Signal Integrity in a GHz World

Editor’s note: This article is Sponsored by Pericom Semiconductor

The convergence of technology dynamics with market dynamics brings the importance of signal integrity into sharp focus.

Figure 1: Intel’s lithography process roadmap. Note that these dates represent technology milestones; devices actually shipping in these technologies followed up to 24 months later. (Courtesy: Intel.)

Two sets of dynamics are on a collision course making an often overlooked technical detail—signal integrity on PCBs, cables, connectors, and IC chip sets—more important than ever. When clocks and switching speeds were in the hundreds of megahertz, digital signals implemented in standards like 10/100 Ethernet, USB 2.0, SCSI, PCI and others behaved as expected and systems functioned per spec.

Spin the dial up into the gigahertz range and signal integrity becomes a major issue for designers. Smaller mobile devices crammed full of the latest high speed processors and I/O interfaces, coupled with die-shrunk multi-gigahertz processors and peripherals, mean signals get attenuated due to losses, cross coupling, and the low-cost PCBs and connectors common at consumer price points.

The silver bullet to solving these technical challenges is the repeater or redriver device. This inexpensive marvel added to the signal channel drives longer traces across FR4, through flex/ribbon cables, across backplanes, and meets eye diagram margins in all high-speed open standard interfaces. Let’s first examine the dynamics creating signal integrity challenges then see how the redriver solves signal integrity problems.

Figure 2: Intel’s CPU roadmap. (Courtesy: Intel.)

The Technology Dynamic

One dynamic is the relentless march of technology markets. CPU and endpoint peripheral controller speeds get faster all the time and GHz clocks are the norm. At the same time, IC’s pack more transistors onto the same size die through process shrinks and smaller feature sizes. Smaller transistors operate at lower voltages, offering lower output drive capability and an inability to move signals down long or high impedance lines.

Intel’s roadmap exemplifies rapid die shrinks, moving from 45nm in 2007 to 22nm today (Figure 1). For example, the Core Duo that replaced popular Pentium processors shipped in 65nm lithography while today’s 4th Generation Core i7 (Haswell) devices ship in 22nm (Figure 2).

As shown in Figure 3, Intel’s very small 22nm FinFET three-dimensional process packs in more transistors per cubic die millimeter, but has reduced operating gate voltages by an additional 0.2V. This is a whopping 20% reduction, and there’s a corresponding reduction in output drive capability for I/O like Ethernet, PCI Express, USB 3.0, SATA 3.0, SAS 2.0/3.0, and DDR memory. In three years, Intel expects 7nm transistors by 2017 and they’ll likely operate at a fraction of today’s Vdd voltages. Output drive capability will suffer even more.

Occurring in parallel with the trend to small, output signals are clocking faster, too. In 2002 PCI Express 1.0 ran 2.5 GT/s (gigatransfers/s). PCIe 2.0 doubled that to 5 GT/s while PCIe 3.0 is 8 GT/s (Figure 4). According to the PCISIG, PCIe 4.0 is targeting 16 GT/s by next year in 2015. Figure 5 shows similar increases in other SERDES standards.

Figure 3: The move from planar transistors at 32nm to much faster three-dimensional 22nm Tri-Gate (FinFET) MOSFETs reduced gate voltages by 20 percent to around 0.8V. (Courtesy: Intel.)

These speeds exacerbate signal integrity problems exponentially. PCI Express 3.0 will tolerate a max channel loss of 20 dB: roughly 18 inches in FR4. With circuitous on-board routing, vias, connectors, risers and backplanes, non-handheld platforms will be quickly out of spec. Put another way, the move from 2.5 Gbits/s to 8 Gbits/s triples attenuation from 0.3 dB/inch to 0.9 dB/inch (FR4). In short: PCIe 3.0 simply won’t work properly in this system example without signal integrity mitigation techniques. In handhelds, trace lengths might not exceed 18 inches, but other PCB and chipset factors become equally significant.

The Platform Dynamic

Figure 4: PCI Express just keeps getting faster. (Courtesy: PCISIG.)

Another dynamic affecting signal integrity is the platform race from desktop computing to mobile where tablets, smartphones, 2:1’s (tablet plus separate keyboard) and wearables are obsessively shrinking platforms and their internal boards. Devices sit closer than ever to each other, multilayer boards with stubs and buried vias struggle to route high fanout SoCs and peripherals, and space economization often mounts non-trivial ICs on flex circuits or even in stacked-chip multichip modules (MCMs)—or wherever there’s available space. In mobile platforms, despite their small external size, long and interrupted trace lengths are possible. Routing takes a back seat to real estate; signal integrity is often third priority if it’s considered at all.

For example, both micro HDMI and USB 3.0 are common in high-end smartphones. Case ergonomics may necessitate external connectors for these standards at non-ideal

Figure 5: Serial standards roadmaps. Faster speeds mean signal integrity challenges.

distances from the respective transmitter peripheral controllers. The connectors might not even be on the motherboard; rather, it’s possible they attach via internal cable or flex circuit. USB 3.0 signals that might be marginal at the case connector will be unacceptable at the other end of a user-connected cable. FR4 dielectric losses are 0.4 dB/inch at 5 GHz and total PCB attenuation in a large “phablet”-sized smartphone could conceivably exceed several dB.

Mobile devices have another unique challenge caused by extremely dense miniature mother and daughter boards routing gigahertz signals on narrowly spaced traces. Cross-coupling will occur due to proximity and skin effects. In such a small space, multiple distributed clock lines will pick up noise since source clocks can’t always be close to transceiver ICs, causing random jitter on data lines. This kind of jitter will run concurrently with data signals, adversely affecting the signal integrity that was already adversely affected by FR4 losses and impedance mismatches!

In other words: small packages like mobile compound signal integrity problems.

It All Adds Up Adversely

Both of these dynamics—platform changes and technology evolution—have adverse affects on signal integrity. Board layouts and mechanical considerations increase jitter, clocks degrade, attenuation builds, impedance problems cascade, and second and third order transmission line phenomena like the skin effect and narrow spacing coupling become non-trivial.

As signals degrade, bit-error rates increase and the margin built into the specs in Figure 5 is wiped out, resulting in poor video quality, data loss, retry errors, and degraded speeds. Even new specs targeting mobile like the PCISIG’s M-PCIe Gear 3 running M-PHY (5 – 5.8 Gbits/s) will encounter signal integrity challenges. As well, Moore’s Law speeds up ICs but they’re less capable of driving long, lossy traces and channels.

Underlying all of these concerns and at the back of designers’ minds is cost. Exotic PCB materials and high-quality RF-like connectors can sometimes mitigate a few of these challenges, but at what cost in platforms designed for low-cost consumer applications? Regardless, choosing an IC leaves little choice: its electrical characteristics are predefined.

Is there a way out of this signal integrity challenge?

Redrivers: The Silver Bullet

Figure 6: PCI Express redriver installed in Tx (two directions).

There’s a simple solution to these signal integrity problems. Redrivers are cost effective active repeater components that restore and boost signal integrity. They condition signals at the Tx and Rx ends of electrical channels to compensate for expected or actual losses. On the transmit side they apply pre-emphasis to signals to match the loss characteristics of the channel. On the receive side, adaptive equalization is applied to open and restore the signal eye. Together, margin is added for specifications like those shown in Figure 5 to compensate for PCB trace, connector, cable, and IC characteristics.

Redrivers are active amplifiers that do not contain more complicated clock, data recovery, or PLL circuits. Hence, no external clock reference components are required—this highlights their simplicity and helps explain their low system cost.

A typical redriver configuration is shown in Figure 6. Here the PCIe redriver is in the Tx route only from Root Complex to Endpoint, and vice versa. Figure 7 shows first the degradation of 8 GT/s PCI Express 3.0 across a server motherboard (test point B) and after a redriver (test point C). The eye diagram clearly shows a much improved signal, providing up to 12 dB of additional margin on the spec and compensating for PCB, connector, via and other losses.

The redriver shown in this PCIe example operates on the PHY layer only (of the OSI model), making it transparent to in-band signals. That is: it deals strictly with signal levels and not data sets or symbols like active clock synchronizers or retimers might. Redrivers reduce deterministic jitter in the channel, compensate for impedance mismatch, and introduce less than a nanosecond of latency.

And redrivers offer many benefits in densely-packed mobile devices, too. Besides opening the eye, one subtle advantage is how redrivers can reduce EMI by proper placement of the redriver “boost point” in the trace.

Figure 7: The real world results of Figure 6 installation. Right eye diagram shows redriver-improved signal now compliant to PCIe 3.0 spec.

Baby, You Can Drive My Channel

It goes without saying that good high frequency design practices should be followed for modern multi-gigahertz standards like USB 3.0, PCI Express 3.0, 10 Gb Ethernet, 12 Gb SAS 3.0, and so on. But there are many factors inherent in today’s designs that still conspire to degrade signal integrity to the point where channels don’t operate to spec, data is lost, or image/audio quality suffers. Redrivers are simple, active devices that are not only cost effective but are technically effective at opening the eye of SERDES signal channels. Using them on both the Tx and the Rx side of the channel assures maximum margin and outstanding signal characteristics.

Redrivers from Pericom Semiconductor are so effective, that some of the open standards we’ve discussed—such as USB 3.1 SuperSpeed—call out specific specifications for signal integrity products. As speeds go up, IC output drive decreases, and whether overall system sizes get larger (servers) or smaller (mobile), redrivers are a key tool in system designers’ toolboxes improving signal integrity in a gigahertz world.

60_ciufo_chrisChris A. Ciufo is editor-in-chief for embedded content at Extension Media, which includes the EECatalog print and digital publications and website, Embedded Intel® Solutions, and other related blogs and embedded channels. He has 29 years of embedded technology experience, and has degrees in electrical engineering, and in materials science, emphasizing solid state physics. He can be reached at

Contact Information


1545 Barber Lane
Milpitas, CA, 95035

tele: 408-232-9100
toll-free: 800-435-2336
fax: 408-232-9100

Share and Enjoy:
  • Digg
  • Sphinn
  • Facebook
  • Mixx
  • Google
  • TwitThis

Tags: ,

Extension Media websites place cookies on your device to give you the best user experience. By using our websites, you agree to placement of these cookies and to our Privacy Policy. Please click here to accept.