DSP vs. All-Analog: A Comparison of 200G Optical Interconnect Architectures - NADDOD Blog

DSP vs. All-Analog: A Comparison of 200G Optical Interconnect Architectures

NADDOD Nathan Optics Application Engineer Jun 29, 2023

The demand for faster and more flexible optical interconnects is increasing in cloud data centers, as well as other industries. While 100G and 400G optical interconnect solutions are available, data center operators are also looking for intermediate options that can provide improved capabilities while being flexible and cost-effective. One such option is the 200G optical interconnect solution, which has become increasingly popular due to its impressive capabilities.

There are currently two mainstream solutions for 200G optical interconnects: the 200G QSFP56 (4x50G PAM4) DSP solution and the 200G QSFP-DD (8x25G NRZ) all-analog architecture solution. What are the differences between them?

200G QSFP56 DSP Solution

Pros

Higher spectral efficiency

To save fiber usage and meet the single-wave rate requirement, achieving a 4x50G optical interconnect is one of the 200G optical interconnect solutions due to the bandwidth limitation of electro-optical devices. However, the channel loss and reflection introduced by the 56G signal are too high, and the tolerance for channel crosstalk is greatly reduced. Currently, it is difficult for NRZ technology to exceed the single-lane 56G transmission rate. Therefore, the industry has introduced PAM4 technology to solve the problem, PAM4 modulation allows for twice the data rate to be transmitted over the same bandwidth as NRZ modulation.

After introducing digital signal processing(DSP), the signal can be directly compressed in the spectral domain on the transmitting side, and the receiving side can recover the signal through an adaptive FIR filter. This method can turn the uncontrollable analog bandwidth impact in the modulation/receiver into a known digital spectral compression, reducing the demand for optical device bandwidth.

Improved signal quality

PAM4 overcomes the weak ability of traditional NRZ modulation at 56G rate and doubles the bit rate without increasing the bandwidth. However, PAM4 sacrifices the signal-to-noise ratio, making the product more sensitive to noise. The introduction of DSP chips exactly compensates for the corresponding disadvantages of PAM4 technology.

DSP is a high-speed digital processing chip that, in addition to providing the digital clock recovery function provided by CDR(Clock and Data Recovery), can perform dispersion compensation, noise removal, and nonlinear interference removal to restore the 200G signal sent from the transmitter. It also supports higher-order modulation formats to improve spectral efficiency and can solve device and channel transmission effects and processing signal-to-noise ratio problems.
Flexible

DSP can adapt to different fiber types and lengths, which makes the 200G QSFP56 DSP solution suitable for a wide range of applications.

Industry standard

QSFP56 complies with IEEE 802.3cd, IEEE 802.3bs, and QSFP56 MSA standards.

Cons

Increased power consumption

Since DSP introduces DAC/ADC and algorithms, its power consumption is higher than that of CDR chips based on traditional analog technology. Currently, the design power consumption of 400G OSFP/QSFP-DD based on 16nm DSP solutions is about 12W, which is a huge challenge for both the transceiver itself and the panel heat design of future switches.

Higher cost

Cost is always a concern for data centers and 5G operators. Unlike traditional optical devices, DSP chips, because they are based on mature semiconductor processes and supported by massive applications, there is expected to be a larger space for cost reduction.

More complex

DSP requires more sophisticated signal processing techniques, which increases the complexity of the 200G QSFP56 DSP solution.

Sum up, the 200G QSFP56 DSP solution provides higher spectral efficiency and improved signal quality, but requires more power and is more expensive and complex than an all-analog architecture.

200G QSFP56 Products

200G QSFP-DD All-Analog Architecture Solution

Pros

Lower latency

Compared with the 200G QSFP56 (4x50G PAM4) DSP solution, the 200G QSFP-DD adopts the 8x25G NRZ modulation method, which increases fiber usage but can flexibly use the all-analog architecture. The delay of an all-analog optical interconnect is only one-thousandth of that of the DSP solution, which is a key advantage in achieving the fastest system and network performance.

Better signal integrity

When the data throughput increases from 100G to 200G, and even higher, signal integrity is a key performance indicator, considering the chain reaction caused by error transmission to the data stream. Without DSP, the reason why 200G can maintain the best signal integrity is largely due to the continuous improvement of clock data recovery (CDR) devices and the underlying signal conditioning technology. The latest generation of analog CDR deployed in the all-analog 200G transceiver proves that it can achieve an extremely low bit error rate (BER) and better performance than Pre-FEC of less than 1E-8, which is comparable to the DSP 200G transceiver overall.

Lower power consumption

Without DSP, the energy consumption of the all-analog 200G optical transceiver is much lower, and the heat generation is significantly reduced. In contrast, the clock input operation of the DSP transceiver may be 2-3W higher. This may not sound like much, but when the power consumption of thousands of optical transceivers in the data center is added up, the final number will be staggering. In this case, saving 2-3W of power consumption per transceiver is extremely beneficial for optimizing operating costs and cooling efficiency.

Lower cost

At the device level, the simplified design of the all-analog 200G optical transceiver reduces the total component count and avoids the development and implementation costs of DSP.

Industry standard

QSFP-DD complies with IEEE 802.3bs and QSFP-DD MSA standards.

Cons

Limited flexibility

The all-analog 200G module is less flexible than a DSP-based solution and may not be suitable for all applications or fiber types.

Limited distance

NRZ modulation is less suitable for long-distance transmission than PAM4 modulation, which can limit the range of the all-analog 200G transceiver.

Limited data rate

NRZ modulation has a lower spectral efficiency than PAM4 modulation, which limits the data rate that can be transmitted over the same bandwidth as PAM4.

Limited compatibility

The all-analog 200G transceiver may not be compatible with all existing 200G optical interconnect solutions, which may limit its adoption in some situations.

Sum up, the 200G QSFP-DD All-Analog Architecture Solution has lower power consumption, lower cost, and lower latency than a DSP-based solution, but is less flexible and has limited distance and data rate capabilities.

200G QSFP-DD Products

Conclusion

Both the 200G QSFP56 and the 200G QSFP-DD solutions have their advantages and challenges. While the 200G QSFP56 (4x50G PAM4) DSP solution provides higher spectral efficiency, it requires more sophisticated signal processing techniques and is more expensive. On the other hand, the 200G QSFP-DD (8x25G NRZ) all-analog architecture solution maintains signal integrity without the need for DSP and has lower power consumption and cost. For data center operators, you should carefully evaluate your needs and choose the solution that best fits your requirements.


Related Resources:
200G Optical Transceiver: QSFP56 vs QSFP-DD
Mainstream for 200G Data Centers, 200G QSFP56 or 200G QSFP-DD?