Keysight Technologies BER Measurement Using a Real-Time Oscilloscope Controlled From M8070A. Application Note

Transcription

1 Keysight Technologies BER Measurement Using a Real-Time Oscilloscope Controlled From M8070A Application Note

2 02 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Real-Time Error Detection above 32 GBd The scope of this application note is to explain the BER (bit error ratio) measurement procedure using the M8045A pattern generator and DSAZ634A Infiniium Z-series oscilloscope when controlled from M8070A system software for BER test solutions. This configuration provides a solution for BER measurement at higher data rates above 32 Gbaud. The real-time scope can be used to measure BER to the higher data rates found in links using PAM4 signaling. The basic differences between this method and the traditional BER measurement approach using the M8040A high performance BERT (bit error ratio tester) are explained, where both use the same M8045A PG to generate the data stream. The hardware and software configurations of the setup using the real-time scope as pattern capture front-end are described. Then the measurement details of BER, including variation of BER under various data rates, time required to complete the measurement, and input sensitivity of real-time scope are discussed. Finally, the differences between the BERT and real-time oscilloscope-based approaches are detailed. Configuration for BER Measurement with Real-time Scope and M8070A Software Prerequisites The following requirements must be ensured for the successful integration of the real-time oscilloscope with the M8070A software. Supported oscilloscope models: DSOZ634A DSAZ634A DSOX96204Q DSAX96204Q DSAZ594A (version 4.5 onwards) DSOZ594A (version 4.5 onwards) Oscilloscope firmware version: or higher Following licenses are required on the oscilloscope N5384A Serial Data Analysis (SDA) N8827A PAM4 Measurement (PM4) N5461A Equalization (DEQ) M8070A software revision: 4.0 or higher For demonstration purpose, following instruments are recommended as pattern generators M8045A pattern generator from M8040A high-performance BERT M8196A arbitrary waveform generator

3 03 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Hardware For the purpose of demonstration, a direct loopback configuration between the pattern generator (PG) and the oscilloscope was used with no Device Under test (DUT) in the data path. Hence the oscilloscope input is directly connected to the pattern output. The hardware components required for this setup are: one of the oscilloscope models mentioned above, M8045A pattern generator (PG) from the M8040A BERT-series instruments, and cable accessories. The differential inputs of channel 1 of the M8045A PG and its remote head M8057A are connected to the RealEdge inputs of the DSAZ634A oscilloscope through the M8057A. Use of proper cables is extremely important for signal quality. We recommend to use a combination of M8045A-801 short cable 1.85 mm (m) to 1.85 mm (m), 11900B female-to-female adapters, and M8046A mm 0.85 m matched cable pair for the connection between the remote head and oscilloscope. Both oscilloscope and AXIe chassis, in which the M8045A PG is housed, must be connected to the controlling PC with M8070A software (version 4.0 or higher) installed on it, using USB or HiSLIP. Device Under Test (DUT) Differential out PC USB2 USB1 Real-time Oscilloscope USB RealEdge 1R RealEdge 3R Direct loopback PG (BERT PG/AWG) Chasis Chasis USB Figure 1. Basic connection setup: Real-time scope controlled from M8070A

4 04 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Software First-time setup Additional steps must be performed to configure the software prior to first use. Once the real-time scope is connected to the PC, from the IO Libraries Suite, open the Connection Expert and check whether the oscilloscope is shown. If connected, it should be displayed. (See screen shot on following page.) Click on that icon on the left side, then all the details of the scope will be shown on the right side. Please change the visa aliases set to this instrument to one of the following: RTS_PROXY, RTS_PROXY_1, RTS_PROXY_2, RTS_PROXY_3 or RTS_PROXY_4. Once this is done, the instrument is ready to be successfully integrated into the M8070A software. This is as shown in Figure 2 below. Click here to change the alias Figure 2. IO Libraries Suite Connection Expert Normal setup The interface is modified in such a way that during the measurement of BER, the user doesn t need to access the oscilloscope s user interface. All the necessary controls are included in the M8070A software. Once hardware connections are made, power up the M8045A PG (AXIe chassis) and oscilloscope. Both the instruments will require a warmup time achieve full accuracy. When they are ready to be used, the M8070A software can be launched. During the initialization process, the software will configure both M8045A PG and oscilloscope. Before we go into the details of BER testing, we ll quickly have a look at the oscilloscope controls that are included in M8070A software.

5 05 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Auto Alignment Figure 3. Infiniium Z-series oscilloscope controls incorporated in M8070A software After the successful launch of M8070A, go to the Modules View tab, which is as shown above. We can see that INF1 is shown as a separate module, treating the oscilloscope like an M8040A module (such as M8045A PG or M8046A ED). As shown in Figure 3, the controls are divided under two tabs: Common and Data In. The Common tab sets the general-purpose measurement parameters. As shown above we can see that the Common tab includes sub-tabs such as Acquisition and Horizontal. The Acquisition sub-tab controls acquisition related parameters such as sampling rate, memory depth and can also reconfigure the oscilloscope through the Reconfigure option. The Horizontal sub-tab has an option of activating or deactivating the external reference clock input if any such reference is applied to the oscilloscope. The Data In tab controls the parameters related to the input of the oscilloscope. It has sub-tabs such as Acquisition, Clock, Equalization, Line Coding, Comparator and CDR (clock data recovery). An important feature to note is the auto-alignment function (shown in Figure 3 beside the BER display). The real-time oscilloscope doesn t have this feature. This feature is not quite the same as the auto-alignment feature of the M8040A-series BERT, where it is used to search for the widest eye-opening with its voltage threshold and sampling point (delay) being stepped over the eye. The auto-alignment feature here optimizes the sampling threshold, delay and equalizer. Before starting the measurement or when changing any of the above measurement parameters (such as data rate, type of line coding etc.), auto-alignment must be done. For example, initially when everything is set up as wanted then by hitting this auto-alignment, the widest eye opening will be automatically found by adjusting the delay and voltage threshold values and the lowest possible BER measurement will be generated. If equalization is performed on the incoming data, then this auto-alignment will first optimize for the unequalized signal to find the widest eye opening, then it will optimize the equalization process with the given set of parameters (such as taps and pre-taps) and finally it will reoptimize the equalized signal to find the widest eye opening. In this manner, auto-alignment ensures the optimum sampling point and equalizer setting. We also have the option of individual auto-alignments in Comparator and Equalization sub-tabs to optimize the eye-opening (delay and threshold) or equalization individually.

6 06 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note BER Measurement Using M8040A BERT Traditional BERT uses a straightforward approach to bit error testing wherein it sends a known pattern of bits and checks the bits that are received in real-time against the expected pattern, counting the errors. The clock for the error detector is then provided either by the pattern generator, the DUT or a clock recovery unit. Using a real-time scope (stand alone) The BER (Bit Error Ratio) and SER (Symbol Error Ratio) measurements for PAM4 signals require clock recovery and at least two error-free copies of an identically repeating bit pattern in acquisition memory. Typically, a PRBS test pattern is used. The oscilloscope stores these copies into its memory and then compares the subsequently measured bit patterns with those saved patterns and determines the BER. This feature is not available for NRZ patterns and the capture memory limits the length of the pattern. The BER measurements are provided in two ways: BER per acquisition which gives the BER for every acquisition of the incoming bit pattern and a cumulative BER which gives combined BER of all the acquisitions until the current one. Using a real-time scope controlled by M8070A This approach uses the M8070A software to control the real-time oscilloscope. In this case the real-time oscilloscope is used only to capture the signal and convert it into symbol levels (as a digitizer) and the comparison with the expected pattern is done by M8070A. The main advantages of this approach as compared to the case where the oscilloscope is used as an error detector are: 1. Comparison against an expected pattern and not against the stored pattern (in case of the real-time scope, the stored pattern may have systematic errors already). The stand-alone oscilloscope can never establish the reference pattern if the link is extremely bad in terms of BER performance because we would never get a single pass of the pattern that is error free. 2. It can work for longer PRBS patterns as well (for example, PRBS 2^31 or QPRBS31) 3. It works for both NRZ as well as PAM4 (real-time scope supports only PAM4) 4. It can use all measurement capabilities in M8070A e.g. jitter tolerance measurement Having discussed the approaches to measure the BER, let s now have a look at how to make a BER measurement with the real-time oscilloscope by using M8070A software. Key settings to measure the BER (per acquisition) Once the M8070A software is successfully launched, we need to follow a proper sequence to enable the BER measurement using the real-time oscilloscope. Step 1 - PG settings: Under the M8045A PG section of the M8070A software Set the pattern to PAM-4 (for this example) Adjust the data rate to 26 GBaud (this is the default rate set) Set the pattern to PRBS 2^15-1 Set the input signal amplitude to 300 mv pp Finally turn on the output amplifiers. To detect the PAM-4 signals, the oscilloscope should also be configured in accordance with the PG in terms of clock and type of line coding.

7 07 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Step 2 - Real-time oscilloscope settings: Reconfigure the oscilloscope (clicking on the reconfigure option under acquisition tab) to refresh the settings Set the parameters such as clock frequency (in this case 26 GHz by default), line coding (PAM4), acquisition depth (128000), loop bandwidth divisor to 5000 (under CDR section of Data In tab, recommended to be set to 1667 in case of NRZ) Select the necessary pattern (in this case PRBS 2^15-1) from the sequence editor and ensure it is the same as that of the PG Turn on the global outputs Set the oscilloscope s acquisition state to ON under the Common tab of the oscilloscope in the M8070A software. Step 3 - Auto alignment: Perform the auto-alignment by clicking on Auto Alignment as shown in Figure 3 on page 5. Once done, we should see the BER 0 at the bottom right corner of the M8070A front screen. Note that this BER 0 (displayed on screen) is the BER per acquisition and not the cumulative BER and it is evaluated at 26 GBaud data rate. When we click on the auto-alignment, we can see a clear open eye on the oscilloscope corresponding to the zero BER and it is also shown on the M8070A screen. As we increase the data rate, we can see that BER is maintained at 0 until 50 GBaud and thereafter it starts encountering errors, causing the eye to become extremely narrow. This data rate value (50 GBaud) is subjected to change if different cables or setup is used. Need for equalization All PAM4 and most of the NRZ links working at very high speed (above 45 GBaud) do not run error free by design. Hence, if loopback is tried out either with ED or with real-time scope, BER is non-zero. To reduce this problem, we can introduce either deemphasis or equalization. Deemphasis is introduced at sending end i.e. PG but equalization is introduced by either ED or real-time scope (at the receiving end). The main purpose of equalization is to correct for the problem caused by the transmission channel. Equalization techniques provide a way to discern the original signal (the signal coming out of the transmitter) given a distorted signal at receiver. The equalization facilities given by real-time oscilloscope are FFE (feed forward equalization), DFE (decision feedback equalization) and CTLE (continuous time linear equalization). Out of these, FFE is included in M8070A. To handle the problem of non-zero BER at higher data rates (above 45 GBaud), it is recommended to use the equalization function above symbol rates of 50 GBaud in a direct loopback configuration. When a real DUT is used, it should be used above 45 GBaud. As explained earlier, this equalization facility is also included in the Equalization tab under the Data In section. This section gives us the liberty of setting parameters such as number of FFE (Feed Forward Equalization) taps and pre-taps. It also has the option of auto-set for the equalizer taps. We can manually set the taps of equalization but if we increase the number of taps, the measurement speed decreases considerably. By default, the equalizer is set to 5 taps and 1 pre-tap. The normal procedure to apply equalization is simple; select the number of taps (in this case they are kept at their default value of 5 taps) and turn on the equalization. Once turned on, the equalization will be applied to the input data and when done we can see the widely opened good quality eye. The auto-alignment must be done to ensure the wide eye opening after the equalization parameters are set. This open eye is seen on the oscilloscope. With equalization, a BER of 0 can be obtained up to 58 GBaud. Above 58 GBaud, signal quality gets too distorted and hence it is not recommended to use above this limit. Below is the eye diagram with and without using the equalization (9 tap FFE with 3 pre-taps) at 56 GBaud of data rate where we can clearly see the signal quality improvement after using equalization.

8 08 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Figure 4. Eye diagram with (the lower one) and without (the upper one) the use of equalization at 56 GBaud Measuring the cumulative BER Before we go into the details of cumulative BER measurement, let s see how real-time oscilloscope differs from BERT in terms of collecting the input data. Unlike a BERT, the real-time scope is a sampled system and does not acquire data continuously. It acquires in acquisitions. As explained, cumulative BER is the BER of all the acquisitions measured through the current one. To make sure that our system should maintain BER above a certain threshold, target BER is defined. Target BER is the specified BER which an instrument (DUT) is expected to achieve. In other words, it s the BER targeted at a particular value; if a measured BER is less than the target BER then the system is compliant to the specified BER. The detailed BER measurement approaches which focus on measuring the BER under various parameters, can be considered as the aspects of the cumulative BER measurement as they all focus on measuring the BER for the specified time intervals or even for the indefinite time interval. There are three approaches to make the detailed BER measurement which are: BER measurement running for full duration, pass/fail type of BER measurement and BER measurement for a given number of bits. All these options are available under the Accumulation End sub-tab. In the full duration BER measurement, the measurement will be made and graph will be plotted for the specified amount of time (accumulation duration). In the case of pass/fail measurement. the target BER is specified and the BER measurement will continue until the specified confidence level is reached for the specified target BER. There is a statistical relationship between the confidence level achieved and the number of bits compared (roughly 95% confidence level with target BER 1E-6 requires 3 million bits to be checked). As the number of compared bits increases, the confidence level also increases until an error is encountered. If an error is encountered, the confidence level decreases and the measurement continues. If the number of errors encountered are so high that the specified confidence level can never be reached (or in other words that the measured BER is greater than target BER), then the measurement is declared as fail. This type of BER measurement can also be made time limited by specifying accumulation duration. In that case if the required confidence level is not reached in the specified time interval, then also this measurement is declared as fail. In the third type of BER measurement, the specified number of bits are compared and a BER is shown.

9 09 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note In this section, the pass/fail type of BER measurement is explained in detail. To initiate the target BER measurement, we need to select the BER measurement from the Measurement menu. Once clicked it opens a new window showing detailed BER measurement parameters. To set the target BER, go to the Accumulation End sub-tab under Acquisition Parameters and select its type as pass/fail. Set the necessary parameters such as target confidence level (95% in this case) and target BER (1E-6 in this case). After this we are now ready to measure the BER with the specified parameters. The target BER measurement window looks as shown in Figure 5. The accumulation interval is fixed to 200 milliseconds. To start the measurement To start the measurement Figure 5. Target BER measurement with 1 E-6 as target BER at 95% confidence To start the test, once all parameters are set, click on the green arrow in the top left to commence the test. Once the test is started, the oscilloscope will start digitizing incoming data signal and M8070A starts checking the received bits and this we can see in the compared bits and errored bits sections of the results. At the interval of 200 milliseconds (each accumulation interval), the oscilloscope updates the compared bits and the confidence level indicating the increase in number of compared bits and confidence level as well. This will continue until the oscilloscope reaches 95% confidence and then the result will be shown. If any error is encountered in this time frame, the confidence level will be decreased (as explained above). We can certainly increase the measurement window of target BER by setting it even lower (<1E-6) but as we go lower, those many more number of bits that must be checked eventually making the BER measurement very time consuming. The direct loopback BER measurements, i.e. per acquisition as well as detailed BER measurements are successful without using equalization until 50 GBaud (45 GBaud in case of a real DUT) only. Thereafter equalization must be incorporated to ensure the BER is maintained at 0. Again, the direct loopback case is ideal and the BER value may be higher with the real DUT in picture. Hence it is advised to incorporate equalization feature at lower data rates. The default FFE parameters are with 5 taps and 1 pre-tap. But as we approach 56 to 58 GBaud, even this is not sufficient and we should increase the equalization window. For example, setting it to 9 taps and 3 pre-taps. Incorporation of equalization ensures that the system BER is maintained at 0 until 58 GBaud but as we increase the equalization level, the oscilloscope needs more time to give out the equalized digitized signal (applying equalization on received signal and then digitizing it) resulting in a considerable decrease in the measurement speed. When using NRZ signals also, we recommend to use the solution up to 58 GBaud even though the over programming allows the symbol rates up to 64 Gb/s without any need for adding equalization in a direct loopback setup. We must change the settings accordingly (line coding to NRZ and loop bandwidth divisor to 1667 which is advisable).

10 10 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Error insertion To make sure that the BER is being measured correctly, it can be cross-checked by using the error insertion capability available in the M8045A (PG). Figure 6 shows the error insertion window in the M8070A software interface for the M8045A PG module. Figure 6. Error insertion window in M8070A software for PG module We can see the error ratio option. We can select the error ratio which we would like to introduce in the data path. If we select the error ratio as 1E-4 and turn on the error insertion, then we can see that the acquisition BER (BER at bottom right corner of M8070A screen) shown on the screen also shows BER in the order of E-4, indicating that it is encountering errors and the same is shown. If we turn the error insertion off, then BER again shifts back to 0. This confirms that after the error insertion, M8070A software detects it accurately and can reliably be used for BER measurement. This error insertion approach is tested for both PAM4 and NRZ line coding. It should be noted that in our case the target BER threshold is set to 1E-6. Hence if we insert an error at the rate which is less than 1E-6, then those will be overlooked and will not be detected by the system. For example, in the above case if we mention the error insertion ratio as 1E-8 (< 1E-6) then M8070A software will not detect it. To enable it to detect the error ratios of the order less than 1E-6, we must decrease the target BER threshold to less than 1E-6.

11 11 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Measurement sensitivity The sensitivity of the real-time scope is the minimum peak to peak amplitude of an input signal that the scope can detect. In case of PAM4, the relevant amplitude is the single-ended peak to peak amplitude of the data input. All the above-mentioned tests for BER measurements are done at the input amplitude of 300 mv pp and this amplitude was kept constant even though the data rates were changed. Sensitivity measurements on the oscilloscope are made with specific target BER. In our case, we have limited the target BER to 1E-6; hence sensitivity measurements which are discussed hereafter are valid for a target BER of 1E-6. In case of PAM4-line coding, when sensitivity measurements were done on oscilloscope, it was found that the oscilloscope has very good sensitivity as low as 50 mv pp (single ended and without any error or jitter inserted and without equalization applied) up to 45 GBaud. Above 45 GBaud the sensitivity requirements are categorized based on the applied equalization. Without any equalization applied, the oscilloscope sensitivity requirements increase with the increase in data rate. At 50 GBaud, the oscilloscope needs 63 mv pp to sense the input data signal. At 56 GBaud it needs 150 mv pp and at 58 GBaud it needs as high as 250 mv pp. Above 58 GBaud it is not recommended to use the oscilloscope for BER measurement. Still if we continue to 60 GBaud, it shows sensitivity requirements of 250 mv pp and thereafter some varied fluctuations such as at 61 GBaud about 450 mv pp and at 62 GBaud 350 mv pp. With equalization (9 tap FFE), it was found that until 56 GBaud, the required sensitivity is as low as 50 mv pp and thereafter it increases. At 58 GBaud it needs 63 mv pp and then increasing sharply to 350 mv pp and 450 mv pp at 60 and 61 GBaud. Like the case where no equalizer was used, in this case it also shows fluctuations in sensitivity above 61 GBaud such as at 62 GBaud it needs around 425 mv pp. Considering both the cases, we can conclude that the sensitivity requirements generally increase with increase in data rate and equalization improves the performance by maintaining the low sensitivity values at higher data rates until 58 GBaud. All the above data is summarized in the Figure 7 shown below: Figure 7. Sensitivity of real-time scope DSAZ634A with and without equalizer

12 12 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note For different target BER such as 1E-7 or even 1E-8, it is found that we need to increase the amplitude at lower data rates. For example, at 55 GBaud we need to maintain the minimum amplitude of 63 mv pp at a target BER of 1E-8 and thereafter increasing sharply to 463 mv pp at 56 GBaud. In case of NRZ, sensitivity as low as 50 mv pp (single ended peak to peak) is possible until 58 GBaud (with over-programming until 64 GBaud but it is not recommended). Jitter tolerance testing In this chapter, basics of jitter tolerance testing and conducting this test with the M8070A software (using BERT and real-time oscilloscope) are discussed. Jitter: This is the deviation from true periodicity of a presumably periodic signal, often in relation to a reference clock signal [4]. In simple terms, it is the variation in the data transition instant (1 to 0 or 0 to 1) with respect to its expected position. Excessive jitter should be avoided because it leads to the eye closure leading to errors in the data stream increasing the BER. There are various types of jitter classified as per their pattern such as random jitter and deterministic jitter. Deterministic jitter is further classified as periodic, data dependent (which is even further classified as duty cycle distortion and inter symbol interference) and bounded uncorrelated jitter. A few examples of sources causing jitter are thermal noise (random jitter), cross talk (bounded uncorrelated jitter), long or short bits (inter symbol interference) etc. The total jitter is the summation of all the individual jitter contributing to data impairments. Total jitter at a bit error ratio Abstract peak-to-peak Random jitter (RJ), σ Unbounded fluctuations Deterministic jitter, Jpp DJ Bounded, peak-to-peak Thermal noise, shot noise, flicker Periodic jitter (PJ) Sinusoidal Data dependent jitter (DDJ) Data smearing Bounded uncorrelated (BUJ) Crosstalk Duty cycle distortion (DCD) Lead/trail edge Inter symbol interference (ISI) Long/short bits Figure 8. Classification of jitter 1. Wolaver, Dan H. (1991): Phase-Locked Loop Circuit Design

13 13 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Jitter Tolerance: Knowledge of system jitter tolerance provides key insights into the digital design circuitry. Each receiver is designed to accept a certain amount of jitter. Jitter tolerance is the ability of the receiver to tolerate or to deal with the jitter on the received signal so as to successfully receive the pattern symbols and interpret them. The most receiver standards, define a certain jitter tolerance with amount of jitter across a certain jitter modulation range. Before we go into the details of jitter tolerance let s have a closer look at the sinusoidal jitter which is introduced into the data stream to carry out the jitter tolerance measurement. Sinusoidal jitter is the periodic jitter in which the jitter function follows a sinusoidal curve. It means the variation of the transition event with respect to its ideal position will be sinusoidal. Following figure shows the jittered NRZ signal seen on the Infiniium sampling oscilloscope DCA-X 86100D at 28 GBaud of data rate. From the histogram, we can clearly see the two peaks at the end resembling the sine function. In this example, 1 MHz of low frequency sinusoidal jitter with 700 mui of amplitude was injected using M8045A PG. Figure 9. Sinusoidal jitter as seen on Infiniium DCA-X Basic jitter tolerance testing approach: The basic approach of testing jitter tolerance includes injection of sinusoidal jitter into the pattern generated by pattern generator and measuring the BER via a loopback path. The target BER can be selected for a jitter tolerance measurement. The BER measurement is carried out for various jitter amplitudes and at different jitter modulation frequencies. The test begins with the injection of sinusoidal jitter of particular frequency (start frequency) and the jitter amplitude is varied (either increased or decreased depending upon the algorithm selected) until the BER exceeds the target BER set and then the measurement shifts to next jitter modulation frequency value (either higher or lower frequency depending upon the algorithm selected). In this way, a jitter frequency vs jitter amplitude graph is plot which is the outcome of jitter tolerance test. Automated sweep of the jitter amplitude and the jitter modulation frequency helps to minimize the duration of a jitter tolerance measurement. Let s look at the jitter tolerance test window in M8070A in more detailed manner.

14 14 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note To begin the JTOL test Figure 10. Jitter tolerance measurement window In this window as shown in Figure 10, we have tabs such as BER Setup, Graph, Instruments and Measurement Setup. We ll discuss each tab in detail. BER Setup: The sub-tabs include Target BER: Sets the desired target BER Confidence Level: It decides the reliability of the BER measurement. The more the confidence level, the more will be the number of bits compared for the target BER set. Higher confidence levels require measuring more data, and hence measurement time is increased Frequency Relax Time: Time duration for which the measurement will wait before hopping to the next jitter frequency Amplitude Relax Time: Time duration for which the measurement will wait before hopping to next jitter amplitude at the same frequency Graph: This tab consists of sub-tabs which are associated with displaying the jitter tolerance results in graphical manner. It has four sub-tabs which are Template Limits, Points, Compliance Limits and Show All Points. The sub-tabs are self-explanatory. Instruments: This tab states the devices involved in the measurement. It has two sub-tabs which are Generator and Analyzer. When the M8045A PG is used, we have for Generator two options to select-either channel 1 or channel 2. These settings can be changed through this tab. Measurement Setup: This is the most important tab and it sets all the necessary parameters required to configure JTOL test. The subtabs are: Start and Stop Frequencies: They indicate the start and stop frequencies of the jitter tolerance measurement. Depending upon the algorithm, we can select either a high-to-low or a low-to-high frequency type of measurement.

15 15 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Number of Points: It is the number of frequency hops that the measurement will take when it is running from start to stop jitter frequencies. Mode: Two modes are possible - characterization or compliance. In compliance, the jitter tolerance will only be done for compliance points whilst in characterization, the test will be performed as per the mentioned specifications. Compliance Margin: Given only in compliance mode, it indicates the acceptable limits of system performance (results). Algorithm: It indicates the algorithm selected for performing the jitter tolerance test. Few examples of algorithm are Up Logarithmic, Down Logarithmic, Up Linear, Down Linear, Adaptive Binary etc. In Up Linear, the amplitudes start from a very low value (0.01 UI) increase until a very high value when the resultant BER is more than the target BER and the increment in amplitude is linear. Binary Step Size: Used for adaptive binary type of algorithm, it indicates the amplitude step value when hopping from one amplitude to the next. Linear Step Size: Used for linear algorithms where it indicates the amplitude step value when hopping from one amplitude to the next. Log Step Size: Used for logarithmic algorithms where it indicates the logarithmic step value in percentage when hopping from one amplitude to the next. CDR LBW Auto: When turned ON, the jitter tolerance measurement takes into consideration the CDR LBW (clock data recovery loop bandwidth) curve of the DUT. This option is by default in ON state. Jitter tolerance test using a BERT and a real-time scope when controlled from M8070A An automated jitter tolerance measurement is possible with the M8070A system software for the M8000 series. Starting from software version 4.0 it can also be used with a real-time oscilloscope together with a pattern generator, such as M8045A (or AWG). A standalone real-time oscilloscope cannot perform a jitter tolerance measurement. We ll now discuss how to conduct the jitter tolerance measurement for both BERT and real-time oscilloscope controlled from M8070A. The high performance BERT is well equipped to carry out the jitter tolerance measurement. We set all the necessary parameters as mentioned in the previous section and start the test by clicking on the green arrow in the top left corner as shown in Figure 10. The target BER value which is set to 1E-9 by default. We can surely change this or alternatively we can continue with the test keeping the same value. But for real-time oscilloscope we recommend using a value of 1E-6 or more (1E-5, 1E-4 etc.), otherwise the measurement would take too long. The time taken by the measurement depends upon the time required to complete the individual BER measurements. BERT being very quick, jitter tolerance measurement targeted at 1E-9 BER takes a few minutes to complete. The jitter tolerance measurement test time also depends upon the factors such as the number of points, the algorithm selected, the selected mode which decides the number of number of points to be tested. But if we keep these parameters constant then the measurement time is directly impacted by the individual target BER measurement which is then affected by various factors in the case of real-time oscilloscope controlled from M8070A as we ll discuss in the next topic.

16 16 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note If we keep the target BER very high such as 1E-3 or 1E-4 then the measurement time is limited by acquisition time but if we keep the target BER very low such as the default target BER of JTOL test measurement 1E-9, the testing becomes impractical for real-time scope because of the time taken to complete the measurement. Refer to the Figure 11 which shows the time taken to complete the measurement at different target BERs. All the lower target BER values (<1E-6) are therefore not practical. We recommend to use the adaptive binary algorithm for jitter tolerance testing when the real-time scope is controlled from M8070A as it saves the measurement time considerably. In this algorithm, typically we start from the highest measurement jitter frequency to the lowest jitter frequency. For the highest jitter frequency, the UI amplitude will be gradually increased as per the binary step size given until it encounters first failure. Once this is confirmed, for the subsequent frequencies, this amplitude is taken as a reference point and from this point, next UI amplitude is checked for compliance. This process continues and the graph is plotted. JTOL measurement testing graphs for real-time scope when controlled from M8070A for different target BERs are as shown in Figure 11 below. Figure 11. Jitter tolerance measurement results when a scope is used and controlled from M8070A

17 17 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Detailed comparison between BER measurement using a BERT and a real-time scope This ppart of the application note will explain in detail the features of both traditional BERT measurement and BER measurement using the real-time oscilloscope. We have already seen that both differ in their approaches, where traditional M8040A high-performance BERT directly compares the received bits against the sent ones whilst the scope initially needs two error free copies of the pattern in its acquisition memory and compares all the received patterns against the stored ones to make BER measurement. But when the scope is controlled by M8070A software, bit comparison and BER calculation is done by the M8070A software. A real advantage of using real-time oscilloscope for error detection purposes is the ability to see the waveform used for loopback. This allows verifying equalization and verifying the various parameters of the waveform used. These approaches can be compared with respect to various aspects. Some of the important aspects are: Measurement time: This is the major aspect when a real-time scope is used for the measurement. The measurement with the traditional BERT M8040A is very fast and quick. It takes about a few milliseconds to complete the measurement and produce the BER result. An important property of BERT is that the time required for measurement is much less even if the symbol rates change (although it s not totally independent of symbol rates; for example, to capture bits takes 30 times longer for 2 Gb/s than for 60 Gb/s). When the real-time scope is controlled from M8070A, the oscilloscope is used to capture and digitize the incoming signal. In this case, as discussed earlier, the target BER set in our case is 1E-6. If we decrease the target BER threshold, then the oscilloscope must digitize many more bits to satisfy the required target BER threshold to a certain confidence level (95% in this case). Without using equalization, the BER measurements at 32 GBaud PAM4 at bits acquisition depth (UI) looks as shown in Table 1. Duration is shown in hours: minutes: seconds. Table 1. BER without equalization at 32 GBaud (applies for DSAZ634A controlled by M8070A) Target BER Confidence Level Duration Compared Bits Confidence at Target BER 1e-4 95% e % 1e-5 95% e % 1e-6 95% e % 1e-7 95% e % 1e-8 95% e %

18 18 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note In Table 1, we clearly see that decreasing the target BER threshold increases the duration of the measurement. Comparing target BER 1E-7 and 1E-8 for example, the duration is 5 minutes and 47 minutes respectively at 95% confidence level. It is thus feasible, but the significant increase in measurement time makes it impractical for the use in typical test lab environments for characterization and validation of designs. If we incorporate equalization, the measurement duration increases even more. Table 2 illustrates that equalization greatly slows down the measurement. Even at a target BER of 1E-7, the measurement takes more than an hour, eventually making it impractical. Moreover, if we increase the extent of equalization, such as using 9 taps instead of 5 taps, then it increases the measurement duration even more. Table 2. BER with 5 tap FFE at 32 GBaud (applies for DSAZ634A controlled by M8070A) Target BER Confidence Level Duration Compared Bits Confidence at Target BER 1e-4 95% e % 1e-5 95% e % 1e-6 95% e % 1e-7 95% e % 1e-8 95% ~ ~3e8 ~95% Another parameter that affects the measurement duration when using a real-time oscilloscope is the symbol rate of the received pattern. The ratio between symbol rate and sample rate defines how many samples are taken per UI. The symbol and sample quantities are related by the following equation: N(Sample) = Sample Rate Symbol Rate * N (Symbols) Where: N(Samples) = Number of samples N(Symbols) = Number of symbols Using the Real Edge inputs of the Infiniium Z series oscilloscope, the sample rate is fixed to 160 GSa/sec i.e. 160 GHz. For a specified target BER, as we have discussed earlier, the oscilloscope must digitize a fixed number of symbols for a specified confidence level. It means in the above equation for a specified target BER, N(Symbols) are also constant which implies that N(Samples) are inversely proportional to the symbol rate. This can be simplified as follows; when we decrease the symbol rate, the oscilloscope needs more time to reach the specified number of symbols to ensure the target BER and confidence level. But as sample rate is constant for the oscilloscope, it will take more number of samples over the increased time interval. The higher number of samples requires greater memory depth, which increases the processing time. Therefore, lower sample rates take longer to process for the same number of symbols (decided by target BER), effectively leading to increase in the measurement duration. For symbol rates below 15 Gbaud the measurement duration increases significantly, so that the attractiveness of the real-time scope for making BER measurements in typical validation labs gets marginal.

19 19 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note The acquisition depth of an oscilloscope also affects the measurement time. If we increase the acquisition depth, then more bits are captured in one acquisition, indicating that we ll need less acquisitions to reach the target bit threshold and the measurement time will be reduced. But with higher acquisition depths, processing and update times will also be longer. Type of line coding used is also responsible for affecting the measurement time. At the constant data rate, time required to reach the target threshold set by target BER at specified confidence level (normally specified at 95%) by PAM4 type of line coding is half as compared to NRZ type of line coding. This is because in one symbol NRZ has one bit whereas PAM4 has two bits making is twice as fast as NRZ type of line coding. Update rate and measurement time: There is another associated with the measurement time which is the update rate. It is the rate at which the error information is updated. It directly depends on the individual acquisitions hence we can say that there is a direct relationship between the time required for one BER per acquisition measurement and the update rate. This eventually affects the measurement duration. For example, one update with one million bits takes less time than ten updates with 10 5 bits. Hence, we can say that lower update rates have less measurement time but the measurement looks like being stopped because of increased acquisition depth. Update rate is affected by acquisition depth, symbol rates, equalization and type of line coding. The increase in acquisition depth decreases the update rate. The lower symbol rates take longer to process the logic behind is same as that of the measurement time. Equalization if introduced reduces the update rate and PAM4-line coding is twice as fast as NRZ in terms of update rate. The following Figure 12 illustrates the dependencies assigning the exact values to them. Figure 12. Update rate dependencies All in all, traditional BERT is very fast, whereas factors such as target BER, amount of equalization incorporated, symbol rate and acquisition depth affect the BER measurement using the real-time oscilloscope.

20 20 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Performance of BER testing: BERT instrumentation is specifically designed for BER testing and hence delivers real-time, true and fully-sampled measurements (against an expected pattern in real-time comparison and 100% of bits compared). Each received bit is compared against the one that is expected without any gaps in the acquisition interval in between. Hence, the ED (error detector) of BERT knows exactly what is correct. Using the real-time oscilloscope as a standalone error detector, if the initial error-free copy encounters an error, all the subsequent comparisons will be shown as errored. Hence, the BER measurement shown by the oscilloscope is just relative to the captured reference pattern with time gaps in between the subsequent acquisitions (explained later in the same section). On the other hand, with the M8070A, the scope just captures and digitizes the received signal. The signal is captured in acquisitions by the scope and sent to the M8070A software where it is compared against an expected pattern and the BER measurement result is provided. The BER testing is affected by the time gap between the successive acquisitions, during which the oscilloscope is blind and does not measure. If we want the measurement to be more reliable, then we must increase the acquisition depth of an oscilloscope to store more bits for subsequent comparisons. If we increase the acquisition depth to a very high value (such as 2 Mbits), then it is possible to collect all the required bits in a single shot (depending on the pattern length) and the measurement can be given directly. But if we lower the target BER, then surely, we need to compare a substantial number of bits that can t be finished in a single shot of acquisition. Also, as mentioned earlier, increase in acquisition depth results in increased processing time and low update rate. This makes the oscilloscope and even the measurement appear stopped (for higher acquisition depths such as 10 Mbits, 1 acquisition takes almost 34 seconds). It is recommended to keep acquisition intervals short for very low target BER threshold. Considering all the things above, we can conclude that the approach in which the real-time oscilloscope is controlled from the M8070A software gives BER information with the requested confidence level, but it is affected by the dead time between the subsequent acquisitions. However, the ability to compare against an expected pattern rather than a captured reference pattern eliminates the risk to overlook the systematic bit errors. The integration approach will detect the errors which are randomly distributed. The probability of not detecting errors only exists when errors are occurring in the dead time when the oscilloscope is blind and this phenomenon is repeated exactly in the dead time interval. This probability is very low hence, we can definitely use the real-time scope when controlling it from M8070A for making BER measurements. Achievable symbol rates: The currently available error detector options at the time of publishing of the BERT M8040A system supports symbol rates up to GBaud for NRZ line coding and 30 GBaud for PAM4. It means that although the M8040A series BERT PG (M8045A) supports symbol rates up to 64 GBaud, the total system performance is limited by its M8046A error detector. If only the real-time scope is used as BER tester, then error-free loopback up to approximately 58 GBaud is possible using the M8045A as pattern generator. (Above 58 GBaud, the loopback is not error free.) Hence, the real-time scope can be used at higher data rates than achieved with the traditional BERT. When real-time scope is controlled from M8070A, like real-time scope standalone, higher data rates are achievable. In the current setup, the M8045A pattern generator and the real-time scope together can deliver measurements until 58 GBaud (can reach until 64 GBaud but this is not recommended). As previously described, for PAM4 coding it is recommended to use real-time oscilloscope up to 58 GBaud and equalization is needed above 50 GBaud for loopback (45 GBaud in case of real DUT). For NRZ line coding also the data rates until 58 GBaud are supported without using the equalization (over-programming can support data rates up to 64 GBaud but this is not recommended). Again, this only applies to the direct loopback case, with certain cables. The distortions introduced by the DUT transmitter used for loopback can change this limit substantially.

21 21 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Sensitivity: We have already discussed the sensitivity of the oscilloscope. In this section, we ll focus on the differences between traditional BERT and real-time scope in regard to BER measurement sensitivity. With the M8040A BERT, the input sensitivity of M8046A error detector is 70 mv pp for NRZ. The same applies for PAM4 but this is per eye. Moreover, this 70 mv pp value is the required inner eye opening. So, for PAM4, the minimum peak to peak amplitude required is at least 210 mv pp. On the other hand, the real-time scope has sensitivity of 50 mv pp single ended for PAM4-line coding until 50 GBaud without equalization (only with direct loopback and without DUT) and until 56 GBaud with 9 tap FFE. This 50 mv value (from the sensitivity diagram shown in Figure 7) is the data out peak to peak amplitude. The inner eye seen by the scope digitizer is at least three times smaller. With 9-tap FFE incorporated, it can perform down to 63 mv until 58 GBaud which is the recommended data rate limit. Hence, we can definitely say that the real-time scope performs better (or has better sensitivity) with low amplitude incoming data signals as compared to traditional BERT. In the case of NRZ line coding as well, the real-time scope performs better as it functions well with sensitivity requirements as low as 50 mv pp (single ended peak to peak) until 58 GBaud data rates. Equalization: This is another parameter where the real-time oscilloscope has an advantage. In case of traditional BERT, software equalization is done by giving the number of levels as a direct input under the DataIn tab for the ED module. The oscilloscope however offers a large extent of adjustability in terms of equalization. We can set parameters such as taps and pre-taps of the FFE equation. On the basis of number of taps and pre-taps, the real-time scope calculates the coefficients required for the FFE equalization process. Type of line coding: The M8040A BERT supports both NRZ and PAM4 signals. The standalone real-time oscilloscope supports only PAM4 signals. But when it is controlled from M8070A, both NRZ and PAM4 line codes are supported for BER measurement. To summarize the comparisons, we can say that the real-time scope offers advantages in terms of achievable data rates, sensitivity and equalization whereas it certainly takes more time to carry out BER measurement and the measurement duration is affected by various factors such as equalization, target BER, symbol rate and acquisition depth.

22 22 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Summarizing the Different Error Analysis Approaches Real Time Scope Standalone Real Time Scope Integrated into M8070A BERT Error Detector No true error information Reported BER/SER can be 0 even for the case that all bits are incorrect True error information True error information without having guard bands (dead zones) in between the acquisitions BER and SER for PAM-4 signals only BER and SER for both NRZ and PAM-4 signals BER and SER for both NRZ and PAM-4 signals No detailed error information Only mean, min, max and standard deviation of BER Data rates supported Up to 45 GBaud without equalization and 58 GBaud with equalization (only for PAM-4) Sensitivity Can sense signals as low as 50 mv pp (single ended peak to peak) for PAM-4 (inner eye seen is at least 3 times smaller) Measurement Time It is affected by factors such as equalization, symbol rate and acquisition depth. Only BER per acquisition and cumulative are given Error counting on software equalized signal which gives a lot of adjustability Detailed error information Counted/erroneous 0/1 bits received/ compared/erroneous symbols Data rates supported NRZ: Up to 58 GBaud without equalization PAM-4: Up to 45 GBaud without equalization and 58 GBaud with equalization Sensitivity Can sense signals as low as 50 mv pp (single ended peak to peak) for both NRZ and PAM-4 (inner eye seen is at least 3 times smaller in case of PAM-4) Measurement Time It is affected by factors such as equalization, target BER, symbol rate, type of line coding, update rate and acquisition depth. Supports advanced measurements such as accumulated BER, jitter tolerance Error counting on software-equalized signal which gives a lot of adjustability Ability to view equalized loopback eye Ability to view equalized loopback eye Histogram only Detailed error information Counted/erroneous 0/1 bits received/ compared/erroneous symbols Data rates supported: NRZ: Up to GBaud PAM-4 Up to 30 GBaud Sensitivity: Requires 70 mv pp inner eye opening for both NRZ and PAM-4 (total single ended peak to peak amplitude for PAM-4 would be min. 210 mv pp) Measurement Time Although measurement time is not independent of symbol rate, measurement is much faster (takes a few milliseconds to give out the measurement) than that of the real-time scope. Supports advanced measurements such as accumulated BER, jitter tolerance Software equalization is included but it s not as flexible as the real-time scope Finally, we can conclude that the real-time oscilloscope when controlled from M8070A software offers advantages in terms of achievable data rates and equalization flexibility but takes more time to complete the measurement. Typical BERs that can be verified using this approach are of the order 1E-6. For lower BER (less than 1E-6) thresholds, it takes much longer to complete the measurement. Reference (*) Wolaver, Dan H. (1991). Phase-Locked Loop Circuit Design. Prentice Hall. p ISBN Literature Infiniium Z-Series Oscilloscopes Data Sheet J-BERT M8020A High-Performance BERT Data Sheet M8040A High-Performance BERT 64 Gbaud Data Sheet EN EN EN

23 23 Keysight BER Measurement Using Real-Time Oscilloscope Controlled from M8070A - Application Note Evolving Since 1939 Our unique combination of hardware, software, services, and people can help you reach your next breakthrough. We are unlocking the future of technology. From Hewlett-Packard to Agilent to Keysight. For more information on Keysight Technologies products, applications or services, please contact your local Keysight office. The complete list is available at: Americas Canada (877) Brazil Mexico United States (800) mykeysight A personalized view into the information most relevant to you. Register your products to get up-to-date product information and find warranty information. Keysight Services Keysight Services can help from acquisition to renewal across your instrument s lifecycle. Our comprehensive service offerings onestop calibration, repair, asset management, technology refresh, consulting, training and more helps you improve product quality and lower costs. Keysight Assurance Plans Up to ten years of protection and no budgetary surprises to ensure your instruments are operating to specification, so you can rely on accurate measurements. Keysight Channel Partners Get the best of both worlds: Keysight s measurement expertise and product breadth, combined with channel partner convenience. Asia Pacific Australia China Hong Kong India Japan 0120 (421) 345 Korea Malaysia Singapore Taiwan Other AP Countries (65) Europe & Middle East Austria Belgium Finland France Germany Ireland Israel Italy Luxembourg Netherlands Russia Spain Sweden Switzerland Opt. 1 (DE) Opt. 2 (FR) Opt. 3 (IT) United Kingdom For other unlisted countries: (BP ) internet-infrastructure/ DEKRA Certified ISO9001 Quality Management System Keysight Technologies, Inc. DEKRA Certified ISO 9001:2015 Quality Management System This information is subject to change without notice. Keysight Technologies, 2017, 2018 Published in USA, March 22, EN