[time-nuts] Rubidium Cells for Sale ?!

[jimlux]

On 5/6/19 7:04 PM, William H. Fite wrote:
> Mother of God, really?
> 
> I had a friend, now of blessed memory, who was lead communications engineer
> for Grumman on the lunar lander. He used to boggle our minds with stories
> of the truly absurd lengths that NASA made them go to have hardware "space
> certified."
> 


This is changing..

Recognize that for NASA, they're usually building just one (or maybe 2 
flight units plus a spare, if you're on a big budget Class A mission 
with redundant strings). So they don't think in terms of MIL-HDBK-217 
kinds of reliability calculations of statistics.. It's more a matter of 
"what could go wrong, and how can we prevent that".

So you wind up with a lot of requirements and tests that may not be 
statistically justifiable. They tend to require large design margins 
(since you're building just one, the unit test campaign is both 
verifying the design AND doing acceptance testing).

For instance, some years ago voltage/frequency stress testing was 
popular - run it at a variety of frequencies and voltages, well beyond 
the design range, and show that you've got margin.

There's also a "if it doesn't meet the datasheet specs under all test 
conditions, it is deemed to have failed" philosophy.  A classic problem 
is optocouplers.  You might choose a part that has a current transfer 
ratio of 100, and your design needs a CTR of 1.  But the data sheet says 
90<CTR<110.  So after radiating it with some dose, the CTR has degraded 
to 85.  The part still works just fine (you need a CTR of 1) but the 
parts engineer says "nope, that part has failed at the dose, you can't 
use it".

A lot of space qualification is paperwork to prove you have 
"traceability to sand" for the parts.  Lot numbers, production dates, 
etc.  So when the GIDEP (http://www.gidep.org/) comes out for a 2N2222A 
(yes I've gotten one), you can go and see if YOUR particular NPN 
transistor is covered.

And then there's testing at many levels (not all of which is valuable)...

Incoming inspection of resistors.  Back in the 60s, someone must have 
gotten a "out of spec" resistor in a box of 5% resistors - so the 
procedure was put in place: Measure each resistor (after assigning a 
serial number, and verifying the color stripes, including the stripe 
width and colorimetric properties) with a calibrated ohmmeter(with 
calibration data recorded), record each measurement, and attach that the 
to the build book.

So, today, you get a reel of 1000 resistors all 100 ohms. Someone in 
incoming inspection in a clean room at an ESD safe (to 50V) workstation, 
pulls each resistor off the tape, takes a picture of it, verifies the 
physical size on a coordinate measuring machine, measures the 
resistance, logs that information, and carefully places the resistor 
into an assigned cell in a waffle pack.   Then, later, someone takes the 
resistors out of the waffle pack, puts them back into a tape, after 
measuring the resistance and dimensions, so that it can be loaded into 
the automated assembly machine.

This is what makes space qualified equipment expensive.