During the past few years, most mainstream high-speed serial data standards have been transmitted at bit rates well below 5 Gbits/s. Today, many signal standards under development are reaching or surpassing this rate. Due to channel loss, higher-frequency content is degraded while lower-frequency content is intact.
A typical dispersion loss value for a high-speed NRZ signal transmitted across a 30-in. server backplane is –4 dB at 1.25 GHz. Using the same backplane, this loss reaches –8 dB at 2.5 GHz (5 Gbits/s). This amount of loss causes significant signal distortion at the far end of the backplane. To compensate for this, receiver equalization or channel improvements can be implemented. At 5 GHz (10 Gbits/s), a –25-dB loss occurs through the channel, making the signal unusable without receiver equalization.
Slower serial data standards (such as 2.5 Gbits/s) could be tested by probing at the receiver at the far end of the backplane. This probed signal then could be tested for eye pattern and jitter measurement compliance to ensure signal quality. Within an equalized system, this is no longer possible.
Typically, the equalized signal occurs only within the receiver chip and cannot be probed. The eye pattern formed at the receiver package pins is severely distorted and often completely blurred, preventing eye pattern and jitter measurements from being performed. To validate this internal signal, a method of viewing the signal inside of the receiver must be used to allow users to analyze the ideally equalized signal.
For fast serial data signals, the channel often attenuates high-frequency content below the measurement noise floor of the system used for characterization. For this reason, probing at the package pins of the receiver and then equalizing the far-end signal within the measurement system will significantly boost high-frequency noise levels, resulting in inaccurate device characterization.
To prevent this problem, virtual probing lets users physically probe at the transmitter package pins, instead of the receiver package pins, and then co-simulate the effect of the backplane. In doing so, the signal has been acquired in an environment in which the signal strength is much greater than the noise.
By co-simulating the backplane, a clean and accurate signal at the far end of the backplane can be obtained. Such signals are acquired far above the noise level and aren't subject to the measurement noise limitations that would occur when physically probing at the receiver pins. This accurate signal can now be equalized using feed-forward and decision feedback equalization for ideal equalizer emulation.
This technique can be used for backplane emulation to test backplane models. Using design software, a backplane model can be designed and saved as an S-parameter model. Inserting this model into the oscilloscope and using live signals, designers can quickly determine the effect of various designs well before prototyping occurs.
In the final phase of design, backplanes can be produced in one location and tested in another without the need for physically transporting any hardware. De-embedding probes can also be performed using this virtual probing technique. Creating a network diagram and inserting an S-parameter model for a system will allow the scope to solve for the system response with or without the probe loading effect. This allows for a user to probe a system and determine the steady system response that would occur if no probe were attached.
Virtual probing also allows the ability to acquire waveforms in circuit configurations different from the configuration actually used in the measurement. Canceling measurement noise, de-embedded probes and fixtures, and providing accurate measurements within inaccessible nodes can be attained today using the combined techniques of receiver equalization and virtual probing.