Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

JWST Solid State Recorder

jwst-docs.stsci.edu

Post reply on HN
Print view View original

21–30 of 40 posts

Re: JWST Solid State Recorder

#21

Post by progbits »

osamagirl69 wrote: post #10

I am sure there are better presentations out there, but this conference proceedings has some great pictures of the control systems in JWST [1] I was also able to find a high resolution picture of the spacewire interface card with its glorious space grade asic packages [2] Sadly the SSR is not shown, I assume that it is basically a gigantic sheet of sram and a fpga. [1] https://www.researchgate.net/publication/2339621…

In [2], especially on the second picture z you can see some solder joints are not consistent or perfect: convex rather than concave fillets and differing amounts of solder.

Now of course it still looks better than most commercial PCBs and far ahead of anything I ever made. But I was expecting a board flying to space to be "perfect", touched up by a gray haired wizard.

How come? Is this actually plenty good, do they cover this in so much conformal coating there is no chance of joint breaking due to vibration and thermal cycling, or what?

Btw the large square packages in the first picture look amazing.

Re: JWST Solid State Recorder

#22

Post by bigiain »

AaronFriel wrote: post #8
xupybd wrote: post #6

How does this storage hold up over time? I'm guessing that could be one of the limiting hardware elements. There must be some redundancy in that right?

Given the capacity, I would be surprised if it wasn't single-level cell NAND (larger write capacity) or redundant. From articles I could find, it was installed in the JWST in 2012[1] and developed by a company called SEAKR which produces "Solid State Recorders"[2]. Certainly doesn't look like any commercial server storage I've seen. I imagine the contracts could be FOIAed here and specs like anticipated write capacit…

> From articles I could find, it was installed in the JWST in 2012

I suspected something like that. I assumed there was a reason there was "only" 65GB of storage up there, when I can get a half terabyte microSD card for under $70 on PrimeDay. Lots of Moore's Law has happened in the last 10 years...

As a RaspberryPi enthusiast, I'm under no illusions about the reliability of microSD cards, but I wonder whether if they were building that data recorder today, that a huge raid 1 with perhaps 20 or 50 mirrors stored on 512GB SD cards might be able to compete on reliability and cost (including launch weight)?

Re: JWST Solid State Recorder

#23

Post by semicolon_storm »

Shared404 wrote: post #4
worewood wrote: post #3

Earlier quoted context omitted.

Surprisingly high IMO

I know people who live within 30 minutes of me who get less than that - and I live in a college town.

Considering that's 2MB/s upload, I get less than that and I live in a major city with a major ISP.

Re: JWST Solid State Recorder

#24

Post by mcnugget »

bigiain wrote: post #22
AaronFriel wrote: post #8

Earlier quoted context omitted.

Given the capacity, I would be surprised if it wasn't single-level cell NAND (larger write capacity) or redundant. From articles I could find, it was installed in the JWST in 2012[1] and developed by a company called SEAKR which produces "Solid State Recorders"[2]. Certainly doesn't look like any commercial server storage I've seen. I imagine the contracts could be FOIAed here and specs like anticipated write capacit…

> From articles I could find, it was installed in the JWST in 2012 I suspected something like that. I assumed there was a reason there was "only" 65GB of storage up there, when I can get a half terabyte microSD card for under $70 on PrimeDay. Lots of Moore's Law has happened in the last 10 years... As a RaspberryPi enthusiast, I'm under no illusions about the reliability of microSD cards, but I wonder whether if they…

Bit flip probability is inversely proportional to density of the storage media so denser might have you run into issues. Also theres no point on building a recorder that's 10x larger than what you need if you're going to be limited by downlink availability. You could probably get away with all those SD cards in LEO though for a short missions, JWST falls into the category of missions where the price of that storage medium is going to be a rounding error vs the price it would incur if it randomly failed during qualification.

Re: JWST Solid State Recorder

#25

Post by pkaye »

bigiain wrote: post #22
AaronFriel wrote: post #8

Earlier quoted context omitted.

Given the capacity, I would be surprised if it wasn't single-level cell NAND (larger write capacity) or redundant. From articles I could find, it was installed in the JWST in 2012[1] and developed by a company called SEAKR which produces "Solid State Recorders"[2]. Certainly doesn't look like any commercial server storage I've seen. I imagine the contracts could be FOIAed here and specs like anticipated write capacit…

> From articles I could find, it was installed in the JWST in 2012 I suspected something like that. I assumed there was a reason there was "only" 65GB of storage up there, when I can get a half terabyte microSD card for under $70 on PrimeDay. Lots of Moore's Law has happened in the last 10 years... As a RaspberryPi enthusiast, I'm under no illusions about the reliability of microSD cards, but I wonder whether if they…

Another reason for the 65GB of storage is that storage is only used to buffer a day worth of transmitted data along with the previous days data till reception is confirmed.

Re: JWST Solid State Recorder

#26

Post by I_Byte »

fnordpiglet wrote: post #16

Reading this stuff always makes me wish I did real computer stuff.

I feel this. As a first year Computer Science student at university I am currently torn between staying as CS or switching to Computer Engineering. I am really interested in the low level aspects of computers from a software perspective (OS development, assembly, compilers) but I feel my education in my area of interest won’t be complete if I don’t also understand things from a hardware perspective.

Re: JWST Solid State Recorder

#27

Post by formerly_proven »

AaronFriel wrote: post #8
xupybd wrote: post #6

How does this storage hold up over time? I'm guessing that could be one of the limiting hardware elements. There must be some redundancy in that right?

Given the capacity, I would be surprised if it wasn't single-level cell NAND (larger write capacity) or redundant. From articles I could find, it was installed in the JWST in 2012[1] and developed by a company called SEAKR which produces "Solid State Recorders"[2]. Certainly doesn't look like any commercial server storage I've seen. I imagine the contracts could be FOIAed here and specs like anticipated write capacit…

> https://www.seakr.com/wp-content/uploads/2016/08/Flash_FMC-G...

Looks like any other SSD from a few years ago, just bigger and in a metal box.

> The GEN3 FMC is a 192 GByte card designed for space applications. It is a standard 6U cPCI form factor composed of 3 cards: the main controller card and two memory mezzanine cards. It supports data and control transfers via the backplane cPCI bus.

Re: JWST Solid State Recorder

#28

Post by kevin_thibedeau »

progbits wrote: post #21
osamagirl69 wrote: post #10

I am sure there are better presentations out there, but this conference proceedings has some great pictures of the control systems in JWST [1] I was also able to find a high resolution picture of the spacewire interface card with its glorious space grade asic packages [2] Sadly the SSR is not shown, I assume that it is basically a gigantic sheet of sram and a fpga. [1] https://www.researchgate.net/publication/2339621…

In [2], especially on the second picture z you can see some solder joints are not consistent or perfect: convex rather than concave fillets and differing amounts of solder. Now of course it still looks better than most commercial PCBs and far ahead of anything I ever made. But I was expecting a board flying to space to be "perfect", touched up by a gray haired wizard. How come? Is this actually plenty good, do they c…

These electronics tend to be hand soldered so there is less uniformity than reflow.

Re: JWST Solid State Recorder

#29

Post by AaronFriel »

formerly_proven wrote: post #27
AaronFriel wrote: post #8

Earlier quoted context omitted.

Given the capacity, I would be surprised if it wasn't single-level cell NAND (larger write capacity) or redundant. From articles I could find, it was installed in the JWST in 2012[1] and developed by a company called SEAKR which produces "Solid State Recorders"[2]. Certainly doesn't look like any commercial server storage I've seen. I imagine the contracts could be FOIAed here and specs like anticipated write capacit…

> https://www.seakr.com/wp-content/uploads/2016/08/Flash_FMC-G... Looks like any other SSD from a few years ago, just bigger and in a metal box. > The GEN3 FMC is a 192 GByte card designed for space applications. It is a standard 6U cPCI form factor composed of 3 cards: the main controller card and two memory mezzanine cards. It supports data and control transfers via the backplane cPCI bus.

That's just a flash card. It's not the solid state recorder.

Re: JWST Solid State Recorder

#30

Post by hoseja »

mcnugget wrote: post #24
bigiain wrote: post #22

Earlier quoted context omitted.

> From articles I could find, it was installed in the JWST in 2012 I suspected something like that. I assumed there was a reason there was "only" 65GB of storage up there, when I can get a half terabyte microSD card for under $70 on PrimeDay. Lots of Moore's Law has happened in the last 10 years... As a RaspberryPi enthusiast, I'm under no illusions about the reliability of microSD cards, but I wonder whether if they…

Bit flip probability is inversely proportional to density of the storage media so denser might have you run into issues. Also theres no point on building a recorder that's 10x larger than what you need if you're going to be limited by downlink availability. You could probably get away with all those SD cards in LEO though for a short missions, JWST falls into the category of missions where the price of that storage m…

It would still be better to have 10-100x in enterprise-grade storage with, say, quadruple redundancy. As Ingenuity has shown.
Post reply on HN