The data sheet for the LT3080 linear voltage regulator suggests
using printed-circuit-board (PCB) traces for ballast
resistors. Although the LT3080’s low offset voltage suits it
well for this technique, it can be used for other ballast applications—
for instance, for a set of bipolar transistors.
For example, in the figure, ballast resistors R1 and R2 can be
short lengths of copper trace. In a practical case, R1 and R2 can
be about 5 mm (0.2 in.) long and just a few mils wide. Obviously,
this is economical.
Low-value current-sense resistors can be expensive, and using
PCB copper for the resistors also saves the expense of mounting
them while increasing reliability compared to soldered
connections. The same benefits can be seen in other designs
requiring low-value ballast resistors, and they may be a practical
alternative to discrete parts even when the PCB traces take considerably
more board area.
But there’s an advantage to using PCB traces that’s not mentioned
in the LT3080 data sheet. It involves the traces’ high
temperature coefficient.
Using copper traces to sense and measure current may not
be a good idea, because the resistance has a temperature coefficient
of almost 0.4% per °C, which causes nearly a 10% change
over 25°C. Although that’s a problem for precision measurement,
it’s actually an advantage in ballast resistors, because
the resistor carrying the highest current will show the largest
self-heating (assuming the same environment for all the ballast
resistors). That self-heating will increase the resistance, thus
lowering the current in that branch compared to a branch that
doesn’t self-heat as much.
For example, assume that the two regulators in the figure have
output voltages that differ by 2 mV and the load current is 2 A.
If R1 and R2 are each 10.00 mO, the higher voltage regulator
will deliver 200 mA more current than the lower voltage regulator.
But if R1 and R2 are narrow traces that self-heat at 2.5°C/
mW, with 10-mO resistance at the board’s nominal operating
temperature, the current difference between the two regulators
will be approximately 150 mA, or 25% less than if the circuit
used temperature-stable resistors for R1 and R2.
The LT3080 data sheet mentions that you can distribute
these regulators around a board. They don’t have to be close
together to share current between two regulators. That’s an
obvious advantage to distributing heat around your board.
However, it also suggests an additional advantage you can
obtain from the temperature coefficient of resistance of copper
ballast resistances.
You’re probably more interested in the temperature of the
regulator than the actual current sharing. Consequently, you
can opt to lay the board out so that R1 picks up the thermal rise of U1. On top of that, you can also have R2 sense the temperature
of U2.
Particularly for a multi-layer board, you can put the ballast
resistors on the layer immediately below the regulator and pick
up a big percentage of the temperature rise caused by the regulator.
And, you can make the ballast resistance traces relatively
large so that their self-heating is small and their thermal rise is
mostly caused by picking up the rise from the nearby regulator.
Thus, you can see that when it comes to ballasting, the high
temperature coefficient of resistance of copper trace is an
advantage, not a liability.