Access Official Technical PDF

Share Summary:LinkedInEmail

In high-speed optical transmitters, a digital-to-analog converter (DAC) helps determine the bandwidth of the electrical signal that drives the optical source. The digital-preprocessed analog-multiplexed DAC (DP-AM-DAC) addresses this constraint by combining digital signal processing, two lower-bandwidth sub-DACs and a high-speed analog multiplexer. The approach can produce an output signal with nearly twice the bandwidth of either sub-DAC, offering a way to extend DAC bandwidth without relying on a single wider-bandwidth CMOS converter.

This architecture is relevant to engineers and technical decision-makers working with optical communications infrastructure, where signal generation, modulation and transmission performance depend on the capabilities of the complete transmitter chain.

Why DAC bandwidth matters in optical transmitters

Digital signal processors (DSPs) support key functions in modern optical transmission, including high-order modulation, pulse shaping, equalization and dispersion compensation. These functions help generate and process signals for high data rates and efficient use of available spectral bandwidth. But the signal still needs to pass through a DAC before it can drive the transmitter’s analog components.

In the CMOS-based transmitter design described in the whitepaper, the DAC is integrated with the DSP. Its analog bandwidth is a limiting factor: the cited commercial CMOS DACs have bandwidths of about 30 GHz. DACs built on compound semiconductor platforms such as indium phosphide (InP) or silicon-germanium can offer greater bandwidth, but may consume more power and pose integration challenges when the DSP remains based on CMOS.

The DP-AM-DAC is designed to work around this trade-off. Instead of requiring one DAC to reproduce the entire wideband signal, it distributes signal information across two sub-DAC paths and recombines their analog outputs.

How the DP-AM-DAC extends signal bandwidth

The architecture has three main elements: a digital preprocessor, two sub-DACs and an analog multiplexer (AMUX). The preprocessor transforms the desired digital signal into two coordinated sub-signals that each fit within the bandwidth of a sub-DAC. The AMUX then selects between the analog outputs in sequence at a clock rate, combining them into a wider-bandwidth output.

Preprocessing and frequency reconstruction

In the type-I approach, the preprocessor separates the target signal into low- and high-frequency components. It flips the high-frequency component around the midpoint of the sub-DAC bandwidth, then combines the components into two sub-signals with specific amplitude relationships and complementary phases. This processing prepares each signal for conversion by a sub-DAC with roughly half the bandwidth required by the final output.

The AMUX alternates between the two analog inputs. In frequency terms, this switching produces the original sub-signals along with shifted copies, or images. The phases of the images from the two paths are arranged so that the desired components combine to reconstruct the target signal across a wider frequency range. A low-pass filter can remove residual image energy outside the target band.

With sub-DAC bandwidth represented as fB/2, the type-I design can produce an output bandwidth of approximately fB. The reported implementation used a symmetric arrangement of the two sub-DAC paths. This can make it easier to balance the branches than an approach based on analog mixers, although the two paths still need to operate together as intended.

Type-I and type-II options

A second version, type II, uses the same hardware configuration but a different preprocessing algorithm. The reported benefits are a clock-frequency requirement reduced by half and suppression of the residual image. These are architectural differences rather than changes to the basic building blocks: both approaches rely on digital preprocessing, two sub-DACs and analog multiplexing.

AMUX performance and transmission demonstrations

The AMUX is central to the system because it must switch between its two inputs at high speed while passing the selected signal. In the reported module, designed using 0.5-micrometer-emitter InP heterojunction bipolar transistor technology, measurements showed a through loss below 3 dB and isolation above 20 dB over the measured range up to 50 GHz. These measurements characterize the AMUX module; they are not a complete-system performance specification.

Transmission experiments demonstrated how the architecture could be used with optical links. In one intensity-modulated direct-detection (IMDD) experiment, two CMOS-based arbitrary waveform generator channels served as sub-DACs, each with an analog 3-dB bandwidth of about 20 GHz. The type-I DP-AM-DAC operated with a 43.3-GHz multiplexing clock to generate an 80-Gbaud PAM4 signal. After transmission over 20 km of standard single-mode fiber, the measured bit error rate was 6.2 × 10-3. The authors reported a corresponding net data rate of 142.9 Gbit/s when assuming 12% overhead hard-decision forward error correction.

A separate type-II demonstration used sub-DACs with an analog 3-dB bandwidth of about 32 GHz and a 37.5-GHz clock. Discrete multitone modulation was used in the experiment. Following 10 km of standard single-mode fiber, the total bit rate was 300.12 Gbit/s, with a reported net data rate of 250 Gbit/s under the stated soft-decision forward error correction threshold. The measurement equipment limited the observed bandwidth to 62 GHz, even though the type-II principle can support signal bandwidth up to twice its clock frequency.

Practical considerations for system teams

For engineers evaluating this method, the key point is that bandwidth extension comes from coordinated operation across digital and analog subsystems. The preprocessor must create suitable sub-signals; the sub-DACs must reproduce them; and the AMUX must switch and combine them with the required timing and signal quality. The output spectrum can also contain residual images, making filtering and receiver-side processing relevant to the overall link design.

Assessment should therefore consider the complete transmitter and receiver chain rather than the nominal bandwidth of an individual DAC. Relevant design questions include:

  • What bandwidth and modulation format does the target optical link require?
  • Can the available sub-DACs and AMUX support the needed signal and clock rates?
  • How will filtering, equalization and error correction fit into the system?
  • Which preprocessing approach—type I or type II—best suits the clocking and image-suppression requirements?

Conclusion

The DP-AM-DAC demonstrates a practical architectural route to DAC bandwidth extension: digitally divide signal information between two sub-DACs, then use an analog multiplexer to reconstruct a wider-bandwidth output. Transmission experiments reported high-speed IMDD results, including a 250-Gbit/s net data rate using the type-II approach, as well as applications in long-haul digital coherent transmission. For optical system designers, its significance lies in using coordinated digital and analog design to address a bandwidth bottleneck, while accounting for clocking, filtering and the performance of the full link.

Access Official Technical PDF

About The Author