Serial I-O for 8051

Dinesh K. Sharma
Electrical Engineering Department
I.I.T. Bombay
Mumbai 400 076
(version 14/10/07)

Motivation

Serial communications means sending data a single bit at a time. But why not send the data
in parallel? Why bother with parallel to serial and serial to paralle conversions?
There are several situations where serial communication is preferable. If the transmitter
and the receiver are remote, it is expensive to lay a parallel cable with multiple wires in
it. Also, it may be difficult to ensure that all bits on a parallel bus have the same delay.
If the difference in delay is comparable to the data rate, a late bit from one transmission
might arrive at the same time as an early bit from the next. The transmission will then get
garbled. Also, if we are using an inherently single wire medium such as a phone line, serial
communication would be required.
How do we send the data serially? At the transmitter, this involves loading data in
parallel to a shift register, and shifting out a bit at a time. At the receiving end, we have
another shift register, which receives this data a single bit at a time and when all bits are
received, the shift register can be read in parallel.
From Transmitter

To Receiver

Shift Register

Shift Register
Serial Line

The transmitter and the receiver must agree on the order in which the bits will be sent. It
is common to send the lease significant bit first.

Synchronization

The picture presented above is somewhat oversimplified. Obviously, the rate at which the
transmitter shifts the data out must be matched to the rate at which the receivers shifts the
data in. Even if the clock frequencies at the transmitter and receiver are perfectly matched,
the phase of the clock at the receiver should be carefully adjusted so that we sample the serial
line when it is stable and not when the data on it is changing. How can this be ensured?

2.1

Synchronous Serial I-O

One way would be for the transmitter to send the data through one wire and the shift clock
through another. The receiver would then use this same clock for shifting the data in.

From Transmitter

To Receiver
Data

Shift Register

Shift Register
Shift Clock

If the transmitter places data on the positive edge and the receiver samples the data at the
negative edge, the data will be stable at the sampling instant. However, this doubles the
wire cost. Therefore, this solution is practical only if the transmitter and the receiver are
located close to each other. This way of sending serial data is called synchronous serial IO.
It is useful for sending data over short distances. The advantage of this method is that it
can achieve very high data rates. The 8051 provides support for this method through one of
the modes defined for its serial IO (mode 0).

2.2

Asynchronous Serial I-O

If the transmitter and the receiver are physically remote, they must use independent shift
clock generators, set nominally to the same frequency. In this case, their clock frequencies
will invariable have a small but non-zero mismatch. Unfortunately, this means that with
time, the sampling time will drift further and further away from the ideal sampling instant.
When the sampling time has shifted by more than a half period, the receiver might sample
the wrong bit on the serial line, leading to errors. Obviously some means of synchronization
must be provided to avoid this and to ensure reliable transfer of data.
Since we dont want to add any more wires, it is clear that the wire carrying data is
the one which must carry the extra information required for synchronization. This involves
sending extra bits on the wire along with the data. The data bits along with the added bits
constitute a frame
To synchronize two clocks, we need to create an edge. Serial transmission often uses non
return to zero or NRZ encoding to minimize noise and power dissipation. If successive bits
are identical, no edge may be generated by the data itself. To ensure that an edge occurs at
the start of every frame, we add a bit before and a bit after the useful data. These bits are
chosen to be different in value, thus ensuring that each frame starts with a value different
from the one with which the previous one ended. These are called start and stop bits. It is
common to use a 1 for a stop bit and a 0 for a start bit. Then each frame begins with a
negative edge.

Write instants

Start

1
1
0
Serial Data

Stop

T/2
T

Sampling Instants
Each new bit is placed on the serial line after time T. (This is called bit time and the corresponding frequency the baud rate). Notice that the first sampling instant is offset from the
writing instant by half a bit time. This is to ensure that a bit is sampled in the middle of
the bit interval T, where it is stable. After sampling the start bit, each bit is sampled a time
T after the previous bit.

Often a Parity bit is also added to the data. The transmitter and the receiver agree to
use either even or odd parity. The parity bit is chosen by the transmitter to be a 0 or a
1 so that the total number of bits in a frame is always even (or always odd). The receiver
counts the number of 1s and ensures that the count is even (or odd) as previously agreed
with the transmitter. If it is not, it can signal an error and the software can take appropriate
action (such as a request for re-transmission). The use of parity bit is optional.
The receiver now knows that a frame will always begin with an edge. It uses a locally
generated clock for synchronization and timing. Because of small mismatches in transmit
and receive clocks, the start of frame edge on the serial data may not coincide with an edge
on the locally generated clock. We take the next edge of the local clock as the effective start
of frame. The error in timing due to this could be as much as a whole cycle of the local clock.
Obviously, the local clock should not operate at the data rate - a timing error of 1 bit time
will then shift the data by 1 bit!
Therefore, the locally generated clock operates at a frequency much higher than the bit
rate. This high frequency clock is divided down to get the shift clock. For example, the
locally generated clock might operate at 32 times the shift clock. A 5 bit counter will be
used to divide the local clock. The worst case error in timing will now be one cycle of the
locally generated clock. Since it is running at a frequency much higher than the shift clock,
this error will be small.
Frame Detect
T/2 T T

Sampling Instants
T
T
T
T

1
0
start

b0 b1
(LSB)

b2

b3

Time

b4

b5

b6

b7 stop
(MSB)

Synchronization occurs at the beginning of each frame. Therefore timing errors do not
accumulate from frame to frame. The counter used to divide the local clock is reset when
the start of frame edge is detected. In the example that we have chosen (internal clock =
32X bit rate), we reset the 5 bit counter when we detect the frame edge. Now, after 16 cycles
of the internal clock, (corresponding to time T/2 where T is the interval between bits) the
start bit is sampled. If it is not 0, a framing error has occurred and should be handled in
software say by a request to re-transmit. If the start bit is correctly sampled, data bits
will be sampled every 32 counts, till all data bits have been sampled. (This might be 8 bits
if no parity bit is appended and 9 bits if parity bit is used). The next bit must be the stop
bit (normally a 1). If it is not the correct value, a frame error has occurred and must be
handled appropriately. If the stop bit is correctly received, we accept the data and wait for
the next start of frame.
Notice that this scheme will work equally well, if additional delay is inserted between the
end of one frame and the beginning of the next. After the last bit has been sampled, no
further timing needs to be performed. The serial line idles at the stop bit (normally 1).
The counter is reset to zero whenever the next frame edge is detected. Therefore it does not
matter how much time elapses between the end of a frame and the beginning of the next.

2.3

Serial IO protocol

Before serial communications can be carried out, the transmitter and the receiver must agree
on a number of parameters.
Whether synchronous or asynchronous serial communication will be used.
3

 At what rate will the bits be sent.

How many bits will be sent at a time (frame size).
Whether parity will be used and if so, which kind of parity (even or odd) will be used.
Whether simultaneous transmission and reception (duplex) will be possible or only one
of these will be carried out at a time (simplex).
These parameters constitute the serial protocol. A circuit which implements asynchronous
serial communication protocol is known as a Universal Asynchronous Receiver Transmitter
(UART). If it can also carry out synchronous serial communication, it is called a USART
(Universal Synchronous Asynchronous Receiver Transmitter). Microprocessors often use dedicated chips like the Intel 8251 USART for implementing serial communication. The 8051
has a built in USART.

Implementation of Serial Port in 8051

The 8051 architecture is designed to make sending and receiving serial data quite simple. It
is capable of duplex communication, that is serial transmission and reception can be carried
out simultaneously. Pins of port 3 perform a dual function. P3.0 can also serve as the serial
data input (RxD) and P3.1 can serve as the serial data output (TxD) when serial IO has
been enabled.
The 8051 contains two registers, one for receiving and the other for sending serial data.
Both of these respond to the same address (99H) in the special function register area. (The
assembler aliases the address 99H to the name Sbuf). The receive register places data on the
internal bus when a read is performed from Sbuf (99H), while the transmit register receives
the data when a write operation is carried out to this address.
Clock Control

TxD

Shift Reg.

Sbuf

RxD

Sbuf

Wr

Rd
Internal Data Bus

The transmit register itself acts as the shift register for outputting serial data through the
TxD pin (P3.1). When the data has been shifted out, a flag called TI is set, inviting the
processor to write fresh data to Sbuf. The serial data input pin RxD (P3.0) is connected to an
independent shift register, whose output is transferred to the receive buffer after a complete
frame has been shifted in. A flag (RI) is set at this time to inform the processor that serial
data is available for reading at Sbuf. Flags RI and TI are parts of a bit addressable register
SCON, which resides at address 98H in the special function register area. SCON is used for
configuring serial communication.

3.1

Configuring the Serial Port

Configuration of the serial port involves choosing the mode of communication, deciding rate
of sending data and managing the parity bit etc. Much of this work is done by the special
function register SCON at byte address 98H in the SFR area. SCON is bit addressable.
Bit No.
Bit Addr
Bit Name

SCON register at BYTE address 98H

7
6
5
4
3
2
9F
9E
9D
9C
9B
9A
SM0 SM1 SM2 REN TB8 RB8
4

1
99
TI

0
98
RI

As described before, two physically distinct registers responding to the byte address 99H are
used as transmit and receive buffers during serial IO. In addition to these resources, the most
significant bit of the PCON register at address 87H (known as SMOD) is used for doubling
the baud rate when set.
The upper nibble of SCON configures the operation of the serial IO, while the lower nibble
is used for auxiliary data arising per frame during communication. SM0 and SM1 decide the
mode of communication. (Note the somewhat non-standard convention of calling the more
significant bit SM0 and the less significant bit SM1). SM0 determines if 8bits of data will be
sent per frame (SM0=0) or 9 (when SM0=1). 9 bits are used when we want to send a parity
bit in addition to 8bit data. 9 bit communication is also used in a special way with SM2
during multicast communication. SM1 determines if fixed or variable baud rate will be used.
SM2 is used for multicast transmission and will be explained later. In normal use, it should
be kept cleared. REN must be set to enable reception of data on the RxD line. (What would
happen if serial reception is permanently enabled?)
If we use a parity bit, we need to send and receive 9 bits of data along with the start
and stop bits in every frame. Sbuf can hold only 8 bits each in the send and receive buffers.
Where can the 9th bit be accommodated? Bits TB8 and RB8 in SCON are used to accommodate the 9th bit during transmission and reception respectively. When all bits of a
frame have been received serially, the hardware in 8051 sets the flag RI. This is an invitation
to the processor to read Sbuf (and if necessary, RB8). Similarly, TI is set when the frame
corresponding to the data written to Sbuf (and if necessary, TB8) have been shifted out of
the TxD line. This is an invitation to write fresh data to Sbuf if more data needs to be
sent out on the serial line. Notice that TI and RI are not automatically cleared and must be
cleared by software which reads/writes Sbuf. It is possible to enable a special interrupt called
Serial Interrupt. This is triggered by the OR of TI and RI. The interrupt service routine is
supposed to examine the flags TI and RI and take appropriate action accordingly.

3.2

Serial IO modes

SM0 and SM1 determine the mode of serial IO.

Mode 0 This mode corresponds to SM0=0 and SM1=0. This is a Shift Register mode:
Data is transmitted as well as received through the RxD pin and the shift clock is
supplied through the TxD pin. Since shift clock is supplied separately, there is no need
for a start and a stop bit. 8 bits of data are transmitted or received (SM0 = 0) at a
fixed rate (SM1 = 0) given by the crystal oscillator frequency divided by 12.
Mode 1 This mode corresponds to SM0 = 0 and SM1 = 1, and is a duplex UART mode.
Data is sent through the TxD pin and received through the RxD pin. A 10 bit frame
frame containing 8 bit data (SM0=0), a start bit (=0) and a stop bit (=1) are used.
The rate of sending and receiving data is set by the overflow rate of timer T1 (SM1 =
1 implies variable data rate). Details of how T1 is used will be discussed later. The
data rate can be doubled by setting the most significant bit of the special function
register PCON at address 87H. This bit is called SMOD. Notice that PCON is not
bit addressable and masking must be used to set or clear this bit without disturbing
the rest of PCON, which controls the sleep and idle modes to reduce power dissipation
during times of low activity. The stop bit sampled on the serial input line is copied to
bit RB8 in SCON in this mode.
Mode 2 This mode is set by making SM0 = 1 (11 bit frame containing 9 bits of data, start
and stop bit) and SMO = 0 (fixed baud rate). This is an asynchronous mode using a
start and stop bit. Serial data is transmitted using the TxD pin and received using the
RxD pin. The baud rate is fixed at fosc /64. The baud rate can be doubled to fosc /32
by setting the SMOD bit.
5

Mode 3 This mode is set by making SM0 = 1 (11 bit frame containing 9 bits of data, start
and stop bit) and SMO = 1 (variable baud rate). This is an asynchronous mode using
a start and stop bit (identical to mode 2) except that the baud rate is determined from
the overflow rate of timer T1. Serial data is transmitted using the TxD pin and received
using the RxD pin. The baud rate in modes 1,2 and 3 can be doubled by setting the
bit SMOD (PCON.7).

3.3

Serial I-O using Interrupts

To do serial I-O under interrupt control, we need to enable serial interrupts by setting bits
4 (ES) and 7 (EA) of the interrupt enable special function register IE at byte address A8H.
This register is bit addressable. (The assembler gives special mnemonics to each bit address.)
Bit No.
Function
Bit Addr
Bit Name

7
IE
AF
EA

SFR IE at byte address A8H

6
5
4
3
2
U
U
SI TF1 Ex1
AE AD AC AB
AA
ES ET1 EX1

1
TF0
A9
ET0

0
Ex0
A8
EX0

The most significant bit of the register is a global interrupt enable flag. This bit must be set
in order to enable any interrupt. Bit 4 (ES) enables interrupt from the serial interface. Once
enabled, the serial interface causes interrupts due to a receive event (RI Flag set) or due to
a transmit event (TI Flag set). The receive event occurs when the input buffer of the serial
line (Sbuf in) is full and a byte needs to be read from it. The transmit event indicates that a
byte has been sent and a new byte can be written to output buffer of the serial line (Sbuf out).
When a serial interrupt occurs, control is transferred to address 0023H of the program
memory, which is the vector address for serial interrupts. The interrupt handler routine is
supposed to be placed here. Since only 8 bytes are available in this region for each handler,
it is conventional to place the interrupt service routine elsewhere and to place an ljump instruction at address 0023H to the handler.
As discussed earlier, there are independent interrupt flags for reception and transmission
of serial data, called RI and TI in the lower nibble of SCON. RI indicates that a byte has been
received and is available for reading in the input buffer. TI indicates that the previous byte
has been sent serially and a new byte can be written to the serial port. A serial interrupt
occurs if either of these flags is set. (Of course the serial interrupt must be enabled for
this to occur). The interrupt service routine should check which of these events caused the
interrupt. This can be done by examining the flags. Either or both of these flags might be
set, requiring a read from or write to the serial buffer Sbuf (or both). Recall that the input
and output buffers are distinct but are located at the same address. A read from this address
reads the input buffer while a write to the same address writes to the output buffer. The
RI and TI flags are not automatically cleared when an interrupt is serviced. Therefore, the
interrupt service routine must clear them before returning. Here is an example handler for
serial interrupts:
Serial_ISR:
PUSH PSW
;
JNB RI,output
;
MOV inchar, Sbuf;
CLR RI
;
output:
;
JNB TI, done
;
MOV Sbuf, outchar;
CLR TI
;
done:
POP PSW
;
RETI
;

Save flags and context

If RI not set, check for TI
Save this character
clear receive interrupt flag
Check if output is required
If no transmit interrupt, leave
Else send the character
Clear Transmit interrupt flag
and flags
and return
6

3.4

Multicast Serial IO

The flag SM2 in SCON is used for multicast communication. When SM2 is set, the receive
interrupt flag is not set if the received character does not have the 9th bit set. The way this
is used is the following:
Several slaves are connected in parallel to a single master. Each slave is given a unique
address. We use mode 2 or mode 3, which uses 11 bit frames containing 9 bit data, the start
bit and the stop bit.
The 9th data bit is not used for parity here. It is used in a special way to enable reception
of data only by selected receivers. The way this is accomplished is that initially, all slaves
have their SM2 bits set. Commands are sent by the master with the 9th bit set. (The master
can easily do this by loading a 1 in its TB8 bit in SCON). Because of this, when commands
are sent, all slaves receive an interrupt and read the command.
To illustrate the procedure with an example, assume that the upper nibble of the command encodes what command it is and the lower nibble contains the address of slave to
whom it applies. (Thus, up to 16 slaves can be managed in this example). Let us say that a
0001 in the upper nibble means become an active listener while 0010 means stop listening.
In order to make a slave (with address = 7) the active listener, the master can send
0001 0111 with TB8 set. All slaves receive this command because the 9th bit was set. Now
slave 7 finds that its address matches the lower nibble. It clears its SM2 bit. The other
slaves, for whom the lower nibble of the command did not match their address, continue to
keep this bit set. Data is now sent by the master with the 9th bit (TB8) cleared.
The selected slave has its SM2 bit cleared. Therefore, it will receive an interrupt whether
the 9th bit is set or not. On the other hand, no interrupt is generated in other slaves (because
SM2 bit is set and their RB8 is not). Thus they automatically ignore the data.
The selected listener can be de-selected by sending the command 0010 0111 with TB8 set.
All slaves receive the command (because RB8 is set). Slave 7 knows that this is a command
and not data, because its RB8 is set. Since the lower nibble matches its address, it has to
act on this command. The upper nibble is 0010, which means de-select. So slave 7 sets its
SM2 bit again and now will be interrupted only by commands (which have the 9th bit set)
and not by data (when the 9th bit is cleared).

Baud rate generation

In modes 1 and 3, the baud rate is generated by the overflow rate of timer T1. While T1
can be used in any mode for this purpose, it is convenient to use it in the auto reload mode
(timer mode 2). This timer mode uses TH1 to store the auto-reload value and uses TL1 as
an 8 bit counter. Once configured, T1 runs continuously and needs no further software effort.
Its input is the oscillator frequency divided by 12. (Interrupts are not required and in fact
must not be enabled for T1).
In this mode, T1 will overflow when TL1 reaches 256. At this time, TL1 will be reloaded
automatically with the value stored in TH1.
Suppose the value n is loaded in TH1 and fc = crystal frequency. TL1 loads the value
n and counts up. It reaches 256 after 256-n cycles. Then it reloads itself with n and starts
counting up again. So, TL1 will overflow and reload the value n every (256-n) inputs.
The frequency of the input to the counter is fc
12

Therefore the frequency of the output is

fc

12(256n)

This constitutes the 32X clock for synchronization.

fc
Therefore the baud rate is: 1232(256n)
Thus, baud rate =

fc

384(256n)

In actual use, we are given the baud rate and must determine n. So
fc
386 baud rate
If SMOD bit is set, the internal clock is just 16X the baud rate. In that case,
n = 256

baud rate =

fc
192(256 n)

fc
192 baud rate
Let us illustrate it with an example. A crystal frequency of 11.059 MHz is often used
because it gives convenient integral values for n for most standard baud rates. Consider a
baud rate of 9600.
so n = 256

11059000
= 256 2.9999 256 3 = 253
384 9600
If we back calculate the baud rate with this value of n,
n = 256

baud rate =

11059000
= 9599.2
384 3

which is close enough.

If we want 19200 baud,
n = 256

11059000
= 256 1.49997 256 1.5
384 19200

In this case we do not get an integral value for n. If we use 256-1 = 255, we get
baud rate =

11059000
= 28799.48
384 1

and if we use 256-2 = 254, we get

baud rate =

11059000
= 14399.74
384 2

neither of which is anywhere close to what we want.

However, in this case, we can set the SMOD bit (PCON.7). Then,
n = 256

11059000
= 256 2.9999 256 3 = 253
192 19200

This gives a baud rate of

baud rate =

11059000
= 19199.62
192 3

Which is acceptable. This is about the highest standard baud rate possible with a standard
8051 which is rated to run at clock frequencies up to 12 MHz.

The lowest baud rate we can get using T1 in 8bit auto reload mode is about 110. A count
of 0 will give
11059000
= 112.4979
baud rate =
384 256
which is marginally acceptable for 110 baud communication. (This is because at such low
baud rates, we can afford to be a fair bit away from the middle of the bit time and still not
run the risk of sampling the wrong bit). However, if we need even lower baud rates this timer
mode will not do.
In these cases, we can either operate the microcontroller at a lower crystal frequency, or
use a 16 bit timer mode and reload the timer in software with a calculated value every time
T1 overflows. Calculation of the count now will require us to include the number of clock
cycles needed to respond to the interrupt and then to reload the counter. Fortunately, the
use of baud rates below 110 is extremely rare.

Putting it all together

Here is the list of things we need to do in order to communicate serially.

5.1

Configuration

For configuring asynchronous serial communication at a given baud rate to be set by T1:
1. Write the interrupt service routine for serial interrupts. (An example has been provided
earlier). Place an ljump to it at the vector address 0023.
2. Set SCON bits to desired values.
Bit No.
Bit Addr
Bit Name

SCON register at BYTE address 98H

7
6
5
4
3
2
9F
9E
9D
9C
9B
9A
SM0 SM1 SM2 REN TB8 RB8

1
99
TI

0
98
RI

For communication without the parity bit, clear SM0. Set SM0 for communication
with parity. Since baud rate is to be decided by T1, set SM1. Clear SM2, as we are
not using multicast transmission. We should (eventually) set REN, but it is better to
do this as the last step to avoid receiving invalid data before configuration is complete.
3. Configure T1 to provide the desired baud rate. This requires setting appropriate bits
in TCON and TMOD registers and loading the appropriate delay value in TH1.
Bit No.
Bit Name
Bit Addr
Bit No.
Timer:
Bit Name

TCON register at BYTE address 88H

6
5
4
3
2
1
TR1
TF0
TR0
IE1
IT1
IE0
8E
8D
8C
8B
8A
89
TMOD register at BYTE address 89H
7
6
5
4
3
2
1
T1
T0
G1 C/T1 T1M1 T1M0 G0 C/T0 T0M1

7
TF1
8F

0
IT0
88
0
T0M0

We need to disable hardware gating for T1 (G1 = 0), use it as a timer (C/T1 = 0),
and put it in 8bit auto reload mode (T1M1 = 1, T1M0 = 0). Since TMOD is not bit
addressable and we dont want to disturb T0 settings, we do it by ORing TMOD with
#20H and ANDing it with #2FH. We set TR1 in TCON to allow T1 to run.
4. Calculate the value to be loaded in TH1 as described earlier. If we need to double the
baud rate to get an integer value for n, set SMOD (PCON.7) by ORing it with #80H.
Else, clear it by ANDing it with #7FH. (This is because PCON is not bit addressable).
Load TH1 with #n.
9

5. Enable serial interrupts and disable interrupts from T1. This is done by manipulating
the specil function register IE.

Bit No.
Bit Addr
Interrupt on
Bit Name

Interrupt Enable Register IE at A8H

7
6
5
4
3
2
AF AE AD AC AB
AA
IE
U
U
SI TF1 Ex1
EA
ES ET1 EX1

1
A9
TF0
ET0

0
A8
Ex0
EX0

We need to set EA (IE.7), ES (IE.4) and clear ET1. IE is bit addressable.

6. Now we are ready: clear RI and TI to remove any stale flags and set REN in SCON to
enable serial I-O.

5.2

Using the serial port

Once the configuration is done, it is easy to use the port. On a receive interrupt, we just
read SBUF and clear RI. On a transmit interrupt, we write the byte to SBUF. If parity is
being used, we can just add 0 to the byte in A and copy the parity flag (PSW.0) to TB8 in
SCON before writing to SBUF.

10