Back in 2002, I was working on getting a PV solar panel system installed on my house.
My initial calculation for the pay-back period was 15 years, due to slightly less-than-optimal roof orientation and some shade from large cedar trees around the house. This was based on the “easy” approach of having a local company handle the entire project, including the rebate.
By paying for consulting to design the system, buying the equipment myself, hiring a contractor to install it, and dealing with the California state rebate program directly, I managed to get the break-even time down to about 7 years. Though at the expense of significant hassle and a few close calls, like the fact that my original inverter couldn’t deal with the (lower) voltage gain from the Kyocera panels I wound up buying.
But back to the title of this post – while working on the design, I searched the net to see if somebody had a way of calculating the optimal angle for the panels on the roof. I found exactly what I was looking for here. Then I noticed that the author of this page is Charles Landau, somebody I had worked with briefly while consulting at Palm. And then I found out that he lives in Nevada City, about a mile away from my brother-in-law.
Now we’re both members of the Nevada City tech lunch group, so we get to talk about solar panels, environmental testing, and open source projects like his CapROS operating system.
Ken's Techno Tidbits 
I’m curious about your output when shade happens. When standard PV panels have shade on just one corner can cut your power by 25% – 50% on that panel, and if that panels are wired in serial (which they almost always are) that panel should cut the total output of the string by the same amount. Shaded cells basically turn the panel into a big resistor.
That will happen unless there are diodes build into the panel to bypass shaded (high impedance) cells.
I’m curious what kind of power loss you see as soon as shade hits it, and at the above pictured shading (I’d expect almost no power). If it’s only proportional to the percent that’s shaded then who makes your panels?
Thought more on this … I supposed it’s possible that each panel’s junction box has one diode to short circuit it as soon as it starts to become resistive, in which case you should see the power drop off incrementally in little steps as the shade moves across, one step for each panel shaded. Have you observed that?
Hi Dave,
Yes, when any of the panels in that group get shade, the entire series takes a similar output hit – which is a problem.
Two things you can’t see from that photo – there’s a similar array of 9 panels to the right, which is also currently shaded. And there’s a huge twin cedar tree in front of the house that’s doing all of this shading.
So we were faced with a dilemma – cut down a beautiful tree (and increase our summer cooling costs) or deal with getting significantly less output when the panels get shaded.
The layout I wound up with (two series of 9 panels) works well in the summer, when we need the power the most to offset A/C usage. When the sun is really high in the sky during June – Aug the shading only happens late in the day, and one of the two series of panels has no shade until 6-7pm.
It took a bit of number crunching to figure out a reasonable compromise, and the help of somebody with a Solar Pathfinder (http://www.solarpathfinder.com/).
I see. You should be aware that forcing current from the sunny panels can (will, if you have 9 in a string) damage the shaded panel over time. Becoming resistors they can get quite hot and damage the EVA (make it cloudy) and increase diffusion at the device level (destroying the PN junction).
I was thinking a possible solution that would only require a combiner box (relatively cheap – you could even make your own), is to rewire them as 6 strings of 3 panels each, combined in parallel in the combiner box. Then when one gets shaded it only knocks out 1/6 of the power until all three panels in that string are shaded. Then when the 4th get’s shaded it only knocks out an addition 1/6, and so on. The important result is that you have much less current flowing through the resistive cells – so it won’t ruin your panels.
Another cheap solution (6 for $12) is a panel-bypass diode between each of first 9 or so panels to get shading. Here’s a link: http://store.solar-electric.com/8ampbypdiod.html You’ll have to know how to solder.
Most panels should already have these built in the junction boxes but for whatever reason they don’t seem to work as well as external ones like these.
Personally, I’d just create 6 strings of parallel, but either method should work.
The only problem with paralleling 6 strings of 3 panels each is that the voltage level for the inverter will be too low.
An example of a small system sized inverter is the Sunnyboy 3000, and its start up voltage is 228V, while paralleling those Kyoceras in 3 module strings will get you a string voltage of somewhere between 75V-100V mpp depending on the module chosen.
The associated problem is that you will have to size up your #12 or #14 AWG coming from the roof to handle the roughly 50 Amps coming down. This also means that there is a larger associated Voltage drop (manageable though) just coming off the roof which only gets worse in the heat from the sun (the prime mover).
Hi Todd,
You’re right that voltage gain is an issue. My original inverter was an SMA Sunnyboy, which (IIRC) needed 250V. My Kyocera panels have a gain of 27V, so 9*27 = 243V.
So I wound up switching to the Sharp SunVista JH-3500U. The input voltage range is 110-350V DC, though the efficiency is lower than the Sunnyboy.
Given the Sharp specs, I think that Dave’s approach of 6 strings would work (6 * 27 = 162V), though I don’t know if that would impact efficiency.
Side note – Just last week I had to get a ground fault fixed. My Sharp control panel was displaying E-46. Checked the ground fault fuse in the box, but that was fine. Turns out three of the panels had slid down on the rails, and one of these wound up slicing through wiring in the process. I guess with all of the heating & cooling, the bolts holding down the brackets get loose.
Actually six strings would be 3 panels per string isn’t it? So that’s 3*27= 81V, which is 19V too low if the minimum startup voltage is 100V.
You could always use 4 strings of 4 panels and have 2 panels left over (sell on ebay) which would give you over 100V per string?
Todd’s point about the increased amperage is correct … cheapest solution is probably to just add 2 more cables in parallel of the gauge as the existing cable.
Hi Dave (& Todd),
Sorry, I’d read Dave’s suggestion as using 3 strings of 6 panels, though the number of times everybody said “6 strings” makes it clear I shouldn’t be writing comments when I’m not awake.
You’re right that 3 panels per string wouldn’t work.
— Ken
I’ve been thinking more about the bypass diode idea and I think it may help, but not by much (because they are in series, not parallel), and in fact the internal diodes may be working correctly already, though they do get ruined by spikes and over heating, etc. If they’ve ever been wired wrong, for example, they could be shot.
Try this: when our first panel starts getting shaded get a ladder and stick your hand on the shaded cells. Are they hot to the touch? Then your diodes aren’t working. If they aren’t hot then they are.
With regard to efficiency … greater parallelism (what I suggested) almost always results in higher efficiencies, because all the panels of a string will assume the lowest voltage of any panel on the string (that’s just how it works).
With that in mind since you can’t do 6 strings of 3 panels each your next best bet is probably to rewire them as 4 strings of 4 (16 panels) and you’ll get more juice than with 2 strings of 9 (18 panels), and it will be easier on your panels. As it is right now when the shade is moving across your panels you are getting a lot of mismatch between the two strings and that’s docking your power by up to 25% more than it would if both strings were always balanced.
Oops … I just hit “submit” on that last reply when I realized I was getting my math mixed up. Ignore the info about mismatch … that occurs if your voltage is mismatched not current (which is your case).
Nevertheless I still think you’ll get more power overall if you wire as 4 strings of 4 than 2 strings of 8, just because you’ll be going down in 2 smaller steps than one huge step. like this:
………#######
….#################
#########################
instead of this:
……..#########
……..#########
#########################
where “#” represents your power. You’ll notice you have less power at the peak of the day, but it’s more than compensated once shade kicks in.
Note you’ll want to double up on your existing wiring coming from the combiner box to the inverter.
Those graphs turned out bad. What I meant was this:
4strings of 4:
——-#######
—###############
#####################
2 strings of 9:
——#########
——#########
#####################
Sorry … wordpress seems determined to shrink my line of periods or dashes. I think you get the idea though. Each graphs is supposed to be symmetric.
The panels look great 🙂
So what you need is called “power point tracking”. Google it. Should cost about $800 but it will solve all your shading problems and based on you image I’m thinking it would pay for itself within 3 years.
Hi Dave,
Thanks for the ref to power point tracking. I’ll definitely follow up on that.
— Ken
{Blue Locktite}
The best value is the Xandex sunMizer MPPT modules. You only put them on the panels that get shaded. they’re $160 each. Looking at your picture though it seems like you have a lot of roof space not being used in bright sunlight. I’d move as many of your panels there as possible, then sell the few left over, and with that money but the Xandex SunMizers, and my guess is that you’ll see a significant improvement right away without it costing you a thing.
Hi Ken,
I’ve been trying to find an email contact for you. I train the use of PV simulation software and this image would be very useful for me.
Cheers,
Chris Nunn from the Solar Design Company