comment on this article

A question of time

The data rate may be fast, but how quick is the design time?

Current fpgas and highly integrated processors generally support a range of high speed interface standards, many of which are LVDS based or which can be configured to LVDS. These interfaces can be used to implement multiple gigabit per second data links. However, as data rates and link lengths keep increasing, practical implementations of these links may require alternative architectures.
Common LVDS type links will run out of steam at data rates of more than 1.5 to 2Gbit/s. Techniques like pre emphasis at the transmit side and equalisation at the receiving end will help keep signal integrity up, but they will go only so far. For higher bandwidths, it may be better to move to a higher speed physical layer, such as current mode logic (CML).
Many designs opt to spread data over multiple lanes, effectively using several high speed links to implement the link. Common implementations are the well known FPD-Link type of interfaces. Here, multiple data lanes and a single clock lane are used to transfer the payload.
A common issue associated with this architecture is the data to clock skew. By definition, one cannot have too much skew between each data and the clock link. This requires careful layout and design, especially if there are many data links. At higher data rates, it limits the maximum cable length.
A more design friendly implementation with respect to skew tolerance – hence ease of layout – would be to use a single data link with the clock embedded in the data stream. By definition, embedding the clock into the data stream eliminates any skew, simplifies layout and eases cable selection criteria. Still, there are architecture issues associated with embedded clock links that have, until now, limited the widespread use of embedded clock interfaces.
For instance, high performance single channel embedded clock links will require a high performance PLL on the receiving side to be able to extract the clock. Many receiver solutions require a reference clock with a tight tolerance around the receive data rate frequency. This is needed for the PLL to be able to lock to the receiving data. Often, it is undesirable to have a reference clock on the receiving side; not just because of the cost associated with it, but also because having a reference clock inherently means the receiver is built for a specific data rate. What if you do not know the data rate of the incoming data? A good example would be a display link. Depending on image resolution, the actual data rate may vary quite a bit. A limited PLL lock range will prohibit any display resolution change.

Ernest Bron

Related Downloads

Comment on this article

This material is protected by MA Business copyright See Terms and Conditions. One-off usage is permitted but bulk copying is not. For multiple copies contact the sales team.

What you think about this article:

Add your comments


Your comments/feedback may be edited prior to publishing. Not all entries will be published.
Please view our Terms and Conditions before leaving a comment.

Related Articles

Adding audio

This whitepaper from SiLabs tells you how to add class D audio to embedded ...

DAQ µModule

Analog Devices (ADI) has introduced a 16-bit, 15 MSPS data acquisition mModule ...

Innovation is the key

Innovation is the key business driver for Analog Devices as Mike Britchfield, ...

Shaping the future

Alexander Everke, the CEO of ams, started his career in the semiconductor ...