News Samsung's Hwaseong Chip Plant Out of Order for 3 Days Due To Power Outage

bit_user

Polypheme
Ambassador
Yesterday Samsung's facility in Hwaseong, North Korea, suffered a one-minute long power outage,
I know South Korean companies used to do some manufacturing in a border town of the North, but I'm pretty sure that's been done away with, for a decent amount of time. And I don't think they ever did anything as sensitive as semiconductor manufacturing, there.

According to Wikipdedia, Hwaseong is both a city in the South, and a county in the North. I'm fairly certain this news is referencing the city.



in 2018 Samsung's Pyeongtaek factory experienced a 30-minute blackout which caused about 50 billion won (US $43.3 million) in losses.
I know they must use a lot of power, to run a facility that big, but I can't help but wonder whether $43.3 M couldn't build an adequate backup generator for it.
 
Last edited:
I know South Korean companies used to do some manufacturing in a border town of the North, but I'm pretty sure that's been done away with, for a decent amount of time. And I don't think they ever did anything as sensitive as semiconductor manufacturing, there.

According to Wikipdedia, Hwaseong is both a city in the South, and a county in the North. I'm fairly certain this news is referencing the city.




I know they must use a lot of power, to run a facility that big, but I can't help but wonder whether $43.3 couldn't build an adequate backup generator for it.
At first I was thinking Tesla Powerwall like in South Australia or an industrial power backup like Generac.
https://www.generac.com/industrial/industrial-solutions/manufacturing

But it may be hard due to being North Korea.

Something much easier to find would be an APC Power Saving Back-UPS Pro 1500 which can be bought from Amazon or at a Bestbuy in countries not named North Korea
APC Power Saving Back-UPS Pro 1500 costs around $224
https://www.amazon.com/APC-Back-UPS-Battery-Protector-BR1500G/dp/B003Y24DEU

$43.3M / $224 = 193,000 APC Power Saving Back-UPS Pro 1500 units
193,000 x 1500 Volt Amps = 289,500,000 Volt Amps x a power factor of 0.8 = 231,600,000 or 231 MegaWatts

Now you wouldn't want all that to be in series which would reduce the maximum megawatts, but that should be enough juice for a minute of power loss.
I'm sure it isn't that simple, but North Korea is highly motivated to make it work.

And you could buy them from Bestbuy (in places not named North Korea) and have them shipped to North Korea much more easily than an industrial sized one.
 
Last edited:
  • Like
Reactions: bit_user

bit_user

Polypheme
Ambassador
Good points. I had been thinking more along the lines of diesel generators, but batteries might do. Especially if you had good reason to think most outages would only be short ones.

And even with diesel generators, some batteries would still be needed to cover the gap until the generators can come online.

And you could buy them from Bestbuy (in places not named North Korea) and have them shipped to North Korea much more easily than an industrial sized one.
Given that they're probably made in China, which is a country sharing a border with North Korea, they'd probably be even cheaper for North Koreans to buy than us Americans.


P.S. Hi dude. Haven't seen you around, much.
 

InvalidError

Titan
Moderator
Good points. I had been thinking more along the lines of diesel generators, but batteries might do. Especially if you had good reason to think most outages would only be short ones.
Diesel would be no good since it takes several seconds to start them and transfer load. You need something to provide a 30-60s buffer between main feed going down and backup getting up to speed. The question here is which costs more? A large-scale power backup with its added maintenance, footprint and liability to (hopefully) weather seconds-long power outages or losing a batch of wafers?

Some power companies use flywheel energy storage to regulate frequency and voltage against transient load, might be a better option than batteries when disruptions rarely last more than a few seconds.
 

bit_user

Polypheme
Ambassador
Diesel would be no good since it takes several seconds to start them and transfer load. You need something to provide a 30-60s buffer between main feed going down and backup getting up to speed.
It's funny how I said that in the very next sentence, which you omitted.

FWIW, the diesel generator at my job doesn't take even 10 seconds to fire up and step in. I couldn't tell you how big it is, though... maybe a couple hundred kW?

I happen to know that backup generators are available with enough battery backup to cover that gap, though they're not cheap. I imagine they might be used for hospitals.

I wonder what datacenters do. They've become fairly mission-critical, and a sudden power outage would be highly disruptive, although that would be a heck of a lot of generation capacity.

The question here is which costs more? A large-scale power backup with its added maintenance, footprint and liability to (hopefully) weather seconds-long power outages or losing a batch of wafers?
This is the very question I posed, in my first post. Is $43.3 M not enough to build that generation capacity? Any further outages they'd have avoided (like this one) could probably cover the ongoing maintenance.

The other question worth pondering is how much equipment would need backup power, simply to avoid batch spoilage and reduce production restart time. There might be a much lower bar for mitigating the impact, which could also be satisfied with cheaper & easier-to-maintain, battery-based solutions.
 
Last edited:

InvalidError

Titan
Moderator
The other question worth pondering is how much equipment would need backup power, simply to avoid batch spoilage and reduce production restart time.
A fab has a ton of ASML lithography machines, so you have to start your list of essential stuff that requires online backup power if you don't want to lose wafers with those, 10+MW right there. Then you have metal vapor deposition furnaces, add some more tens there - doubt 7nm leaves much tolerance for a wafer getting the wrong amount of metal deposition due to plasma bias voltage being too low or the plasma source extinguishing altogether. Wafer coating, etching and rinsing stations can't be off by seconds either when dealing with coating/etching/rinsing nanometer-scale layers.

The only figure I could find at a glance is about 100MW for a large fab... back in 2013. Likely much higher today.