Based on the results of my recent work and studies, I have summarised the following methods for applying timing constraints, listed in order from easiest to most advanced:
0. Core Frequency Constraint
This is the most basic constraint, so it is numbered 0.
1. Core Frequency Constraint + Timing Exception Constraints
Timing exception constraints include False Path, Multicycle Path, Max Delay, and Min Delay. However, this is still not a complete set of timing constraints. If only these constraints are applied, it indicates that the designer's thinking is still limited to the internal logic of the FPGA chip.
2. Core Frequency Constraint + Timing Exception Constraints + I/O Constraints
I/O constraints include pin assignment locations, drive strength for unused pins, external trace delays (Input Delay, Output Delay), pull‑up/pull‑down resistors, drive current strength, etc. Timing constraints that include I/O constraints are complete timing constraints. As a component on the PCB, the FPGA is part of the overall system timing convergence. As part of the PCB design, the FPGA must be treated by the PCB design engineer like any other COTS device – its I/O Timing Diagram must be read and analysed. The difference between an FPGA and a COTS device is that the FPGA's I/O timing can be adjusted within a certain range late in the design cycle. Nevertheless, it is still best to give full consideration to I/O timing during the early PCB design phase and document it properly.
Precisely because the FPGA's I/O timing may change during the design process, accurately constraining it is an important factor in ensuring a stable and controllable design. Many problems where an FPGA's operation with external devices becomes unstable after a re‑compile can often be traced back to this.
3. Core Frequency Constraint + Timing Exception Constraints + I/O Constraints + Post‑fit Netlist
The introduction of a Post‑fit Netlist starts from a successful timing closure result. A specific set of logic (Design Partition) is fixed in terms of its placement and routing results (Netlist) on the FPGA, ensuring that these results can be reproduced in a new compilation – and thus the timing closure of that logic is also guaranteed. This process of preserving the previous compilation result is Incremental Compilation. The type of netlist retained and the degree of preservation can both be configured, and are not limited to Post‑fit Netlists, allowing the corresponding retention strength and optimisation effects to be achieved. With the strong support of EDA tools, even though this is a fine‑grained, gate‑level constraint, the designer only needs to perform a series of configuration operations without concerning themselves with specific placement and routing details. Because gate‑level constraints are too numerous to be saved in the qsf file, the retained netlist can be exported as a Partial Netlist to a separate qxp file, which, together with the coarse configuration information in the qsf file, completes the incremental compilation.
4. Core Frequency Constraint + Timing Exception Constraints + I/O Constraints + LogicLock
LogicLock is a placement constraint applied at the FPGA device floorplan level. LogicLock constraints are coarse‑grained: they only specify the placement boundaries and sizes (LogicLock Regions) within which the top‑level module or sub‑modules can be adjusted. A successful LogicLock implementation requires the designer to anticipate possible timing closure targets, consider the timing impact of the spatial relationship between specific logic resources (pins, memory, DSPs) and the LogicLock Regions, and refer to previous successful timing closure results. This process of weighing and planning the FPGA's physical layout is FloorPlanning. LogicLock gives the designer more control over placement locations and ranges, effectively conveying design intent to the EDA tool and preventing it from blindly optimising non‑critical paths due to a lack of placement priority information. Because the placement of modules in each compilation is confined to an optimal, fixed range, the reproducibility of timing closure results is higher. Due to its coarse‑grained nature, LogicLock constraints are not numerous and can be retained in the qsf file.
It should be noted that methods 3 and 4 are often used together – that is, the LogicLock Region specified by FloorPlanning can be treated as a Design Partition for Incremental Compilation. This is the reason why these two methods are often confused.
5. Core Frequency Constraint + Timing Exception Constraints + I/O Constraints + Register Placement Constraints
Register placement constraints are fine‑grained placement constraints that go down to the register or LE level. The designer applies precise control to achieve reliable timing closure. Manually constraining the placement of every register in the design to guarantee timing closure is a monumental task. It represents the ideal of having complete control over the physical implementation of the design – an ideal that is impossible to achieve within a limited time frame. The usual practice is to apply register placement constraints to specific parts of the design and run the placement and routing tools to obtain timing closure information, iterating several times to approach the desired timing target.
I recently saw a design where every register in a sub‑module had a specific placement constraint. Consequently, timing closure for that module was guaranteed in every re‑compile. Upon analysis, the sub‑module's design and constraints were originally developed in a schematic, and after achieving timing closure, the design was converted to HDL, with the corresponding constraints also saved in the configuration file.
6. Core Frequency Constraint + Timing Exception Constraints + I/O Constraints + Specific Path Delay Constraints
Good timing constraints should be "guiding" rather than "forcing." By specifying the timing delay ranges of the critical paths in the design, the detailed work is left to the EDA tool to implement freely within those constraints. This is also an ideal goal. It requires the designer to be fully aware of every timing path, to distinguish which paths can be closed with core frequency and simple timing exception constraints, and to identify those paths that require MaxDelay and MinDelay constraints – without missing a single one. It also demands strong and "intelligent" support from the EDA tool. Setting path‑delay constraints is indirectly setting placement and routing constraints, but it is more flexible than the methods described in 3, 4, and 5, while maintaining accuracy. Achieving timing closure through timing constraints – rather than explicit placement and netlist constraints – is the true essence of timing constraint methodology.
I recall a fellow engineer once saying, "Good timing is designed, not constrained." I have always taken this as my guiding principle for logic design and timing constraints. Good constraints must be based on a good design. Without a good design, no amount of effort in applying constraints is meaningful. However, correct constraints can also be used to check the quality of the design. The timing analysis report can identify areas where timing considerations are insufficient, allowing them to be corrected. Through several iterations of "analysis – modification – analysis," the design can be perfected. In essence, the design is the foundation of the constraints, and the constraints are the guarantee of the design – the two complement each other.
About Maxipcb