
According to long-time signal integrity guru Donald Telian, there are two schools of thought on signal integrity, and these two schools converge at the PCB and the I/O buffer on a chip. Over time, old school signal integrity has merged with new school signal integrity, highlighting that not all signal integrity problems can be addressed on the PCB.
First, let's look at the old way of thinking about signal integrity. Old school signal integrity refers to problems that are isolated on the PCB, and thus the solutions generally exist only on the PCB. This concerns signal losses during propagation between a transmitter and receiver, driving innovations in materials and interconnect analysis tools to the point we are at today. Designers targeting the fastest interfaces in a PCB must now address these loss issues at very high bandwidths. Typically, small issues in the PCB layout cause the most problems for wideband signal propagation. These can include:
- Ball out and fan out routing into a BGA
- Neck down or tapered routing into components or connector pins
- Design and construction of vias for vertical transitions
- Proximity to other traces on the board, i.e., crosstalk
- Excessive mode conversion in differential interfaces
In the new school of signal integrity, the focus shifts to what happens in the interface standard and what is placed on the chip. Due to the issue of losses at very high frequencies bringing SNR levels close to 1 on long interconnects, innovations in the I/O buffer were needed to recover signals from very fast bit streams.
This led to innovations in simulation that attempted to more accurately account for the behavior of transmit and receive buffers in an interconnect, leading to IBIS and IBIS-AMI modeling. The other major innovation that has helped overcome the limitations of PCBs is equalization, implemented in the transmit and/or receive circuits. Equalization essentially accounts for the losses in a long interconnect by pre-amplifying a signal at the transmitter or amplifying and filtering a signal at a receiver, which may be nonlinear.
Watch the clip below with Donald Telian or watch the full Altium OnTrack podcast episode here.
The Takeaway
Interconnect designers must consider these factors when planning layout and routing as part of system design. Newer generations of common interfaces are implementing equalization to account for higher losses at higher frequencies, which will continue to enable higher data rates in the future. Both approaches to signal integrity are needed to fully characterize high-speed digital channels targeting the fastest interface standards.
Transcript
Zach: Now, one of the things that you talk about, and I think your experience justifies your perspective on this, is the concept of new school versus old school signal integrity. What do you mean by that?
Donald: Oh, yeah, thanks. I think that's a really relevant question and a good place to start. In the late eighties, nobody believed that PCB traces were transmission lines. They thought everything was just a capacitive load and you just waited for it to charge. There's a picture in my book that shows, even back then, that was not true. That's almost the definition of old school. There were many things that went along with that, like terminations, topologies, and ground return. These are really important concepts.
As I work with customers now, no one is debating those things. I haven't been asked in 20 years to remove ground planes. Everyone is convinced that we need ground planes and shields.
Zach: You would be surprised what some students come up with sometimes regarding grounding because a new generation seems to need convincing again.
Donald: Yes, and so, this is the baseline. In higher speeds, nobody's asking about that. As people enter the field, they need to have that baseline, which I refer to as old school. As I work with customers and teach these concepts, we take a lot of this as given. There's a lot of good material out there that will show, persuade, and teach you that. This allows us to talk about what's happening right now.
There was a time we overlooked transmission lines and ground planes, but what are we overlooking right now? With my book, I think one of the most interesting aspects is what I don't talk about. It's more about what the new school entails. I have a great slide that shows this, but we won't get into it now.
The premise is, what are the leading edge, tier one professionals persuaded of now? What can we leave behind and what new discussions can we get into? Every speed jump has leveraged a different piece of technology. I have to talk about new things and what's relevant now. That's all it really means.
Zach: Sure, I understand. As enabling technology continues to shift, there's always a new thing. The most recent one enabling faster speeds is equalization. There are different types of equalization, and it seems like every so often, a PhD student figures out a new way to do it and gets us over the next hump.
Donald: Yes, it's a fascinating migration. The ball keeps bouncing and changing. What a great field to be in, though. It's always reinventing itself and changing.