DDR3 FPGA Read Leveling With Delay Calibration

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

The increasing data transfer rates between memory devices and FPGAs outpace the memory devices' capacity, leading to data recovery errors due to suboptimal timing between phase-shifted DQS signals and system clocks in DDR3 memory interface standards, which conventional techniques like first-in-first-out memories fail to adequately address.

Innovation Solution

Implementing leveling elements, such as delay elements or intermediate registers, between input and output registers to calibrate and adjust the delay of input signals, ensuring error-free data transfer by averaging minimum and maximum delays for optimal phase alignment of DQS and DQ signals with the system clock.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional techniques like first-in-first-out memories are used to retiming signals, then data transfer between memory devices and FPGA is enabled, but the circuit size and complexity increase significantly

Engineering Contradiction:
Improvedata transfer reliabilityVSAvoidcircuit complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent extracts the essential retiming function from complex FIFO memory circuits and implements it using simple delay elements and phase-shifted clocks. By taking out only the necessary timing adjustment capability, the solution achieves reliable data transfer without the bulk and complexity of conventional FIFO implementations.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The patent changes the timing parameters of the clock signal by introducing phase-shifted versions of the system clock. By adjusting the phase relationship between clocks and data signals through delay elements, the patent achieves proper synchronization without requiring complex memory-based retiming circuits.

Inventive Principle:
Principle #35Parameter changes

2Stability of the object's composition

If the system clock is routed using fly-by topology to memory devices, then signal integrity is improved, but the DQ and DQS signals become asynchronous with the system clock

Engineering Contradiction:
Improvesignal stabilityVSAvoiddata recovery reliability
Core Design Contradiction:
Stability of the object's compositionVSReliability

Solution Approach 1:

The patent applies preliminary phase adjustment to the clock signals before they are used to sample the asynchronous DQ and DQS signals. By pre-shifting the clock phases to account for the fly-by topology delays, the patent ensures that data sampling occurs at optimal timing points, preventing recovery errors before they can occur.

Inventive Principle:
Principle #10Preliminary action

Solution Approach 2:

The patent introduces phase-shifted clock signals as intermediary timing references between the fly-by routed system clock and the data sampling operations. These intermediary clocks act as mediators that bridge the timing gap created by the fly-by topology, enabling reliable data recovery despite the asynchronous nature of the signals.

Inventive Principle:
Principle #24Intermediary (Mediator)

3Productivity

If data is transferred at high speeds between memory devices and FPGA, then data throughput is improved, but timing errors and data recovery errors increase

Engineering Contradiction:
Improvedata throughputVSAvoiddata recovery accuracy
Core Design Contradiction:
ProductivityVSReliability

Solution Approach 1:

The patent implements dynamic timing adjustment by using multiple phase-shifted clock signals that can be selectively applied based on the actual timing requirements of each data group. This dynamic approach allows the system to adapt to varying timing conditions at high speeds, maintaining data recovery accuracy while maximizing throughput.

Inventive Principle:
Principle #15Dynamics

Solution Approach 2:

The patent uses periodic phase-shifted clock signals to sample data at multiple timing points throughout each clock cycle. By periodically sampling data with different phase offsets, the system can reliably capture high-speed data transitions regardless of their exact timing, thereby maintaining data recovery accuracy at increased throughput rates.

Inventive Principle:
Principle #19Periodic action

Data Source

PatentUS7593273B2Read-leveling implementations for DDR3 applications on an FPGA
Publication Date: 2009.09.22 ALTERA CORP
  • US7593273B2 patent drawing
  • US7593273B2 patent drawing
  • US7593273B2 patent drawing

AI summary

Circuits, methods, and apparatus for transferring data from a device's input clock domain to a core clock domain. One example achieves this by using a retiming element between input and core circuits. The retiming element is calibrated by incrementally sweeping a delay and receiving data at each increment. Minimum and maximum delays where data is received without errors are averaged. This average can then be used to adjust the timing of a circuit element inserted in an input path between an input register clocked by an input strobe signal and an output register clocked by a core clock signal. In one example, an input signal may be delayed by an amount corresponding to the delay setting. In other examples, each input signal is registered using an intermediate register between the input register and the output register, where a clock signal is delayed by an amount corresponding to the delay setting.