You need to test, we're here to help.

You need to test, we're here to help.

01 June 2020

USB4 Electrical Testing: Where Are We?

Preliminary drafts of the USB4 Compliance Test Specifications (CTS) were released to USB-IF work group members in early 2020 covering the electrical through protocol layer testing of USB4 router assemblies captive cable devices. These CTS documents define the required testing for USB4 hosts, hubs, and peripherals that make up the USB4 ecosystem.  Thunderbolt 3 (TBT3) is also supported in USB4 end products and is tested using the existing TBT3 Host/Device CTS.

While the electrical test methodologies are similar to previous USB 3.2 compliance tests, USB4 is largely based off the TBT3 physical layer, and these test methods have been adopted. Here, we summarize the USB4/TBT3 test approach called out in the aforementioned documents.

USB4 Transmitter Testing

Figure 1. Test points TP2 and TP3 for USB4
Electrical Compliance Testing (Source: USB-IF)
Presently, there is no SIG-TEST software for performing transmitter (Tx) testing, although it is anticipated that one will be released by the USB-IF in the future. All Tx tests performed by QPHY-USB4-TX-RX utilize Teledyne LeCroy methods, with measurements made directly by the oscilloscope.

Figure 1 shows the Test points defined in the CTS for physical layer testing. Testing of the Tx-side is  similar to those performed for USB 3.2 in that the signal is captured at the connector (TP2), where many measurements are defined. Then, a passive cable model is embedded and equalization is applied to result in TP3 measurements.

Noteworthy are three new SSC tests at TP2: Down Spread Range & Rate, Phase Deviation and Slew Rate.

The Tx Frequency Variation Training test measures that, as one Tx clock transitions to another or a clock transitions to its compliance pattern, the SSC tracks within the amount of time the CTS specifies.

The Eye Diagram & Mask Test is performed both near-end and far-end at TP2, the output of the connector of the DUT, and at TP3, which is TP2 with a 2 m or 0.8 m cable embedded (using .s4p files) and a CTLE and 1 tap DFE applied to achieve the specified eye opening. To do that first requires a process of scanning CTLE and DFE settings to find the best combination to use to create the optimized eye height and width.

Among the Router Assembly Tx tests, there is now a Wireless Band Conducted Emissions test, which looks at how many conducted emissions can be expected from the DUT at different frequencies. In QPHY-USB4-TX-RX, this is done as a spectral analysis, taking a slice of a frequency range and measuring the noise seen at those frequencies to confirm the DUT is within band required.

Figure 2. Teledyne LeCroy jitter breakdown.
Most notably, USB4 introduces a new jitter formulation for jitter measurements. While it utilizes the familiar Dual-Dirac methodology to extrapolate Total Jitter (Tj), unlike the “classic” formulation that first breaks out bounded Deterministic Jitter (Dj) from unbounded Random Jitter (Rj), the new formulation first break outs correlated Data Deterministic Jitter (DDj) from Uncorrelated Jitter (Uj). The Teledyne LeCroy adaptation of this new jitter breakdown, used by our jitter measurements, is shown in Figure 2.

The principal ways in which this changes our jitter measurements are:
  • Bounded Uncorrelated Jitter, BUj (aka UDj in the USB4 specification) includes periodic components, meaning BUj will include whatever residual SSC is not tracked out by the PLL in use.
  • Other Bounded Uncorrelated Jitter (OBUj) does not include periodic components, therefore OBUj does not include residual SSC.
  • Random Jitter is different from Rj in the “classical” formulation, in that: a) it is fit to uncorrelated CDF, not total CDF; b) this Rj is not used in the Tj calculation – the assumption is that Tj (BER) is expected to come from classical Dual-Dirac.
  • What the USB specification formerly referred to as Duty Cycle Distortion (DCD), we refer to as Asymmetric DCD (Asym), whereas what the specification now refers to as DCD, we call Odd/Even Jitter (OE). 
These measurements are all made using convolution or vector math, so it cannot be expected that all the jitter components will add up to Tj, but knowing the changes in terminology will help to understand the measurements as made by the software. The QPHY-USB4-TX-RX manual explains these measurements in greater detail.

USB4 Receiver Testing

There is a three-step process to receiver (Rx) testing:
  1. Calibrate TP3’ (Case 1), the short-channel receiver
  2. Calibrate TP3 (Case 2), the long-channel receiver
  3. Perform the receiver tests on both channels.
To do the TP3’ calibration, we connect directly from the MP1900A through DC blocks into the oscilloscope, then perform the calibrations for SSC, ACCM, random jitter, periodic jitter, total jitter and eye height/width.

For the TP3 calibration, the cables are disconnected from the oscilloscope and connected through a calibrated ISI channel, the receptacle test fixture, a USB-C cable with a specified amount of loss, and a calibration fixture before finally reconnecting to the oscilloscope.

Figure 3. Receiver TP3 calibration setup.
A necessary first step in this calibration process is to characterize the cables and channel, looking first at the a) insertion loss of the specified cable, then b) the insertion loss of the total channel, both of which must meet the specification. For this, you can use the WavePulser 40iX to get the S-parameter files used to de-embed the cables and characterize the channel.

Currently, there are no USB4-specific fixed adjustable ISI boards available for USB4, but a PCIe 4 ISI board or an variable ISI generator from Artek can be used for early market testing. Once calibrated, the calibration fixture is replaced by the actual DUT, and a microcontroller is connected to the receptacle test fixture. Both the calibrated TP3’ and TP3 undergo the BER test.

Figure 4. Receiver test setup.

USB4 Electrical Test Equipment

USB4 has two, 20 Gb/s lanes sampled at 80 GS/s. USB4 Tx and Rx testing require a real-time oscilloscope with at least 21 GHz bandwidth and 80 GS/s sample rate. Our LabMaster 10 Zi-A oscilloscope can meet this specification on all four channels, enabling simultaneously multi-lane testing. The WaveMaster 8 Zi-B can do so in two-channel mode, with each lane tested sequentially. Both oscilloscopes host the SDAIII-CompleteLinQ software, which provides the necessary eye diagram and jitter measurement capabilities, and can be used to further debug causes of failure.

A high-speed BERT that can test from 10 Gb/s to 20.625 Gb/s is also required for receiver testing. Teledyne LeCroy has partnered with the Anritsu Corporation to integrate their SQA-R MP1900A BERT (which supports data rates from 2.4 to 32.12 Gb/s for many standards) into our USB4 test solutions.

Teledyne LeCroy’s QualiPHY test software, QPHY-USB4-TX-RX, was the first USB4 Tx and Rx compliance solution. It fully automates transmitter testing and receiver calibration and testing, coordinating the oscilloscope, BERT and device under test (DUT). QPHY-USB4-TX-RX supports testing USB4 router assemblies, captive devices, and TBT3 hosts/devices to these standards.

Test fixtures used to break out signal from the DUT to the test instruments, uControllers for USB-C cable testing, and variable ISI boards are also used. These can be acquired from Wilder Technologies.

The cables and fixtures needed to access the signal must be de-embedded from the Tx test measurements and the receiver calibration. De-embedding requires S-parameters that characterize these channels, which can be measured using the WavePulser 40iX, then imported into the QPHY-USB4-TX-RX software.

Teledyne LeCroy also offers USB4 protocol test solutions incorporating the Voyager M4x USB4 protocol analyzer.

To learn about our USB4 electrical test solution in more detail, see the on-demand webinar, USB4 Physical and Protocol Layer Testing, Part 2 (physical).

No comments:

Post a Comment