There are several new enhancements associated with the 16.6 PCB SI Channel Analysis (CA).
Read on for more details …
SigXplorer has been enhanced to provide greater flexibility associated with AMI model management. Experimentation with buffer and AMI combinations, until this release, required joining them at the library level. No manipulation could be done in SigXplorer. Now, the 16.6 release provides canvas-level editing options, allowing you the ability to join any packaged IBIS buffer to any existing AMI model. Additionally, the influence from the AMI portion of the buffer models can be suspended (ignored) without deleting the AMI model entirely. For convenience, the AMI enable/disable as well as the AMI add/remove are all directly accessible from the SigXplorer canvas.
Adding AMI models
To add an AMI model to an existing buffer, right click on the buffer and select Add AMI Model:
Similarly, you can remove an AMI model by right clicking on the buffer and selecting Remove AMI Model:
Enabling and disabling AMI models
SIgXplorer provides you with a command to disable and enable an AMI model on an IBIS device, Tx or Rx. After disabling the AMI model on a device, you can run simulations without taking the AMI model into account. The Disable AMI command, which appears on the pop-up menu when you right-click a device, can be used when you want to establish a baseline simulation result with the analog buffers and the channel but without the AMI model.
When you disable an AMI model, a slanted line appears across the AMI model icon graphic indicating that it is disabled:
You can re-enable the AMI model on a device using the Enable AMI command, which appears on a device on which AMI was previously disabled:
GUI overhaul
The user interface has been completely updated in the 16.6 release. On-screen guidance is included both in the form of the task description block and as a 1-line “reminder” when the mouse hovers over a command.
All of the commands have been organized by function, with each function represented as a tab on the form. Settings can be saved and retrieved for all tabs simultaneously using a configuration file. Each option has been assigned a meaningful default value to enable quick simulations from the Action buttons along the base of the new GUI. Incremental improvements have been added in every functional tab. Characterization has been enhanced to allow import from third-party simulators. The outputs from Channel Analysis are now configurable and include new options for Time Domain and Voltage BER (bathtub) curves. Quick launch buttons combined with “.cacfg” file, and “.top” files yield three-click net checks.
Characterization tab
This tab contains options to set up the characterization. You are now able to import a characterization using the Import Characterization option. The characterization is represented as an impulse response. The input file is an ASCII file with two columns of data. The first column contains a time and the second column contains a voltage:
Stimulus tab
Preferences tab
Output tab
You are now able to see a snapshot of the time domain waveform after it has been processed by the AMI models. The input stimulus waveform along with the clock ticks will also be saved for the time duration specified. You will be warned if you select more than 5000 bits at a time as selecting a longer bit duration will adversely impact both the simulation time and the size of the files. The start bit must be a number between 0 and the total number of bits. The end bit must be a number between the start bit and the total number of bits. The output waveforms will be saved in text files and saved in the results directory.
Vertical bathtub curves that show the voltage (vertical axis) opening of the eye have been added to the horizontal bathtub curves (representing the eye width in the UI). The bathtub curves predict BER up to the desired limit (default 1/10e15):
Advanced features tab
Crosstalk implementation
One of the most important requirements of high-speed simulation is to be able to perform accurate and reliable crosstalk simulation for a topology/layout with multiple driver/receiver pairs that are electrically coupled to each other. The presence of AMI models on each of the drivers and receivers mean that the effect of equalization (or any other algorithmic effect) must be accurately captured in the simulations on a particular victim receiver from adjacent aggressors.
Channel simulation with multiple electrically coupled Tx/Rx pairs is possible with reasonable accuracy. Each channel (both aggressors and victim) is characterized individually. Simulations are performed to obtain the waveform results for the victim Rx. To perform the simulation for the victim Rx, the impulse matrix for the channel constitutes the impulse response of the primary victim as well as the crosstalk impulse responses from the aggressors. The crosstalk impulse responses of the aggressors are subjected to the Init functions of the aggressors and they can be modified by the Init functions of the aggressors. The crosstalk for AMI models/simulation is illustrated in the following example with three Tx/Rx pairs:
In this example, the middle channel (Tx2/Rx2) is the victim channel, while the other two channels (1 and 3) are aggressor channels. IR1_1, IR2_2, and IR3_3 represent the primary channel impulse responses for the three channels, while IR1_2 and IR3_2 represent the impulse responses of the crosstalk channel from Tx1 to Rx2 and Tx3 to Rx2, respectively. The impulse matrix presented to the init function of Tx1 will have IR1_1 and IR1_2 in the first and second columns respectively. Similarly, the impulse matrix for the init function of Tx3 will have IR3_3 and IR3_2. Depending on the AMI model, the init functions of the Txs may or may not modify the IRs presented to them. Assuming that they do, the crosstalk IRs will change to IR1_2' and IR3_2'. These two IRs are used to construct the final IR for the victim Rx which will have IR2_2', IR1_2', and IR3_2' in the first, second, and third column of the impulse matrix. Any AMI simulation with this IR will incorporate the crosstalk effects of the AMI models as well as the traditional crosstalk.
I look forward to your feedback from using these new capabilities.
Jerry “GenPart” Grzenia