The secret weapon to help you calmly deal with noise detection problems

Today’s noise problem has moved into the next area: high-speed serial backplanes. In fact, it’s all about noise and jitter, which is strictly an application technology. Pavel Zivny of Tektronix defines noise as “all undesirable phenomena in the vertical (voltage or power for optical signaling) direction” and defines jitter differently as “all undesirable phenomena in the horizontal direction (i.e., time direction)”. The problem is that it is difficult to separate the two issues because both phenomena work together to affect bit error rate (BER) performance.

It is now possible to make a lot of serial data measurements, or at least to represent them for analysis, as shown in the eye diagram in Figure 1. After the advent of the memory tube analog oscilloscope, engineers began using eye diagrams for qualitative analysis. They looked at two transitions on a repeating bitmap and found that if the data was jittery, the overlapping waveforms would flatten out.

Early digital storage oscilloscopes ( DSOs ) had color displays but relatively small memories, and added a third qualitative relationship by using color to show the relative distribution of jitter events. Today’s DSOs have faster speeds, memory that can store long data sequences, and DSP engines that can perform data analysis, allowing qualitative analysis of eye diagram measurements.

The oscilloscope also simplifies setup and provides standard image masks and images for go/no-go testing (see Eye Diagram Sample Size and Timing).

To understand why noise in a broad sense affects serial data transmission, here are a few eye diagrams. Figure 2a is an eye diagram of a good signal. The receiver will have a short period of time to detect the transmission. In Figure 2b, the voltage in the eye diagram is fixed, so the problem is not with the voltage, but the timing edge jitter is very bad. In Figure 2c, the timing edge is fine, but the signal has bad noise in the vertical plane, which may come from the power supply.

Figure 2d shows a signal with both noise and jitter issues. The receiver comparator will have to compete with low noise and thus control a narrower voltage range between high and low levels. In addition, as the transmission moves back and forth, the timing edges used by the circuit to understand the information are gradually reduced.

Bit Error Rate

The actual purpose of these measurements is to evaluate the impact of different jitter and noise on the bit error rate (Figure 3). Jitter has many components. Noise, on the other hand, is mainly divided into random noise and deterministic noise.

Random noise is what we usually refer to as noise or RMS noise when we observe a signal using an oscilloscope or spectrum analyzer. Deterministic noise is divided into periodic noise, which has a clear spectrum distribution independent of the signaling bit rate. For example, crosstalk from the power supply is a type of periodic noise. Periodic noise also appears on the processor clock when serial data is not running on the processor clock.

Data-dependent noise (DDN) captures the impairments caused by the bit rate in a way that depends on the bit pattern, such as a lonely-low pattern (in which many ones are interspersed with zeros). DDN is usually caused by ISI (intersymbol interference), which is the physical mechanism by which energy is coupled from one bit to an adjacent bit due to conductive or capacitive coupling, losses, or transmission line effects.

Eye diagram measurement

Mike Hertz of LeCroy says that because a complete data record is available in the instrument's memory, the location of individual bits can be determined by comparing the interval of each bit in the original waveform to a preloaded mask. When the mask test is turned on, the entire waveform is scanned bit by bit and compared to the mask.

Once a mask collision is detected, the bit code is saved and a table of bit values ​​is generated. This table is sorted, starting with the first bit in the waveform. It can be used to index back into the original waveform to display the waveform of the bad bit. Certain eye diagram measurements are specified as required tests in many standards. Basic eye diagram tests include amplitude and timing.

The eye amplitude is the difference between the distribution of a zero-level signal and the average of the distribution of a signal with a level of 1. The eye amplitude measurement is performed by distributing the amplitude values ​​around the center of the eye (usually 20% of the distance between zero-crossing times).

Eye height is a signal-to-noise ratio measurement that is very similar to eye amplitude, except that the standard deviation between zero and one levels is subtracted from the eye amplitude.

The eye width reflects the total jitter of the signal. The time between crossings is calculated based on the mean of the histogram at two zero crossings in the signal, and the standard deviation of each distribution is subtracted from the difference between the two means.

Optical signals on fiber require another eye diagram measurement called "extinction ratio". This measurement is needed because the laser transmitter is not completely turned off during data transmission. The definition of extinction ratio is simple, it is the ratio of the optical power of the laser when it is in the on state to the optical power when it is in the off state. Making laser power measurements is more complicated than voltage measurements because it requires an optical-to-electrical converter in front of the measurement equipment.

For timing measurements, the eye crossing point is the point where the level transitions from zero to one and from one to zero reach the same amplitude. It is expressed as a percentage of the eye amplitude. The measuring instrument displays horizontal slices of the eye diagram and cuts the slices at the minimum histogram width.

If you take vertical slices instead of horizontal slices, you can measure the average power, which is the average of the entire data stream. Unlike the eye amplitude measurement, which separates the ones and zeros histogram, the average power is the average of the histogram. If the data encoding is working properly, the average power should be 50% of the total eye amplitude.

Bit Error Rate Extraction

The basic part of the BER calculation is the time interval error (TIE), which is the difference between the data edge and the recovered clock edge. The measurement of the TIE histogram can determine the probability of the jitter value exceeding a given maximum value.

To obtain the BER, the TIE of the data sample is plotted as a histogram of the TIE value versus the number of times that value occurs. This is done to determine the probability that data transmission occurs simultaneously with data sampling. The histogram will conditionally indicate the probability that a data edge will occur at a given time within the bit period if the data is also being sampled at the same time. The bathtub curve (product failure curve) graphically displays this relationship (Figure 4).

Systems often specify bit error rates in the 10 ^{-12 range. Measuring events with a probability of less than one in the 10 -12} range will produce too many edges to acquire, and too many edges to be stored in today's instruments. This will require summing up the histogram from smaller measurements.