Let’s start with the truth: There’s no such thing as zero latency. Really. It’s just physically impossible for something to go from A to B in zero time. It can go super-fast (low latency) or even super-super-super-fast (near-zero latency), but zero latency is not real.
Latency is the time—or delay—it takes for data (such as video, audio, controls, etc.) to go from a transmitter to a receiver. From a consumer’s perspective, latency is the delay between what’s coming out of the source and what’s arriving at the receiving end (e.g., display). For some applications, latency doesn’t have much of an impact, but sometimes it can affect performance and quality. In some cases—say, data transmitted within the car—latency can have tragic consequences and even be life-threatening.
Several elements affect latency, such as bandwidth, different interfaces, the cable/wire being used, transmission distance, compression, processing time, and the specific mechanisms implemented to handle electromagnetic interference (EMI).
For connected cars to function effectively and for the autonomous age to become a reality, achieving the transmission of ultra-high-bandwidth data with near-zero latency is critical. As more devices and applications are added, more data must be moved among the many elements and compute units in the car. Without the proper channels and bandwidth, the data can be severely delayed, harming overall performance.
Several technologies enable higher bandwidth transmission, but they demand a much higher quality cable and/or limit transmission distance considerably. Valens has developed a solution that enables uncompressed, multi-Gig data transmission over unshielded wires, converging different interfaces (such as PCIe, USB, audio) over the same link. Compression, although often considered as an option to transmit high data rates, isn’t feasible as it can add considerable latency.
Another important factor to consider is the noisy automotive environment. Not only in terms of the amount of interference—too many cables, connectors and devices, all very close to each other—but also the speed in which EMI attacks occur. Given the size and speed of such attacks, it’s inevitable that data will be lost.
In general, lost data can be retransmitted, but it first needs to be buffered, which adds latency. From that point, it gets more complicated. If data is buffered only on noise attacks, we add jitter (packet delay variance) to the transmission, which means that each application will have to add its own buffer to compensate for that jitter. Again, more buffering, more latency. Some applications are particularly vulnerable to jitter, such as video and audio. Other applications are not affected by jitter, such as PCIe and USB, but are actually more sensitive to latency.
The solution developed by Valens incorporates mechanisms that can address both types of applications:
- Fixed Delay enables jitter-sensitive applications to minimize overall buffers, by adding small internal buffers, and therefore, lower latency.
- Minimal Delay allows latency-sensitive applications to receive the data with minimal delay, but with additional jitter that can be handled by native mechanisms, such as flow control.
The key here is to look at the entire picture, considering the many variants and tradeoffs involved, and to activate the right mechanisms to guarantee higher performance, with the highest bandwidth and most resilience against EMI. On this front, Valens has worked to build in flexibility to address an application’s specific needs, while also delivering performance and security. There’s no “one size fits all” solution, and the ability to identify the specific challenges and requirements of each interface will be decisive in achieving the ultimate connected and autonomous car.
Daniel Schwartzberg is Director of Technical Pre-sales at Valens.