jump to navigation

Space power? June 12, 2009

Posted by Maury Markowitz in solar power satellites.

A few weeks back the science interwebs were alight with the news that Pacific Gas & Electric, California’s largest power supplier, had inked a deal to buy solar power beamed down from space. The fact that the post was on the NEXT100 blog, not an official PG&E press release, didn’t stop anyone from blogrolling it to the other end of the Earth in moments. Nor did the fact that the company, Solaren Corp, didn’t even have a web page seem to trigger any concern. If you googled it up, you could only find this company from Armenia, and even today you have to poke about a bit to find this page, which is entirely content-free and may not even be the same company (it’s listed in Washington).

Space power, eh? Ain’t gonna happen. Let me show you why…

Building a solar panel on Earth is pretty easy, but you only get sunlight during the day, and there’s weather to consider too. Poking about on the net, I found that we get a little under 2000 hours of “bright direct sunlight” a year here in Toronto. Since there’s 8544 hours in a year, that means we get useful sunlight about 1/4 of the time.

And even then we don’t get all of the power we could; in Earth orbit, with no atmosphere to filter the sunlight, there’s about 1,366 Watts of solar energy for every square meter, but by the time it gets down here, we get about 1,000 W/m2.

The high frontier

Ok, so put those solar panels in space. Five times as much power, right there – four times the sunlight because there’s no night or clouds, and another boost because of the greater power of that light due to the lack of the atmosphere.

Of course you have to get the power down to the planet, but we know how to do that. There’s a system called a “rectenna” that beams the power down as microwaves, and it’s pretty efficient, upward of 90%. This will be attenuated somewhat by the atmosphere, so maybe the overall effect is on the order of 85%. Now put the satellite in a geostationary orbit (GEO) so it sits above your target market, like California, and beam the power down to the grid.

So we’re looking at getting maybe four times as much power as the same panel on Earth, and it’s almost 24 hours a day (there’s the odd eclipse). That’s a lot more useful than day-only power like a panel here on Earth. So what’s not to like?

Show me da money!

Uhhh, how about the $12,000 a pound shipping costs? Here, follow along as I do a little math…

A really good solar cell for space applications is the ZTJ cell from emcore. It’s about 30% efficient, which is way higher than a traditional cell like you see on a rooftop, which might get you about 12% if you’re lucky. At 30%, you’re looking at getting around around 400 W/m2 of the 1,366 available. The Next100 press release stated they were looking at 200 MW, so that’s 500,000 m2 of cells. How much do they weight? emcore has them at 84 mg/cm2. That’s 5 billion cm2 we have to cover, so that’s 420 billion mg, or about 925,000 pounds.

At $12,000 a pound that’s $11 billion just for the launch cost of the cells.

Ok, you know, $11 billion isn’t all that much in the grand scheme of things. OPG is planning on spending over $20 billion CAD on the Darlington B expansion, for instance. But the difference is the power it’s going to produce. Solar cells are generally rated for about 20 years, so we’re looking at 200 MW being generated 24/7 for 20 years. That’s about 35 billion kWh in total. How much is a kWh? Well for baseload, here in Ontario you’re looking at about 4 cents and in California it peaks out at about 6 cents. Let’s split the difference and call it 5 cents.

So over 20 years we get $1.75 billion worth of electricity. Anyone else see the problem?

Now there are ways you can improve on this, slightly. Solaren’s patent shows a system that uses concentration, which boosts you into the 40% efficiency range and lets you replace the solar cells with mirrors. Those mirrors can be based on lightweight inflatable structures, so the launch mass goes down.  But solar cells aren’t really that heavy in the first place, so the savings aren’t huge – they’re nowhere near the 100 times reduction you’re going to need to make the numbers work out. And then there’s the small problem that every attempt to make a lightweight space mirror has failed miserably.

Basically if you just set up those exact same panels on Earth you get some amount of power, let’s call it 1 solaren. Now if you set them up in space you get 5 solarens. The difference is 4 solarens. Unless the shipping costs are less than the price you get paid for those 4 solarens, just set it up here on Earth and use the money you save to buy four more of them. And the simple fact of the matter is that electricity is cheap, rockets are not.


I need to point out that there’s nowhere near this many high-efficiency solar cells in the world, and no way to make them fast enough either. emcore’s the only company that’s actually signed deals for concentrated PV, and their production is only a few hundred kW a year. A 850 kW plant was going to take 15 months to supply. 200 MW represents decades of worldwide production, and there’s no sign that either emcore or Spectolab has any sort of deal with Solaren.

And forget about actually launching this stuff up there. The entire worldwide capacity to GTO is about 70 launches a year, with an average of about 5 tons of payload. Most of these are used by the military, and they’re not going to sell them to you. To get this stuff up there, you’ll need to buy the spare capacity of the entire world for several years.

And they say they’ll be operational in 2016? Hmmm.

And did I say 20 years of power? Good luck with that. There’s this looming problem called the Kessler Syndrome that we’re right on the leading edge of running into. Basically all the junk that we’ve been putting up there for the last forty years is still floating around, bumping into things. We’ve already lost two satellites in GEO, and they were way smaller .

Sweet mother Earth

I don’t know what to think of the Solaren story. The company really exists, at least there’s a phone number and people who answer it. They really did file a patent, but anyone can do that. But the plan is so grandiose, and so near-term, that there’s just no way I can see this being remotely possible. If they said they wanted to launch a 2 kW testbed in 2016 that’s one thing, but a 200 MW production system?

Hear that beeping? That’s my BS detector going off.


1. Nathan - July 10, 2009

I’ve found my way to your blog from your postings on Ars:




I’ll probably post this on both this blog and the newer Ars discussion.

I think your judgement about the PG&E plan seems right. But your posts on Ars also seem to be attacking the PowerSat plan. So it’s in that context that I speak:

“Where are the numbers?” Agreed. The 2nd Ars post didn’t give us any numbers, that’s for sure.

There are three primary weaknesses to your discussions with regards to PowerSat and SSP in general:
1) You don’t consider the trade-offs between panel effeciency, panel weight, and panel production cost and capacity. You pick a point based on effeciency alone. It may not be the best point, and PowerSat clearly disagrees, going with thin film CIGS cells.

2) You dismiss ion thrusters far too readily and without substantiating your arguments. In the first Ars article discussion you say: ‘Quick, name a production ion thruster with more than, say, 10 N of thrust. That’s what they’d need to place 10 tones in GEO in the “months” they quote. Hint: there isn’t one.’ You don’t support that 10 N number.

But if we stipulate it’s correct there are still other problems. The thrust per thruster (1/er?) is mostly irrelevant. The 3 stats that really matter assuming you have a reliable thruster (and we do, look at the DS1 spare thruster test) are specific impulse (Isp), power required and thrust per unit mass. Isp is definitely high, as is power required, but that will clearly come from the solar panels (and put an upper limit on thrust) and thrust per unit mass.

The latter two are the limiting factors on the thrust available to the vehicle. For a fixed specific impulse (which is just exhaust velocity divided by earthside acceleration due to gravity’s magnitude), power required should scale linearly with thrust (either by adding more thrusters or by increasing mass throughput of a given thruster). If we look at the HiPeP, assuming an Isp of 9600 seconds, a power required of 39 kW and a thrust of 670 mN, we get 17mN/kW (also 0.058 kW/mN). For 10 N, that means we need 582 kW. PowerSat says they can get 17MW, so power available is not a limiting factor.

That leaves the unit mass of such a thruster. I’ve had trouble find thrust to mass ratios for ion drives –just the drive units. The best figure I have is from “An Overview of the HiPEP Project” by Elliott, Foster and Patterson, all NASA-Glenn engineers, that sets the thruster mass target at less than 3 kg/kW. I’m not sure what they achieved, but that does push the mass for a 10 N system to 1750 kg. That’s quite high, but it’s well within the total mass allowance for the PowerSat module. So, it’s definitely conceivable that the PowerSat design could have 10 N worth of HiPEP type thrusters.

Fuel will also take up another significant fraction of the mass. Wikipedia, “Delta-V budget”, says that the delta V between Equatorial Low Earth Orbit and Geosynchronous orbit is 3.9 km/s. Neglecting gravity losses and applying the rocket equation, we get a propellant mass of 0.45 tons. Turner* says that non impulsive (constantly thrusting) systems suffer gravity losses that reduce the effective specific impulse by a factor as high as 2.3. That gives us a fuel mass of 1.07 tons. Assuming everything is in US tons (907 kg/1 US ton), that’s a propellant and engine mass of 2720 kg out of a module mass of 9070 kg. That still fits.

Another problem with your analysis is that you keep using $12000 per pound as a launch cost. (Let’s go with $26400/kg to avoid mixing unit systems). Even using the Futron paper you quote “Space Transportation Cost Trends in Price Per Pound 1990-2000,” that number only applies to GEO orbits, but you keep using it even when it comes to the PowerSat program, which is supposed to use ion drives to go from LEO to GEO. I guess this is because you’d written off ion drives, but they do seem viable. So, looking at the Futron paper, we see average costs of about $11000/kg for launchers in the 10 ton range already using Western launchers and half that for non Western launchers. Using that data suggests that PowerSat should get a similar or lower price because it plans to build the sats first and then get a launch contract to leverage bulk pricing and because it can launch to the cheapest LEO orbits because it’s using ion drives.

3) You’ve mentioned a couple of collisions in GEO, but haven’t substantiated them. I haven’t found them, either. The only ones I’ve found are in LEO.

You’ve also mentioned the Kessler Syndrome, the increase in space debris in LEO until it becomes unusable. It is a problem, but I consider that an argument for getting Space Solar Power up as soon as possible. Orbitting solar arrays driving lasers and masers can be used to vaporize and/or push space trash out of LEO and out of earth orbit as discussed in the Wikipedia article “Laser broom”.

Furthermore, Kessler Syndrome applies to LEO, not GEO. MEO to GEO space is much bigger and much more empty than LEO space. While the PowerSat modules will pass through LEO, they will end up in GEO, so they won’t contribute as much to Kessler syndrome and they won’t be vulnerable to the current LEO debris environment for very long.

I appreciate you making your posts and stimulating my interests in the discussion. It’s been very informative.


*Turner’s “Rocket and Spacecraft Propulsion” seems to be a good a reference all around for this sort of thing. A lot of it is available on Google books.

Maury Markowitz - July 11, 2009

I hope you don’t mind if I only reply to the second two points here, as the former will be the topic of a separate post. I’ll start with the last issue first.

I stated that there were two collisions in GEO, a statement you question. You’re really not aware of this? It’s pretty widely known and commented about. There’s been plenty of articles on the topic in a variety of sources, and it a recurring theme in space debris studies. And you might want to check the history link on any Wikipedia articles you quote, because I probably wrote it.

While it’s true that the Kessler Syndrome refers specifically to LEO, I’m generous with the term. And laser brooms are just as much unobtainium as the rest of the technologies being mentioned as the Great White Hope of SPS’s. NASA planned on testing one, and that went exactly nowhere. Using one for targets in GEO is slightly harder, for the simple reason that we don’t know where most of them are. Whatever hit the Russian comsat is still unidentified.

You also asked about the thrust needed to boost the system from LEO to GEO. You followed with an analysis, but one that I don’t entirely follow your argument. Here’s mine; 10 tonnes to GEO at 3.9 km/s is 15 N continuous for a one-month burn, ignoring all losses. I was being generous when I said 10. The DS1 thruster you mentioned provides less than 100 mN of thust, can operate at those peak levels for very short periods of time and could no way be considered to be anything like a 10 N thruster. The highest power existing thrusters are around 2N and only operate in single-pulses. There are designs for MPDT thrusters in the 20 to 60 N range, but these have been demonstrated only in single pulses, and these designs suffer erosion problems that make traditional ion thrusters look positively long-lived.

But this leads into another complaint about my quoting the (use your dimensions) $25k per kg cost for GEO launches. Why would I use such a number? Because that’s the real number. If ion thrusters could make GEO launches so much less expensive, as you claim, why are there exactly zero examples of such a system? If it saves so much money in a SPS launch, it’s just as useful for a comsat, after all. So why is this the case? Because high-thrust ion thrusters don’t exist.

So sorry, I’m not just pulling this out of my butt. My comments stand. In order for any of this to become even remotely practical, several bits of technology have to be developed, perfected, and each provide a reduction of costs or weights around an order of magnitude. I’m not holding my breath.

Thomas Chiasson - January 2, 2012

I can’t find any records of a verified GEO collision ever taking place. All the links you’ve provided refer to the possibility of a GEO collision, or to a Russian satellite being hit by ‘an unknown object’.

And it is strongly suspected that the “unknown object” was debris. There are dozens of such references:


I’m not saying that collisions in GEO aren’t a possibility, I’m just saying that given the tight management of GEO, collision is much less of an issue than it is at LEO.

Let’s put numbers to this, shall we? The chance of collision in GEO is 1/10th that of LEO:

Click to access SWF%203FEB12%20Hazard%20from%20Orbital%20Debris.pdf

Further, the probabilities are based largely on geometry. A single SPS like the recent designs I’ve seen are larger than every object ever launched, put together. Look at the daily movements in the document above, and consider a gigantic object spanning that space.

When people talk about problems with space debris and collisions, they’re usually talking about LEO

Most modern studies on this issue consider both cases, as in the example above. I recommend you check out the Wikipedia’s article for more links and data.

the possibility of GEO becoming unusable is still pretty remote.

Space debris is the #1 threat to manned spaceflight today (at least to Shuttle missions, back in the day). 1/10th of that is hardly “safe”. And when you consider that (a) SPS’s are larger than anything ever built in space before, and (b) that the SPS has to get from LEO to GEO, this should not be dismissed lightly.

Here’ let me once again quote Dr. Kessler:

Some of the most environmentally dangerous activities in space include large constellations such as those initially proposed by the Strategic Defense Initiative in the mid-1980s, large structures such as those considered in the late-1970s for building solar power stations in Earth orbit, and anti-satellite warfare using systems tested by the USSR, the U.S., and China over the past 30 years.

In the last 10 years we’ve done only one of these, and it was considered a disaster. SPS use calls for doing the other two. This is not a good idea, and you should not dismiss it with a no-numbers-presented wave of the hand, SHOW ME THE NUMBERS!

Thomas Chiasson - January 2, 2012

I can’t find any records of verified collisions at GEO, including in the links you’ve provided. The closest I see is that Russian satellite being hit by an unknown object, but there’s no telling if it’s man made or not. GEO is certainly getting crowded, but the danger of it becoming unusable seems pretty remote, especially in comparison with LEO.

Maury Markowitz - July 11, 2009

> GEO space is much bigger and much more empty than LEO space

This needs comment. GEO space is actually much smaller than LEO. Unlike LEO, GEO objects are confined to a single inclination and altitude. If all LEO orbits were confined to a single ring, say the 110 nm 28.5 degree launch from Kennedy, that would indeed result in less space than GEO. But they’re not, and the density of objects in GEO is considerably higher than most LEO orbits.


Nathan - July 11, 2009

Good point as far as GEO being constrained to a ring. That said, it’s a ring of huge circumference. So I can still see there being enough room. The 180 slots per comm band does add an additional constraint, I notice. Also, I’ll take a look at those collision risk assessment papers when I get the chance, though I don’t know when that will be. Thanks.

Also, there’s no reason the power sat can’t also be comm sats as well. If they’re going to transmit in a given band/slot that’s already taken, PowerSat may have to buy that slot and/or piggyback a comsat with the SPS.

Some variation on a laser broom will work — they’re not unobtainium. The necessary components exist already and baring a change to the laws of physics, you’re going to impart momentum to whatever particles you lase. They may be slow, but they’ll still work. The key component to it is the power source and the SPS provides that. Admittedly, they’re tangential to the SPS debate, but as some sort of space garbage collection is going to become necessary to protect both current satellite assets and future access to space, they are highly valuable, perhaps moreso than SPSs, and have the same technology base as SPSs. So the risk of the investment in either is hedged by the dual purpose tech.

We definitely need more info about the debris environment, but once again, the best enabling tech I can imagine is the SPS. Use them to drive lidar/radar and actually find out what’s in the area. Power transmission to earth aside, giant solar arrays in space have a lot of other good uses.

I also think that a well designed SPS would be fairly robust against collisions. You might lose some functionality or generation capability, and that’s another cost that needs accounting for.

I didn’t realize that the NSTAR test had been run at such a low power. Thanks. Still, NSTAR is 10 years out of state of the art and thus a bad example on my part. HiPEP seems like a better reference point. It seems

Regardless, you don’t need a single thruster of a given thrust capability.You don’t need a high thrust thruster. That doesn’t matter because you can always just add more thrusters. What matters is the specific thrust, which my analysis showed produces a thruster mass no greater than between 300 kg and 1800 kg depending on how much travel time you want 8 months down to 1 months for a HiPEP type unit. The thrust per thruster is irrelevant. As it turns out, ion thrusters are light enough that we can just keep adding extras if we need more. Thruster density might be an issue, but they don’t look that large to me. You’d only

My analysis of thrust required for a given transfer time is just a back of the envelope a=F/m and a*t=delta-v, keeping mass fixed so we have a worse case scenario. We get F = m*delta-v/t. This is also what we’d get if Turner’s Isp correction for low, constant thrust drives applies to the time of transfer approximation shown on Wikipedia, which I haven’t had a chance to check.

Our numbers aren’t far apart, and I think the difference may be you using metric tonnes (1000 kg) and me using US tons (907 kg). I gathered from the article the 10 tons meant US tons, but I could be wrong. I wish everyone would just use SI units.

Anyway, PowerSat was talking more like 6-8 months transfer times, which looks a lot more realistic from both of our analyses. In fact, it fits the HiPEP data so well that it actually gives me a good deal of hope.

My overall point here is that, while I haven’t done the analysis on the structures or solar panel tech, the ion thruster components of the system could be done with cutting edge tech. Not necessarily very mature or well tested, but stuff that exists. And furthermore, the 6-8month transfer time fits so well with a 1-2 N system (probably several smaller thrusters, a single thruster of that thrust does seem harder to do). I’ll see if I can’t find out more about the maturity of the HiPEP system. I definitely need to shore that case up.

I’d like to see your analysis done with CIGS cells instead of silicon. I’ll try to do it myself if I get the chance. I’ll keep you posted, if you’d like.

2. Nathan - July 10, 2009

Also, should you reply, could you reply to this comment or to my email so I get the comments in my email. I can do rocket science but I have trouble using webpages, apparently.

3. Here we go again with the SPSs. « Energy Matters - September 1, 2009

More on this topic, which just won’t die.

4. Interesting Reading #359 – The Blogs at HowStuffWorks - November 10, 2009

[…] Japan eyes solar station in space as new energy source – “It may sound like a sci-fi vision, but Japan’s space agency is dead serious: by 2030 it wants to collect solar power in space and zap it down to Earth, using laser beams or microwaves…” Counterpoint: Space power? […]

5. Al Globus - April 1, 2011

The core of this argument is that if you put the same solar panel on Earth and in space, the Earth-bound one wins. The obvious solution is to use different solar panels in space, ones that are optimized for the space environment. For example, take a look at the design in http://space.alglobus.net/papers/SSI2010SSPpaper.pdf The solar power generating hardware is thin-film 0.025 mm thick on an 0.0075 mm backing (this has been proven in orbit by the Ikaros satellite). Such flimsy materials work in space because there is no wind. On Earth, the slightest breeze would destroy it.

Maury Markowitz - April 1, 2011

> The core of this argument is that if you put the same solar panel
> on Earth and in space, the Earth-bound one wins.

Sort of. More accurately, launch costs will be greater than the economic value of the electricity produced. So it’s really a power-to-weight calculation, not a comparison of the panels.

You mention a number of 0.8 kg/kW (kWp I assume) for PowerFilm at around 4% arial efficiency (sounds about right). So then, let’s compare that with the ZTJ cell example I used, which is 84 mg/c^2, which is 0.84 kg per m^2. At 30% efficiency, that’s around 550 Wp/m^2, so about 1.5 kg/kWp.

So the technology you propose reduces the cost/weight by a factor of 2. As I noted in my article, you need to reduce it by at least 100, and likely as much as 1000 to make the numbers work.

So, where’s the other 500 times coming from?

6. Keith Henson - April 2, 2011

Another way is to work the problem backwards and see what you need for launch cost.

I favor thermal cycles, the GE engines on Boeing 777 aircraft turn out 10 kW per kg. Counting collector, radiators and transmitters, I think 5 kg/kW (on the ground) is a reasonable number (that’s half as much mass on the power satellite because of the 50% transmission loss).

For a ten year payback and 2 cents per kWh, you can spend $1600 per kW. If a third is transport to GEO, the cost needs to be $100 per kg.

Because of the rotten mass ratio, I don’t think this can be achieved at all with chemical fuels.

But laser heated hydrogen will give up to 9.8 km/sec and a Skylon derived vehicle will give 10.5 km/sec equivalent exhaust velocity till it runs out of air at around 26 km and 2 km/sec.

To get an average of 9 km/sec to LEO takes 6 GW of lasers heating hydrogen. The lasers (at $10/watt) will cost $60 B, the rest of the system, vehicles, bounce mirrors in GEO, ground facilities perhaps $40 B. The cost to run it at a million tons per year to GEO will be $100 B.

But the cash flow from selling power satellites at $1.6 B per GW would be $320 B per year.

If 1 GW nuclear power plants are expected to cost $10 B, then for ten times that much money you can set up a factory to crank out 200 GW per year.

Keith Henson
(L-5 Society Founder)

Maury Markowitz - April 3, 2011

> The lasers (at $10/watt) will cost $60 B, the rest of the system, vehicles, bounce mirrors in GEO,
> ground facilities perhaps $40 B. The cost to run it at a million tons per year to GEO will be $100 B.

… or we can save that $100 billion and install the systems in Nevada using 100% known technology and zero development cost.

7. Johhny Pedant - August 2, 2013

There are 365.25 * 24 = 8766 hours in a year. You seem to have used 356 days in a year.

Maury Markowitz - August 2, 2013

Indeed, it was a typo in my calculator. I have left it in there so no one accuses me of fudging with the article.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s