In this post, we'll break down the essential components of a GMSL Channel Specification, explore the critical design considerations you need to know, and provide practical guidance to help you avoid common pitfalls. Whether you're designing your first GMSL-enabled system or optimizing an existing one, this guide will give you the clarity you need to move forward with confidence.
The GMSL Channel
To put the GMSL channel in perspective, consider the lengths: a few millimeters of bond wire and a few centimeters of PCB traces make up an extremely small portion of the overall pin-to-pin channel, and then there are multiple meters of cable. From the pin of the serializer to the pin of the deserializer, over 99.9% of the channel is out of our direct control.
Figure 1: An Example of the GMSL Channel with the ICs Highlighted
Figure 2: The GMSL Channel Definition from the Channel Specification
What we can control are the limits of which we can guarantee robust performance. The limits are the losses we can tolerate, known as Insertion Loss and Return Loss.
Insertion and Return Loss
These two requirements are the key factors in determining the length of cabling you can use for a GMSL system. These are explained in many ways, but mnemonically, this is how I’ve always thought of them.
- Insertion Loss: The amount of signal lost when components are INSERTED into a channel. The cable typically accounts for most of this loss, as it is the longest element in the channel.
- Return Loss: The amount of signal that is reflected and RETURNS through the channel. This is more of a concern on shorter channels, like a short cable or PCB trace, where the reflected energy doesn’t attenuate as much.
Figure 3: Insertion Loss is Denoted as S12 and S21 and Return Loss is Denoted as S11 and S22
Physical cables will unavoidably have insertion loss due to cables dissipating the energy of the signal. And return losses are caused by any impedance mismatch, which is also inevitable due to manufacturing tolerances.
Cables used in GMSL systems should be measured and evaluated prior to implementation to ensure the system insertion loss does not exceed the requirement in the GMSL Channel Specification. Too much insertion loss due to the cable will cause the GMSL signal to degrade beyond our deserializer’s ability to receive the signal.
Return loss may not be as obvious, but the source is primarily the connectors and PCB routing. Cables are well manufactured to the correct impedance, but PCB manufacturing and too many connectors can add multiple sources of impedance discontinuities. Too much return loss could cause enough energy to be reflected at the GMSL transceiver, which would interfere with both the transmitted and received signals.
It is always imperative to know these parameters prior to starting your design, and they can be easily and accurately measured with a Vector Network Analyzer. We’ve detailed the procedure at the end of the GMSL2 Hardware Design and Validation Guide.
I did a webinar on EngineerZone that walks through an example cable calculation.
We also have some great GMSL U videos going even further in depth on Insertion Loss and Return Loss topics.
Crosstalk
Additionally, there are two other channel spec requirements that should be understood. The GMSL2 crosstalk specification places limits on the permissible parasitic coupling from high-speed links (aggressors) and/or noise sources onto a GMSL2 link.
Proper layout planning and well-shielded cables can help to greatly mitigate the noise sources away from the GMSL signal. The main perpetrators are clock sources, power supply noise, and traces on secondary layers.
I always recommend laying out the GMSL trace first and placing some design constraints around the trace as a reminder when laying out the rest of the board. For further explanation, there is another GMSL U video walking through the Crosstalk topic.
Link Margin
The link margin specification is important to ensure that the link has enough budget to accommodate signal amplitude fluctuations. This is an additional requirement on top of the insertion and return loss requirements to guarantee link robustness.
So even if you comply with the static insertion loss measurement, we want to ensure the link has margin if the signal fluctuates.
The measurement is straightforward, where the transmitter (serializer) incrementally drops the transmit amplitude and the receiver (deserializer) checks for any type of error that may have occurred. As soon as an error happens, the current transmit amplitude is subtracted from the starting amplitude and returned as the margin. For further information, you can refer to the GMSL U video describing the Link Margin.
Conclusion
With a solid GMSL Channel Specification as your foundation, you can navigate the complexities systematically. By following the guidelines we've outlined, you'll be well-positioned to create GMSL systems that perform reliably. The key to success lies in planning ahead.
Assuming all these requirements are met, the link should be error-free but that’s not always how the physics play out. Next month, we will start looking at some errors that could occur in the signal chain.
See all the blogs in the GMSL Debugging series.