🚀 go-pugleaf

In the nine responses I've seen here, the closest I've seen to my
favorite solution to interfacing TTL to bipolar transistors is
reference to the concept that TTL generally sinks current much
better than it sources it.  First, I would have used a 2N4401 or
2N4403 instead of a 2N2222/2N3904.  The 4400-4403 series is
designed to handle currents much higher than the 2222 et al.  As
I recall, the 4400 series betas peak at about 150 mA collector
current.  They are about as cheap as 3904s.

Second, I would look for a way to turn the transistors on when the
TTL output is low:  to do this, consider TTL output to R to
base of PNP.  Emitter of PNP to +5.  Resistor shunting base to
emitter.  Proper choice of resistors can guarantee turnoff and
substantial base current to the PNP in the on state.  The PNP
can then drive a load directly, or drive the base of an NPN
power transistor, at in excess of 100 mA easily, and that's
assuming very low betas.

I'd generally try to use an integrated power driver in a commercial
application, but for hobbiest hacking where it's what's in the
junkbox that counts, I've become quite fond of the ttl --> pnp
solution; it uses the TTL the way it was intended w.r.t. output
currents.

Another note: several folk suggested a Darlington connection.
Be aware that the drop across a Darlington won't get much lower
than a volt for silicon devices at room temp and moderate current.
A cleaner output (lower drop) results from running the first of the
two transistors from, say, +5 on its collector; ya gotta limit current
somehow, probably most easily with a resistor from first emitter to
second base.  But the problem with all this sort of thing is that the
output voltage vs current for TTL in the high state is not well-controlled.
They guarantee some minimums, but your circuit should be able to
handle TTL output anywhere between Voh(min) and Vcc when the current
out is no more than the Ioh rating.  By contrast, the Vol range is
very limited, and the available current is much higher.  Thus, my
TTL -- pnp -- npn solution has the same nominal available net current
gain as a ttl -- darlington, but has much more available output current,
since it's "on" when the TTL output is "low".  Especially if this is
going to be used in an automotive environment, you should design for
the extremes you will likely encounter.