[SOLVED] EECS 151/251A Lab 3: Rotary Encoder and Debouncer, Finite State Machines…

EECS 151/251A FPGA Lab
Lab 3: Rotary Encoder and Debouncer, Finite State Machines,
Synchronous Resets, Synchronous RAM, Testbench Techniques
1 Before You Start This Lab
Before you proceed with the contents of this lab, we suggest that you review these documents that
will help you better understand some concepts we will be covering.
1. labs fa16/docs/Verilog/verilog fsm.pdf
Goes over concepts of FSM in Verilog. Provides an example of implementing FSM’s in Verilog
and pitfalls to watch out for.
2. http://www.labbookpages.co.uk/electronics/debounce.html Read the ”What is Switch
Bounce” section to get idea of why we need a debouncer circuit. Read the ”Digital Switch
Debouncing” section to get a general overview of the circuit, its parts, and their functions.
3. http://www.xilinx.com/products/boards/s3estarter/files/s3esk_rotary_encoder_interface.
pdf
Read slide 5 (Rotary Encoder and Signals) to get an idea of how the encoder works and the
signals it generates. You can read the next few pages to get a better idea of how to use the
signals. You will be implementing the circuit described in these slides in this lab.
2 Lab Overview
In this lab, we will learn about circuits to take the signals generated by the buttons and rotary
encoder on the ML505 board and convert them into a digital signal we can use in our FPGA design.
You will be using the LED’s to confirm that your input conditioning circuits are working correctly.
We will discuss how to use synchronous resets to reset our circuits to a known initial state. We
will be creating a basic FSM in the music_streamer that uses the buttons and rotary encoder to
change states and alter the music playback.
Run git pull in your git cloned labs_sp17 directory to fetch the latest skeleton files for this
lab.
3 Synchronizer, Debouncer, and Rotary Encoder
3.1 Synchronizer
In Verilog (RTL), digital signals are either 0’s or 1’s. In a digital circuit, a 0 or 1 corresponds to a
low or high voltage. If the circuit is well designed and timed (fully synchronous), we only have to
worry about the low and high voltage states, but in this lab we will be dealing with asynchronous
signals.
The signals coming from the push buttons and rotary encoder on the ML505 board don’t have an
associated clock signal. Thus, when those signals are put through a register, the hold or setup time
constraints of that register may be violated. This may put that register into a metastable state.
2
Figure 1: The ’ball on a hill’ metaphor for metastability. If a register’s timing constraints are
violated, its output voltage oscillates and after some time unpredictably settles to a stable state.
In a fully synchronous circuit, the timing tools will determine the fastest clock frequency under
which the setup time constraints are all respected and the routing tools will ensure that any hold
time constraints are handled. Introducing an asynchronous signal that isn’t changing with respect
to a clock signal can cause a register to go into a metastable state. This is undesirable since this will
cause a ’mid-rail’ voltage to propagate to other logic elements and can cause catastrophic timing
violations that the tools never saw coming.
We will implement a synchronizer circuit that will safely bring an asynchronous signal into a synchronous circuit. The synchronizer needs to have a very small probability of allowing metastability
to propagate into our synchronous circuit.
This synchronizer circuit we want you to implement for this lab is relatively simple. For synchronizing one bit, it is a pair a flip-flops connected serially. This circuit synchronizes an asynchronous
signal (not related to any clock) coming into the FPGA. We will be using our synchronizer circuit
to bring any asynchronous off-FPGA signals into the clock domain of our FPGA design.
Figure 2: 1-bit 2 Flip-Flop Synchronizer
Edit the lab3/src/synchronizer.v file to implement the two flip-flop synchronizer. This module
is parameterized by a width parameter which indicates the number of one-bit signals to synchronize.
3.1.1 Testing in Simulation
The testbenches to be run are stored in lab3/sim/tests. Each .do file in this directory is run
when you run make in the lab3/sim directory. If you only want to run one testbench, you can
rename all the other .do files in this directory to have a different file extension. Alternatively, you
can leave the file extensions alone, and specify the test you want to run like this:
3
cd lab3/sim
make CASES=”tests/sync_testbench.do”
We have provided a testbench for your synchronizer called sync_testbench in lab3/src/sync_testbench.v.
Take a look at the code for this testbench and run it; the testbench shouldn’t print any failure
messages and you should inspect the waveform before you move on. For details on the
constructs/techniques/syntax used in this testbench, refer to the ’Testbench Techniques’ section of
this lab.
3.2 Debouncer and Edge Detector
Recall this graphic from the prelab debouncer reading. It is an overview of the debouncer circuit
which includes the synchronizer circuit.
For this lab, the debouncer circuit will take a button’s glitchy digital input and output a clean
signal indicating a single button press. The reason we need a circuit for this is shown in the figure
below.
When we press the button, the signal doesn’t behave like a perfect step function. Instead the
button signal is glitchy due to mechanical ’bounce’. A debouncer turns this waveform, which shows
a single button press, into a clean signal with a single voltage transition.
4
Take a look at lab3/src/debouncer.v. This is a parameterized debouncer which can debounce
width signals at a time. Your debouncer receives a vector of synchronized 1-bit signals and it
outputs a debounced version of those signals. The other parameters reference the constants used
in the circuit from the prelab reading.
The debouncer consists of:
1. Sample Pulse Generator – Tells our saturating counter when to sample the input signal.
It should output a 1, every sample_count_max clock cycles. By default sample_count_max
is set to 25000.
2. Saturating Counter – This is a counter that counts up to pulse_count_max. The saturating
counter should increment by one if both the sample pulse and the input signal are high at
a clock edge. At any clock edge, if the input signal is 0, the saturating counter should be
reset to 0. Once the saturating counter reaches pulse_count_max, it should hold that value
indefinitely until the input signal falls to 0, upon which the saturating counter should be
reset to 0. The debounced_signal of your debouncer should be an equality check between
the saturating counter and pulse_count_max.
You should use the same sample pulse generator for all input signals into your debouncer, but you
should have a separate saturating counter per input signal. You will likely need to use a 2D reg in
Verilog to create the saturating counters. You will also likely need to use generate-for.
Here is an example of creating a 2D array and using a generate-for loop:
reg [7:0] arr [0:3]; // 4 X 8 bit array
arr[0]; // First byte from arr (8 bits)
arr[1][2]; // Third bit of 2nd byte from arr (1 bit)
genvar i;
generate
for (i = 0; i < width; i = i + 1) begin:LOOP_NAME
always @ (posedge clk) begin
// Insert synchronous Verilog here
end
end
endgenerate
3.2.1 Edge Detector
The debouncer will act to ’smooth-out’ the button press signal. It is then followed up with an edge
detector that can take the high-to-low transition of the debouncer output and use it to generate a
1 clock period wide pulse that the rest of our digital design can use.
Create a variable-width edge detector in lab3/src/edge_detector.v.
5
3.2.2 Testing in Simulation
We’ve provided a testbench to test your debouncer and edge detector circuits in lab3/src/debouncer_testbench.v
and lab3/src/edge_detector_testbench.v. Run the testbench, make sure it passes, and inspect
the waveforms before FPGA testing. Make sure there are no undefined (red line) signals.
If you are seeing issues where certain registers are red lines (X’s), make sure you give them an
initial state. For a 2D reg initialization, use the following initialization code in debouncer.v:
integer k;
initial begin
for (k = 0; k < width; k = k + 1) begin
saturating_counter[k] = 0;
end
end
The debouncer testbench has 2 tests:
1. Verifies that if a glitchy signal initially bounces and then stays high for less than the saturation
time, that the debouncer output never goes high.
2. Verifies that if a glitchy signal initially bounces and then stays high for more than the
saturation time, that the debouncer goes high and stays high until the glitchy signal goes
low.
The edge detector testbench tests 2 scenarios, when the signal_in is a pulse 10 clock cycles wide
and a pulse 1 clock cycle wide and verifies that the edge_detect_pulse output goes high twice,
both times with a width of 1 clock cycle.
3.2.3 Testing on the FPGA
We have created a top level module called debouncer_fpga_test that will create a 8-bit register and
will use button presses to add and subtract from it. This module will use both your debouncer.v
and edge_detector.v.
Pressing any compass button will cause the register to increment by 1 and the LEDs will show the
current value of the register. Pressing the rotary wheel inward, will cause the register to decrement.
Pressing the CPU_RESET button (near the SATA port), will cause the register to reset to 0.
make TOP=debouncer_fpga_test
make TOP=debouncer_fpga_test report
make TOP=debouncer_fpga_test impact
Make sure that your report gives you zero warnings for synthesis. You must fix any and all
warnings before your debouncer will as expected on the FPGA.
Show the TA the debouncer working before moving on. It is critical that your debouncer works properly.
6
You will discover when playing with your debouncer that the buttons have a way that they like
being pressed to minimize bounce; get a good feel for them.
3.3 Rotary Encoder
The rotary encoder is a device that has two switches that close as you rotate the wheel. Recall this
diagram from the prelab reading.
Our main concern is finding out which direction the wheel turned. The following oscilloscope
waveform from the prelab reading illustrates how we will do so.
If the pulse from B (bottom wave – blue) happens before the pulse from A (top wave – orange), it
indicates that the wheel has been turned left. If the wheel is spun in the opposite direction then
A’s pulse will occur before B. We will take advantage of the fact that we can examine the logic
level of wave B at the rising edge of wave A to determine direction of wheel movement.
Open up lab3/src/rotary_decoder.v. This module takes the synchronized and debounced A
and B signals, and a clock input. It outputs a rotary_event pulse (one clock cycle wide) when
a wheel spin has been detected and rotary_left (when rotary_event is high) indicates whether
that spin was to the right or the left.
You need to implement the circuit from slide 8 of the prelab reading. In that slide rotary_q1
refers to input A and rotary_q2 refers to input B. You don’t need to use the rst input, but it is
recommended for future use of this module.
7
3.3.1 Testing in Simulation
We have provided a rotary decoder testbench for you in lab3/src/rotary_decoder_testbench.v.
This testbench isn’t self checking. You will have to inspect the waveform manually.
Proceed to the FPGA test once you have confirmed expected behavior in simulation.
3.3.2 Testing on the FPGA
There is a test top level module called rotary_decoder_fpga_test. It allows you to spin the
rotary encoder to increment and decrement a 8-bit counter whole value is shown on the GPIO
LEDs. Pushing the rotary encoder button will cause the counter to reset it to 0. Run it as such.
make TOP=rotary_decoder_fpga_test
make TOP=rotary_decoder_fpga_test report
make TOP=rotary_decoder_fpga_test impact
Show the TA the rotary encoder working before moving on. It is critical that your
rotary encoder works properly.
Congratulations! You just built four highly useful and practical digital circuits. Now let’s integrate
them into our larger music streamer design.
4 Testbench Techniques
There are several testbenches included in this lab for your synchronizer, edge detector, rotary
encoder, debouncer, and music streamer that introduce you to some useful Verilog testbench constructs.
• @(posedge ) and @(negedge ) – These are a different type of delay statement from what you have seen before. #10 would advance the simulation by 10 timesteps.
These commands will advance the simulation until the rises or falls.
For example:
@(posedge signal);
@(posedge signal);
Simulation time will advance until we have seen two rising edges of signal.
• repeat – it acts like a for loop but without an increment variable
For example:
repeat (2) @(negedge clk);
repeat (10) begin
8
@(posedge clk);
end
The simulation will advance until we have seen 2 falling clock edges and will then advance
further until we have seen 10 rising clock edges.
• $display – acts as a print statement. Similar to languages like C, if you want to print out a
wire, reg, integer, etc… value in your testbench, you will need to format the string. It works
like printf() in C.
For example:
$display(“Wire x in decimal is %d”, x);
$display(“Wire x in binary is %b”, x);
• tasks – tasks are subroutines where you can group and organize some commands rather than
haphazardly putting them everywhere. They can take inputs and assign outputs.
task wait_for_n_clocks();
input [7:0] num_edges;
begin
repeat (num_edges) @(posedge clk);
end
endtask
• fork/join – Allows you to execute testbench code in parallel. You create a fork block with
the keyword fork and end the block with the keyword join.
For example:
fork
begin
task1();
end
begin
$display(“Another thread”);
task2();
end
join
Multiple threads of execution are created by putting multiple begin/end blocks in the fork-join
block. In this example, thread 1 runs task1(), while thread 2 first $displays some text then
runs task2(). The threads operate in parallel.
• Hierarchical Paths – you can access signals inside an instantiated module for debugging purposes. This can be helpful in some cases where you want to look at an internal signal but
don’t want to create another output port just for debug.
For example:
9
tone_generator tone_gen ();
$display(“Signal inside my tone_generator instance, clock_counter: %b”,
,→ tone_gen.clock_counter);
5 Synchronous Resets In Design and Simulation
Begin by copying your tone_generator and music_streamer from lab 2. Do not change the
port declaration of the skeleton files, only copy over your implementation.
Now that we have a debouncer that can give us a pulse for a press of a button, we have a way of
explicitly resetting our circuits! You will recall that in the previous lab, we set the initial value of
registers as below so that our simulation would have defined signals.
reg [23:0] clock_counter = 0;
Now that we have a reset signal tied to the CPU_RESET push button, we can do this instead.
always @ (posedge clk) begin
if (rst) begin
clock_counter <= 24’d0;
end
else begin
clock_counter <= clock_counter + 24’d1;
end
end
Unlike what we did before, this Verilog is synthesizable for all deployment targets, FPGAs, ASICs,
and CPLDs alike. Go ahead and modify your tone_generator and music_streamer to use the
provided reset signal to get your registers to a default state.
After doing this, run the tone_generator_testbench again using make in the lab3/sim/ directory.
View the waveform using ModelSim and see how we used a reset in the testbench to bring all the
registers to a defined state without specifying a default value.
6 Music Streamer Tempo Control
Let’s use the new user inputs we now have access to. You will recall that your music_streamer by
default chooses to play each tone in the ROM for 1/25th of a second. Extend the functionality of
the music_streamer so that spinning the rotary encoder changes the tempo of the notes. Pushing
in the rotary encoder should reset the tempo back to the default value. Use the rotary_event,
rotary_left, and rotary_push inputs in the music_streamer.
You should implement this by using a register to hold the number of clock cycles per note. Instead of
this number being hardcoded in Verilog to represent 1
25 th of a second, you can change it at runtime.
10
Spinning the rotary encoder once should add or subtract a fixed number from this register which
should alter the time each tone is played. You get to choose this number; find something reasonable.
Try this out on the FPGA and verify that you have control of your music_streamer’s tempo using
the rotary encoder. You should be able to speed up and slow down the music you are playing.
7 Music Streamer FSM
Now, you will implement a simple FSM in the music_streamer.
The FSM will have 3 states: PAUSED, REGULAR_PLAY, REVERSE_PLAY. Here is the state transition
diagram:
start REGULAR PLAY
REVERSE PLAY
PAUSED Center button
South button Center button
1. Your initial state should be REGULAR_PLAY.
2. Pressing the center compass push button should transition you into the PAUSED state from
either the REGULAR_PLAY or REVERSE_PLAY states. Pressing the center compass push button
while in the PAUSED state should transition the FSM to the REGULAR_PLAY state.
3. In the PAUSED state, your ROM address should be held steady at its value before the transition
into PAUSED and no sound should come out of the piezo speaker. After leaving the PAUSED
state your ROM address should begin incrementing again from where it left off and the speaker
should play the tones.
4. You can toggle between the REGULAR_PLAY and REVERSE_PLAY states by using the south
compass button. In the REVERSE_PLAY state you should decrement your ROM address by 1
11
rather than incrementing it by 1 every X clock cycles as defined by your tempo.
5. If you don’t press any buttons, the FSM shouldn’t transition to another state. Also, the
rotary encoder wheel can be used to change tempo regardless of which state you are in.
Your music_streamer takes in user button inputs that it can use to transition states. You should
drive the compass LEDs in this fashion corresponding to the three states:
LED Value
Center current_state == REGULAR_PLAY
East current_state == PAUSED
South current_state == REVERSE_PLAY
North 0
West 0
You can run the testbench in lab3/src/tone_generator_testbench.v to test out your state
machine. Take a look at the code to see what it does and inspect your waveform to check that your
FSM is performing correctly. Verify that you don’t have any unexpected synthesis warnings.
Put your design on the FPGA with make and make impact and try transitioning states. For
checkoff be able to demonstrate your state machine working and the tempo control with the rotary
encoder.
8 Building a Music Sequencer FSM
Here is a new state transition diagram for our music sequencer we will build inside the music_streamer
module.
12
start REGULAR PLAY
REVERSE PLAY
PAUSED
PLAY SEQ
EDIT SEQ
Center
South
Center
South
North
North
Center
We have added two new states PLAY_SEQ and EDIT_SEQ. You should wire up the west compass LED
to current_state == EDIT_SEQ and the north compass LED to current_state == PLAY_SEQ.
You should implement the skeleton of this state machine before proceeding with the complete
explanation.
Our goal is to create an 8-tap music sequencer that we can use along with our regular music
streamer. Our music sequencer will have a place (RAM) where it stores the tone_switch_periods
of 8 notes. When we are in the PLAY_SEQ state, the music streamer will play the 8 notes, one after
the other, in a continuous loop. Each note will play for a set amount of time as determined by
the sequencer tempo. While in the PLAY_SEQ state we can change the sequencer tempo using
the rotary encoder (the sequencer tempo is different from the ROM tempo you modified
earlier)
We can edit these 8 notes on the fly by moving into the EDIT_SEQ state. In this state, the LEDs
will show which of the 8 notes we are currently editing. By spinning the rotary encoder, we can
select a new pitch for this note. By pushing in the rotary encoder, we can save the selected pitch
in the RAM location for this note. We can use the east and west buttons to edit a different note.
You will find some skeleton code in lab3/src/music_streamer.v to help you implement the
sequencer. This is a complicated circuit, so you should add features slowly and modify the
music_streamer_testbench to test out your sequencer. Here is a list of requirements for your
sequencer:
1. You should have two tempo controls.
13
(a) The ROM tempo affects the duration of each note in the ROM and can be modified
in either the REGULAR_PLAY or REVERSE_PLAY states by the rotary encoder.
(b) The sequencer/RAM tempo affect the duration of each note in the sequencer RAM
and can be modified only in the PLAY_SEQ state by the rotary encoder.
2. The GPIO leds should be used for the following functions in different states.
(a) In the REGULAR_PLAY, PAUSED, and REVERSE_PLAY states, the 8 GPIO LEDs should
display the top 8 bits of the ROM address.
(b) In the PLAY_SEQ state, the 8 GPIO LEDs should display the note being played. For
example, if note 5 is currently being played, LED 5 should be on and the rest off.
(c) In the EDIT_SEQ state, the 8 GPIO LEDs should display the note being edited. For
example, if we are editing note 6, LED 6 should be on and the rest off.
3. In the EDIT_SEQ state, the piezo speaker should play the current pitch of the note. Spinning
the rotary encoder should increase and decrease the pitch by some amount per click that you
determine. Pressing the rotary encoder button should save the current pitch of the note into
the sequencer’s RAM.
4. In the EDIT_SEQ state, pressing the east or west buttons should change the note being edited.
As you change notes, the piezo speaker should play the pitch of the new note being edited.
5. You can use the same clock_counter for all parts of this state machine including the sequencer and the regular music streamer.
6. You can use the same sequencer_address for the two sequencer states.
7. You can choose what the reset values for the sequencer RAM entries are. A sensible default
is provided in the skeleton code.
Please ask the TA if you have any questions about specifications for the sequencer or FSM in
general.
9 Checkoff
1. Show the TA your working design with the FSM. Be able to transition states by clicking on
the north and center buttons and show that your music_streamer matches the spec.
2. Show the tempo control working by spinning the rotary encoder to speed up and slow down
the music.
3. Demonstrate that hitting the CPU_RESET button resets the ROM address back to 0 and puts
the FSM into the REGULAR_PLAY state.
4. Demonstrate that you can transition into the SEQUENCER state and that you can edit your
tones and play them back.
5. Show the TA your Verilog RTL for all the components you designed for this lab (synchronizer,
debouncer, rotary decoder, FSM) and briefly explain the design of each of them.
14