Solar System

Playing with solar in the year 2023

Solar, not practical here

Where I live, the back of my small ranch house points directly south, but there are many trees that keep my back yard and roof in the shade for most of the day. Ironically we get more sun in the winter when there are no leaves. And decent sun in the morning on the eastern facing side. But that's not gonna keep me from playing. In 2021 my friend Larry had a boat, and the three 120W solar panels on it failed. He bought new ones and I got the old bad ones. Two of the 3 connection boxes had partially melted. The panels seemed to be in decent shape. The diodes had fried in two of the panels. I measured the raw panels and they indeed put out  > 100W.  Removing and replacing the diode boxes was pretty simple.

These 3 panels are nominally 12V (closer to 18V). They have 4 x 9 = 36 cells each, which is typical for 12V panels. On his boat they were wired in series to a charge controller which charged the 12V batteries.

Why did they fail? The diodes clearly overheated, but why?  There is a large Radar unit near the panels which partially shades the panels depending on the boat angle to the sun. We suspected a shading problem. Each panel puts out about 18V at 7A, But when one panel is shaded, it's output drops significantly, and the 7A that the other panels put out need to go somewhere. It goes into the diodes of the shaded panel. The diode boxes typically have three diodes in series. This puzzled me. Since they are getting constant current from the other panels in series, what's the sense of using 3 diodes? That is 3x the voltage drop (wasted voltage) and 3x the power dissipation of a single diode. Is it for voltage rating?

Bypass Diodes

Since I was puzzled, I built a LTSpice simulation of a solar panels to see what was happening. A simple model for one solar cell is a current source in parallel with a single forward-biased diode. Here is 2 panels in series, each with 36 cells. Each panel has 3 bypass diodes in series.

sim1

Each cell is outputting 5A, except for the upper right one which is simulating a shaded panel that is outputing 1A. The first panel outputs 27V (blue), and the 5A cells of the second panel raise that to 55V (red), but the single 1A cell is back biased with the full 26V of that panel, effectively reducing the output of that entire panel to 0 A!  The 3 bypass diodes conduct, and the total output is reduced by 3 diode drops to 25V (green).

By connecting a bypass diode D79 across the 1A cell, that cell's back-bias voltage is reduced to only one diode drop, and nearly the full panel output is output. Bypass diodes significantly reduce the effects of partial shading.

In the early days of solar panels, this shading effect was identified, and many panels brought out the 1/3 and 2/3 connection taps. This explains why junction boxes have 3 diodes in series. If the intermediate taps were connected, a partially shaded panel would only lose 1/3 of it's output, not 100% But many modern, low-cost panels do not bring out the 1/3 and 2/3 taps. I assume to save money.

junction1
Ideally each cell would have its own bypass diode. Some companies such as Optivolt have technology that helps shaded cells. I assume that's what they do.

But, using 3 diodes on a panel without any taps is bad. The 3 diodes dissipate 3x the power of a single diode, and also drop 3x 0.7V or 2.1V so the string is outputting 1.4V less than with a single diode. All bad!  The box is dissipating 3x the power that it should. At 10A, this means 3 x 0.7V X 10A = 21W vs. 0.7V x 10A = 7W. And the box is mounted to the back of a hot solar panel, so add another 40C temperature rise, and you have a recipe for diode failure. This is likely what caused the diodes on Larry's boat to fry. Another mitigating factor is that the mounting didn't allow any air flow behind the panels. 

Bypassing 2 of the 3 diodes would improve efficiency and reliability. It won't help get more power out of a shaded panel though.

BTW I notice that 3 diode junction boxes are hard to find. The current ones on Amazon  parallel 2 diodes and have 6 diodes as shown above. I'll bet this is because of failures. Plus the fact that modern panels output  higher currents.

Y&H WVC600 MicroInverter Teardown and Testing

inverter1

I bought a low-cost grid-tie inverter from Amazon. WVB-600, 120V option. The design looks good. It is designed to accommodate two 24V, 300W panels and has two set of MC4 connectors. These are tied together in parallel inside the converter. Each panel has it's own 15A fuse, so be careful if you tie the panels together externally,

As far as regulatory approvals, it comes with a small CE sticker. It has very few actual specifications. Not even efficiency. It is designed for 28 to 50V, and has two solar panel connectors. It is designed to accommodate two nominally 24V, 300W panels. Each panel connection is separate and isolated. There is a single MPPT controller for the two panels.

Included is a simple low-cost AC panel meter that shows how much power is being generated. Cost is $140.

Clearly the AC panel meter s not IP67 rated. To achieve the IP rating, the panel meter would need to be replaced with a straight AC power cord.
The enclosure has pretty good gasketing, and the connectors are gasketed. It is intended to be used outside. It has an IP67 rating. Ideally it should be mounted in a cool, dry location. Its thermal design is OK.  Also mounting an inverter outside behind hot solar panels and on a hot windless day exposes it to another 20-30C or so temp rise. It is 'rated' for 600W, but gets quite warm when I use it with my 240W panels. I measured efficiency at 240W and it is about 88-90%. If the efficiency is the same at 400-600 600W, then the box will dissipate about 60W when running two 300W panels. This is a lot of power without forced-air cooling.

There are reports of these shutting down due to over-temperature at high power. I recommend that above 300W or so it should be used indoors and have some forced-air cooling. A small  80mm, 12V,  2W fan should do pretty well. Gluing some finned heat sinks to the cover wouldn't hurt.

I thought I would do some testing, a tear-down and some design analysis. See if anything can be done to reduce temperature rise and maybe improve safety and reliability. Maybe improve efficiency?

I powered it from a 24V lab supply. It drew about 5A current draw did not increase beyond 5A. It really wants 28V minimum.

From a 28V, 250W, 9A switching supply, it draws the full 10A, causing the supply to current limit at about 10A. It outputs 250W. It shuts down and cycles ON/OFF due to the power supply going into current limit when the output hits 250W. The measured efficiency at 28V is about 250W out / (10A X 28V in ) or about 90%. Not bad.

Here is a decent description of a grid-tie inverter. It is a decent representation of the WVC-600. Except the WVC-600 doesn't have the input boost converter. DC goes directly to the H-Bridge.

inverter
        annotated


Solar Panel Simulator

To properly test this inverter, I need a solar panel simulator. Something that will power the inverter without the variations of an actual solar panel. But real solar panel simulators are quite expensive. Can I build a low cost one that works?

A power supply is one solution. But power supplies have a sharp current limit. And switching supplies aren't intended to operate continuously in current-limit.

A solar panel acts as a voltage source with a source resistance, and then transitions to a current source as the load is increased. Maximum Power Point controllers depend on this soft 'knee'. They work by gradually increasing current draw to find the maximum power. When the power begins to drop due to too much current draw, they back off the current. Then they continuously increase and decrease the current to find and stay at the maximum power point. If there is no current limit, they may continuously increase until they reach their current limit. If the current limit is too abrupt, they may not be stable.

Here is a typical I vs V curve and Power vs V of a solar panel. Power supplies are typically specified with Voltage on the Y axis. The curve is pretty similar if you swap the axes.  I will use I on the X axis and V on the Y axis.
vi
 
A switching power supply will provide the raw power, but some type of external current limit is needed. To get a soft 'knee' maybe the current limit circuit should have a low gain? Ideally the voltage should be about 36-48V, and the current limit should be about 1-10A.

The ideal curve equation is hard to model in straight analog. Easier in digital. But I'm an Analog Guy so here goes my analog attempt.

In EE101, I learned that to deliver maximum power, the load resistor is equal to the source resistor. So a simple solar panel simulator could be a Thevenin voltage source with a power resistor in series. Unfortunately to deliver 500W, the source resistor also dissipates 500W. And the power supply must deliver 1000W. A simple thevenin is not practical at these power levels. There needs to be a current monitor that then drops the output voltage in a smooth manner. When the current exceeds the MPP point, the voltage decreases.

Here is a LTSpice simulation of a Solar panel simulator. At least in Spice it looks good :-). It uses a LDO PNP stage that I developed for the Triple Power Supply. The high-side current sense uses a diff amp, and has imperfect common mode rejection. Using 0.1% resistors for the 10K, 100K and 20K values will help.

Some of the resistors are higher power: R7 and R2 dissipate 5W at 10A, and so are are 10 or 25W chassis mount. Q1 dissipates about 30W at the maximum power point. R30 dissipates about 3W (9V, 300mA) so should be 5W or more. R21 is 30V so about 1W dissipation. Use a 3W. Q3 is 20V * .3A so about 6W. It should be a TO-220 mounted the the heat sink. The heat sink will see about 50W total. It should be large and have a fan. Q1 should be a TO-247 or TO-3. If the output is shorted to GND, the power of Q1 can get very high.

The max current is controlled by V4. The negative slope of the V curve is controlled by the current feedback gain, R18. 

Here is V4 stepped through 4 different values to set the max current and therefore the max power. Pretty good.

sim3

 I didn't feel like building a big breadboard or a PC board for a one-off, so I tried for something simpler:

sim4

Here is a much simpler circuit. It uses a single transistor to sense the current. I built it and it works, but initially the inverter would not stabilize, I thought that maybe the cutoff slope was to steep. Turns out the inverter won't stabilize unless the Vin is > ~25V. I initially used a 28V power supply turned up to about 31V. After current limit the voltage drops to about 24V, not enough for the inverter. I tried a 36V power supply (24V 250W in series with a 12V 250W) and it works fine. The inverter finds the maximum power point at both 4A and 8A. But even at 4A the power transistor gets very hot. 10V drop at 4A runs about 100C temperature with a TO-3 transistor on a medium sized finned heat sink (12x8cm). Definitely need a fan, larger heat sink, and more than 1 transistor for 8A. At 8A with a single transistor, the transistor shorted out.  I have some old TO-3 transistors, but these are hard to mount on a flat heat sink. I'm thinking 2 modern TO-247 transistors on a large, air-cooled heat sink. TIP35C look line a good candidate. 140W power, low cost. The TO-220 driver transistor and the power resistors will mount to the heat sink.

real 0r18

Here is the measured output of a 4A (single 0.18 ohm resistor). The transistors I used are TIP41C for the driver, and a single 2N3773. Since TO3's are a pain to install and wire, I suggest two TIP35C in TO-247,  and to run each one at 5A or less.

sim5
To vary the power, you can use either a single or two transistors.To reduce the current from 8A to 4A, turn off Q2. This can be done with a low-current SPDT switch by switching the base of Q2 from the driver to its own emitter.

Here is the messy prototype using a single power transistor. The 36V supply consists of a 12V 250W and a 24V 250W power supplies stacked and wired in series. The TO3 heat sink has a 2N3773 transistor and a TIP41C driver transistor on the back. The two power resistors are 0.18 ohm 25W. The small breadboard has the 2N3904 and the two 1K resistors. A low-cost panel meter indicates the voltage and current up to 32V and 10A. The beefy silver heat sink is for when I get some modern TO-247 transistors. The fan runs off the 12V power supply. 

bread

Ideally I'd use a 48V 10A supply instead of these old, non-enclosed ones. They are pretty cheap. I'd need to find 12V for the panel meter and fan.



  Dave's Home Page


Last Updated: 10/20/2023