Fix a calibration adder that turns a sensor sample of -1 into 65,535.
Hardware context: A signed sensor sample widened into a calibration adder, and a byte negator for a motor driver. Repair the adder that zero-extends a negative 16-bit sample into 20 bits.
A pressure sensor sends each sample as a 16-bit two's-complement number. The calibrate block adds a 20-bit offset and passes the level on. In the lab every level was right. In the field the tank ran slightly below its zero mark, the sample became -1, and the display jumped to 65,535.
This is a Fix problem. The module below compiles and is right for every sample that is zero or positive. Find why a negative sample goes wrong and repair it.
The check applies four pairs of sample and offset, one per nanosecond, and records the level after each.
Signals in the Timeline
| Signal | Width | Role |
|---|---|---|
script | 144 | input of each check |
Full specification
module calibrate (
input logic [15:0] reading, // the sample: two's complement, 16 bits
input logic signed [19:0] offset, // the calibration offset: two's complement, 20 bits
output logic signed [19:0] level // offset + reading
);
// the logic to repair
endmodule
| Signal | Meaning |
|---|---|
| reading | The sample, from -32,768 to 32,767 in two's complement. The port itself is declared without the signed keyword. |
| offset | The calibration offset, a signed 20-bit number. |
| level | The sum of the offset and the sample, as a signed 20-bit number. The test offsets keep the sum in range. |
The result fields are named level_step0 to level_step3, one per step, shown as signed decimal numbers.
Examples
Example 1
The Bug Report- Input
reading ffff (-1), offset 0; reading 0000 (0), offset 0; reading 0001 (1), offset 0; reading fffe (-2), offset 0
- Output
level_step0 = -1, level_step1 = 0, level_step2 = 1, level_step3 = -2
With no offset, the level must equal the sample: -1, 0, 1 and -2. The broken block shows 65,535 for the sample -1.
Example 2
Positive Samples- Input
reading 0064 (100), offset 50; reading 7fff (32767), offset 0; reading 0001 (1), offset -1; reading 1000 (4096), offset 1000
- Output
level_step0 = 150, level_step1 = 32767, level_step2 = 0, level_step3 = 5096
For a sample that is zero or positive, the four new bits are 0 either way. This is why the lab tests passed.
Example 3
Negative Samples with an Offset- Input
reading ff9c (-100), offset 50; reading 8000 (-32768), offset 0; reading ffff (-1), offset 1; reading f000 (-4096), offset -1000
- Output
level_step0 = -50, level_step1 = -32768, level_step2 = 0, level_step3 = -5096
The sample -100 with the offset 50 must give -50. The most negative sample, -32,768, must come through unchanged when the offset is 0.
Constraints and Contract
- Inputs are known 0/1 values. Each pair is held for one nanosecond and level is sampled at its end.
- Keep the module name and the ports, including the unsigned declaration of reading.
Hints
Hint 1
Write the sample -1 as sixteen bits. Then write the 20-bit number that the adder really adds. Which four bits were invented, and with what value?
Hint 2
A value is widened with copies of its top bit only when it is signed. Look at how the port reading is declared.
Hint 3
Tell the adder to read the sample as signed. The $signed function does it. So do four copies of bit 15 in front of the sample.
What the grader checks: Eight directed and eight seeded cases of four steps each run on the block. Every level is compared with a model written in JavaScript. A sum outside the 20-bit range is outside this problem.
16 checks on Icarus · browser samples / server checks. Run the examples first, then check the full regression. Full checks use the server judge. Downloads contain public samples only. A passing result establishes the tested cases, not a proof for all possible inputs.
See all 3 steps and the final capstone