As the industry moves to adopt the MIPI Alliance's M-PHY standard , designers are encountering some significant challenges related to oscilloscope measurements and, more specifically, probing. These challenges include strict requirements such as bus termination and input return loss, as well as the need to minimize common mode loading on the device under test (DUT) and signal fidelity requirements such as wide bandwidth, low noise, and high sensitivity.
The intent of this article is to provide information that will increase your chances of accurate and repeatable test results to ensure compliance with the standard. We will first review the requirements of the M-PHY standard relevant to oscilloscope probing, discuss the tests required in the M-PHY Physical Layer Conformance Test Suite (CTS), and provide practical examples of M-PHY probing with currently available oscilloscopes and probes.
M-PHY bus operation/modes
The MIPI M-PHY standard supports different speed modes, high-speed (HS) and lower speed PWM and SYS modes, each with different data rates or GEARs. In HS mode, the official specification is released for GEAR 1 (~1.5 Gb/s), GEAR 2 (~2.9 Gb/s), and GEAR 3 (~5.8 Gb/s).
To increase the likelihood of MIPI M-PHY designs from different manufacturers working when used together, the MIPI Alliance recommends that designs be tested against the M-PHY Physical Layer Conformance Test Suite (CTS). With the development of the GEAR 4 (~11.6 Gb/s) specification proceeding, testing conformance is becoming more of a challenge.
Unterminated and terminated modes
The voltage measurements specified in the M-PHY standard assume that the bus has a known reference load (RREF in Figure 1 ) connected between the positive (TXDP) and negative (TXDN) outputs of the M-PHY transmitter (M-TX). The reference load varies if the LINE is either not terminated (NT) or resistively terminated (RT). When the LINE is operating in NT mode, RREF_NT is specified as a minimum impedance of 10kO between the TXDP and TXDN pins. The NT mode is most often used for low-speed communications, since driving a high speed signal into a high impedance transmission line can be challenging.
Figure 1. The voltage measurements specified in the M-PHY standard assume that the bus has a known reference load (RREF in the image) connected between the positive (TXDP) and negative (TXDN) outputs of M-TX.
High speed and terminated mode
When the M-PHY LINE is operating in high speed mode (HS-MODE), it will most likely be terminated to minimize reflections and mimic a receiver connection with a known transmission line impedance. For terminated mode of the LINE, the M-PHY standard defines RREF_RT as a floating 100O impedance across the TXDP and TXDN pins.
While the M-PHY specification defines both RT and NT states, conformance testing for HS-MODE is typically only defined for RT. Further, the M-PHY CTS only specifies high speed tests in RT mode, stating that high speed data measurements such as jitter on non-terminated signals are typically not practical or even possible.
When needed, non-terminated signals can be measured using either a high impedance active or differential probe. These oscilloscope probes can meet or exceed the specification with 10kO or higher impedance for signal frequencies that are less than a few MHz. Measuring the LINE in high-speed terminated mode with an oscilloscope requires that either the scope or its probe be as close to the desired 100O differential termination as possible.
A floating 100O termination is the ideal, but difficult to realize in practice due to parasitic elements in a design. Therefore, several alternative approaches have been proposed to achieve a nominal 100O differential termination along with sufficient return loss and high common mode impedance that minimizes DC and AC current draw on the transmitter. Table 1 lists the ideal termination and four other practical alternatives.
Since the ideal floating 100O cannot be achieved with real components, engineers testing M-PHY HS signaling must decide on one of the practical alternatives. Each alternative has pluses and minuses. Note that the floating 100O termination has a common-mode impedance that varies widely over M-PHY’s frequencies of interest due to parasitic elements. Common mode current flow and common mode input impedance are two areas where the alternatives differ significantly. All the alternate termination schemes have the differential properties of the ideal 100O, but exhibit other undesirable characteristics like 25O common-mode input impedance at high frequency.
Connecting the M-TX LINE directly to two channels of a scope is equivalent to the dual single-ended approach (#2). This alternative has the disadvantage of drawing common-mode current from the transmitter, which may cause problems for the transmitter. In alternative #3, adding DC blocking capacitors to the dual single-ended termination addresses the DC common-mode current, but the capacitors must be chosen carefully to ensure the termination meets the M-PHY return loss specification. Also, using blocking caps prevents DC measurements on the TX required by the CTS.
The two final alternatives mimic the ideal 100O resistor by drawing no DC common-mode current, but have low common-mode impedance (25O). This approach, however, allows for the common-mode impedance to be controlled better than the floating 100 O termination. If the capacitor (alternative 4, Cterm) used with the termination is large enough, the AC common mode impedance will be nominally constant at 25O. 100nF is needed for this to start at 1MHz.
The DC voltage termination (alternative 5, Vterm) capability includes a DC source inside the probe or scope, and provides an adjustable DC termination reference level. Setting Vterm to the TX’s common-mode voltage results in no CM current and eliminates the need for DC-blocks or bias-tees. Unlike a DC block, however, the signal’s DC voltage is still present at the probe or scope input. Using Vterm minimizes the DC loading. By setting the termination voltage equal to the DC bias voltage of the input signal, the probe DC loading is nulled out.
The termination of the M-PHY bus at GEAR 3 speeds and higher is further specified by the return loss for the reference load ZREF_RT according to the figure and values shown in Figure 2. High return loss limits unwanted reflections from the LINE that can impact the quality of the measurements.
Figure 2. Termination of the M-PHY bus is further specified by the return loss for the reference load ZREF_RT as shown here.
One of the issues that can prevent a probe or fixture from meeting the return loss requirement is capacitive loading across the 100O termination. The return loss specification sets some practical limits on the amount of series and parallel loading allowed across the ideal 100? termination, as shown in Table 2 for GEAR 3.
These constraints can limit the choices for line termination with a probe or scope. Figure 3 shows two termination approaches and input return loss. The first one is a probe with 50O differential inputs (blue trace). The second is a high impedance differential probe attached to a 100O termination fixture (red trace). In this example, the probe shown in blue meets the M-PHY specification by remaining below the limit line. The high impedance fixture and probe does not meet the spec as it exceeds the limit at 2.7 GHz and above.
Additional transmitter test requirements
M-PHY transmittertests that cover HS-MODE GEARs 1-3 include: intra-lane skew, pulsewidth, common mode , total and deterministic jitter. M-PHY transmitterscan support different drive strengths in an effort to save power. TheM-TX measurements are defined in the CTS with the M-TX running in bothLarge Amplitude (LA) and Small Amplitude (SA) modes. The low amplitudesignaling and the need to measure jitter result in a requirement thatthe probe and scope acquire low peak-to-peak voltage signals with fastedges.
In the power-saving Small Amplitude mode, the M-PHYtransmitter’s output peak-to-peak voltage is reduced to a maximum of 280mV. Slew rates must be controlled in the M-TX to reduce EMI noise.However, minimizing jitter at high data rates requires an edge speedthat is a fraction of a UI.
Even when the TX is running in largeamplitude mode, the amplitude of M-PHY signals is relatively small at0.5Vpp. Acquiring these low amplitude signals requires a high bandwidthmeasurement system with high sensitivity. Also, low additive noisewithin the probe and scope are key to making accurate measurements.
Howcan engineers reduce the noise added to measurement by the probe andscope? Comparing published noise specifications is an importantparameter, as is the probe attenuation factor. High impedance probesattenuate the signal as it enters the probe to limit loading on the DUT.Probe loading is not an issue with M-PHY as it is expected that the buswill have a low impedance (100Ω) differential load when running interminated mode. The attenuation factor of the probe works against thegoal of having a low noise input by reducing the signal-to-noise ratioof the measurement by a factor equal to the probe’s attenuation value.
Probe type comparison
Nowlet’s compare two probe types in example measurements of 5.8 Gb/s, lowamplitude (~200 mVpp) data signals. The first probe uses 50Ω, SMA-styleinputs. Since the loading of the probe is not a concern, the SMA-styleprobe can be used to acquire M-PHY TX signals. The second probe is ahigh impedance differential probe with a 100Ω termination fixtureattached to the TX outputs and then attached to the probe inputs, asshown in Figure 4 .
Figure 4. In this example, a high impedance differential probe with a 100Ω termination fixture is attached to the TX outputs
SMA-styleprobes have the advantage over high impedance probes in that theirattenuation settings are much lower than the high impedance probes. Forexample, available SMA-style probes can be set to have an attenuationfactor of 1X or even <1X for small amplitude signals. A highimpedance probe with similar bandwidth performance has a minimumattenuation of 5X. With lower attenuation in the probe, the measurementresults are cleaner with lower noise. This difference can be seen in aneye diagram. Figures 5 and 6 compare the same 5.8Gb/s signal whenacquired with a 50Ω low attenuation probe and a high impedance probewith 100kΩ DC impedance and minimum 5X attenuation.
Figure 6. The eyediagram from a 5.8Gbit/s signal using a high impedance probe with 100kΩDC impedance and minimum 5X attenuation.
You cansee the difference in noise between these two eyes by comparing the‘fuzz’ in each of the two eye diagrams. The data acquired with theSMA-style low-attenuation probe is cleaner and less fuzzy than the eyediagram acquired using the high impedance probe. Measured parameters forboth eye diagrams confirm that the eye diagram is more open in Figure 5 than the one shown in Figure 6 . For example, the eye opening measurement reduced by 18% when using the high impedance probe due to its higher noise.
Oscilloscopeand probe noise performance are critical for accurately measuring thecharacteristics of the M-PHY signals. Without a low noise floor, othercapabilities of the scope and probe are useless since the noise can hidekey characteristics of the signal. With M-PHY, the noise performance ofthe measuring system is critical because of tight amplituderequirements defined in the specification. In small amplitude (SA) mode,the M-TX peak-to-peak output voltage ranges from 160mV to 280 mV. Thedifference between logic levels is relatively small if the scope andprobe do not have low noise and do not have sufficient sensitivity(12-15 mV/div, 120 mV full-scale).
Be sure to pay closeattention to scope specifications that increase sensitivity by zoomingthe signal display when the limit of the front-end hardware has beenreached. Zooming on a signal increases the visibility of a signal but itdoes not increase the sensitivity. The zoom operation also increasesthe noise shown on screen.
The M-PHYspecification for the minimum TX rise time (20%/80%) is 0.1*UI,equivalent to 17.2ps in GEAR 3. However, the practical rise time of theTX signal at the measurement point is limited by capacitance in thepackage and the circuit board traces. Although the actual capacitance ofa device may not be known, if its capacitance is as high as the maximumspecified value for CPIN_RX of 1.5pF, then the rise time for GEAR 3signals will be increased from 0.1*UI to a value closer to 0.4*UI(>70ps).
Ideally, a probe and scope should have very lowcapacitance and a rise time 3X faster than the signal being measured.Using the rise time value for GEAR3, it is difficult to find a scope orprobe that has a rise time 3X faster than 17.2 ps (0.1*UI). It is,however, an easier task to find a probe with a rise time that is 3Xfaster than 70ps (0.4*UI).
Since in practical terms the signal’srise time at the probe point will likely be much slower than 17.2 ps,it can be a challenge to verify that the TX output does not exceed thelimit of 0.1*UI specified in the M-PHY specification. The CTS recommendsmaking the rise time measurement and then using a de-embedding processto remove any signal loss from PCB traces, cables and connectors. Usingde-embedding, it is possible to determine what the rise time of thesignal is at the M-TX pin and determine whether or not the TX meets therequirements of the standard.
ACand DC common mode output from the M-PHY TX is an undesirable source ofEMI. Therefore, measuring the AC and DC common mode signals at the TXoutputs are part of M-PHY CTS measurements. The floating 100Ωtermination of the line is intended to provide a termination with nopath to ground where the common mode current can flow. As noted, thisideal termination is not possible with today’s PCB materials andcomponents.
All practical termination approaches will havecommon-mode loading at M-PHY frequencies. This loading attenuates thecommon-mode signal level and must be de-embedded for accurate commonmode measurements. The more consistent the common-mode load, the easierit is to de-embed the loading effect. Figure 7 shows two examples ofcommon mode input impedance for the low attenuation SMA probe and highimpedance probe and fixture compared earlier.
Theblue line shows a common mode impedance that does not varysignificantly above or below 25Ω from DC to 5 GHz (M-PHY frequency rangeof interest). Whereas the second probe (red trace) has an impedancethat varies widely from DC to 5 GHz. Although DC current will flow inboth cases, the de-embedding of the probe loading for a common modemeasurement with the probe shown in the blue trace is simply a scalingoperation. Considering the voltage divider ratio of the TX’s sourceimpedance and the 25Ω input impedance of the probe, the measured commonmode voltage using this probe should be multiplied by 2. The common modede-embedding task for the high impedance fixture is a more complexoperation.
Measurement requirements summary
Table 3 summarizes the requirements for measurement equipment used to test conformance of an M-PHY transmitter running in HS mode.
Table3. These are the requirements for measurement equipment used to testconformance of an M-PHY transmitter running in HS mode
TheMIPI M-PHY standard presents significant challenges for oscilloscopesand probing. These challenges result in stringent requirements forHS-MODE measurements. 50Ω, SMA style probes have been shown to yieldsuperior results compared to high impedance probe approaches,particularly for HS-G3 and G4 speeds. Therefore, the newest version ofthe M-PHY CTS has been updated to list support for “SMA probe” typeprobing solutions.
Chris Loberg is a Senior Technical Marketing Manager at Tektronix responsible for Oscilloscopes in the Americas Region. Chris has heldvarious positions with Tektronix during his more than 13 years with thecompany, including Marketing Manager for Tektronix’ Optical BusinessUnit. His extensive background in technology marketing includespositions with Grass Valley Group and IBM. He earned an MBA in Marketingfrom San Jose State University.
MIPI Alliance DRAFT Specification for M-PHY Version 3.1 Revision 03 – 17 February 2014
MIPI Alliance DRAFT M-PHY Physical Layer Conformance Test Suite Version 3.0, 2014