Determinate jitter measurement with oscilloscope

  • Detail

Determinate jitter measurement with oscillograph

there are various special test equipment designed for jitter measurement, but the price is very high, and it is too professional for some applications. In fact, by using ordinary digital sampling oscilloscope, designers can also accurately and quickly measure the deterministic jitter, and the cost is much lower than expected. This paper introduces how to use oscilloscope to measure jitter for short test sequence

accurate jitter measurement is a challenge for Ethernet and optical path components. Jitter can be divided into two categories: random (unbounded) jitter and deterministic (bounded) jitter. Deterministic jitter (DJ) includes all jitters that can be reproduced under controlled conditions, and can be subdivided into periodic jitter (PJ), jitter caused by test sequence template (PDJ), pulse width distortion (DCD) and limited uncorrelated jitter

the main causes of deterministic jitter are:

1. the reference level is fuzzy. It is caused by the low-frequency cut-off point of the system, which will produce jitter near a long series of continuous same digits (CID)

2. the system bandwidth is insufficient. This will prevent some pulses from reaching the steady-state level and cause jitter on the separated pulses (such as.. 010.. Or.. 101.. Data sequences)

3. amplifier bias. It can cause pulse width distortion during each data conversion

4. nonlinear amplifier effect. It can produce unpredictable jitter effects, often causing jitter after a long string of CID

5. power supply noise and crosstalk. Jitter independent of data input (sometimes referred to as finite uncorrelated jitter) may occur

in optical fiber communication system, jitter will accumulate on each component. At the receiving end, the clock and data recovery circuit (CDR) analyzes the data and decomposes the serial rate clock. The jitter on the CDR shows small frequency changes in the clock rate. Slow changes (small frequency jitter) are easy to track, while fast changes (high frequency jitter) are not so easy. If the receiver has too much high-frequency jitter, the clock cannot be decomposed, and a large number of errors occur in data communication

consideration of jitter allowance

in order to prevent the above situation, system designers should consider using jitter allowance. Note that deterministic jitter increases linearly (in the worst case, all jitter sources are added together), while random jitter increases in geometric multiples (square sum and square root) at the delivery ceremony. Here, it is assumed that the noise sources causing random jitter are independent. Separating the dithering components will make the dithering generated by each component more random, which has several advantages, including longer link distance and lower component cost

it should be noted that the deterministic jitter measurement error always makes the deterministic jitter of components appear larger, and it is directly subtracted from the margin. The random jitter measurement error is not so serious, so it is more necessary to accurately measure the deterministic jitter

measuring deterministic jitter requires knowing a sequence of data templates. K28.5 is a sequence template usually specified for measuring the jitter of fibre channel and Ethernet systems operating at 1gb/s to 3.125gb/s. This sequence is a special character in the 8b/10b decoding table and often represents the beginning or end of a frame. The repeated k28.5 sequence (consisting of k28.5+ and k28.5- in turn) contains a data string, which has five consecutive ones and five consecutive zeros (the longest consecutive same number in 8b/10b decoded data), and it also contains separate sums

the repeated k28.5 sequence can be used to measure the deterministic jitter caused by the fuzzy reference level, low bandwidth and bias, and other sequences are more suitable to measure the jitter caused by the nonlinear influence of the self created special die

jitter measurement

there are usually three methods to measure deterministic jitter with an oscilloscope, namely eye diagram method, average eye diagram method and average cross measurement method (Table 1)

when the eye diagram method is used, the oscilloscope will display multiple crossed waveforms at the same time, and you can quickly see the overall jitter (deterministic jitter and random jitter are combined). The main advantages of eye diagram method are fast speed and convenient setting. Figure 1 shows the typical settings of test equipment and DUT. However, eye diagram can neither separate random jitter from deterministic jitter, nor remove the jitter caused by the test system

when using the eye diagram method, it should be clear that the trigger method will mask most of the deterministic jitter. For example, suppose that the sequence generator provides a trigger every 10 clock cycles. If the sequence length is even, the oscilloscope will not trigger on odd digits, which will mask some conversions. This problem can be avoided by using the odd length sequence or the trigger on the clock output of the sequence generator

if the device under test contains a time regeneration circuit (clock recovery or re timer), a good phase-locked loop shall be used to recover the oscilloscope trigger clock, which is specific to a specific protocol. In addition, if there are optical elements in the device under test, it is also necessary to add appropriate optical converters (optical electric or electro-optical converters). If necessary, it can be assumed that the optical converter or PLL has been included in the test equipment

the average mode provided by some oscilloscopes can remove the random jitter in the eye diagram. Compared with the basic eye diagram, this mode has higher accuracy and repeatability, but there are also problems related to triggering

the average cross measurement method can accurately measure most sources of deterministic jitter on k28.5 sequence. The steps are as follows:

1. connect the tested device as shown in Figure 1, and set the oscilloscope trigger as the sequence template

2. display the whole k28.5 sequence on the oscilloscope screen

3. calculate the expected crossover time of each conversion on the screen (it is better to skip the unit time interval without conversion only because the steel may cause clamping due to thermal expansion)

4. repeat the average operation

5. tabulate the crosses and calculate the jitter. There are 10 crosses in the repeated k28.5 sequence ()

6. ensure that the signal fills at least 2/3 of the vertical plane, so as to optimize the digital performance of the oscilloscope

7. set the main position of the oscilloscope as small as possible (make the delay leave the trigger, and the display as short as possible to reduce the trigger jitter)

8. use the maximum horizontal resolution (multiple horizontal points)

9. repeat averaging (64 to 1000 times on average) to reduce random jitter

note that periodic jitter and limited uncorrelated jitter cannot be captured by the averaged oscilloscope waveform. Other methods should be used for measurement. If it is safe to assume that the tested device will not produce a large number of such jitters, the average cross measurement can be used

for systems using scramblers, this technique is too difficult to use because the test sequence of such systems is usually very long. For example, the length of the pseudo-random binary sequence commonly used to test the scrambled data system is more than 8million bits. The average cross measurement of such a long graph is slow and inaccurate

enhance accuracy

nowadays, the deterministic jitter of many components is almost or less than that of template generator and oscilloscope. Therefore, to accurately determine the jitter of these components, it is necessary to find out the jitter introduced by the test system, and then adjust the measurement. The system error can be measured by removing the tested device from the test device and determining the jitter of each conversion of data graph. Then, put the tested device back into the test system and measure the jitter value again. The jitter generated by the tested device can be determined by arithmetic method

the average cross measurement method introduced here is very suitable for deterministic jitter measurement of Ethernet and fibre channel components using 8b/10b coding system. Accurate and repeatable jitter measurement can be carried out only with common test equipment. (end)

Copyright © 2011 JIN SHI