Spehro Pefhany said:
I bet they drift like crazy over 100 hours or so with DC..
It's not rocket science?
Here is a version of the printed data I'm looking at.
http://rocky.digikey.com/WebLib/Yageo America/Web Data/5% Chip Resistor.pdf
See for example, page 9 (0603).
Best regards,
Spehro Pefhany
Figures 9 to 13 show the classic K=I^2*t on the right side, and the
classic flattening (rather than approaching infinity) on the left
side. The physical mechanisms for this are interesting and account
for otherwise perplexing differences in resistor series and even
ranges of values. For example, failure testing reveals that a 3R3
resistor might tolerate much higher pulse power than a 4R7 in the same
product series, while the reverse might be true in another product
series. For another example, some 0603s might be more tolerant of
high pulse power than some 0805s.
The 'element' of film resistors, both thin and thick film types, are
made by depositing a layer of material on a substrate and connecting
the film to pads or leads. To control resistance, these are the
available variables:
1. Element formulation (what its made of)
2. Element thickness
3. Element width
4. Trim, i.e., adjustments to element geometry, usually width, that
are used to reach the desired Ohms value
It's important to understand how big a role Trim plays in resistor
production: Most manufacturers produce a very limited range of
Formulation,
Thickness, and Width product, and Trim is used to produce the
incredible range of resistance values. For example, suppose that a
particular formulation of carbon-based thickfilm resistor 'ink' is
screened onto ceramic substrates with a particular geometry, including
a fairly uniform thickness. Suppose that this produces 1 Ohm
resistors plus or minus 20%. A trimming operation can be used to
adjust resistances to any value greater than 1R2 Ohms. In practice,
these same raw resistors might be trimmed for the range of 1R2 to 4R7,
a ratio of 4:1. Other formulations and/or geometries would be used
for higher value resistors.
Now consider how most film resistors are trimmed: Some of the element
is 'burned' away by a laser or other ablative tool under feedback
control. This is usually done in the 'J' pattern, as shown in this
diagram (fixed width font here):
-----------------------
| || |
| || |
| || |
| \\ // | ____
| == | ^
| | w current crowding region.
----------------------- ----
That's supposed to show a typical planar element, with the 'J' shaped
feature the laser cut. Electrode connections would be on the left and
the right. They use the J-shape, because it provides more precise
control of resistance than a straight '|' shape: resistance variation
slows down as you enter the curve. (The reasons are left as an
exercise for the reader. ;-) Anyhow, the important thing to realize
is that current flow ends up crowding into the region below the J,
which has width w. The region immediately below the J is where film
resistors fail, because it is also the region of max power density.
For very high pulse amplitudes and short pulses, the power density in
this narrow region reaches a failure point, even if the average power
is low. The amplitude of this failure tends to be fairly constant,
regardless of pulse duration, hence the flattening of the I-squared-t
curve. (Basically, this is because the failure occurs virtually as
soon as the critical amplitude is reached.) And, the critical
amplitude is strongly influenced by the value of w, the remaining
width after trim. Following one of the examples above, suppose that
we compare a 4R7 resistor and a 5R1 resistor. The 4R7, having been
trimmed from 1R0 (nominal), will have a very small w, approx 25% of
the untrimmed width. The 5R1, having been trimmed from a nominal 4R6
(or something like that), will have a large w. Consequently, it's not
unusual to see test results having strange steps and discontinuities
when you plot failure point versus resistance. Similar strange things
sometimes happen when you compare product lines and package sizes.
These results can generally be explained in terms of how the resistors
are designed and processed.
The actual failure mechanism varies with resistor technology.
Thickfilm resistors fail when the element cracks due to differences in
expansion rates of the element, substrate, and passivation layers.
Thinfilm resistors may fail when the element literally vaporizes.
Perhaps the oddest result of all is this: Many low value thickfilm
resistors will tolerate very high levels of steady state power without
failing. You can literally melt their solder connections without
destroying the resistor itself or much affecting its value.
Paul Mathews
