Supercharging PICs with advanced design
Design tools that combine flexibility and precision can speed up design cycles and reduce the number of iterations needed, accelerating PIC technologies to market.
By Martin Fiers and Chiara Alessandri from Luceda Photonics
When a photonic circuit designer sends a drawing to a foundry, they are making a significant commitment. Depending on various factors, such as the volume and whether the wafer run is shared or dedicated, they will typically have to pay somewhere between €10,000 - €1 million and wait 4-9 months to get the manufactured circuit back. If they find there is an error in the prototype, then it will be an extremely expensive one; they will have to correct their design and go through another iteration, incurring the same cost in time and money again.
And beyond the expense to the company, there is a broader cost to the PIC industry and society as a whole. From photonic biosensors for diagnostics to LiDAR for autonomous vehicles, we know that PICs could be instrumental in achieving multiple exciting technologies, with the potential to transform healthcare, transport, computing and more. Anything we can do to reduce the time to market brings us closer to realising these benefits in people’s daily lives.
But these circuits are highly complicated devices, requiring extremely precise design and fabrication – even a seemingly tiny error in a circuit’s physical layout can have a decisive impact on how the chip functions. To illustrate this point, Figure 1 shows an example of two real designs for a PIC wavelength demultiplexer. The designer initially sent the one on the left to the foundry, but when he received the chip back, he discovered its performance was significantly poorer than he expected.
Figure 1. The two PICs might appear identical, but the one waveguide in
the one on the left is about 200 nm too long. This significantly impairs
its performance as a wavelength demultiplexer.
It turned out that one of the waveguides in the drawing is just fractionally – about 200 nm – too long, and this is enough to lead to an extra π phase shift that significantly impairs the circuit’s ability to separate the frequencies. The design on the right is correct and performs much better. Yet, to the human eye, they appear identical. This is just one example, but there are many different types of errors that are hard to spot, even after a careful design process.
So, rather than asking designers to pore even more painstakingly over the drawings, we could instead create tools that are specifically designed to make it much easier to generate the correct layouts, and which flag errors that have cropped up as automatically as possible. This is where the Luceda Photonics Design Platform comes in. With its flagship product IPKISS, the platform supports designers throughout the whole design flow, from the initial ideation phase, through component and circuit design, to functional validation and tape-out preparation.
Figure 2. IPKISS has a library of common PIC components. Users can also write code for custom components, which then become reusable IP blocks.
Flexible component design
We often draw comparisons between electronic circuits and photonic circuits, but, when it comes to design, there are some important differences. In electronics, designers can usually work with a set of standardised building blocks: transistors, resistors, and capacitors. In photonics, on the other hand, there is a much larger number of possible components – and more variations of each component – that designers can choose from.
IPKISS has a library of many of the most commonly used elements for designers to quickly add to their circuits, some of which are shown in Figure 2.
However, many PIC designers need to create custom components specifically tailored to their technology and use case. IPKISS therefore uses a flexible code-based approach, based on the Python programming language, which is more expressive than a graphical user interface. With IPKISS, designers create PIC devices and circuits by writing IPKISS code based on Python, from which the software can generate the layout. In addition, their custom building blocks become reusable IP blocks, that can be employed throughout the lifecycle of the product, or even be applied to different products and projects.
A major advantage of using Python in particular is that it has become widely adopted as the standard programming language in many science and engineering disciplines, and is being taught as part of many technical degrees. PIC designers are therefore likely to already have the coding knowledge needed, making it maximally accessible and convenient. It also comes with many libraries for visualisation, analysis, and integration.
Additionally, Python is a superb language to automate the design flow, offering the possibility of fine-tuning and adjusting the code to each organisation’s unique requirements, as well as giving the user control over the tiny details that can make or break a photonic device or circuit.