Thank you for your interest in our webinar on design strategies for minimizing jitter and insertion loss. It will be presented by Mike Schnecker of Rohde & Schwarz and Amit Bahl.
Missed the webinar? Click the link below and watch the recording!
You can ask your questions on this thread.
| Questions | Answers |
|---|---|
| Can the impedance calculator include temperature effects? (the impedances of the same transmission line at different temperatures—10m K, 4.2K, 77K, and 300K) | No, however, the impedance calculator does take into account the trace’s trapezoidal shape and the effect of multiple dielectric materials. |
| How much improvement in transmission is gained by the use of differential pairs? | Differential pairs offer significant advantages in signal transmission, primarily by effectively rejecting common-mode noise. This is crucial because many sources of interference, such as switching power supplies and electromagnetic interference (EMI), exhibit common-mode characteristics. Differential signaling significantly enhances the signal-to-noise ratio (SNR) and reduces jitter by utilizing the difference between two signals on a pair of wires. Quantifying the exact improvement in decibels (dB) is challenging as it heavily depends on the specific environment and noise sources. However, differential signaling is generally the preferred method for high-speed data transmission, especially at rates exceeding 16 gigahertz. |
| I’m trying to choose between stripline and GCPW for a 16 GHz clock signal (sine wave, not square wave). In my design, I have a probe sitting right above these clock paths, and the signal data rate in the probe is very sensitive to any crosstalk. Now, the concern is: 1. If I use GCPW, then I am concerned about the crosstalk via radiation between GCPW on PCB and the probe. 2. If I use stripline, then radiation gets substantially suppressed, but extra via transition may introduce extra reflections and potentially degrade the jitter. But since this is a sine wave, I think reflection shall not be a huge concern unless I accidentally made the transmission line quarter-wavelength trace. But in this case, between the b/w stripline and GCPW, what other factors shall I consider, and what is the typical practice?? |
Reflections can occur with sine wave signals as well. To minimize reflections, Proper termination is crucial. Determine the load impedance (IC terminal impedance) and ensure it matches the transmission line impedance. If necessary, use a termination resistor at the load. Source impedance matching is also important. Add a small series resistor at the source to match the source impedance to the transmission line impedance and reduce ringing. For stripline: Maintain impedance continuity throughout the signal path. Design vias with a characteristic impedance that closely matches the transmission line impedance. For GCPW: Employ a large ground plane with ground stitching around the trace to effectively reduce radiation. |
| If we use 112 Gb/s, what is the frequency at which we are looking at insertion loss, and by how much? | For 112 Gb/s NRZ signaling, the fundamental frequency is approximately 56 GHz. Insertion loss at this frequency is critical. While the rule of thumb suggests analyzing up to five times the fundamental frequency (280 GHz in this case), examining the insertion loss up to 100 GHz should provide sufficient information for most practical applications. However, a more detailed analysis at the fundamental frequency of 56 GHz is highly recommended. |
| Does the accuracy of the step response method you showed depend on the load matching on the receiver side? Is that a potential weakness compared to the normal pattern recovery method? | Yes, the accuracy of the step response method is dependent on the load matching at the receiver. Typically, these measurements are performed with a 100-ohm differential load to match common transmission line impedances. However, this load-matching dependency is not unique to the step response method. The traditional pattern recovery method also inherently includes the reflection effects of the measurement equipment (e.g., scope input). The difference lies in the ability to directly observe and analyze the source signal and its interaction with the load in the step response method. A significant advantage of the step response method is its exceptional ability to accurately measure signals with severe distortion (stressed eyes). This is primarily because the method utilizes the entire waveform data for analysis. |
| How often do you want to stitch the guard trace to GND on the PCB? | Stitch vias are typically placed at intervals of approximately lambda/4, where lambda represents the wavelength of the highest frequency signal on the PCB. |
| What happens if the signal rise time exceeds 6 inches on the PCB? Will it lose one cycle? | When the signal rise time exceeds 6 inches on a PCB, it suggests operation within the lumped element regime. In this regime, distributed effects such as reflections are less prominent. While this simplification can make design easier, it doesn’t directly imply cycle loss. Cycle loss is typically associated with factors like excessive jitter, signal distortion, and inadequate equalization. |
| At what bit data rate, do the traces need to be externally routed versus internal layers? | Data rate itself does not directly determine whether traces should be routed externally (on the top layers) or internally (on buried or inner layers) of a PCB. Both external and internal layers can accommodate high-speed signals. |
| For via fencing, what is your recommended via-to-via spacing relative to the wavelength of interest? | For via fencing, a common guideline is to space the vias approximately lambda/4 apart, where lambda is the wavelength of the signal. However, this is a simplified rule and the optimal spacing depends on factors like signal frequency, PCB stackup, trace width, and design complexity. |