The Solid-State Shift: Reinventing the Transformer for Modern Grids
24 comments
·April 12, 2025Animats
sightbroke
> Then they'll make them "cloud-enabled" and they will be hacked.
It's worse:
https://arstechnica.com/security/2025/01/could-hackers-use-n...
metalman
the integration of the SST's is years out, anything right now is test phase, and work at bringing awareness to the grid operators, who by the way are all conversant with the peculiaritys of working with steam, ie: a conservative bunch who you can bet, will be running code on terminals so esoteric and unpublished as to invite madness for the unwary.........that, and power companys have been embedding there own private fibre inside the core's of the HVTL, so they have unusual options. It's the grid fer fucks sake, give them some credit. Powermag looks to be extra solid,bookmarked.
fc417fc802
It's fine provided that the link from the equipment that feeds it data is optically isolated to only go in one direction.
A public internet connected web server that enables remote equipment control is indeed scary.
Animats
> It's fine provided that the link from the equipment that feeds it data is optically isolated to only go in one direction.
Then people get two of them, one for each direction.[1] Can someone explain why this is supposed to be secure? It's apparently a real product.
mppm
Amazing. Maybe I should pitch them my idea of MIL-spec acoustic relays for communicating with air-gapped facilities?
fc417fc802
Wut? That reads like satire. The equipment is inside the security boundary, the LAN is outside. What is the purpose of enforcing one-way control signals when the thing sending the control signals is by necessity within the same security boundary as the destination for those signals?
I want to extend the benefit of the doubt and assume my own ignorance but I'm really struggling with this one.
hristov
This is very interesting and cannot happen fast enough, considering the current worldwide transformer shortage.
I have a question for people more familiar with these. What exactly happens at the isolation stage. They say it includes a high frequency transformer (HFT). But its input and out put is DC. And classic transformers operate on AC. So in order to get the transformer working, one would have to chop up the incoming dc power into a square wave or a sine wave. But what transistors can you use to do this, considering you are dealing both with very high power and very high frequencies?
itcrowd
The keyword you are looking for is IGBT (insulated-gate bipolar transistor) -- the type of transistor used in such DC-DC converters
hristov
I don't think IGBTs can handle frequencies up to "several megahertz" as the article says.
pwg
The diagrams on the page are nothing more than the typical switch mode power supply [1] layout, the only difference being handling grid power levels and conversion back to AC at the output (most switch mode PSU's output DC).
bacon_waffle
There are techniques for "stack"ing transistors so that the individual swtiching devices see potentials that are within spec and much lower than the voltage switched by the overall circuit.
cyberax
You need to use nuclear-powered transistors!
No, I'm not joking. For these kinds of voltages, you need to use highly homogenous doped silicon, and the only way to produce it is to irradiate silicon with neutrons. It transmutes some of the silicon atoms into phosphorus: https://nrl.mit.edu/facilities/ntds/
yjftsjthsd-h
I assume that if people are going to the trouble to literally irradiate the material in order to get what they need, they can't get the results by just mixing in phosphorus. Could somebody who actually understands this tell me why that is?
CorrectHorseBat
https://www.waferworld.com/post/the-complete-guide-to-doping...
It's apparently possible for boron doping.
I think because phosphorus is bigger than silicium you'll get too many defects in the crystal, while with the smaller boron it is not an issue.
cyberax
Because there is no other way to produce homogenous enough thick wafers of doped silicon. Other methods rely on diffusion from the surface, which is not enough for this case.
And doping during crystal growth doesn't produce homogenous enough silicon.
Gibbon1
Power transistors can be had that can switch up to maybe 3000V at the most. But maybe 1200-1500V is more common.
You can stack power transistors to switch higher voltages on the primary side. On the secondary side you just need an H-bridge. Which can be made up of transistors in parallel.
We've had high power high voltage transistors for about 40 years now. A lot of this isn't technical but rather economic. As the price falls the applications where they are cheaper increases. It's notable for instance Toyota started work on their hybrid drive in the mid nineties when inverters for 10-100 HP motors became cheap enough.
algo_trader
Ex-tesla Drew Baglino is starting a new company Heron for solid state transformers
Of topic - dealing with medium voltage e.g. 6kv-10kv - are there existing be-spoke solid-state solutions?
e.g. PV/battery 400V converted to 6kv without inverter/transformer
Calwestjobs
figure no. 1 is missing green color on HVAC transmission line. i do not have ExTwitter and i do not know other way to let them know.
it is interesting to think about human made object in terms of how much materials they use, for example big old transformer contains 2 tons of iron, new solid-state transformer with same capacity uses only 300 kg of silicon (Si), 120kg kg of plastics and 50 kg of copper.
JumpCrisscross
E-mail the author spatel@powermag.com [1].
wbraun
Where are your weight figures coming from, they are not in the linked article.
Packaged semiconductors are going to be more metal interconnect / plastic encapsulation / ceramic insulation than silicon by weight.
These systems will also have a significant weight fraction in magnetic materials, either ferrite ceramics or amorphous metals.
Still a huge weight savings, but the weight fractions you are giving see off and are missing some important materials.
mppm
Why do solid-state transformers need so much silicon?
LargoLasskhyfv
Because they contain long cascades of these https://en.wikipedia.org/wiki/Insulated-gate_bipolar_transis... to be arranged into these https://en.wikipedia.org/wiki/Power_electronics
mppm
As a sibling comment implies, high-power IGBTs are still built from small and thin silicon dies with most of the module being heat spreader, insulator, interconnect and packaging. Even MW-class modules probably contain a gram or so of Si.
Then they'll make them "cloud-enabled" and they will be hacked.
It's scary to see protective relays for power systems with embedded web servers. "IEEE C37.118 synchrophasor measurement, DNP3 Outstation, Modbus TCP/RTU, Telnet, FTP, Simple Network Time Protocol (SNTP), built-in web server, and IEC 61850" [1]
[1] https://selinc.com/products/351/#