This is a (somewhat rambling) description of our residential, 4 kilowatt, grid-tied, photovoltaic power generation system. If you do better with pictures only, go here.

The site

The site, our house, is a standard, wood-frame, residential home built in the 1950s. It is located in the foothills near Los Angeles about 10 miles from Pasadena, CA (and thus CalTech/JPL, where I work). There are good things and bad about this site with respect to building and installing a PV system.

The Good

The Bad

Design overview

Ultimately we decided to mount the solar panels rigidly on the south-facing sections of our house roof. We have 40x100-watt modules (of 36 cells each) attached to a sturdy frame about 8-inches above, and parallel to, the roof. Each panel is nominal 17V open-circuit, and we have paired them up to give us a "24V" system. We have divided the total array into 4 sections, each having 4-6 24V module pairs, depending on the size of that section of the roof. The wires from each module pair run to a junction box for that section, and from the junction boxes down to the inverter cabinet on the ground.

Once in the inverter cabinet the wires from the 4 sections go through a ground fault interrupter and then through a set of DC breakers and finally to the batteries. The breakers allow us to disconnect the array sections from the batteries (for servicing, etc). Between the batteries and the inverters are 2 more DC breakers, one per inverter, that likewise allow us to shut down the inverters.

Also connected to the inverters is the AC input from the public power grid, as well as the AC output to the house loads. The inverter manages the power from these sources, keeping the DC voltage in a range that makes the batteries happy and supplying AC power to the house loads. If the loads require more power than is available from the panels, enough is pulled in from the public grid to make up the difference. If the loads are not fully using the solar power, the "extra" is pumped onto the public grid, running the meter backwards.

In normal operation the batteries in our system are not used. Only when the grid power goes away completely do the inverters draw on the batteries. This keeps the batteries from constantly going through the discharge/charge cycle, increasing their life and minimizing maintenance, while at the same time protecting us (OK, our computers) from brownouts and (most) blackouts.

The solar panels and associated wiring

The physical connections

First, of course, there are the solar panels. Here's a couple pictures of all 40 of them stacked on our driveway/work area just after delivery. Each panel is rated at 100 watts output @ 17 volts. Our system is designed for 24 volts so we put pairs of panels together into 24V "mini-arrays". Yes, I said 24V. Even though the panels are rated for 17V each (which would seem to mean that 2 panels would put out around 35VDC), that's open-circuit voltage. In operation you get somewhat less than that once you actually start trying to push electrons through useful loads.

In fact, the open-circuit voltage of one of the mini-arrays floats at almost twice that - around 40V (regardless of light level, oddly enough) - but as soon as you attach them to a reasonable load (or a battery) they lose 6-8 volts. They'll lose even more if there are clouds or high temperatures. Once installed in the circuit with the inverters, the inverters will draw down this voltage even further to maintain correct voltage for the batteries. Thus this configuration is correct, even though the numbers don't seem to add up initially.

In the middle of this picture (on the sawhorses) are two modules in the process of being wired into one of these mini-arrays and having some pieces of angle-aluminum bolted on. These brackets not only hold the arrays together, but they also serve to attach the mini-array to the mount points on the roof. Here's a couple pictures of the first of these mini-arrays up on the roof. The arrays are connected to the roof by parallel lengths of angle-aluminum.

Here's another picture I took the next day that shows the scheme a little better. We have arrays on 3 sections of our roof, the front (west) of the house, the south, and the east. The east array is actually two arrays, but since they're all connected in parallel to the inverters the distinction is mostly semantic. Once the panels are all installed, the conduit and combiner boxes attached, and all the wires are connected everything looks like this and this. (That wire running from the lower-left corner of the second picture down to the roof is the incoming grid-power cable coming down from a telephone pole.)

Once the arrays have been bolted in place, the wires coming from the mini-arrays on each segment are routed into a combiner box. The combiner boxes connect all the power-carrying wires together so that the output of all panels in a segment comes out of the combiner box as one large cable carrying up to (in our case) ~1K watts @24VDC. In our design we have 4 combiner boxes, so we have 4 of these wires plus their corresponding ground wires. Also going into the combiner boxes is an isolated chassis-ground wire from the metal frame of the panels (actually it's attached to the rails, which are attached to the frame on the panels). All the wires go down through conduit to the inverters' electrical panel.

Once they get to the electrical panel, the first thing that happens is that the power and frame-ground goes through a ground-fault protector. Then the 4 power-carrying wires each go through their own independent breakers so we can shut them down individually if we need to (to work on them for example). After the breakers they get combined into one line that goes into the batteries and to the DC side of the inverters.

Each Trace 4024 inverter puts out 120VAC. We have two because we need 240AC. We need 240VAC for two reasons. One is because we have two 240VAC loads that we want to power, one of which is our air-conditioning (and there is something quite satisfying about powering your AC with the same sun that is making things so hot). The other (and main [sorry again]) reason is because we have 240VAC coming in from the public power grid.

The inverters have the ability to synchronize their outputs so that they can drive 240VAC loads, in addition to 120VAC loads. By connecting the inverters together we can connect one inverter to each of the two legs of the 240VAC input from the public power grid. This way we can drive both our 240VAC loads and our 120VAC loads. Connecting the inverters in this way also gives us the ability, if the batteries are fully charged and the solar panels are putting out more power than the house loads require, to push our excess power onto the public grid. (Yes! This will run the meter backwards.)

If that's all clear as mud, maybe this will help:

The inverters are packed with other features, some of which we will either not use or will use only very rarely in our installation (such as a gasoline generator input). One thing we will do in "stage II" after we get the basic system up is attach something to the "remote control port". Initially we will check status and control the operation of the unit(s) from the front panel, but this port allows us to connect to and manage the functions of the unit from a computer serial port. We are hoping to link this capability to our household web server to display the inverters' status on any computer in our house, or even on remote computers over the web.

UPDATE 08/2002:
The remote control ports of the inverters turned out to be unsuitable for use with a computer as we had initially hoped. This, combined with the fact that once we had the inverters set up we really didn't have to mess with them means that this project has been shelved.

HOWEVER, we are currently considering building our own microcontroller-based data logging system to keep tabs on the whole system, instead of just the inverters. More as this plan(?) develops.

The original electrical service

Before the solar panels and inverters we simply had the 240VAC connected to the meter, and from there to the circuit breaker panel. The 240VAC came into the house as 3 wires, 2 "hot" wires carrying 120VAC, each 180 degrees out of phase with each other, and a neutral reference wire. These wires came up to the house, through the meter, and into the breaker panel. From there the 240 was connected to a couple 240V breakers for the loads that required it, and then each of the two 120V lines were attached to half of the 120V breakers to feed the rest of the house.

The modifications for solar input

Besides the physical installation of the panels described above, what we did was basically to insert the inverter in between the meter and the existing circuit breaker panel. This required minimal (basically no) rewiring of the house and minimal modification to the existing circuit breaker panel. The mains power from the grid goes through a 60A circuit-breaker into the original panel, then from that panel to the garage and the "AC-in" on the inverters. The house loads are connected to a second circuit-breaker panel that is then connected to the "AC-out" on the inverters.

The DC power from the solar panels is connected to the batteries and to the DC-in on the inverters. This puts the inverters at the center of the system where it can automatically manage the various power sources and sinks.

Still having problems? Check out this diagram.


If the panels are producing more power than we can use then the inverter forces the extra back through the meter and onto the grid. If the panels aren't producing enough, we suck power from the grid through the meter and the inverter and into the house loads to make up the difference. If the panels are producing no output and the batteries are charged (as will typically be the case at night, for instance) then the inverters will basically be a no-op. If the grid goes out - not that that ever happens here in California, nooo ;) - then we pull power from the panels and/or the batteries.

Under normal circumstances we will have access to grid power so we will not use the batteries, and need not use the solar panels, though of course we will to offset our electric bill. Basically we are using the grid itself as a giant bank of batteries. The "local bank" of batteries is just there to stabilize the solar output voltage for the inverters, and to provide a sort of whole-house UPS (with 36ms transfer time) in case the grid goes out. The inverter automatically manages this potentially (sorry) complex mix of power sources and sinks so we can behave just as we did before installing the solar system. We like the inverters.

update 04/2002: In informal (and sometimes unintended <G>) tests we have found that we easily have the capability to remain powered through moderate outages, such as the infamous 2-hour "rolling blackouts" of 2001, with no disruption to our normal routine. We also have the ability to keep the refrigerator and freezer operating through the night if we need to by shutting down (or not running) non-essential electrical appliances such as TVs and washing machines. During the daylight hours we typically do not have a problem getting enough power from the panels.

One thing that we have done to facilitate managing the power consumption in this situation is to go through the house and locate all the "phantom loads" - the "wall wart"-type power supplies and appliances such as VCRs and TVs that have electronics that continue to draw power even when "off". (Our entertainment center alone draws 1.3A @ 120v even when everything is supposedly off!) These we have put on power strips so that we can shut them off completely if necessary.

Needless to say, in the future we will pay more attention to such things when purchasing electronic equipment to try to avoid adding any more of these sorts of phantom loads to our system in the first place. Manufacturers take note!

Net Metering

We can do this project because our power company has a "net metering" plan for people who generate at least some of their own power. What this means is that we pay for our electricity on a yearly "net" basis. When we are producing a net surplus of power, we push the extra onto the grid, running the meter backwards. When we are producing less than we use, we pull the difference from the grid, running the meter forwards again. The power company sends us a yearly bill for the net power we have used for the year (or bill us for $0, and "pocket" the extra if we produce more).


By using the grid as our storage system in this manner, we can keep the battery bank small, cheap, and easy to maintain. This should lower our operational costs as well, since one of the major costs of solar power is the maintenance (and replacement) of the power storage system that is required to power things when there is no sun. Even so, running only on batteries we have enough power to easily coast through a 2-hour "rotating blackout", even with no help from the solar panels. If the panels are getting sun and/or we cut back on our power usage then we will be able to last correspondingly longer. If there is a major outage of a day or more then we will likely run the batteries out of power, but such a major outage is unlikely. Hopefully we will not have run the batteries through too many charge/discharge cycles so they will last longer than a set of batteries would normally as well.

(In the past there has only been 1 outage lasting more than 1 day in the 20+ years that we have lived here and it was caused by a substation power transformer literally blowing up. Even so, we do have a gasoline generator that we can use to keep food in the freezer from thawing if things degrade to that degree, and of course we will also have several hours of solar input every day, even on cloudy days and in the winter.. at least enough to handle critical loads.)

Actually, one of our biggest concerns at this point is overcharging the batteries. Under normal conditions, i.e. with the public grid powered and inverters connected to it, any excess power is dumped to the grid. However, when there is a grid failure the inverters are required to automatically disconnect themselves from the grid for safety reasons. This means that there is no place to dump any excess, and if the panels remained connected to the batteries in this situation the batteries could easily be overcharged.

Remember, the battery output and solar panel output are both attached in parallel to the inverters' DC side. The only way that the inverter(s) can control the voltage in this situation is by drawing off any excess and dumping it somewhere. The only places where it can be dumped is into the house loads and onto the grid. We could deal with this "manually" by running around and turning on (yes, on) appliances during a grid power failure, or by manually disconnecting one or more arrays at the input to the electrical panel that feeds the batteries and inverters, but this would be a PITA not to mention comical.

Fortunately the inverters also have several general-purpose relays that we can use. They are triggered by battery voltage so we can potentially rig them so that if the batteries are full, either power is shunted to a water heater or something, or else the panel output is simply disconnected. When the battery voltage falls sufficiently the relay(s) can reconfigure things to charge the batteries from the panels again.

Did I mention we like the inverters?

Currently we simply have the relays rigged to sound an alarm when the battery voltage rises above a certain threshold. At that point we can manually disconnect one or more arrays to bring the solar input in line with house loads. As a final fail-safe measure we also have a solenoid connected in such a way as to (mechanically) trip the GFI breaker if the battery voltage continues to rise. Reconnecting the solar inputs in this case requires manually resetting the switch.

UPDATE 03/2002:
We now have 4 of the general-purpose relays connected to bi-directional solenoids rigged to flip the panel disconnect breakers (switches). The inverters have been set to drive the relays to disconnect each section of panels at different voltage levels on the DC (battery+panels) side, and reconnect them once the voltage drops back down.

The inverters actually have two settings for each relay, the voltage to trip on at and a "hysteresis" value that can be set independently to trip off. This is good because it means that we can set the relay to disconnect a panel section above a certain voltage, say 26V, but not go on again until the voltage has dropped to a couple volts below that. This keeps the relays from "chattering" if the voltage fluctuates in a range around the threshold, which it will do as soon as the section gets disconnected for example, or as clouds move in front of the sun. This in turn keeps the wear on the mechanical parts to a minimum.

As I mentioned, we have set each of the 4 relays to trip at slightly different voltages. This means that as the house loads and/or solar power from the panels varies some of the panel sections will remain connected. Ideally we would like the power output of the solar panels to exactly offset the house loads so that the batteries are neither charging nor discharging. Practically speaking the relay arrangement is somewhat less than ideal, but as a fallback capability for exceptional situations it does the job. It's not nearly as efficient at matching loads to power production as would be the case if the grid were connected, but it's a lot better than the (still in place) all-or-nothing GFI breaker solution we had originally.

Also, whereas the GFI needed to be manually reset each time, this new configuration is completely automatic. If the grid loses power, the inverters automatically disconnect from the grid, the battery voltage starts to climb, and the relays trip the first panel section off. If the battery voltage continues to climb the relays will disconnect another section, and another, and finally the last section, at which point no power will be getting to the batteries.

If the house loads draw down the voltage enough or the sun goes behind a cloud, the relays reconnect each section one by one to bring the voltage back up. Because the voltage thresholds are different for the different relays, it's entirely possible to have, say, two panel sections connected and the other two disconnected, and for sections to be connected and disconnected as necessary to maintain battery voltage as the grid outage continues, and the solar input and house loads fluctuate.

When Edison fixes the problem and the grid powers back up, the inverters automatically reconnect to the grid and the inverter pulls the DC voltage back down by shunting power to the grid. If the DC voltage remains nominal, the relays pulse all the solenoids to flip the breakers on, and the breakers reconnect all the panels, enabling the maximum solar output to be shunted to (or "stored on") the public power grid.

We do like the inverters. :)

UPDATE 07/2007:
We have installed a transfer switch to swap the house between the inverters and a direct-to-grid configuration. We did this mostly so that we could work on the PV system without having to shut down the house (except for a momentary outage as we flip the switch). The switch affects only what feeds the house - grid or inverters - and does not affect the grid-interactivity of the inverters. If we flip the house over to the direct grid connection the PV system still remains connected to the grid as well, and can push power back to the grid.

This is something we probably should have included in the original design but didn't. It hasn't caused us any real issues, but it might have been a problem if we had had trouble with the inverters. If that had happened and we'd had to shut down the system then the house would have gone dark because the house was hardwired to the inverter output. This even if the grid was working fine! Not a situation you want to be in. This way if the grid has trouble we put the house on the solar system, and if the solar system has problems we flip the switch over to the direct-to-grid connection. Of course the nominal configuration is to leave the switch in the position that connects the house to the solar system.

UPDATE 07/2009:
We just added some dedicated charge controllers to the system to protect the batteries. This was something that we considered for the original design but decided against because of cost. Since then there have been newer products released that are both cheaper and can handle higher power installations. The previous conficuration (above) had the inverters drive relays to disconnect the panels from the batteries if the voltage coming off the panels drove the DC voltage too high. This worked well, (and is still in place, actually) but being mechanical we were concerned about the system eventually failing and/or the relays being a maintenance issue.

The charge controllers are Schneider Electric (formerly Xantrex) model C60 controllers, and we put one on each of the 4 array sections. They were about $120 each, and we had to add another electrical cabinet to hold them, so that plus wiring, conduit, and electician's time added about $800 to the system cost (maybe 1 year to the payback time), but being solid-state they should be much more reliable.

(As an aside, I noticed while shopping that prices for PV parts have really come down in the last few years. I haven't done any rigorous design pricing, but at a guess I'd say I might be able to put together a system with the same specs and output as ours for at least 30% less, maybe more. Factor in 10 years of inflation and it might be closer to half the cost in real terms. At least something seems to be going the right direction!)

The Bottom Line

So what was the price of this whole setup? By far the biggest expenses were the panels. The panels individually cost $425, and we have 40 of them for a total of nearly $17,000. The inverters together cost nearly $5,000. Hired labor (electricians, etc.) was another $5-6k. The remainder of the $31870.63 price includes the DC disconnect boxes, the GFI, batteries, conduit, combiner boxes, wire, fuses, various mechanical parts (mounting hardware, etc.), permit/inspection fees, and warranties.

Fortunately California has a program that rebates up to half of the price of a solar installation such as ours. Thus, of the nearly $32k we initially paid, we applied for and received just over $15k back from the state. This brought our total cost for the system down to just under $17k.

If you would like to see a more detailed breakdown of costs (and suppliers) you can check out this PDF.

UPDATE 2012:
To the above you have to add the price of one set of replacement batteries, the transfer switch, and the charge controllers. The installed price for these items were about $850, $150, and $800, respectively, bringing the total system price to $33670.63.

That's just the initial cost of the system though.. what about the recurring and TCO costs? Those are tricky questions because there are a number of unknowns. Let's start with the knowns.

First of all is our electric bill. According to the last electric bill before we switched to net metering, the "marginal $/KWhr" (the cost of an additional KWhr) of electricity was as high as $0.23-0.26/KWhr, depending on the time of year (winter rates are lower than summer rates), and which "tier" we fell into. (Here is our actual last month's electric bill from SoCal Edison before we switched.)

Second is the output of our solar system. For monitoring purposes we have installed two additional electricity meters, similar to the one that the power company reads to determine our bill. One monitors how much the house loads take, and the other is for the garage loads. Since the "official" power company meter shows the net power actually coming into the system, the production of the solar panels can be obtained by adding the garage and house meters, then subtracting the power company's meter. (Technically this also includes inefficiencies for things like charging the batteries and inverter DC->AC conversion, but for cost calculations this is typically what you want.)

Being solar, our production varies based on time of year, the weather, etc., but for last year our actual overall production was 4404.9 KWhr. Multiplied by $0.23-0.26/KWhr, this gives an estimate of between $1013 and $1145 that we saved last year.

Unfortunately it isn't that simple. The true amount saved will actually be a bit less, since even though most of our "top tier" usage is in the expensive summer months, some of it occurs during the winter and gets the lower rates. On the other hand, the solar output is less in the winter, so we have to rely more on "imported" electricity from the grid, and it is thus easier to reach the more expensive rate tiers in winter. (Even though the most-expensive tier in winter is less than half what it is in summer.)

Confused? So were we. The desire to get away from that sort of thing was one of the factors in our decision to install the PV system. But I digress.

Anyway, sorting all that out is a bit of a trick, but on average we wind up saving about $900-1000 per year. One might be tempted to divide the $16,660 cost of our PV system by $1000 and come up with an estimated "break-even" time of around 17 years.

The Future

How long it actually takes to pay off our system however depends on a number of unknown factors. The "known unknowns" are things like the cost of maintenance of the system, the future price of electricity, and our future energy usage.

Since there are very few moving parts to our system, maintenance costs are basically limited to periodic replacement of the batteries. The battery bank cost nearly $800, but since we don't often cycle them they should last much longer than would be typical for an off-grid PV system. We're estimating that we will have to replace them once every 7-10 years. If new power storage technology comes out, maybe that can be extended. If grid power becomes more unreliable and we have to cycle them more we might have to replace them more often.

(NOTE: we did replace the batteries once, in 2007(?). It cost about $850.)

The solar panels will degrade over time, but they are guaranteed by the manufacturer to produce at least 80% of their initial output for at least 25 years. The inverters are solid-state and should last for many years as well. They are only warranted for 5 years however, so it is possible we would have to pay to replace them roughly that often. Trace had a very good reputation however, and Xantrex seems to be keeping up the tradition. We would be somewhat surprised if the inverter model we're using didn't last at -least- 10 years or more.

As I type this, the system has been operating for nearly 4 years. In that time we have not had a significant failure of any part of the system. We had a couple "operator errors" in the early days that caused the inverters to shut down temporarily, as they were designed to do, but on the whole the system has performed well. The battery capacity is degrading as expected. The solar panel output seems as good as it was when the panels were installed, though it would be difficult to detect minor degradation due to the highly variable nature of solar power.

The price of electricity is in all probability going to go up in the future. Even assuming no natural disaster taking out major power plants, it is hard to imagine either demand going down or the cost of any form of energy production going down over the long term. This is especially true of anything based on fossil fuels. True, a breakthrough in something like fusion power could occur, but I would estimate the chances of that happening within the next 20 years or so are slim. The joke in the nuclear community is that fusion is a technology that is coming in 50 years, and has been for 50 years.

On the other hand, as I type this Southern California Edison, our power company, is pushing for a 14% rate hike in electricity, Southern California Gas Co. wants a 33% hike in natural gas prices, crude oil is over $50/barrel, and gasoline is near $3/gal and rising fast. Over the last 30 years, rates for residential electricity customers in California have increased at an average of 6% per year. I see no reason to expect this to change for the better, and it appears quite probable that that rate of increase will itself increase.

Our future electricity usage is uncertain. While we have no intention of radically changing our usage, to a certain extent we are at the mercy of what the manufacturers of commercial products produce. Recently wide-screen plasma TVs have become popular, some of which use 500W or more. Computer manufacturers are bringing out systems with 500W power supplies, 200W graphics cards, and multiple 100W+ CPUs.

While we would probably not buy those sorts of products, if say, Toyota were to produce a GO-HEV version of something like the Camry then we would certainly be interested in that, even if it increased our electricity usage by a significant amount (because in that case any higher electric bills would be offset by lower gasoline bills). In the end, we have enough control over our energy usage that this should not be a problem.

Resource links...

Get in touch!

If you have any questions or would like to send me feedback about this project please feel free to email me at (or spamdump4242-solar at with a subject of "solar project" to get through my anti-spam filters.

If you do not have javascript enabled, I don't blame you, but you won't be able to see the above email address as a clickable link (I'm sorry, but I just get too much spam sent to this address otherwise). I'm also occasionally on IRC though, so you might try looking there if you want to get in touch with me. Look for the nick "spaceguy" or the "#solarpowered" channel on undernet.

Another option is via my facebook page.

This page intentionally graphics-deficient.
Last updated December 15 2012 03:03:04