| |
 There
are two kinds of digital integrated circuits in common use. There are more
varieties, some quite esoteric, but these two handle most of the load, and
are the only ones easily available for study and use. The two kinds are
bipolar ("TTL") and CMOS, and both are about as old as integrated circuits
themselves. Their circuit properties are quite different, although modern
devices can actually be used together (this is not usually a good
practice, however). Bipolar devices, somewhat inappropriately called TTL
(transistor-transistor logic), early became the most popular choice,
together with NMOS large-scale devices that imitated TTL properties. CMOS
with its rugged flakiness was used when power consumption had to be kept
to a minimum,
and for certain devices where its peculiarities were valuable, as in
ripple counters and the 4046, for example. The TTL integrated circuits
were given numbers 74XX, while the CMOS circuits had numbers 4XXX.
The two states of TTL logic were a low state near 0 V produced by a
saturated transistor switch, and a high state of rather indefinite
voltage, provided either by a pull-up resistor, or by the
 "totem-pole"
output of a circuit, which pulled up the output near 5 V. These two states
are very easily distinguished, and insensitive to noise, especially the
low state. The effect of noise depends not only on the voltage outputs,
but on the output resistance as well, so comparison of voltage levels is
not a reliable guide. 4000 CMOS suffered from high output resistance and
high input resistance as well, which allowed small extraneous charges to
cause problems. However, the saturated transistors of TTL took a lot of
current, and also a lot of time to get out of saturation (though they were
much faster than CMOS).
TTL was modified by a Schottky diode between collector and base that
prevented transistors from going into deep saturation, saving both current
and time. The circuits were also redesigned to use less current, and were
called LS, or low-power Schottky. The numbers were now 74LSXX. There have
been further developments (e.g., "fast" or 74FXX, "low-power" or 74LXX, "Schottky"
or 74SXX--same power as TTL but much faster, etc.) but LS is excellent,
inexpensive and superior for general use.
CMOS was also greatly improved by giving the outputs a lower resistance,
and making the circuits work more like TTL. The result was the 74HCXX
series, which has the same functions as TTL, even the same pinouts, and
can replace TTL in most applications. There is a variation in which the
inputs are made to mimic TTL inputs, called HCT, with part numbers
74HCTXX. These are occasionally useful, because of the wide use of LSTTL,
but have no advantage over the normal HC chips, which should usually be
chosen. Although HC chips resemble TTL in look and function, they are true
CMOS, with the peculiarities of that family.
It's important to understand the characteristics of the
inputs and outputs of the LS and HC families, which are quite different.
The HC family is in many ways simpler, since the inputs carry only a very
small leakage current, less than a microampere, no
 matter
what is done to them. They have a capacitance of 10 pF, which governs the
amount of charge that has to be moved to change their states. The outputs
provide ("source") current when high, and accept ("sink") current when
low. When high, they act like a small resistance connected to the positive
supply, and when low like a small resistance connected to GND. The small
resistance is around 50Ω, and the high and low states are relatively
symmetrical. An HC output can be made to give or accept 20 mA, though this
is something of a strain. They will easily drive an LED at 10 mA or so
whether high or low.
An LS input handles current. When you hold it high, it wants
something around 20 μA, and when you pull it low, you have to accept 0.4
mA. If you leave the input disconnected, or "open," it can't get the
current it wants at any voltage, and acts like a high or H state. For
experiments, one can simply leave unused inputs alone, and they will act
as highs. In actual circuits, it is best to connect unused inputs to +5 or
GND by wires. Some people use 4.7k resistors instead of a plain wire to
pull up inputs, but it really doesn't make much difference with LS. With
HC, inputs MUST be given a definite state. If left alone, they will float
to some intermediate value and now be one thing and now another. In 4000
CMOS, they put things in an intermediate state where the circuits drew
excessive power. With HC, I have not noticed this, but the uncertainty of
state still remains. When experimenting with HC, it is good practice to
wire all unused inputs to +5 or GND.
An LS output is also different from an HC output. It pulls up
to something over 3.5 V, but usually not to +5 if there is any load at
all. Most LS chips give around 4.7 V when high. In this state, it will not
furnish more than about 0.4 mA without dropping below 3.5 V. The pull down
is quite different. Here the output will absorb up to 8 mA without rising
above 0.5 V. This is the advertised value, and most chips will do even
better. There is no problem sinking 20 mA if you can stand the output
voltage's rising a bit. (the HC's output will be up to 1 V by this time).
The output is, therefore, asymmetric, as are the inputs. One LS output
will handle about 20 LS inputs, as far as static current is concerned. The
characteristics of LS and HC inputs and outputs are compared in the figure
below.
The number of HC inputs that can be supplied from one HC
output (the "fan-out" as it was once called) is not limited by current,
but by capacitance and the speed required. 10 HC inputs have about 100 pF
capacitance, and the output has a resistance of about 50ohms, so the time
constant is 5 ns, which is fast enough for most purposes. The inputs of
both LS and HC do not like voltages that change too slowly. A rise time of
over 500 ns can lead to problems, especially with complex chips. Although
we study these chips with DC voltages, they are actually high-powered
racing cars, not family sedans, and require special handling if they are
to work at their accustomed speed.
The rise and fall times of inputs and outputs should not be
confused with the propagation time of a signal across the chip from input
to output. Propagation times for inverters are 15-20 ns for either LS or
HC, which is fairly typical of digital logic.
The supply voltage for LS is 5.00 V (4.75 V to 5.25 V), and
must be held within this range by a properly-regulated power supply. A 5.1
V Zener can be used for this purpose. The supply voltage for HC is much
more flexible. It can be anything between 2 and 6 V (7V is the maximum,
and it is not good to shave this too closely). It works satisfactorily on
3 V, but is normally used with the same 5 V supply as for LS. HC draws
very little operating current when operating at slow speeds, only
microamperes, so it is the obvious choice for battery-powered devices. |
|
|
