Design of FSM-controlled digital systems

Giuliano Donzellini, Domenico Ponta

Asynchronous Serial Transmitter and Receiver, on FPGA

050430

 

v1.80

In this laboratory you are asked to complete the design of an asynchronous serial communication system (see the figure below). After the simulation of the project, you'll realize a prototype on FPGA to verify the actual operations.

The system is composed of two identical units, A and B, interconnected between them, at a distance, through two serial communication lines. Each unit includes a transmitter (TX) and a receiver (RX). Each unit works with a nominal clock frequency of 10 MHz. The clock generators of the two units are physically different and separated: due to manufacturing tolerances, the actual frequencies of both will be a little different. Thus, the communication between the two units must be considered asynchronous.

The two units communicate with each other by sending and receiving serial packet sequences (see figure below):

Each packet is composed of a START bit at '1', followed, in order, by the bits D0, D1, D2, D3, D4, D5, D6 and D7, and terminated by a STOP bit at '0'; the "bit time" is equal to 1.6 µS, corresponding to 625 Kb/s (10 MHz /16). Note that the clock is not transmitted to each other: the result is then an asynchronous serial transmission.

In the following, first you will design, separately, the transmitter (TX) and then the receiver (RX), verifying their operations through simulation. Then, you will integrate the two modules in one unit, including both the transmitter and the receiver. After the simulation of the unit, you will implement it on an FPGA-based prototype board. You can test the communication system using two separate boards (units A and B), connected by a cable
.


The serial transmitter TX

The following figure shows the schematic of the transmitter only, of which we will design the controller, after the analysis of the functionality of the other system elements. Task of the transmitter is to send on the SER_OUT line, when the user presses the !TX_GO button, a serial packet containing the data D7..D0 (at the TX_DATA input), according to the format and timing described above. At the end of the transmission, it activates the output TX_END for the duration of a bit time.

In the figure we observe, highlighted by colored boxes: a bit time counter [A, in red], a shift register [B, in blue], a bit counter [C, in red], and the transmission controller [D, in brown], based on a finite state machine (FSM).

 Bit Time Counter  [A]
The TX bit time counter, based on the Cnt4 component, is in charge of dividing by 16 the frequency of the clock CK (10 MHz), in order to transmit the data packet at the given bit/rate (625 Kb/s = 10 MHz /16). This part of a serial transmitter is sometimes referred to as baud rate generator. The counter is set up to count down cyclically; every time the outputs Q3..Q0 reach the number '0000', the output TC (terminal count) is activated, resulting in a pulse every 16 clock cycles. The cyclic activation of TC is used by the controller to synchronize the transmission of each data bit, at the given bit rate. The LD output of the controller, when activated, presets the counter to the maximum value '1111' (set on the inputs P3..P0).

 Shift Register [B]
The TX shift register (based on the Univ4 and Univ8 components) serializes the parallel data (TX_DATA) on the output line SER_OUT. S1 and S0 control the register mode of operation. When the controller drives S1 = S0 = '0', the register state will not change; instead, when S1 = S0 = '1', on the next positive clock edge, the register loads in parallel the eight inputs D7..D0 (in order: P0 of Univ4 and P7..P1 of Univ8), the Start Bit (P0 of the Univ8) and the Stop Bit (P1 of the Univ4). The unused register bits are set to '0'.

Serial transmission starts when data is loaded, since the loading operation sets the Start Bit on the line SER_OUT. The controller then obtains the transmission of the successive bits of the package, one after the other, by shifting the register to the right (setting S1 = '0' and S0 = '1'), for the duration of a clock cycle, at the end of each bit time.

 Bit Counter [C]
The TX bit counter (also based on a Cnt4) counts the number of bits to be transmitted. The controller initializes the counter to '1010' (= 10 bits to be transmitted), by setting the LD line. Every time the controller activates the EN line, the count is decremented by one. The controller, then, by reading the line TCN (end of count), generated by the counter, will know when to end the operations of transmission.

 Controller [D]
The TX controller, that we must design, takes care of sending and/or receiving the appropriate signals to/from the system components.

Design specifications of the transmission controller

You can download here a ASM diagram, to be completed, of the transmission controller, where the state variables X, Y and Z have already been defined, as well as the outputs LD, S0, S1, END and EN, and the inputs GO, TCB and TCN (click on the picture to open the diagram to be completed in the d-FsM):

The controller receives the command TX_GO (on the line GO): its falling edge initiates the transmission of data. In the diagram provided, the states (a) and (b) have the responsibility not only to wait for the falling edge, but also to initialize the transmission, activating the line LD.

LD initializes the bit time counter; from which, the controller will periodically receive the signal TCB, used to output the bits on the line on a regular basis (at each bit time). The activation of LD also prepares the bit counter, that will inform the controller when the serial transmission must be ended.

The controller also handles the transmission shift register: as described above, to start the transmission, it order the register to load the parallel data TX_DATA, along with the start and stop bits. The controller then enters a loop in which transmits the other bits by imposing a right-shift step to the register, on each activation of TCB.

Each time a bit is transmitted, the controller activates the line EN, for the duration of a clock cycle, updating the counter of the bits to be transmitted. Note that, if we were not use a counter for this purpose, we would greatly complicate the algorithm of the controller, providing repetitive sequences, almost identical to each other, for every bit to be transmitted. In this way, instead, the sequence is unique, and the cycle ends at line activation TCN of the counter (when the number of remaining bits reaches zero).

Finally, the controller activates the output TND, to signal the end of transmission. To ensure that the duration of the output TND will be equal to a bit time, you should wait for the last time the signal TCB, and then go back to waiting for the falling edge of GO.

Using a d-FsM simulation, verify that the functionality of the designed FSM will match the specifications. Then, import the new component in the d-DcS, to complete the schematic available here. In the given schematic, some control outputs have been added, useful for a timing simulation (a suitable test sequence "TxSequence" is available in the timing diagram window).


The serial receiver RX

The figure below shows the block diagram of the receiver only, whose job is to decode packets arriving on the serial line SER_IN, taking into account the format and timing of the transmission. At the end of a packet decoding, the receiver produces in parallel, on the outputs U7..U0, the 8 bits that there were contained, and signal their availability, activating the output RX_RDY, for the duration of a bit time.

Even in this case, we will design the controller only, after analyzing the functionality of the other elements. To do this, we observe in figure: a bit time counter [A, in red], a shift register[B, in blue], a counter of the number of bits [C, in red], and the receiver controller [D, in brown].

Bit Time Counter  [A]

The operation of the RX bit time counter is almost identical to the one of the transmitter, but with a significant difference in its initialization. Even in this case, the counter generates a pulse every 16 clock cycles (1.6 µS, the bit time) on the output Tc. This line, renamed TCB at the controller input, is used to synchronize the acquisition of the data bits, one at a time, in the center of the time interval reserved to each one. Activating LD, the controller initializes the counter to the value '0110', and not to the maximum value, so that, starting the count from the start bit reception, the activation of the first Tc will occur in the middle of the bit time.

 Shift Register [B]
The RX shift register (the Univ8 component) de-serializes the data packets incoming from the line SER_IN. The working mode of the register is set by the RX controller by S1 and S0: if both are maintained at '0', the state of the register does not change. Serial data input (InR) is connected directly to the line SER_IN. The input of a data bit in the register and, at the same time, the right-shifting of the other bits, will occur when the controller sets S1 = '0' and S0 = '1' (in the middle of each bit time, and for the duration of a clock cycle). By repeating this process for each bit of the received packet, all the data bits will be available at the outputs RX_DATA.

 Bit Counter [C]
The RX bit counter (also based on a Cnt4) counts the number of bits to be received. The controller initializes it to 8, setting the LD line; then, every time the controller activates the EN line, the count is decremented by one. By controlling its output Tc (TCN for the controller), we are able to know if we have received all the bits.

 Controller [D]
We'll design the RX controller, the block that controls the timing of the components described above.

Design specifications of the receiver controller

In summary, the controller, on idle, waits for the arrival of a packet on the serial line SER_IN. When the start bit is detected, its validity is verified and, thanks to this, all the operations of the shift register are synchronized, in order to correctly capture the next packet bits, and finally, to perform the verification of the stop bit (in the following these operations will be examined in more detail). As in the case of the transmitter, we can contain the algorithm of the controller in a reasonable number of states, as timings and counts are delegated to external devices.

Here is available a ASM diagram of the controller, to be completed. In the file, the state variables X, Y and Z have already been defined, the outputs LD, S0, S1, RDY and EN, the inputs LIN, TCB and TCN and, finally, the states (a) and (b ) are already set (click on the figure will open the file in the d-FsM):

Looking at the ASM diagram, in the wait state (a) the controller checks the serial line (at the input LIN), awaiting the start bit. In state (a), the activation of LD initializes the counters of the bit time and the number of bits: coming out from this state, the counters will then be free to count, starting from the initialization value. On the occurrence of the start bit, we switch in the state (b), awaiting the middle of the bit time (i.e. awaiting TCB activation). If, in the meantime, LIN returns low, the start bit format is considered invalid, and then we return to the initial state, awaiting a new start bit.

Let us remember that, as said before, the bit time counter has been initialized so that it activates TCB, the first time, in the middle of the bit time: while we are in the state (b), the bit time counter is decreasing its value. Upon arrival of TCB, we declare valid the start bit, and continue beyond.

Awaiting TCB in state (b), also, allows us to synchronize, with respect to the start bit, the acquisition of successive data bits. In fact, after the first activation of TCB, which occurred at the middle of the start bit, the bit time counter will continue to activate TCB cyclically every 16 clock positive edges, and thus always at the middle of the time of all subsequent bits.

The acquisition of the 8 data bits is handled by a cycle of states in which the controller:

  1. waits until TCB is active (because when TCB is activated, a bit time elapsed and we are in the middle of that range);
  2. if TCB is active, but TCN is not yet (i.e. we have at least one or more bits to be received), it orders the register to acquire the bit and to right-shift its previous content, and requires to the bit counter to decrease the number of bits still to be received;
  3. returns to step 1.

If TCB and TCN are both active, there are no more data bits to be received, and that's when the stop bit is available on LIN. It is not necessary to memorize this bit, but only to verify its value. If the stop bit is properly at '0', we just have to activate RDY for the duration of a bit time (awaiting TCB again) and then back again awaiting a new packet in state (a). However, if the stop bit value is wrong, we wait, without generating RDY, that the line LIN goes back to '0', than we return in the state (a).

The functionality of the FSM can be verified, preliminarly, using the d-FsM simulator; then, the component obtained must be imported into the d-DcS, in the schematic, to be completed, available here. Again, a few control outputs have been added to the schematic, to make your timing simulation more readable (you can use the test sequence "RxSequence", defined in the timing diagram window).


The complete unit (TX + RX)

The following figure shows the schematic of the complete unit (transmitter and receiver):

The transmitter and receiver are exactly those defined above. The clock CK and the !RESET have been unified, and are visible at the bottom of the schematic. A multiplexer "2-1" was added: in the diagram, it is highlighted with a red frame. The new TEST input, when activated, allows us to connect the receiver directly to the transmitter, to allow stand-alone testing, without the need to link two different systems. In this way, in simulation we are able to send a packet and receive it directly in the same system.

To simulate the entire system of serial transmission and reception, we provide here a schematic, to be completed with the two FSM previously designed. The schematic, as in the previous cases, has been enriched with some control outputs, to make more complete the timing simulation (in the diagram window time is available the test sequence "TestSequence"). The added outputs will help also the experimentation of the system on the FPGA board. We recommend to use this schematic without delete or modify the input and output terminations because they have been set for the export of the project on the FPGA board.


After the simulation of the entire system, let's move to its physical implementation on the FPGA board. The general procedure is described in the introductory tutorials:

   Circuit Prototyping on Altera DE2 Board
   Sequential Circuit Testing on Altera DE2 Board
   Microcomputer Testing on Altera DE2 Board

The "Test on FPGA" command (of the d-DcS) will open the dialog window visible in the next figure:

The associations between the d-DcS schematic and the FPGA board (inputs and outputs devices), have been already defined, so you do not need to change them. The clock CK is set to 100 Hz (rather than 10 MHz), to allow visual examination of the behavior of the system. The "Slow Clock Mode", set as shown in the figure above, will be activated at run time by the switch SW[17], and it is defined to work in manual mode (controlled by the push-button Key[01] and displayed by the red LED LEDR[17]). According to this setting, if the switch is zero, the clock CK will work normally at 100 Hz, but if the switch is at one, the clock will be generated a pulse at a time, at each activation of the push-button.

To optimize the testing of the serial communication system on the FPGA board, associations are highlighted in a summarized way in the figure below (the "control panel" of our system):

We can test the system stand-alone by activating the test mode (switch TEST upward), or by using two separate boards, interconnected by means of an appropriate cable, to exchange data in both directions:

We make the connection using one of the two 40-pin connectors available on each FPGA board. As seen in the figure above, we use a cross-connection between the transmitter and the receiver, in both directions, for a total of 3 wires (including the ground reference GND). Here the pattern of the connection schematic, which refers to the expansion connector "GPIO_1" present on the board.

Even without activate the TEST switch, we can perform preliminary tests with a single board, using a simple jumper inserted into the connector, as shown in the following figure (yellow arrow), in order to connect with the transmitter and receiver:

Be extremely careful in placing and removing the jumper, to avoid mechanical damage to the connector pins. To avoid possible damage to the FPGA chip, whose pins are directly connected to the connector, it is necessary to carry out this operation when power is off.

à