A scope hands you the timing error of every edge. Pull the duty-cycle distortion out of it, and the peak to peak with and without it.
Hardware context: The digital blocks of a serializer/deserializer: line encoder and decoder, scrambler, CRC and error-correcting code. Keep the line balanced and busy, and detect or correct the errors a fast serial link makes.
A jitter analyser measures the time interval error (TIE) of each edge, in milli-UI. It also says whether each edge was rising or falling.
Write jt_edge(tie, rising) to take one edge, and jt_reset() to start again. Then write two results.
jt_dcd() returns the duty-cycle distortion: the mean TIE of the rising edges minus the mean TIE of the falling edges. jt_pp(remove_dcd) returns the TIE peak to peak, over all edges as measured, or after taking each edge's own direction mean off.
Signals in the Timeline
| Signal | Width | Role |
|---|---|---|
test_number | 8 | input of each check |
Full specification
function automatic void jt_reset();
function automatic void jt_edge(input int tie, input bit rising); // tie in milli-UI
function automatic int jt_dcd();
function automatic int jt_pp(input bit remove_dcd);
| Item | Rule |
|---|---|
| Means | Each direction's mean is its sum divided by its count, rounded toward zero (SV integer division). |
| DCD | Rising mean minus falling mean. It may be negative. |
Peak to peak, remove_dcd = 0 | Largest TIE minus smallest TIE over all edges. |
Peak to peak, remove_dcd = 1 | The same, after subtracting the rising mean from each rising edge and the falling mean from each falling edge. |
| Edges | At most 4,096 per run. Rising and falling edges alternate. |
The tests build TIE values from a DCD, a random part, a periodic part and a constant offset, all bench values. A constant offset must not change any result.
Examples
Example 1
DCD Only- Input
1,000 edges, DCD 30 milli-UI
- Output
dcd_wrong = no, pp_wrong = no, pp_without_dcd_wrong = no, your_dcd = 30
Rising edges sit at +15 and falling edges at -15. The DCD is 30; the peak to peak is 30 as measured and 0 with the DCD removed.
Example 2
DCD with Random Jitter- Input
2,000 edges, DCD 40, random +-25
- Output
dcd_wrong = no, pp_wrong = no, pp_without_dcd_wrong = no, your_dcd = 39
The random part nearly averages out of each mean: the DCD comes back as 39, not 40. Removing it narrows the peak to peak by about that much.
Example 3
Negative DCD with an Offset- Input
DCD -20, offset 300
- Output
dcd_wrong = no, pp_wrong = no, pp_without_dcd_wrong = no, your_dcd = -19
The offset moves both means by 300 and cancels in their difference. The DCD comes back as -19: falling edges are late.
Constraints and Contract
- Keep the four function names and their arguments.
- Keep your state in module-level variables, and clear it in
jt_reset(). - Results are integers in milli-UI.
Hints
Hint 1
Keep a sum and a count for each direction. The DCD needs nothing else.
Hint 2
The peak to peak without DCD needs both means, and they are known only at the end. Store the edges.
Hint 3
A constant offset moves both means by the same amount. Why does that leave the DCD unchanged?
What the grader checks: The fixture feeds six edge sequences to your functions and compares the three results with a reference computed the same way. It does not model a scope, a clock recovery reference or random jitter with Gaussian tails.
6 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 1 steps and the final capstone