PAM Data Eye Reference Voltage Adaptation for Decoding Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
In graphics memory systems transitioning from non-return to zero (NRZ) to pulse amplitude modulation (PAM) communication, the unsymmetrical multiple data eyes lead to direct current (DC) offsets, causing errors in eye decoding due to differing power distribution paths between devices, resulting in non-optimal eye decoding and degraded performance.
Innovation Solution
A system that trains each data eye separately to find the optimal reference voltage (Vref) for pulse-amplitude modulation (PAM) signaling, adjusting Vref levels to center them vertically within each eye, allowing for independent setting of voltage references for improved decoding accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single common reference voltage is used for all data eyes, then the system is simple to implement, but decoding accuracy degrades due to DC offsets from unsymmetrical data eyes and different power distribution paths
Solution Approach 1:
The patent divides the single common reference voltage into multiple separate reference voltages, one for each data eye. This segmentation allows each reference voltage to be independently optimized for its specific data eye, compensating for DC offsets caused by unsymmetrical eye shapes and different power distribution paths, thereby improving decoding accuracy without excessive complexity
Solution Approach 2:
The patent applies local quality by providing each data eye with its own tailored reference voltage that is specifically optimized for that eye's characteristics. Instead of using a uniform reference voltage for all eyes, each reference voltage is locally adapted to match the specific DC offset and symmetry properties of its corresponding data eye, improving overall system performance
2Reliability
If separate reference voltages are trained for each data eye, then decoding accuracy improves, but the system complexity and training requirements increase
Solution Approach 1:
The training process is segmented into separate training operations for each data eye. Each reference voltage is trained independently using its own training sequence, allowing the system to optimize each eye's reference voltage without interference from other eyes. This segmented approach improves decoding reliability while keeping the training methodology systematic and manageable
Solution Approach 2:
The patent performs preliminary training actions by establishing separate reference voltages for each data eye before actual data transmission begins. This preliminary configuration of eye-specific reference voltages ensures that the decoding system is pre-adapted to the specific characteristics of each data eye, improving reliability from the start of operations
3Reliability
If reference voltages are centered at the vertical center of unsymmetrical data eyes, then bit error rate performance improves, but power distribution matching requirements become more stringent
Solution Approach 1:
The patent changes the reference voltage parameter from a single common value to multiple eye-specific values. By adjusting each reference voltage to be centered at the vertical center of its corresponding unsymmetrical data eye, the system improves bit error rate performance. This parameter change compensates for variations in power distribution paths, reducing the stringency of power delivery matching requirements between different devices
4Reliability
If multiple separate voltage references are used for different data eyes, then eye margins improve, but the power delivery network complexity increases
Solution Approach 1:
The power delivery network is effectively segmented by providing separate reference voltages for different data eyes. This segmentation allows each reference voltage to be optimized for its specific eye, improving eye margins and reliability. The segmentation approach manages power delivery network complexity by organizing the multiple references in a systematic manner that corresponds to the segmented training and decoding processes
Data Source
AI summary
Inter-device communication with a pulse-amplitude modulation (PAM) signal can have at least two data eyes with different Vref levels. A physical interface (PHY) can be trained for the PAM signal by training a first data eye separately from the second data eye. The training can include adjusting the first Vref level separately from the second Vref level to center each reference voltage on its respective data eye.