8051 Tutorial: Timers

The 8051 comes equipped with two timers, both of which may be controlled, set, read, and configured
individually. The 8051 timers have three general functions: 1) Keeping time and/or calculating the
amount of time between events, 2) Counting the events themselves, or 3) Generating baud rates for the
serial port.

The three timer uses are distinct so we will talk about each of them separately. The first two uses will
be discussed in this chapter while the use of timers for baud rate generation will be discussed in the
chapter relating to serial ports.

How does a timer count?

How does a timer count? The answer to this question is very simple: A timer always counts up. It
doesn’t matter whether the timer is being used as a timer, a counter, or a baud rate generator: A timer is
always incremented by the microcontroller.

Programming Tip: Some derivative chips actually allow the program to configure whether
the timers count up or down. However, since this option only exists on some derivatives it is
beyond the scope of this tutorial which is aimed at the standard 8051. It is only mentioned
here in the event that you absolutely need a timer to count backwards, you will know that you
may be able to find an 8051-compatible microcontroller that does it.
USING TIMERS TO MEASURE TIME

Obviously, one of the primary uses of timers is to measure time. We will discuss this use of timers first
and will subsequently discuss the use of timers to count events. When a timer is used to measure time it
is also called an "interval timer" since it is measuring the time of the interval between two events.

How long does a timer take to count?

First, it’s worth mentioning that when a timer is in interval timer mode (as opposed to event counter
mode) and correctly configured, it will increment by 1 every machine cycle. As you will recall from the
previous chapter, a single machine cycle consists of 12 crystal pulses. Thus a running timer will be
incremented:

11,059,000 / 12 = 921,583
921,583 times per second. Unlike instructions--some of which require 1 machine cycle, others 2, and
others 4--the timers are consistent: They will always be incremented once per machine cycle. Thus if a
timer has counted from 0 to 50,000 you may calculate:
50,000 / 921,583 = .0542
.0542 seconds have passed. In plain English, about half of a tenth of a second, or one-twentieth of a
second.

Obviously it’s not very useful to know .0542 seconds have passed. If you want to execute an event
once per second you’d have to wait for the timer to count from 0 to 50,000 18.45 times. How can you
wait "half of a time?" You can’t. So we come to another important calculation.

Let’s say we want to know how many times the timer will be incremented in .05 seconds. We can do
simple multiplication:

.05 * 921,583 = 46,079.15.

This tells us that it will take .05 seconds (1/20th of a second) to count from 0 to 46,079. Actually, it
will take it .049999837 seconds--so we’re off by .000000163 seconds--however, that’s close enough
for government work. Consider that if you were building a watch based on the 8051 and made the
above assumption your watch would only gain about one second every 2 months. Again, I think that’s
accurate enough for most applications--I wish my watch only gained one second every two months!
Obviously, this is a little more useful. If you know it takes 1/20th of a second to count from 0 to 46,079
and you want to execute some event every second you simply wait for the timer to count from 0 to
46,079 twenty times; then you execute your event, reset the timers, and wait for the timer to count up
another 20 times. In this manner you will effectively execute your event once per second, accurate to
within thousandths of a second.

Thus, we now have a system with which to measure time. All we need to review is how to control the
timers and initialize them to provide us with the information we need.

Timer SFRs

As mentioned before, the 8051 has two timers which each function essentially the same way. One timer
is TIMER0 and the other is TIMER1. The two timers share two SFRs (TMOD and TCON) which
control the timers, and each timer also has two SFRs dedicated solely to itself (TH0/TL0 and
TH1/TL1).

We’ve given SFRs names to make it easier to refer to them, but in reality an SFR has a numeric
address. It is often useful to know the numeric address that corresponds to an SFR name. The SFRs
relating to timers are:

SFR Name Description SFR Address

TH0 Timer 0 High Byte 8Ch
TL0 Timer 0 Low Byte 8Ah
TH1 Timer 1 High Byte 8Dh
TL1 Timer 1 Low Byte 8Bh
TCON Timer Control 88h
TMOD Timer Mode 89h

When you enter the name of an SFR into an assembler, it internally converts it to a number. For
example, the command:

MOV TH0,#25h
moves the value 25h into the TH0 SFR. However, since TH0 is the same as SFR address 8Ch this
command is equivalent to:
MOV 8Ch,#25h
Now, back to the timers. First, let’s talk about Timer 0.

Timer 0 has two SFRs dedicated exclusively to itself: TH0 and TL0. Without making things too
complicated to start off with, you may just think of this as the high and low byte of the timer. That is to
say, when Timer 0 has a value of 0, both TH0 and TL0 will contain 0. When Timer 0 has the value
1000, TH0 will hold the high byte of the value (3 decimal) and TL0 will contain the low byte of the
value (232 decimal). Reviewing low/high byte notation, recall that you must multiply the high byte by
256 and add the low byte to calculate the final value. That is to say:

TH0 * 256 + TL0 = 1000

3 * 256 + 232 = 1000
Timer 1 works the exact same way, but it’s SFRs are TH1 and TL1.

Since there are only two bytes devoted to the value of each timer it is apparent that the maximum value
a timer may have is 65,535. If a timer contains the value 65,535 and is subsequently incremented, it
will reset--or overflow--back to 0.

The TMOD SFR

Let’s first talk about our first control SFR: TMOD (Timer Mode). The TMOD SFR is used to control
the mode of operation of both timers. Each bit of the SFR gives the microcontroller specific
information concerning how to run a timer. The high four bits (bits 4 through 7) relate to Timer 1
whereas the low four bits (bits 0 through 3) perform the exact same functions, but for timer 0.

The individual bits of TMOD have the following functions:

TMOD (89h) SFR

Bit Name Explanation of Function Timer
When this bit is set the timer will only run when INT1 (P3.3)
7 GATE1 is high. When this bit is clear the timer will run regardless of 1
the state of INT1.
When this bit is set the timer will count events on T1 (P3.5).
6 C/T1 When this bit is clear the timer will be incremented every 1
machine cycle.
5 T1M1 Timer mode bit (see below) 1
4 T1M0 Timer mode bit (see below) 1
When this bit is set the timer will only run when INT0 (P3.2)
3 GATE0 is high. When this bit is clear the timer will run regardless of 0
the state of INT0.
When this bit is set the timer will count events on T0 (P3.4).
2 C/T0 When this bit is clear the timer will be incremented every 0
machine cycle.
1 T0M1 Timer mode bit (see below) 0
0 T0M0 Timer mode bit (see below) 0
As you can see in the above chart, four bits (two for each timer) are used to specify a mode of
operation. The modes of operation are:
TxM1 TxM0 Timer Mode Description of Mode
0 0 0 13-bit Timer.
0 1 1 16-bit Timer
1 0 2 8-bit auto-reload
1 1 3 Split timer mode

13-bit Time Mode (mode 0)

Timer mode "0" is a 13-bit timer. This is a relic that was kept around in the 8051 to maintain
compatability with it’s predecesor, the 8048. Generally the 13-bit timer mode is not used in new
development.

When the timer is in 13-bit mode, TLx will count from 0 to 31. When TLx is incremented from 31, it
will "reset" to 0 and increment THx. Thus, effectively, only 13 bits of the two timer bytes are being
used: bits 0-4 of TLx and bits 0-7 of THx. This also means, in essence, the timer can only contain 8192
values. If you set a 13-bit timer to 0, it will overflow back to zero 8192 machine cycles later.

Again, there is very little reason to use this mode and it is only mentioned so you won’t be surprised if
you ever end up analyzing archaeic code which has been passed down through the generations (a
generation in a programming shop is often on the order of about 3 or 4 months).

16-bit Time Mode (mode 1)

Timer mode "1" is a 16-bit timer. This is a very commonly used mode. It functions just like 13-bit
mode except that all 16 bits are used.
TLx is incremented from 0 to 255. When TLx is incremented from 255, it resets to 0 and causes THx to
be incremented by 1. Since this is a full 16-bit timer, the timer may contain up to 65536 distinct values.
If you set a 16-bit timer to 0, it will overflow back to 0 after 65,536 machine cycles.

8-bit Time Mode (mode 2)

Timer mode "2" is an 8-bit auto-reload mode. What is that, you may ask? Simple. When a timer is in
mode 2, THx holds the "reload value" and TLx is the timer itself. Thus, TLx starts counting up. When
TLx reaches 255 and is subsequently incremented, instead of resetting to 0 (as in the case of modes 0
and 1), it will be reset to the value stored in THx.

For example, let’s say TH0 holds the value FDh and TL0 holds the value FEh. If we were to watch the
values of TH0 and TL0 for a few machine cycles this is what we’d see:

Machine Cycle TH0 Value TL0 Value

1 FDh FEh
2 FDh FFh
3 FDh FDh
4 FDh FEh
5 FDh FFh
6 FDh FDh
7 FDh FEh

As you can see, the value of TH0 never changed. In fact, when you use mode 2 you almost always set
THx to a known value and TLx is the SFR that is constantly incremented.

What’s the benefit of auto-reload mode? Perhaps you want the timer to always have a value from 200
to 255. If you use mode 0 or 1, you’d have to check in code to see if the timer had overflowed and, if
so, reset the timer to 200. This takes precious instructions of execution time to check the value and/or
to reload it. When you use mode 2 the microcontroller takes care of this for you. Once you’ve
configured a timer in mode 2 you don’t have to worry about checking to see if the timer has overflowed
nor do you have to worry about resetting the value--the microcontroller hardware will do it all for you.

The auto-reload mode is very commonly used for establishing a baud rate which we will talk more
about in the Serial Communications chapter.

Split Timer Mode (mode 3)

Timer mode "3" is a split-timer mode. When Timer 0 is placed in mode 3, it essentially becomes two
separate 8-bit timers. That is to say, Timer 0 is TL0 and Timer 1 is TH0. Both timers count from 0 to
255 and overflow back to 0. All the bits that are related to Timer 1 will now be tied to TH0.

While Timer 0 is in split mode, the real Timer 1 (i.e. TH1 and TL1) can be put into modes 0, 1 or 2
normally--however, you may not start or stop the real timer 1 since the bits that do that are now linked
to TH0. The real timer 1, in this case, will be incremented every machine cycle no matter what.

The only real use I can see of using split timer mode is if you need to have two separate timers and,
additionally, a baud rate generator. In such case you can use the real Timer 1 as a baud rate generator
and use TH0/TL0 as two separate timers.

The TCON SFR

Finally, there’s one more SFR that controls the two timers and provides valuable information about
them. The TCON SFR has the following structure:
 TCON (88h) SFR
Bit
Bit Name Explanation of Function Timer
Address
Timer 1 Overflow. This bit is set by the microcontroller when
7 TF1 8Fh 1
Timer 1 overflows.
Timer 1 Run. When this bit is set Timer 1 is turned on. When
6 TR1 8Eh 1
this bit is clear Timer 1 is off.
Timer 0 Overflow. This bit is set by the microcontroller when
5 TF0 8Dh 0
Timer 0 overflows.
Timer 0 Run. When this bit is set Timer 0 is turned on. When
4 TR0 8Ch 0
this bit is clear Timer 0 is off.
As you may notice, we’ve only defined 4 of the 8 bits. That’s because the other 4 bits of the SFR don’t
have anything to do with timers--they have to do with Interrupts and they will be discussed in the
chapter that addresses interrupts.

A new piece of information in this chart is the column "bit address." This is because this SFR is "bit-
addressable." What does this mean? It means if you want to set the bit TF1--which is the highest bit of
TCON--you could execute the command:

MOV TCON, #80h

... or, since the SFR is bit-addressable, you could just execute the command:
SETB TF1
This has the benefit of setting the high bit of TCON without changing the value of any of the other bits
of the SFR. Usually when you start or stop a timer you don’t want to modify the other values in TCON,
so you take advantage of the fact that the SFR is bit-addressable.

Initializing a Timer

Now that we’ve discussed the timer-related SFRs we are ready to write code that will initialize the
timer and start it running.

As you’ll recall, we first must decide what mode we want the timer to be in. In this case we want a 16-
bit timer that runs continuously; that is to say, it is not dependent on any external pins.

We must first initialize the TMOD SFR. Since we are working with timer 0 we will be using the lowest
4 bits of TMOD. The first two bits, GATE0 and C/T0 are both 0 since we want the timer to be
independent of the external pins. 16-bit mode is timer mode 1 so we must clear T0M1 and set T0M0.
Effectively, the only bit we want to turn on is bit 0 of TMOD. Thus to initialize the timer we execute
the instruction:

MOV TMOD,#01h
Timer 0 is now in 16-bit timer mode. However, the timer is not running. To start the timer running we
must set the TR0 bit We can do that by executing the instruction:
SETB TR0
Upon executing these two instructions timer 0 will immediately begin counting, being incremented
once every machine cycle (every 12 crystal pulses).

Reading the Timer

There are two common ways of reading the value of a 16-bit timer; which you use depends on your
specific application. You may either read the actual value of the timer as a 16-bit number, or you may
simply detect when the timer has overflowed.

Reading the value of a Timer

If your timer is in an 8-bit mode--that is, either 8-bit AutoReload mode or in split timer mode--then
reading the value of the timer is simple. You simply read the 1-byte value of the timer and you’re done.

However, if you’re dealing with a 13-bit or 16-bit timer the chore is a little more complicated. Consider
what would happen if you read the low byte of the timer as 255, then read the high byte of the timer as
15. In this case, what actually happened was that the timer value was 14/255 (high byte 14, low byte
255) but you read 15/255. Why? Because you read the low byte as 255. But when you executed the
next instruction a small amount of time passed--but enough for the timer to increment again at which
time the value rolled over from 14/255 to 15/0. But in the process you’ve read the timer as being
15/255. Obviously there’s a problem there.

The solution? It’s not too tricky, really. You read the high byte of the timer, then read the low byte,
then read the high byte again. If the high byte read the second time is not the same as the high byte read
the first time you repeat the cycle. In code, this would appear as:

REPEAT: MOV A,TH0

MOV R0,TL0
CJNE A,TH0,REPEAT
...
In this case, we load the accumulator with the high byte of Timer 0. We then load R0 with the low byte
of Timer 0. Finally, we check to see if the high byte we read out of Timer 0--which is now stored in the
Accumulator--is the same as the current Timer 0 high byte. If it isn’t it means we’ve just "rolled over"
and must reread the timer’s value--which we do by going back to REPEAT. When the loop exits we
will have the low byte of the timer in R0 and the high byte in the Accumulator.

Another much simpler alternative is to simply turn off the timer run bit (i.e. CLR TR0), read the timer
value, and then turn on the timer run bit (i.e. SETB TR0). In that case, the timer isn’t running so no
special tricks are necessary. Of course, this implies that your timer will be stopped for a few machine
cycles. Whether or not this is tolerable depends on your specific application.

Detecting Timer Overflow

Often it is necessary to just know that the timer has reset to 0. That is to say, you are not particularly
interest in the value of the timer but rather you are interested in knowing when the timer has
overflowed back to 0.

Whenever a timer overflows from it’s highest value back to 0, the microcontroller automatically sets
the TFx bit in the TCON register. This is useful since rather than checking the exact value of the timer
you can just check if the TFx bit is set. If TF0 is set it means that timer 0 has overflowed; if TF1 is set
it means that timer 1 has overflowed.

We can use this approach to cause the program to execute a fixed delay. As you’ll recall, we calculated
earlier that it takes the 8051 1/20th of a second to count from 0 to 46,079. However, the TFx flag is set
when the timer overflows back to 0. Thus, if we want to use the TFx flag to indicate when 1/20th of a
second has passed we must set the timer initially to 65536 less 46079, or 19,457. If we set the timer to
19,457, 1/20th of a second later the timer will overflow. Thus we come up with the following code to
execute a pause of 1/20th of a second:

MOV TH0,#76;High byte of 19,457 (76 * 256 = 19,456)

MOV TL0,#01;Low byte of 19,457 (19,456 + 1 = 19,457)
MOV TMOD,#01;Put Timer 0 in 16-bit mode
SETB TR0;Make Timer 0 start counting
JNB TF0,$;If TF0 is not set, jump back to this same instruction
In the above code the first two lines initialize the Timer 0 starting value to 19,457. The next two
instructions configure timer 0 and turn it on. Finally, the last instruction JNB TF0,$, reads "Jump, if
TF0 is not set, back to this same instruction." The "$" operand means, in most assemblers, the address
of the current instruction. Thus as long as the timer has not overflowed and the TF0 bit has not been set
the program will keep executing this same instruction. After 1/20th of a second timer 0 will overflow,
set the TF0 bit, and program execution will then break out of the loop.

Timing the length of events

The 8051 provides another cool toy that can be used to time the length of events.

For example, let's say we're trying to save electricity in the office and we're interested in how long a
light is turned on each day. When the light is turned on, we want to measure time. When the light is
turned off we don't. One option would be to connect the lightswitch to one of the pins, constantly read
the pin, and turn the timer on or off based on the state of that pin. While this would work fine, the 8051
provides us with an easier method of accomplishing this.

Looking again at the TMOD SFR, there is a bit called GATE0. So far we've always cleared this bit
because we wanted the timer to run regardless of the state of the external pins. However, now it would
be nice if an external pin could control whether the timer was running or not. It can. All we need to do
is connect the lightswitch to pin INT0 (P3.2) on the 8051 and set the bit GATE0. When GATE0 is set
Timer 0 will only run if P3.2 is high. When P3.2 is low (i.e., the lightswitch is off) the timer will
automatically be stopped.

Thus, with no control code whatsoever, the external pin P3.2 can control whether or not our timer is
running or not.

USING TIMERS AS EVENT COUNTERS

We've discussed how a timer can be used for the obvious purpose of keeping track of time. However,
the 8051 also allows us to use the timers to count events.

How can this be useful? Let's say you had a sensor placed across a road that would send a pulse every
time a car passed over it. This could be used to determine the volume of traffic on the road. We could
attach this sensor to one of the 8051's I/O lines and constantly monitor it, detecting when it pulsed high
and then incrementing our counter when it went back to a low state. This is not terribly difficult, but
requires some code. Let's say we hooked the sensor to P1.0; the code to count cars passing would look
something like this:

JNB P1.0,$ ;If a car hasn't raised the signal, keep waiting
JB P1.0,$ ;The line is high which means the car is on the sensor right now
INC COUNTER ;The car has passed completely, so we count it
As you can see, it's only three lines of code. But what if you need to be doing other processing at the
same time? You can't be stuck in the JNB P1.0,$ loop waiting for a car to pass if you need to be doing
other things. Of course, there are ways to get around even this limitation but the code quickly becomes
big, complex, and ugly.

Luckily, since the 8051 provides us with a way to use the timers to count events we don't have to
bother with it. It is actually painfully easy. We only have to configure one additional bit.

Let's say we want to use Timer 0 to count the number of cars that pass. If you look back to the bit table
for the TCON SFR you will there is a bit called "C/T0"--it's bit 2 (TCON.2). Reviewing the
explanation of the bit we see that if the bit is clear then timer 0 will be incremented every machine
cycle. This is what we've already used to measure time. However, if we set C/T0 timer 0 will monitor
the P3.4 line. Instead of being incremented every machine cycle, timer 0 will count events on the P3.4
line. So in our case we simply connect our sensor to P3.4 and let the 8051 do the work. Then, when we
want to know how many cars have passed, we just read the value of timer 0--the value of timer 0 will
be the number of cars that have passed.

So what exactly is an event? What does timer 0 actually "count?" Speaking at the electrical level, the
8051 counts 1-0 transitions on the P3.4 line. This means that when a car first runs over our sensor it
will raise the input to a high ("1") condition. At that point the 8051 will not count anything since this is
a 0-1 transition. However, when the car has passed the sensor will fall back to a low ("0") state. This is
a 1-0 transition and at that instant the counter will be incremented by 1.

It is important to note that the 8051 checks the P3.4 line each instruction cycle (12 clock cycles). This
means that if P3.4 is low, goes high, and goes back low in 6 clock cycles it will probably not be
detected by the 8051. This also means the 8051 event counter is only capable of counting events that
occur at a maximum of 1/24th the rate of the crystal frequency. That is to say, if the crystal frequency
is 12.000 Mhz it can count a maximum of 500,000 events per second (12.000 Mhz * 1/24 = 500,000).
If the event being counted occurs more than 500,000 times per second it will not be able to be
accurately counted by the 8051.