TDECQ 5G networks using 100G standard data centers are still upgrading to 200G / 400G. The reason for the upgrade is related to the continuous growth of traffic. Future data centers and network infrastructure will support complex high-traffic applications such as ultra-high-definition video streaming, new Features like video game and virtual reality content, artificial intelligence, autonomous driving and mobile 5G networks.

1. "Past Life" of TDECQ - TDP and TDEC

2. Birth and testing of TDECQ

Let's first take a look at the official title and explanation:

TDECQ

The full name is Transmitter and Dispersion Eye Closure for PAM4 (transmitter dispersion eye closure cost), which is a very important parameter to measure the quality of PAM4 optical signals.

There is an eye diagram in the name, which seems very familiar and familiar, but TDECQ really has nothing to do with the eye diagram.

Since the IEEE802.3bs/cd specification for the PAM4 optical interface cancels the mask margin, which is the eye diagram test we are familiar with, instead TDECQ has become a parameter that must be tested in both production and R&D.

So what exactly is a TDECQ?

This also starts from its "previous life", but before the beginning, you still need to have a preliminary concept. Since the TDP, TDEC and TDECQ mentioned in the text are each other's past and present lives, they are actually one thing. The physical meaning represented is the same, that is, the difference between the sensitivity in two states: the difference between the sensitivity of the ideal reference transmitter and the sensitivity of the transmitter under test and the optical link.

The "Past Life" of TDECQ - TDP and TDEC

To better understand the physical meaning of TDECQ, let's first review TDP and TDEC.

1. TDP

(Transmitter and Dispersion Penalty)

The official title and explanation of TDP is:

TDP (Transmitter and Dispersion Penalty transmitter dispersion cost) first appeared in the transmitter specification specification for 10GBase-SR/LR/ER in IEEE802.3ae released in 2002, and the subsequent IEEE802.3 chapter specifications include 100GBase-LR4 /ER4, etc. have extended this parameter.

So why does the parameter TDP always exist in the standard specification, but people rarely test it or even hear about it?

If you simply answer this question, you can understand that although TDP is a parameter that can most intuitively reflect the performance of an optical transmitter in terms of physical meaning, it is very difficult to test TDP from a testing point of view.

For an example of mountain climbing, you may be more clear:

On our way forward, we will constantly encounter mountains that block our way. If we want to continue to move forward, we can only choose to either walk around or boldly turn over. If you can take a detour, of course, it is best to take a detour. Although you can see all the mountains and small mountains when you stand on the top of the mountain, we all understand this, but it is also very difficult. This is just like the TDP test. In the NRZ era, although everyone understands This test parameter can better reflect the real optical transmitter performance, but it is very difficult to realize this test.

But how difficult is it to build a test? You can learn about it from the "TDP Building Block Diagram" required by the specification:

Figure 1 below is a block diagram of "TDP" in the IEEE802.3 specification:

Figure 1 Test block diagram for TDP in IEEE802.3 specification

The "TDP test process" is as follows:

A For an ideal reference transmitter (Reference Transmitter in Figure 1), use Optical Attenuator, Reference Receiver, Bit Error Tester (BERT), etc. to measure the bit error rate under 1e-12 Sensitivity test, record the sensitivity at this time, that is, the minimum power, P0;

B Replace the previous reference transmitter with the optical transmitter under test (DUT), and add optical fibers, polarization controllers, optical splitters, optical reflectometers, etc. and return loss meet the following table requirements:

Table 1

C On the basis of this optical link, carry out the test of the sensitivity of the bit error rate under 1e-12 again with the same reference receiver and BER as in step A, and record the sensitivity PDUT at this time;

D Finally calculate TDP = PDUT - P0.

As can be seen from the above process, TDP is the difference between the sensitivity in two cases:

? One is the sensitivity of the ideal reference transmitter,

? The other is the sensitivity in the case of transmitter under test + optical link.

So in terms of physical meaning, TDP is very clear and intuitive.

But on the other hand, from the above process, it can be seen that there are many "pits" in the test, that is, the reason why this indicator is so good but everyone does not use it is:

How to find the ideal reference transmitter?

The IEEE802.3 specification does specify the basic parameters of an ideal reference transmitter, but in reality it is difficult to ensure that all requirements are met at the same time, even with an instrument-grade reference transmitter.

How to build an optical link that meets the requirements?

The optical link looks complicated, and it is not an easy task to fully satisfy the dispersion, loss, and return loss.

If you need to test the sensitivity under 1e-12, you can only use the bit error tester, but this will increase the cost of the instrument, and the test time is not short.

For example, to achieve a 95% confidence level, it will take more than ten minutes to test the 1e-12 level for a 10Gbps bit error rate test.

In a word, TDP looks very beautiful, but it is a big head when tested, so there are very few people who have the conditions and the courage to climb the big mountain of TDP. In actual tests, everyone prefers to use another parameter - "mask margin" as one of the performance indicators for evaluating optical transmitters. Although the template margin is simple to test, the standard protocol does not specify the specific margin requirements, which can only be negotiated by the industry.

But it is precisely because of the perfect and intuitive physical meaning of TDP that people still haven't given up the pursuit of it, and are still looking for opportunities to make it practical. The so-called infinite scenery is on the dangerous peak, why do you want to climb to the top? Because only standing at the top can see a wider world.

2. TDEC

(Transmitter and Dispersion Eye Closure)

With the emergence of 100G-SR4, this opportunity appeared in the standard specification of IEEE802.3bm released in 2015, but this time, people changed the name to distinguish it.

TDEC (Transmitter and Dispersion Eye Closure).

At this time, some changes have taken place in the requirements of the standard, which has brought new opportunities, specifically the requirements on the bit error rate have changed, and it seems that the spring of TDEC and TDP will be ushered in.

In 100G-SR4, the optical link no longer requires error-free transmission, but can tolerate a bit error rate of 5e-5. Of course, FEC (forward error correction code) technology is used to ensure the optical transmission performance of the entire system. . The problem of FEC is not considered here, so it can be considered that the 5e-5 bit error rate is the requirement of the IEEE802.3 standard for 100G-SR4.

Why is it said that after the optical link bit error rate has changed from 1e-12 to 5e-5, it has ushered in the spring of TDP (here should be called TDEC)?

That is because to test the bit error rate of 1e-12 level, at least the number of bits of the order of 1e-13 must be compared. The oscilloscope cannot collect so much data in a short time, so it is powerless and can only rely on the bit error tester; and for 5e-5 The oscilloscope only needs to collect the number of points of the order of 1e6 (1M level), which is easy to do in a short time, so there is no need for a bit error tester, and a sampling oscilloscope can tested.

Of course, the question that still needs to be solved is how to obtain the ideal reference transmitter?

This time, people have made great use of their ingenuity, created a series of algorithms, and used mathematical methods to build an ideal reference transmitter model. All testing processes rely on algorithms, which greatly simplifies the construction of the testing environment. .

The "Test Block Diagram of TDEC" is shown in Figure 2 below:

Figure 2 Test block diagram of TDEC

? Compared with the previous TDP block diagram, it has been simplified a lot. The Optical spliiter+Variable reflector in the figure is still in order to meet certain return loss requirements;

? Multimode fiber is not shown in the figure, so the effect of chromatic dispersion is not considered? actually not. The influence of intermodal dispersion of multimode fiber has been converted into the cost of bandwidth,

? It should be noted here that TDEC is only for the SR4 standard, so for the 100G-SR4 TDEC test, the general test filter bandwidth of 25G signals is not the traditional 19.3GHz, but becomes 12.6GHz.

The following is a detailed explanation of "how is TDEC obtained"?

A For the optical signal entering the oscilloscope, draw a histogram in the blue box area (0.4UI and 0.6UI, each box width is 0.04UI) in Figure 3, which is also the area specified in the standard for testing TDEC. It is assumed that the calculated bit error rate BER in this area is not higher than the target bit error rate of 5e-5.

image 3

B Use the algorithm to add noise to the signal in this area to introduce bit errors, and keep adding noise until the bit error rate reaches 5e-5, such as the red mark in Figure 4 below.

Figure 4

C The oscilloscope constructs an ideal signal model based on the amplitude and period of the previous input signal, and uses an algorithm to add noise to the same area of the ideal signal until the bit error rate reaches 5e-5, as shown in green in Figure 5 below.

Figure 5

D Finally, take the ratio of the added noise represented by green to the added noise represented by red, and take the unit of dB to obtain TDEC, as shown in Figure 6.

Image 6

As can be seen from the above process, the TDEC test is very simple. The instrument only needs a sampling oscilloscope (of course, CDR is also required in accordance with the standard specification), and the test speed is very fast, which can be completed in about 2s.

So far, TDEC is a parameter that has both perfect and intuitive physical meaning and simple test method. However, due to the existence of mask margin in 100G-SR4 test, people are still used to mask test, which makes TDEC's existence still relatively Low. However, this laid a solid foundation for the birth of the later TDECQ.

Birth and testing of TDECQ

The transition from NRZ signal to PAM4 signal is by no means a simple quantitative change, but a qualitative change.

Because of the characteristics of the PAM4 signal format, it has brought about great changes in test parameters and test methods. I will not repeat the characteristics and test challenges of PAM4 here, but focus on one problem: the mask margin test commonly used in the previous NRZ era, in the PAM4 era. Does it still apply? If it is not applicable, how can it be replaced?

This is not such an easy question, and even the IEEE802.3 Association took several years to gradually find the answer and is still in the process of perfecting it. We don't need to repeat the exploratory phase of the past, we just need to keep up with the changes of the times and understand the latest progress of it.

First of all, among all the current open standards and specifications, a consensus is that the template margin is no longer suitable for PAM4 testing. At this time, the mountain in front of us has no way around it, we can only turn over it, and we need to use new parameters to test Characterize transmitter performance.

This new parameter is TDECQ (Transmitter and Dispersion Eye Closure for PAM4).

1. The Great Birth

Why TDECQ?

That is because physically, the most direct parameter to measure the performance of an optical transmitter is the aforementioned TDP, whether it is an NRZ transmitter or a PAM4 transmitter. It seems that TDP is still the most favored parameter by standards associations. In order to show the difference, for PAM4, it is renamed TDECQ, and Q is the first letter of four-level Quaternary. Therefore, the physical meaning of TDECQ is exactly the same as that of TDP, so it will not be repeated.

What about the calculation method of TDECQ?

Obviously, it is similar to the previous TDEC. Figure 7 below is the IEEE802.3 standard organization's test block diagram for "TDECQ":

Figure 7 IEEE802.3 standard organization's test block diagram for TDECQ

Does it look familiar? Almost the same as TDEC's. only:

? For single-mode signals, it is still necessary to add fibers to the link to meet the requirements of loss and dispersion;

? For multimode, fiber optics can be omitted.

2. TDECQ test

First review the two test premise of TDECQ: test pattern and test receiver.

The test pattern requirements of IEEE802.3bs for PAM4 signals are shown in Table 2:

Table 2 IEEE802.3bs test pattern requirements for PAM4 signals

in:

PRBS13Q

It is a four-level pattern with a length of 8191 obtained after Gray coding (0-00, 1-01, 2-11, 3-10) of two PRBS13 patterns, which can be used for ER/OMA of the transmitter. test;

PRBS31Q

It is also a four-level pattern with a length of 231-1 obtained after Gray coding (0-00, 1-01, 2-11, 3-10) of two PRBS31 patterns. Note that this pattern is only used for To test the sensitivity of the PAM4 system;

SSPRQ (Short Stress Pattern Random Quaternary)

It is a completely artificially constructed new code pattern, which is composed of 4 segments selected from the traditional PRBS31 code pattern and spliced and encoded for the code pattern with relatively large transmitter pressure. The length is 216-1. The transmitter exerts enough pressure to get closer to testing its performance in real business, and has the characteristics of short patterns, so that the sampling oscilloscope can capture the entire pattern for signal processing such as equalization. SSPRQ is the pattern for TDECQ testing.

For the test of PAM4 optical signal (including parameters such as TDECQ), the IEEE802.3bs specification requires that the reference receiver of the test instrument must meet two requirements:

a. Ideal fourth-order Bessel-Thomson low-pass filter frequency response;

b. The 3dB bandwidth of the low-pass filter frequency response is half of the symbol rate of the signal under test (13.28GHz for 26.56GBaud, 26.56GHz for 53.125GBaud).

Note: For multimode 26.56GBaud PAM4 optical signal testing, the bandwidth becomes 11.2GHz.

After the test pattern and test receiver are determined, we continue to look at the "TDECQ test".

In the TDECQ test, the requirements for CDR are loop bandwidth 4MHz, slope 20dB/dec, 1st order, no peaking.

Figure 8

Unlike NRZ, due to the complexity of the PAM4 signal itself, an equalizer needs to be used at the signal receiving end to open the eye diagram. Therefore, the TDECQ test instrument requires an equalizer (reference equalizer), and the standard has stipulated that the equalizer is 5 tap/T spaced FFE equalizer, but the specific coefficient of the equalizer is determined by the software algorithm according to the input signal.

In the strict standard specification test, it is necessary to build an optical link according to the test block diagram of TDECQ, but it is obviously impractical to build such a complex optical link in the actual PAM4 production test, so in many cases people omit it. The TDECQ of the optical transmitter output signal is directly tested by the optical fiber link.

The "TDECQ test process" is the same as the TDEC described earlier:

A For the optical signal entering the oscilloscope, draw a histogram in the blue box area of Figure 9 (0.45UI and 0.55UI, each box width is 0.04UI, for convenience, only half is displayed), this area is also specified in the standard area of the test TDECQ. It is assumed that the calculated symbol error rate SER in this region is not higher than the target symbol error rate 4.8e-4 required by the specification.

Figure 9

B Use the algorithm to add noise to the signal in this area to introduce bit errors, and keep adding noise until the bit error rate reaches 4.8e-4, such as the red mark in Figure 10 below.

Figure 10

C The oscilloscope constructs an ideal signal model (virtual ideal transmitter) based on the amplitude and period of the previous input signal, and adds noise to the same area of the ideal signal with an algorithm until the bit error rate reaches 4.8e-4, as shown in Figure 11, dark blue Color dotted line identification.

Figure 11

D Finally, take the ratio of the added noise represented by dark blue to the added noise represented by red, and take the unit of dB to get TDECQ, as shown in the following figure.

Figure 12

Note: If the symbol error rate SER of the signal under test (after equalization) is higher than 4.8e-4, the calculation of TDECQ cannot be continued. At this time, it is necessary to find a way to improve the quality of the signal under test.

It can be seen from this process that the calculation process of TDECQ is straightforward and easy to understand, and it is basically realized by algorithms. Compared with TDP test items, it greatly reduces the investment of instruments and the cost of test time.

However, it is precisely because the TDECQ test is mainly implemented by algorithms, and the standard organization is constantly optimizing the algorithm and optimizing the test results of TDECQ, so that the test is as consistent as possible with the real PAM4 system environment. Therefore, from IEEE802.3bs to IEEE802.3cd, optimizations have been made at the algorithm setting level.

◆ For example, the area of the blue box above is not limited to 0.45UI and 0.55UI, as long as the interval between the two boxes is 0.1UI, and the width of each is 0.04UI (increased by IEEE802.3bs);

◆ For another example, the judgment threshold level can allow a fluctuation range of 1% OMA (increased by IEEE802.3cd) based on the ideal position. The work of standards bodies to optimize TDECQ has not stopped and may be supplemented in other specifications in the future. Therefore, keeping up with the development trend of the standards of the times is a necessary condition for standard testing.

The last "TDECQ test scheme block diagram", as shown in Figure 13:

Figure 13

Keysight can provide the industry's most standard-compliant TDECQ test solutions (N1092 series + N1077/8A series and N1092A/B built-in CDR series)

The blue one is the PAM4 signal before the equalizer

Yellow is the PAM4 signal after the equalizer