Can custom designed MMICs reduce microwave systems costs?
Every design engineer wants to achieve the best possible results for the lowest cost, and using a standard product IC seems the obvious route. But can you save money by using custom designed MMICs? To many people this would seem to defy logic – surely custom devices are expensive, so how could using them possibly save money?
Why would anyone try to develop their own custom MMICs (monolithic microwave integrated circuits) when virtually all of the building blocks required in any microwave system are already available commercially?
Nevertheless, being able to specify the exact function of a custom MMIC really can help achieve both performance improvements and significant unit cost savings versus normal supplier pricing – as long as the volume requirements are sufficient.
Designing a custom MMIC incurs many of the same expenses as commercial suppliers: wafer costs, evaluation and qualification costs, test and packaging costs and design team salaries. However commercial suppliers will also be making a significant expenditure on marketing, advertising, after sales support and the operational cost of selling and distributing their devices to customers.
All of this needs to be recouped and will be factored into the component pricing, along with a profit margin. With a custom MMIC these costs can be avoided, and the effective unit cost will be significantly reduced.
As an example, let’s take an E-band low noise amplifier (LNA) IC, which is currently available commercially at $110 for volumes of up to 10,000. If a custom design with the same die area were developed, the costs savings would be huge. To estimate the unit cost we need to determine how many die could be fabricated on each wafer.
If we use a process with 150mm diameter wafers and then factor in the area lost to partial die at the edges of the wafer and also the area that is lost to PCM (process control monitor) sites, typically 4 – 6%, then we can expect to achieve around 3,100 untested chips from each wafer.
On top of this there are the ‘failed die’ to consider. Yield is a complex issue, and the number of failed die will depend on factors such as defect density, process control, total gate periphery, design centering and the actual pass/fail criteria specified.
Assuming a typical RF yield for the E-band LNA in production to be above 80%, this gives us around 2,500 working die per 150mm (6”) diameter wafer.
The actual price of the wafers will depend both on the process used and on the wafer diameter. Processes suitable for operation at E-band tend to use e-beam defined gates and are more expensive than processes optimised for use at lower microwave frequencies.
Modest volume pricing for 6-inch diameter wafers fabricated using such a process is likely to be somewhere between $15,000 and $20,000. Even assuming the higher end of this range, the unit die cost (excluding test costs) is around $8.
As our bare die parts will almost certainly require RF-on-wafer (RFOW) testing, this will incur further costs per die whether they pass or fail. A realistic cost estimate for testing would be around $500 per wafer, which adds just $0.20 on to our die costs.
Of course, there are also a number of one-off costs such as design fees, evaluation and qualification costs, prototype wafer run costs and custom mask set costs. These would typically be amortised over the first three years of production, with the effect on die costs directly related to the volume/number of parts used.
If we consider someone using just 5,000 E-band LNAs per year then the potential cost saving of using a custom part is over $1.5million in the first 3 years. This is a clear demonstration of return on investment.
In addition to the potential cost savings, custom MMIC devices bring a number of other clear advantages too, such as the user’s ability to specify the exact functionality required, rather than compromising on the closest standard component available in a catalogue.
The custom MMICs can be produced in preferred geographical locations, which can assist in avoiding export restrictions or in addressing security concerns.
A custom MMIC also means that you control the pad-out and/or package style and that you, rather than your supplier, can control when your component becomes obsolete.
My conclusion is that, surprising as it may sound, a custom MMIC can indeed mean lower costs, as well as greater control for the user.
Liam Devlin, CEO, Plextek RF Integration