More OT & re at least a related technology: Long ago, I was involved
with the DEC core memory design and core making. Core making was an
"interesting" operation since the only the "bakers" knew the magic for
getting good cores. Actually only about 30% of every batch was any
good. The good news was the bad cores got ball milled with the next
batch of raw materials, then isostatically pressed, fired, ground up
again, pressed again, fired, and finally ground into press-able powder,
pressed in carbide dies, fired, and 100% tested. We made billions all
for DEC systems and were the last systems supplier to offer core memory,
since real time systems customers loved never having to wait and wait
for a reboot after a power failure. The last generation cores were
0.012" OUTside diameter and the all women core stringers in Taiwan put
three wires through each of those tiny donuts. And our core memory was
cheaper to make than DRAMs until after the MOSTEK 64k chip hit its
volume pricing valley. Our ferrite "bakers" were really world class.
DEC bought (cheap) the RCA core memory operation when RCA exited the
computer business in 1971.
When one considers ferrite core memory technology it is a wonder it
worked at all - half select noise, temperature sensitivity, aging,
stress effects, magnetostriction, etc were large engineering and process
control challenges. At least cosmic rays didn't bother it and it
didn't forget and it was the technology that made computers as we know
them possible.
Grant KZ1W
On 7/9/2016 1:21 AM, Ian White wrote:
Several good points there, about the variability of ferrite cores.
Ferrites are, quite literally, "bakery products". Just like bread and
cakes, the properties of ferrites depend on the correct ingredients
measured out in precise quantities, on the precise manner in which those
ingredients are mixed, and also - most critically - on the
temperature/time profile of the baking and cooling.
Just like baking, the manufacture of ferrite materials is a complex
blend of science and know-how. Once a specific product has been
developed, consistency can only be achieved by repeating exactly the
same processes for every batch.
It is very easy to see how QC problems could appear from outsourcing
those critical processes to an offshore company that lacks the original
manufacturer's in-house know-how, with a language barrier that prevents
that information being accurately transferred.
73 from Ian GM3SEK
-----Original Message-----
From: TowerTalk [mailto:towertalk-bounces@contesting.com] On Behalf Of
Jim Brown
Sent: 09 July 2016 00:23
To:towertalk@contesting.com
Subject: Re: [TowerTalk] AES SK
It's good that you raised this, Jim. Fair-Rite is yet another example
of
a great small business that was owned and run by engineers who happened
to be married. He was the Chem E, she was the EE. I met them in their
booth at an IEEE EMC engineering conference in Chicago in 2005. Not
long
after that they sold the business and retired. That's probably when mfg
moved off shore. Several years ago, I heard from a local EE working mfg
that his company was having serious QC issues with their #61 cores of
the same sort you described.
My measurements of coax chokes were mostly done in 2007, the bifilar
chokes in 2009-10. Measurements that produced the families of data for
1-14 turns of the five different materials were done in a well known
lab
in 2002-3 by my collaborator who has chosen to remain anonymous to
avoid
"issues" at work.
I would NOT, however, solely blame QC for the problem with getting
consistent measurements on chokes, simply because the circuit Q of
practical chokes is quite low, typically around 0.5. Rather, I think
much of it is a measurement problem. It is VERY well known that
reflection-based impedance measurements have increasingly poor accuracy
for values of Z that vary by more than about 5:1 from the system
impedance of the measurement system (usually 50 ohms). This is because
the equation for Z involves the sum and difference of S11 and 1, so
very
small errors in S11 result in large errors in Z.
This error is in addition to the stray C of the measurement fixture,
which can cause significant errors in the resonance of the choke. This
is significant with #31 and #43 chokes that are resonant above about 10
MHz, and huge errors in higher Q materials like #61. In both cases, the
actual resonance of the choke is higher than the measured value.
73, Jim K9YC
On Fri,7/8/2016 3:12 PM, Jim Thomson wrote:
Well you could work for Fair rite as a type 31 sales manager. Or
better
yet,
visit their new plant in China, where they make all these products,
and
figure out
why they have such extremes and variations in their type 31 cores
since
the chinese
plant opened. N3RR bought over 700 of em, 2.4 inch od cores, and
found they
are all over the map, and even sent samples to Fair rite. Bill
ended up
devising a
simple 1 turn test, then graded all 700 of em into various sub
groups. No
wonder the
initial CMC results were not repeatable.
_______________________________________________
_______________________________________________
TowerTalk mailing list
TowerTalk@contesting.com
http://lists.contesting.com/mailman/listinfo/towertalk
_______________________________________________
_______________________________________________
TowerTalk mailing list
TowerTalk@contesting.com
http://lists.contesting.com/mailman/listinfo/towertalk
_______________________________________________
_______________________________________________
TowerTalk mailing list
TowerTalk@contesting.com
http://lists.contesting.com/mailman/listinfo/towertalk
|