • build-or-buy?


Alternatively, the behavior of a couple of potentially useful device arrangements could be described in some amount of detail, then prototyped using hacked commercial products as a guide for power supply design. During evaluation and development, the architecture of the device arrangement would be iteratively modified to improve the most desirable parameters. The product would be a power supply design compatible with device arrangements similar to the prototype.