Probing Techniques and Tradeoffs (Part XI): Non-Ideal Situations

Figure 1: VP@Rcvr builds a transmission-line model
to virtually move less-than-ideal probing points

In using an oscilloscope to investigate transmission-line performance, we often encounter situations in which we don't have the luxury of probing at the ideal location. Fortunately, there are software tools that enable us to virtually move our probing point close to the receiver.

Figure 2: LPDDR2 with reflections
as shown here make accurate timing
measurements difficult to obtain

Teledyne LeCroy's Virtual Probe@Receiver (VP@Rcvr) math operator is one such tool. VP@Rcvr, which is enabled by the EyeDoctor II software package, is designed to quickly compensate for signal reflections due to a termination impairment. The tool does not require S-parameters of either the DUT or the probes, but rather builds a transmission-line model to virtually move the probing point closer to the receiver (Figure 1).

If we were to look at a typical LPDDR2 signal, we might see nice, clean transitions in some portions of the waveform. But in others, we might see something that shows evidence of reflections, as in Figure 2. For DDR memory, or any serial-data protocol, for that matter, this sort of non-monotonic behavior makes it difficult, or even impossible, to make accurate timing measurements.

Figure 3: The transmission line between a memory controller
and DRAM shows reflections from an open termination

Were we to examine the eye diagrams of the read and write signals, we might see, for example, read signals that look good but write signals that look not so good. What's really happening in such situations?

What we're looking at is communication between a memory-interface controller and some DRAM. Between the two lies the transmission line, and let's say that in this case, the transmission line has a characteristic impedance of 50 Ω. At the DRAM, we have an open termination.

Figure 4: With VP@Rcvr, we can create a model of the
transmission line and compensate for reflections

If we probe at the controller interface, we first see the incident signal, so we see an incident rise at T₁ (Figure 3). At a later time T₂, the signal hits the open termination and it reflects back to V_A, which is where we actually probed. Then, at T₃, we see the reflected pulse. If we use the oscilloscope as a time-domain reflectometer, we can measure that propagation delay. At a point halfway between the incident and reflected rises, we measure a propagation delay of 822.5 ps.

So 822.5 ps is the round-trip time, but what we're interested in is the one-way trip time, which is 411
ps. Now that we know that we have an open termination as well as the total propagation delay, we can use VP@Rcvr to create a simple model of the transmission line. We populate these parameters into our model and we can now use VP@Rvcr to compensate for the reflection.

Figure 5: The traces at top show DQS and DQ before
application of the VP@Rcvr tool; the bottom traces show
the aftermath of the tool's use

A real-world example demonstrates the utility of the VP@Rcvr tool. In Figure 5, the top-left grid shows the DQS signal before applying VP@Rcvr while the top-right grid shows the unprocessed DQ signal. Both of these signals had reflections occurring around the reference voltage (circled in black in the two upper images). These reflections on the strobe and data lines would pose problems in making JEDEC compliance measurements, which use V_ref for many of the measurements.

In this case, the probe was not placed at the receiver (housed in a BGA package). As a result, the reflection seen at the probe point is caused by the receiver. It would not have been seen by the receiver itself if the probe could have been placed there.

After modeling the circuit, the reflections are removed (as seen in Figure 5) and there are now clean rising and falling edges on the data and strobe lines. These signals accurately depict how the signal appears at the receiver. As a result, we've cleaned up the signal fidelity, enabling us to make accurate measurements and view the signal as it would appear if we probed at the ideal location.

Previous posts in this series:

Probing Techniques and Tradeoffs (Part I)
Probing Techniques and Tradeoffs (Part II)
Probing Techniques and Tradeoffs (Part III)
Probing Techniques and Tradeoffs (Part IV)
Probing Techniques and Tradeoffs (Part V): Probe Loading
Probing Techniques and Tradeoffs (Part VI): Dynamic Range
Probing Techniques and Tradeoffs (Part VII): More on Dynamic Range
Probing Techniques and Tradeoffs (Part VIII): Gain/Attenuation vs. Noise
Probing Techniques and Tradeoffs (Part IX): Best Practices
Probing Techniques and Tradeoffs (Part X): More Best Practices