Oxford PV sets new solar panel efficiency world record
oxfordpv.com
oxfordpv.com
If you want to buy solar panels, you go for the Levelized Cost of Electricity (LCOE). It looks at the cost of electricity of the entire system over its lifetime. It takes interest on loans, cost of panels, inverters, installation cost, labour cost, insurance and the geographic location into account.
It turns out that in most cases a lower efficiency panel built cheaply is better, for example a bifacial panel upright on the ground instead of on a roof (lower labour cost) wins. Even old inefficient second hand panels can outperform a high efficiency solar panel over its lifetime.
We build $0,04 single chip solar cell MPPT inverters (72-122 per solar panel) that for various reasons yield 30% more electricity while also lowering manufacturing (less glass, smaller panels, cheaper inverters with higher efficiency, thinner wires) and installation cost.
We expect solar electricity cost to keep dropping far below 1 cent per kWh. Eventually we'll grow solar panels (for example from captured CO2) over the next decades with nanotech manufactoring methodes (atom by atom) until they will be virtually free, like plants. In this light, the Oxford efficiency metric for future solar panels might be bypassed by other cheaper systems that are less efficient.
There’s never a single metric that you can use to decide “better” - it depends on the needs of the project.
I agree that there is no single metric to decide "better". But in most cases properly calculated LCOE, the cost of a kWh over the lifetime of the solar panel system is the single metric that matters. Because why pay more for electricity?
Only in special projects, like solar panels in space, cost is not the most important metric.
For example utility scale solar can be less than 30% the cost of a residential solar installation, but also adds profit margins, interest rates and distribution costs to a residential customer.
In the later Ringworld books Larry Niven introduced solar 'panels' sprayed onto any surface and wirelessly broadcasting power. The DIY installation costs would now be almost zero.
A similar argument can be made for almost any good, over the lifetime of the good, within an application category, second hand will often be better. However, in general, it is still useful better technology, so that newer models can be more efficient, which eventually trickle down to older models becoming more efficient.
For more than four decades, the price of solar panels declined by 20% with each doubling of global cumulative capacity[2], but you should check that with many more data sources, especially reputable or scientific sources.
Who is the “we” that builds these tiny MPPTs? I’m curious.
We are chip designers and made a 1 dollar cent 64 bit 8 core microprocessor with MPPT and bidirectional buck/boost flyback inverters built-in. This nanoinverter can take the <0.5V of a single solar cell and convert it to (for example) 6V. Because these nanoinverters are networked over the DC conducters to each other, they can aggregate their output voltage to 48V, 110V or much higher so the conductors become much thinner. With all the power transistors, opamps, capacitors and diodes integrated into the chip you save more than $50 on discreet component MPPT and microinverters per panel while also boosting the MPPT per cell thus mitigating almost all shading losses.
The same chip can also charge/discharge (lithium and other) batteries. You can get 20 times more cycling charges out of Li-ion this way but you also save on the charging inverters. A single nano-inverter chip can take the output of one or more solar cells and charge li-ion cell, saving you half the number of nano-inverter chip in a total system.
I still need $50K investment to start the mass production of the new model 180nm chips, that will drop the manufacturing cost to 1 cent. Sadly Ycombinator won't fund me as they dislike single founders and dislike funding hardware startups.If you know anyone who can help us with sales or funding please send me an email.
A competitor already manufactors solar panels with microinverters for strings of cells (6-8 per panel) but their panels cost so much more that it is not worth it. Our nanoinvertors are two orders of magnitude cheaper.
[1] https://vimeo.com/731037615
[2] https://www.researchgate.net/profile/Merik-Voswinkel/publica...
The older white paper is very much related, it describes the large power router chips, the MPPT nano-invertor microcontroller chips for individual solar cells are just a recent adddition to our system of line chips.
The video presentation is about the wafer scale integration, a million core giant version of the same microprocessor as the tiny MPPT chips, so they are very much related.
[1] Morphle MNI003 datasheet:
input voltage: 0.01V-6.01V (solar cell or battery input)
output voltage: 0.6V-12V (aggregated or single output)
can power up from solar cell or battery, 0,02W
External inductor (30-100 μH) needed.
Builtin 8 opamps, 4 power mosfets, 3 capacitors, 3 diodes
Forms a programmable reconfigurable MPPT inverter or a Li-ion (all 6 chemistries) battery charger and discharger.
180 Mhz 64 Bit Microprocessor made from 8 bit ALU slices. Can be configured at runtime as 8 x 8b it, 4 x 16 bit, 2 x 32 bit and 1 x 64 bit processors. Microcode processor executes X86-64, ArmV8, Risc-V instructions and Python and Squeak bytecodes natively.
4K OTP memory
16K SRAM
SPI interface to external flash
one-wire network over power input and output wire two-wire 10baseT ethernet to wire all inverters into networks up to 10240 microcontrollers
networked MNI003 chips can aggregate output voltages up to 240V DC
One of the common issues with solar micro-MPPT is EMI emissions. Have you tested for compliance with emitted radio-frequency noise standards?
- We could package the chip, for example wrap it entirely with the metal inductor. That works as a faraday cage and block the emissions. But will this raise the price?
- You put the chip between the solar cell and the (flat) wires. That works as a faraday cage and block the emissions. Will the solar panel maker do this?
- You could have solar cells without the panel, think fo a CD size solar cell encased in plastic just as a CD is a metal foil encased in plastic. Now you can wrap the chip and block the emissions. You can hang the cell from the wires (in a tree or on the distribution wire like christmas lights)
There are 6 ADC's on board.
The 8 bit cores are very much like the I/O processors of the RP2040 Raspberry Pico chip
Attaching an MCU to an inverter isn’t exactly new technology. It’s not clear why you need a bidirectional buck/boost for solar, let alone an 8-core 64-bit MCU on every cell.
Do you have any technology that couldn’t be replicated by another chip designer company by combining their IP blocks and sending it through their fab queue?
The problem with a “1 cent” chip (or $0.04 or a dollar depending on which comment I read) is that it’s a negligible part of the overall cost. The price of the packaging, magnetics, testing, and integration into the panel will eclipse that by orders of magnitude. If a single-founder company could do it with only $50K investment, then any of the big players could have done it years ago unless you’re in possession of some patents that can’t be worked around.
The above claims of producing “30% more” energy also don’t hold up, given that it would require current systems to be less than 77% efficient (a number that was surpassed long ago)
No, its nothing special. Texas Instruments, Analog Devices and many others sell such chips for decades for several dollars each. What makes our chips (Microcontroller SoC) special is the $0,01 price and the high speed network between them so you can aggregate/balance/coordinate the inverter output voltage and the fully programmable inverter with all components in the chip except the inductor (a small coil).
The bidirectional buck/boost is to charge/discharge battery cells individually, for solar cells we just need a unidirectional boost inverter.
>8-core 64-bit MCU Sorry for the confusion, I could not edit my mistake. It is an single core 64 bit processor (ARM, X-86, Risc-V, bytecode and microcode) but reconfigurable to 8x8bit (for I/O processors like the RP2040 Raspberry pico), 4x16bit or 2x32bit or combinations like 4x8bit plus 1x32bit). The reconfiguration of the 8 bit slices is done with our special kind of FPGA fabric that we call Morphle Logic [1]
>Do you have any technology that couldn’t be replicated?
Only special physics knowledge. Our software. But nothing that they could not reverse engineer in a few weeks.
>then any of the big players could have done it years ago
Yes they could and they have. Their designs are pretty bad though, especially the microcontroller parts. And they just charge way too much for them. The Chinese companies that make the panels have not yet cloned their chips.
>a negligible part of the overall cost. The price of the packaging, magnetics, testing
Wrong. Our package is the cheapest with just two or 6 solder balls on the bare die, no package no pins. Testing, magnetics are no factor.
>and integration into the panel will eclipse that No, the chips are in the place where the solar cell wires attach to each other already, the silicon sliver of material makes the standard wireing slightly cheaper.
>claims of producing “30% more” energy also don’t hold up
This is rounded estimated number from several science papers about maximum MPPT yield improvements and separately from wiring cells in parallel. Its certainly not my claim, to calculate this number scientifically you have to do this for every individual cell and panel on every different geographic location.
[1] https://github.com/fiberhood/MorphleLogic/blob/main/README_M...
Rarely does a LCOE calculation cover all these costs, so the published numbers are still hard to compare.
This is a total guess, but 150 days outdoors might translate to a few years of mostly-indoor use. But even then, there is probably need for more QC, etc. I'm just pointing out that there are other potential markets for PV tech that doesn't necessarily make sense for 30 years outdoors. TBD if that will work for perovskites (and honestly, TBD for dye-sensitized cells being a commercial success).
The “record holder” lasted about 150 days outdoors before dropping below 80% of original efficiency. But there are a number of samples that are still going and catching up.
For reference, 80% of “new” output is close to what many solar module warranties cover.
Doing that lets you get more energy out of the sunlight.
But critically, the mix of the two sets of colours really matters. The cell needs equal numbers of photons in each set, or efficiency drops rapidly. That means "red sky in the evening" will be a big problem for these cells.
One way to work around this is to have a grid of wires between the cells, allowing you to take out or inject current between the cells. You would need a specially designed inverter to be able to do this, but then your cell would be able to handle varying colors of daylight.
You don't notice it much, because your eyes auto white-balance what you see. But to the cell it really matters.
The reason everyone wires solar panels and batteries in series was that no micro- and nanoinvertors where available or too expensive. That has changed recently, so we now should wire everything in parallel and get 20-50% improvement.
This discussion is about cells within the panel, not panels within the system. So far, nobody has designed inverters that can work on the 0.5 volt output of an individual cell - although I agree that if such a thing existed, it would indeed help to boost efficiency when cells are mismatched or panels are part shaded.
The challenge with 0.5 volt inverters is that currents are large and voltages are small for a given amount of energy, meaning all the parts of the inverter need to be very low resistance, thick wires, and expensive (due to more copper).
With 0.5 volt currents and voltages are small so you need only thin cheap wires. Our MPPT nano-inverters than aggregate the 6V output of one cell with the 60 cells to any multiple of 6V and thereby lowering the current even more than standard panels with cells in a string. See my other descriptions of the nano-inverters in this HN discussion.
[edit]
I see you partly answered this question in this thread already. Partly because that answer does not fit with the 'wire everything in parallel' mantra since that would necessitate all panels to produce exactly the same voltage to avoid power loss.
You can program the MPPT nano-inverter to output 6V AC and then aggregate to 360V AC. In this situation its better to output 110-120V AC in the US or 240V AC in most other parts of the world and feed this directly into the grid.
The tricky part is the aggregation of the inverter outputs. The inverters each have different input voltage and current but must all output the exact same voltage to be agregated. The microcontroller have a network between all 60 inverters (122 in larger panels) so they can coordinate/balance their outputs. This is done at 100Kbps to 10 Mbps. The mosfet power transisters in the chip will switch at 100KHz to 1 Mhz to achieve the same ouput voltage.
In a future world of plentiful solar, electricity prices are almost zero at midday, and quite high at sunrise/set. That means that even though you might not make many kwh at those hours, most of your profits might come from the shoulder hours.
Top results: "lithium-ion battery prices hit historical low of $140/KWHr"
https://www.nrel.gov/docs/fy23osti/85332.pdf
$482/kwh of battery capacity, installed to grid
15 year lifetime
85% round trip efficiency,
And it costs $88/MWh to store electricity, and adding in charging from the average $40/MW, that's $135/MWh all in for battery costs in 2022-2023. Today will be cheaper, tomorrow cheaper yet. Mix in direct wind production at a 40% capacity factor, annd stored electricity in lithium ion batteries, and wind can become fully dispatchable baseload for less than $90/MWh, at today's prices.
For comparison, the cost of Vogtle 3, the most recent nuclear power plant in the US, is $180/MWh.
Edit: and the main impediment to plentiful solar is policy & regulations that slow it down. The technology is being deployed at massive rates [1] in the US, 33GW in 2023 compared to 22GW in 2022, and ever accelerating. (Capacity factor adjusted, that's roughly 7GW compared to average US electrify draw of ~500GW, for a single year). But the US is way behind China, which installed 217GW in 2023 [2]
[1] https://www.canarymedia.com/articles/solar/chart-the-us-inst...
[2] https://www.bloomberg.com/news/articles/2024-01-26/china-add...
Turns out today's stationary batteries deliberately keep their state of charge in a narrow range in the middle (ie. 40-60% charged), which reduced storage kWh, which in turn increases cost per kWh stored, all in the name of increasing cycle life, which impacts the capital cost per kWh cycled..
Note just how big of a difference there is between the cell price of <$150/kWh and the installed price of >$450kWh. Some of that is labor, land, concrete slab, inverters, but a good chunk is excess kWh.
Medium to large installations have multiple strings anyway. Why not set up the different types of cells as their own strings on separate inverters?
In dense housing situations or areas with heavy tree cover the limiting factor starts to become available roof space. Spending a little extra to squeeze out more efficiency would make sense.
You solve heavy tree cover and other sources of shade with individual MPPT (maximum power point tracker) 4 cent single chip nano-inverters per solar cell, 72-122 per solar panel. For $2.88 extra your solar panel can get 20%-30% more electricity if any part of your panel has shade part of the time.
For installations with string inverters you get even more savings when you replace them with nano-inverters.
Sorry to pick on something said in passing, but nothing saves the world by itself, nothing and nobody. People love to say, 'my vote [or whatever they do] won't change things'. Nobody can change things by themselves, yet things are changed enormously by people - how is that? :)
> Produced in collaboration with the Fraunhofer Institute for Solar Energy Systems, the panel achieved a record 25% conversion efficiency, a significant increase on the more typical 24% efficiency of commercial modules.
> Oxford PV, a spin-out of the University of Oxford, is a world leader in the development of perovskite-on-silicon tandem solar cells, which have a theoretical maximum efficiency of over 43%, compared to less than 30% for silicon solar cells.
I'm wondering if it's just sloppy reporting. At the end they claim their record breaker was 28.6%.