Categories

Analysis of Timing Analysis and Simulation Strategies in High‑Speed PCB Design

In the field of network communications – including ATM switches, core routers, Gigabit Ethernet, and various gateway devices – system data rates and clock rates continue to increase, and processor operating frequencies are rising accordingly.
Aug 20th,2026 66 Views

In the field of network communications – including ATM switches, core routers, Gigabit Ethernet, and various gateway devices – system data rates and clock rates continue to increase, and processor operating frequencies are rising accordingly. Transmission speeds for data, voice, and video have already far exceeded 500 Mbps, and backplanes operating at hundreds of megabits or even several gigabits are becoming common. The increase in digital system speed means that signal rise and fall times are becoming as short as possible. The high‑speed design issues arising from higher digital signal frequencies and faster edge rates are also becoming increasingly prominent. When the interconnect delay of a signal exceeds 20 % of the edge‑signal transition time, the signal traces on the board begin to exhibit transmission‑line effects, and the design is considered a high‑speed design. The emergence of high‑speed issues brings greater challenges to hardware design. Many designs that are logically correct can fail entirely if handled improperly during the actual PCB layout – a situation that is even more pronounced in the ever‑faster network communications field. Experts predict that in future hardware circuit design costs, the expense associated with logic‑function design will shrink dramatically, while costs related to high‑speed design will account for 80 % or more of the total. High‑speed issues have become a critical factor in the success or failure of system design.

Signal over‑shoot, under‑shoot, reflection, ringing, and crosstalk caused by high‑speed issues can severely affect the normal timing of the system. Reduced timing margins force designers to pay close attention to phenomena that affect the timing and quality of digital waveforms. When higher speeds tighten timing constraints, no matter how thoroughly the system principles are understood beforehand, any oversight or simplification can have serious consequences for the system. In high‑speed design, timing issues are even more critical. This article focuses on timing analysis and simulation strategies in high‑speed design.


1. Timing Analysis and Simulation for Common‑Clock Synchronous Systems

In high‑speed digital circuits, data transmission is generally controlled by clocks that send and receive data in an orderly manner. Chips can only transmit and receive data according to specified timing; excessive signal delays or poor delay matching can cause timing violations and functional failures. In low‑speed systems, interconnect delay and ringing can be ignored because signals have sufficient time to settle. However, in high‑speed systems, faster edge rates and higher system clock rates shorten both the propagation time between devices and the setup time. The equivalent capacitance and inductance of transmission lines also introduce delay and distortion in digital transitions. Combined with factors such as signal‑delay mismatch, these effects can affect the chip's setup and hold times, preventing the chip from correctly sending and receiving data and thus causing the system to malfunction.

In a common‑clock synchronous system, the driver and receiver on the bus share the same clock source. Under the action of the same‑phase clock output from a common clock buffer, data transmission and reception are completed. Figure 1 shows a typical common‑clock synchronous data‑transmission scheme. In Figure 1, the crystal oscillator (CRYSTAL) produces the output signal CLK_IN, which reaches the clock distributor (CLOCK BUFFER). After distribution and buffering, two in‑phase clocks are generated: one is CLKB, used for the DRIVER's data output; the other is CLKA, used for sampling and latching the data sent from the DRIVER to the RECEIVER. Clock CLKB reaches the DRIVER after a flight time Tflt_CLKB. The internal data of the DRIVER is latched by CLKB and appears at the DRIVER output after a delay of TCO_DATA. The output data then travels through a flight time Tflt_DATA to the RECEIVER input. At the RECEIVER input, the other clock CLKA generated by the CLOCK BUFFER (with a delay equal to the CLKA flight time Tflt_CLKA) samples and latches this data, completing the data transfer for one clock cycle in the common‑clock system.

The above process shows that the data arriving at the RECEIVER is sampled on the rising edge of the next clock cycle. Two necessary conditions must be satisfied for correct data transfer:

① The data at the RECEIVER input generally has a required setup time Tsetup, meaning that the data must be valid a minimum time before the clock edge. The data signal must arrive at the input early enough relative to the clock. From this, an inequality for the setup‑time requirement can be derived.

② To successfully latch data inside the device, the data signal must remain valid at the receiver input for a sufficient period to ensure correct sampling and latching by the clock – this period is called the hold time. The delay of CLKA must be less than the invalid time of the data (INVALID). From this, an inequality for the hold‑time requirement can be derived.

1.1 Timing Analysis for Data Setup Time

From the first condition, the data signal must arrive at the receiver before clock CLKA to correctly latch the data. In the common‑clock bus, the first clock cycle latches data to the DRIVER output, and the second clock cycle latches data into the RECEIVER. This means that the data signal must arrive at the RECEIVER input sufficiently early relative to clock CLKA. To meet this condition, it is necessary to determine the delays of the clock and data signals to the RECEIVER and ensure that the receiver setup‑time requirement is met. Any additional time beyond the required setup time is the setup‑time margin Tmargin.

In the timing diagram of Figure 1, all arrowed paths represent delays of data and clock signals either inside the chip or on the transmission lines. The lower arrow path represents the total delay from the first valid clock edge to the data arrival at the RECEIVER input. The upper arrow path represents the total delay of the receiving clock CLKA. The total delay from the first valid clock edge to data arrival at the RECEIVER input is:

TDATA_DELAY = TCO_CLKB + Tflt_CLKB + TCO_DATA + Tflt_DATA

The total delay for the receiving clock CLKA in the next cycle is:

TCLKA_DELAY = TCYCLE + TCO_CLKA + Tflt_CLKA

To satisfy the data setup time, the following must hold:

TCLKA_DELAY_MIN – TDATA_DELAY_MAX – Tsetup – Tmargin > 0

Expanding and considering factors such as clock jitter Tjitter, we obtain:

TCYCLE + (TCO_CLKA_MIN – TCO_CLKB_MAX) + (Tflt_CLKA_MIN – Tflt_CLKB_MAX) – TCO_DATA_MAX – Tflt_DATA_SETTLE_DELAY_MAX – Tjitter – Tsetup – Tmargin > 0 (1)

In equation (1), TCYCLE is one clock cycle period. The first bracketed term is the maximum phase difference between the output clocks CLKA and CLKB from the clock buffer chip – commonly referred to in datasheets as output‑to‑output skew. The second bracketed term is the maximum delay difference between the two clock outputs CLKA and CLKB from the clock buffer chip as they travel to the RECEIVER and DRIVER, respectively. TCO_DATA is the time interval measured under specified test load and test conditions, from the clock trigger edge to the moment the data appears at the output port and reaches the test voltage Vmeas (or VREF) threshold. The value of TCO_DATA is directly related to the internal logic delay, output buffer characteristics, and output loading conditions, and can be found in the chip's datasheet.

From equation (1), the only adjustable terms are essentially two: (Tflt_CLKB_MIN – Tflt_CLKB_MAX) and Tflt_DATA_SETTLE_DELAY_MAX. For meeting setup time, Tflt_CLKA_MIN should be as large as possible, while Tflt_CLKB_MAX and Tflt_DATA_SETTLE_DELAY_MAX should be as small as possible. In other words, the receiving clock should arrive later, and the data should arrive earlier.

1.2 Timing Analysis for Data Hold Time

To successfully latch data inside the device, the data signal must remain valid at the receiver input for a sufficient period to ensure correct sampling – this is the hold time. In the common‑clock bus, the receiver buffer latches data on the second clock edge, while at the same time, the driver latches the next data onto the transmitting end. To satisfy the receiver hold time, the valid data must be latched into the receiver flip‑flop before the next data arrives. This requires that the delay of the receiving clock CLKA be smaller than the delay of the data signal.

TCLKA_DELAY = TCO_CLKA + Tflt_CLKA

Data delay:

TDATA_DELAY = TCO_CLKB + Tflt_CLKB + TCO_DATA + Tflt_DATA_SWITCH_DELAY

To satisfy the data hold time, the following must hold:

TDATA_DELAY_MIN – TCLKA_DELAY_MAX – Thold – Tmargin > 0

Expanding and considering clock jitter Tjitter and other factors, we obtain:

(TCO_CLKB_MIN – TCO_CLKA_MAX) + (Tflt_CLKB_MIN – Tflt_CLKA_MAX) + TCO_DATA_MIN + Tflt_DATA_SWITCH_DELAY_MIN – Thold – Tmargin – Tjitter > 0 (2)

In equation (2), the first bracketed term is still the maximum phase difference between the output clocks from the clock buffer chip. The second bracketed term can again be understood as the maximum delay difference between the two clocks CLKA and CLKB as they travel to the RECEIVER and DRIVER, respectively. To satisfy the hold time, the only adjustable terms are again two: (Tflt_CLKB_MIN – Tflt_CLKA_MAX) and Tflt_DATA_SWITCH_DELAY_MIN. From the perspective of meeting hold time, Tflt_CLKB_MIN and Tflt_DATA_SWITCH_DELAY_MIN should be as large as possible, while Tflt_CLKA_MAX should be as small as possible. In other words, to satisfy hold time, the receiving clock should arrive earlier, and the data should become invalid later.

To receive data correctly, both setup time and hold time must be satisfied simultaneously – i.e., both equations (1) and (2) must be met. Analysing these two inequalities shows that there are only three adjustment paths: transmit clock delay, receive clock delay, and data delay. The adjustment strategy can be as follows: first assume that the transmit clock delay exactly equals the receive clock delay – i.e., (Tflt_CLKA_MIN – Tflt_CLKB_MAX) = 0 and (Tflt_CLKB_MIN – Tflt_CLKA_MAX) = 0 (the timing deviation caused by this assumption will be considered later). Then, through simulation, determine the feasible range for data delay. If no solution exists for data delay, return to the two equalities and adjust either the transmit clock delay or the receive clock delay.

An example is the GLINK bus common‑clock synchronous data transfer in a broadband network switch. First assume that the transmit clock delay exactly equals the receive clock delay, then determine the data delay range. Substituting the parameters, equations (1) and (2) become:

1.5 – Tflt_DATA_SETTLE_DELAY_MAX – Tmargin > 0

0.5 + Tflt_DATA_SWITCH_DELAY_MIN – Tmargin > 0

Guided by these inequalities and combined with the actual PCB layout, we determine Tflt_DATA_SETTLE_DELAY_MAX < 1.1 and tflt_data_switch_delay_min > –0.1, leaving a 0.4 ns margin allocated to the clock skew and Tmargin. The topology is extracted in SPECCTRAQUEST, and signal‑integrity simulations are performed to determine the segment lengths and topology. A full‑scan simulation over all 12 combinations yields Tflt_DATA_SETTLE_DELAY_MAX = 1.0825 ns and Tflt_DATA_SWITCH_DELAY_MIN = –0.0835004 ns, which are within the specified ranges of 1.1 and –0.1. From this, the constraint rules for the GLINK data bus can be derived: ① the delay from the termination resistor to the driver should not exceed 0.1 ns; ② data lines must be matched to within 0.1 ns – i.e., each data line must fall between 0.65 ns and 0.75 ns. These constraints guide the routing process.

Now consider the impact of the rigid assumptions Tflt_CLKA_MIN – Tflt_CLKB_MAX = 0 and Tflt_CLKB_MIN – Tflt_CLKA_MAX = 0. Pre‑constrain the transmit and receive clocks to be exactly equal in length (matched to within 0.02 ns in practice). Under the CADENCE environment, clock simulation yields |Tflt_CLKA_MIN – Tflt_CLKB_MAX| < 0.2 and |tflt_clkb_min – tflt_clka_max| < 0.2. Thus, the remaining margin for Tmargin is 0.2 ns.

The final simulation results are: ① the delay from the termination resistor to the driver should not exceed 0.1 ns; ② data lines must be matched to within 0.1 ns, i.e., each data line must fall between 0.65 ns and 0.75 ns; ③ the transmit and receive clocks must be matched to within 0.02 ns for equal length; ④ Tmargin = 0.2 ns. With the topology template and constraint rules established, they can be imported into SPECCTRAQUEST or ALLEGRO via CONSTRAINS MANAGER. Once these design constraints are set, the auto‑router can be used for rule‑driven automatic routing, or manual routing can be performed with the constraints in place.


2. Source‑Synchronous Timing Relationships and Simulation Example

Source‑synchronous means that the clock strobe signal CLK is sent by the driver along with the accompanying data, unlike common‑clock synchronous where a separate clock source is used. In source‑synchronous data transfer, the data is first sent to the receiver, and after a slightly shorter delay, the strobe clock is sent to the receiver to sample and latch the data. A schematic diagram is shown in Figure 2. Timing analysis for source‑synchronous is simpler than for common‑clock synchronous; the analysis method is very similar. The analysis formulas are given directly below:

Setup time:
Tvb_min + (Tflt_clk_min – Tflt_data_settle_delay_max) – Tsetup – Tmargin > 0

Hold time:
Tva_min + (Tflt_data_switch_delay min – Tflt_clk _max) – Thold – Tmargin > 0

where Tvb is the driver's setup time, indicating how long before the clock edge the driver data becomes valid; Tva is the driver's hold time, indicating how long after the clock edge the driver data remains valid. The other parameters have the same meanings as before.

The following uses the TBI interface, which is very common in communication circuits, as an example to illustrate the source‑synchronous timing analysis and simulation process. The TBI interface mainly includes one transmit clock and 10 bits of transmit data, as well as two receive clocks and 10 bits of receive data. RBC0 and RBC1 are the two receive clocks. In Gigabit Ethernet, these two clocks operate at 62.5 MHz with a 180° phase difference, and the rising edges of the two clocks are used alternately to latch data. Using the timing parameters from the data sheet and substituting into the above equations gives:

2.5 + (Tflt_clk_min – Tflt_data_settle_delay_max) – 1 – Tmargin > 0

1.5 + (Tflt_data_switch_delay min – Tflt_clk _max) – 0.5 – Tmargin > 0

Following the analysis method described earlier: first assume that the flight times of clock and data signals are exactly equal – i.e., the clock and data are perfectly matched – and then analyse the impact of mismatch. The equations become:

1.5 – Tmargin > 0

1 – Tmargin > 0

It can be seen that there is a large margin for both setup and hold times. After simulation, it is found that even when the clock and data are exactly matched in length (matched to within 0.02 ns), there is still a 0.3 ns difference:

(Tflt_clk_min – Tflt_data_settle_delay_max) < 0.3

(Tflt_data_switch_delay min – Tflt_clk_max) < 0.3

Taking Tmargin = 0.5 ns gives a clock‑to‑data matching requirement of 0.2 ns – i.e., the length matching between clock and data should not exceed 0.2 ns.

In the actual simulation, signal‑integrity analysis is first performed for both clock and data signals, and proper termination matching is used to obtain good received waveforms. Figure 3 compares the different simulation waveforms for a clock line with passive termination and with active termination. This shows the necessity of performing signal‑integrity simulation first.

In common‑clock synchronous systems, data transmission and reception must be completed within one clock cycle. Device delays and PCB trace delays also limit the maximum theoretical operating frequency of common‑clock buses. Therefore, common‑clock synchronous is generally used for transfer rates below 200 MHz to 300 MHz. For rates higher than this, source‑synchronous techniques are generally adopted. Source‑synchronous works in a relative clocking scheme, with data and clock transmitted in parallel; the transfer rate is mainly determined by the timing difference between data and clock signals, allowing the system to achieve higher transmission rates.

Through signal‑integrity analysis, timing analysis, and simulation of the host and sub‑cards of a broadband Ethernet switch, the product design cycle was greatly shortened. The analysis and simulation effectively addressed signal‑integrity and timing issues encountered in high‑speed design, fully ensuring both design quality and design speed, truly achieving first‑pass success of the PCB boards. The main board and sub‑cards have passed debugging and have been successfully transitioned to mass production.

About Maxipcb

Maxipcb empowers innovators to turn cutting-edge technologies into reality.We offer one-stop solutions for design, simulation, testing, PCB manufacturing, component procurement and SMT assembly, enabling efficient development, rapid deployment and risk control across the full product lifecycle.
Serving the world in communications, industrial automation, aerospace, automotive, semiconductor and beyond, we build a safer, more connected future together.