GROUND FAULT INTERRUPTERS
This was the first IC design containing my design work.
The design ultimately became the LM1850 and later the LM1851.
This IC is commonly still found in GFI today such as is shown below.
I designed two other GFI ICs before the LM1850 went into production.
My first independent IC design was a GFI whose die photo is shown below.
The ICs in those days were designed to be small cheap.
There was a lesson to be learned when running GFIs in production.
Up until this time, analog ICs where normally tested to DC specs.
This work catch 99.9% of the defected die.
But in the GFI market, having 1 part in 5000 fail was a concern.
The normal reaction was to tighten the specs for testing the parts.
Now the failure rate was one part in 50000.
It was discovered that bad parts could sometimes pass a DC tester.
On rare occasions, several defects could cancel out to pass a test.
The solution was to actually connect the IC in test to 60Hz 110VAC.
This worked so well, that we tested GFI this way all the time.
When Leviton found out about this additional testing, they complained.
When Raytheon became a second source for GFIs, Leviton quickly learned.
One of National Semiconductor's customers had an "unwritten spec".
The IC was expected to work in the intended application.
The production experience from the previous two GFIs had taught a lesson.
Weird and strange failures happen after running 100,000 parts in production.
The solution was to test IC exactly how it was to be used.
In a system design, one's concern with everything it takes to get an IC to work.
This is where analog design fits right in.
Silicon area inside a chip was much cheaper that external components.
The idea was to "integrate" external components into silicon.
GFIs work off the principle that all current is accounted for.
A net in/out imbalance in current will generate a magnetic H field around it.
This is true even when there are 15Amps flowing through a socket.
All the current that flows out of a socket is expected to flow back in.
A magnetic coil around the two wires can detect a magnet field.
__\_
H| / |
_______________|_____________
()_____________\|/____________)
--> I | V
|_/_
\
In real life, there can be something switching power on and off the line.
This produces large voltage spikes along with RF radiation.
So the trick is for the chip to only detect current difference as small as 5mA.
Ignoring one thousand volt voltage spikes in and RF radiation was the challenge.
Some analog signal processing was used to address the problem of false tripping.
My first unpatented invention involved a "3 to 1 pull down".
It can be found in most GFIs on the market today.
At 1ma is about the beginning of sensation.
At about 10mA, a hand will become locked around a wire and cannot let go.
At about 30mAs breathing stops.
At 120 volts, the dry resistance of skin may limit the current.
But with the skin wet, the current may get as high as 500mA.
The UL spec defines a spec of current versus time to prevent false tripping.
Current goes through the heart at the right time can cause ventricular fibrillation.
Based on studies using dogs, the timing curve below was derived.
The UL spec requires trip time to be from 5seconds at a 5mA faults and 16msec at 500mA.
