Measuring HSIC USB without disturbing the system

High Speed Inter-Chip USB (HSIC USB) has recently become popular for connecting application processors to wireless modems and other peripherals in smartphones, tablets, and other mobile products. The specification adapts high-speed USB 2.0 for chip-to-chip use by changing the PHY layer to reduce power and make the PHY more compatible with standard digital logic processes. However, changing the PHY introduces test and measurement problems. Current USB analyzers don’t work because the new PHY is not interoperable with the standard USB 2.0 PHY in standard USB analyzers, and the bus is difficult to measure using standard scopes and probes.

Background
USB 2.0 was designed for upward compatibility with USB 1.1 devices, hosts and cables. The PHY is designed to drive 5 meter cables, using 3.3V signaling. Three speeds are supported: 1.5 Mbps, 12 Mbps and 480 Mbps. The lower two speeds use differential voltage-mode signaling. The high-speed uses differential current-mode signaling, with a terminated bus. The combination typically requires an analog-compatible process, and a PLL for clock recovery.

HSIC USB has no need for backward compatibility, because it’s chip-to-chip. The PHY is designed to drive 10 cm traces on the PC board, using 1.2V signaling. The PHY supports only high-speed USB. The data clock is explicitly signaled (on a signal named STROBE), and data is sent single-ended. DDR techniques allow the STROBE to operate at 240 MHz for 480 Mbps signaling. Legacy speeds (1.5 Mbps, 12 Mbps) are not supported. The resulting PHY is easy to implement without using analog cells.

[Get a 10% discount on ARM TechCon 2012 conference passes by using promo code EDIT. Click here to learn about the show and register.]

This approach results in a simple, yet flexible interconnect system. Only two pins are needed on the host or device. If the PHY is running at 100% duty cycle, the power consumed for signaling will be only about 9.6 mW; if the PHY is idle, then only leakage currents consume power (in contrast, the typical USB 2.0 PHY consumes 50 mW while active, and tens of mW while idle between transactions.) Best of all, standard USB system software can be used unchanged.

Measurement issues
The HSIC PHY bus can be difficult to measure. To see why, let’s examine a typical HSIC USB transaction.


Click on image to enlarge.

USB data is transferred in units called “transactions.” Each HSIC USB transaction begins and ends with the bus in idle. At these times, weak pull-up/pull-down resistors keep STROBE high, and DATA low. When the transaction begins, the outputs of the transmitting PHY are enabled, and the signals are driven with low-impedance LVCMOS drivers. At the end of the transaction, the idle state is actively signaled for two strobe periods; then the transmitting PHY disables the drivers, and the weak pull-up/pull-down resistors passively retain the idle state.

It’s critical to bus operation that the idle state be maintained. If an HSIC USB device detects STROBE low and DATA low for two Strobe periods, it treats this as a USB reset condition. This will reset all aspects of the USB device controller, causing communication with the host to be disrupted.

The high frequency of HSIC USB, combined with the high impedance of the idle state present a measurement challenge. The pull-up resistors may be sourcing as little as 20 µA of current. The receiver at the far end of the bus, fabricated in a normal logic process, has some leakage current. Almost any measurement technique will consume additional current. If the measurement circuit takes even 10 µA of current, the STROBE signal may be pulled down, causing a reset.

Approaches to measuring HSIC USB
Making measurements with an oscilloscope is not easy. HSIC uses relatively high frequencies. The fundamental frequency for STROBE is 240 MHz, and at least 1.5 GHz bandwidth is needed to see timing relationships between STROBE and DATA clearly. Inexpensive high-frequency (GHz) probes are passive. Normally, they will have input impedances of 50 KΩ or less, and will disturb the measurement. Some active probes (such as the Tektronix TAP1500) can make the measurement, but they’re not cheap - a pair will cost roughly $5,000.

Most measurements of HSIC USB are made to help debug software problems, and most software engineers find scopes hard to use. Thus, there’s a need for specialized test equipment that captures HSIC USB data and presents it to the software engineer in a meaningful form.

Digital measurements for HSIC USB test equipment are also not easy. Although the leakage current of typical LVCMOS inputs are low, the input capacitance is troublesome. HSIC USB only allows a total capacitive load of 14 pF, including the PC board trace and the capacitance of the receiver. This generally leaves little margin for a measurement. In addition to the input capacitance of the sense circuit, any measurement circuit attached to the system under test will add capacitance for the connector and the cable that connects to the test equipment. Complicating the problem, HSIC USB systems are typically very small, dwarfed by typical test equipment.

Although primarily a software company, MCCI produces a HSIC USB analyzer/generator, called the Catena 1910. MCCI’s solution to the measurement problem was to create a special Passive Capture Adapter (PCA). This adapter contains a specialized pre-amplifier on a small board that can be physically close to the system under test. The pre-amplifier drives two differential pairs over a cable to the test equipment, where the signal is reconstituted and captured. Critically, the first stage of the pre-amplifier is AC-coupled to the system under test, using a very small capacitor. This effectively isolates the system under test from the capacitances and leakage currents of the amplifier. The pre-amplifier design is optimized for HSIC USB signaling, rather than general purpose logic sensing.

Once the data acquisition problem is solved, one must still physically attach to the system under test. There’s no standard connector for HSIC USB, because it’s a chip-to-chip standard. Even micro-coax connectors might be too large for the available real-estate.

MCCI’s approach has been to support three modes of connection. For test boards, the test equipment can connect to the system under test using a standard Mictor connector. In addition to allowing for capture, this approach allows the test equipment to act as a host or device. For systems with room for MXHP micro-coax connectors, MCCI supplies short micro-coax cables that can directly mate to the system. Finally, for production systems without room for micro-coax connectors, MCCI supplies micro-coax cables that are tinned and stripped (of course, this assumes that the HSIC USB bus is accessible - if the traces are routed on an inner layer without test points, making the measurement can be extremely difficult).

Conclusion
HSIC USB provides a fast, low power connection from chip-to-chip, but it’s challenging to measure. Oscilloscopes with active probes will do the job, but they’re not easy for software developers to use. Specialized test equipment, such as MCCI’s Catena 1910, must use a variety of approaches based on the characteristics of the system under test. However, they can take advantage of the known characteristics of HSIC USB to optimize the capture of data while minimizing disturbance to the system under test.

About the author
Terrill M. Moore is Chief Executive Officer of MCCI Corp., a company he founded in 1995. He has 40 years experience designing and implementing computer system software and holds six hardware and software patents. He received a BA in Philosophy from the University of California, Irvine. His MCCI blog is "Making Connections."