🚀 go-pugleaf

On Tuesday, January 23, 2018 at 3:18:36 AM UTC-5, AndyW wrote:

> There was also emphasis on preventing failures, often by the incredibly 
> expensive techniques of pre-stressing or 'hardening' components that 
> gave you confidence in terms of reliability but increased the cost by 
> orders of magnitude. Now we design systems to work around failures or 
> else to identify impending failures and schedule replacement in 
> down-time so as not to impact the mission.

>From reading various early 1950s writings, it seems the cost
of failure in industrial electronics was high enough to justify
the pre-stress approach, especially for vacuum tubes.  For an
early computer, tracing a faulty component could take a lot of 
time, especially if the failure was not total, but rather a degrade
in performance.  Computers were located at customer sites, which
meant skilled technicians had to travel out (though in the earliest
days I think some techs were resident full time).  There wasn't only
the cost of diagnosis and repair, but also the high cost of downtime.
Early on, they were desperately trying to improve downtime.

>From the various writings it seems there was finger pointing
between the design engineers, builders, and component manufacturers.

In a way, the debate 60 years ago reminds me of the debate today
among software developers--there are some who use the latest bells
and whistles, such as in web design, while others who hold back a bit.
I suppose some engineers in the past (and the writings suggest this)
wanted to use the latest and greatest circuits, perhaps to squeeze 
out some additional speed or miniaturization.

(I think the early AFIPS write-ups on bitsavers cover this.)

On Thursday, January 25, 2018 at 2:31:32 AM UTC-5, AndyW wrote:


> > This might be due to the consumer mindset, which eschews maintenance.
> > Buy a gadget, run it like hell until it breaks, throw it out.  Lather,
> > rinse, repeat.  Under such conditions I can see a bathtub curve as a
> > combination of infant mortality and wear out - both of which would be
> > eliminated by conscientious installation and maintenance.
> 
> The modern mindset is to accept that the failure will occur despite 
> preventive maintenance and instead work on making the failure 
> convenient. You can prevent failure by throwing money at the problem 
> (which is a diminishing returns situation) or you can detect an 
> impending failure and just schedule a replacement at the next docking 
> period, production line shutdown, holiday period etc.

When the Bell System brought out electronic switching, the devices
had dual processors for reliability.  How often the active device
failed and they had to switch to backup I don't know.  One book
suggests that while overall ESS was more reliable than relays,
if ESS failed, it all failed, as opposed to merely slowing down.
The book also said if an ESS got overloaded it would fail, while 
relays would slow down (giving a warning first).  ("Telephone" by
John Brooks).

In Jan 61, Bell tried out a new telephone set in a trial.  They
found that, as they put it, "the mylar foil capacitor and diodes
did not 'perform as expected' and needed to be redesigned".