by Brian Baxter
If you attended IMS this year you saw an important product announcement, in a completely new category, from Marki: a switch. A switch doesn’t seem like a big deal, but this switch is. It is a good product on its own, but it represents a step function increase in Marki’s core capabilities that took more than a million dollars and 18 months to develop. Before we can explain why this product is so important, we need to introduce you to the product itself and then explain why it is special.
The Marki MSW2-1001ELGA is a broadband single pole, double throw (SPDT) switch built using a silicon-on-insulator (RF-SOI) technology. The part is packaged in a compact 2.25 x 2.25 mm 12-pin QFN package. The MSW2-1001ELGA exhibits excellent RF performance by simultaneously providing wide bandwidth, low insertion loss, high isolation, high IIP3 and high power handling capability.
Figure 1: MSW2-1001ELGA functional block diagram
The MSW2-1001ELGA’s RF performance would be impossible without the excellent RON*COFF figure of merit (FOM) exhibited by the underlying RF-SOI semiconductor technology. This is apparent when comparing the performance of the MSW2-1001ELGA to GaAs pHEMT switches offered from competitors, as shown in Table 1.
Table 1: MSW2-1001ELGA vs. GaAs pHEMT Competitors
Note that the MSW2-1001ELGA is superior or equivalent to competing parts in every metric. In the few cases where a GaAs competitor has similar or higher bandwidth, notice that a tradeoff has been made in one of the other metrics that allows the wider bandwidth. For example, competitor A has extended its bandwidth up to 50GHz. This bandwidth extension comes at the expense of higher insertion loss (2.0dB) and lower isolation (-32dB). Compare this result to competitor B that has improved insertion loss and isolation (vs. competitor A) but its bandwidth has been significantly reduced. Finally, competitor C shows improved insertion loss and bandwidth (versus competitor B) but has degraded isolation. None of the competitor parts exhibit the balance of performance found in the MSW2-1001ELGA.
Understanding the RON*COFF Figure of Merit
A transistor used as a switch is very different from a transistor used as an amplifier or mixer. In a switch application, the transistor is either turned ‘ON’ and appears ‘CLOSED’ or it is turned ‘OFF’ and appears ‘OPEN’.
Figure 2: RON and COFF switch equivalences
The RON*COFF FOM is used to compare switch technologies for use in RF applications and is expressed in femtoseconds (fs). RON is the resistance of the switch path while in the ON or CLOSED state. COFF is the capacitance seen when the switch is in the OFF or OPEN state. RON directly impacts insertion loss and COFF is related to bandwidth and isolation. Thus, for a given switch throw count, such as SPDT, the designer’s trade space is bandwidth, insertion loss and isolation.
Circuit design techniques such as the addition of shunt switch branches will add COFF. This simultaneously improves isolation while degrading insertion loss and reducing bandwidth. Alternatively, the designer might choose to increase device size, which will lower RON and increase COFF, resulting in redcued insertion loss, isolation and bandwidth.
Figure 3: Shunt branch switches improve isolation but add capacitance to the ON branch
The results of this trade space are apparent when reviewing a comparison of the Marki MSW2-1001ELGA versus competing GaAs pHEMT SPDT switches. RF-SOI has a FOM of ~80fs while GaAs pHEMT has a RON*COFF FOM value of ~200fs. As such, the Marki RF-SOI switch is expected to have an advantage in bandwidth, insertion loss and isolation.
Until now, we have discussed the inherent advantage of RF-SOI versus GaAs pHEMT in terms of its RON*COFF FOM. There is a second and equally important performance advantage in RF-SOI devices: linearity. In basic terms, the linearity of a device is a measure of its ability to accurately pass signals from input to output without adding distortion. Distortion comes in many forms but in switches it is most often benchmarked by the Input Third-Order Intercept Point, or IIP3. RF-SOI has an advantage in linearity over GaAs due to the improved linearity of the silicon transistor capacitances versus GaAs. The result of this can be seen in Table 1 by comparing the IIP3 values for the Marki MSW2-1001ELGA to GaAs pHEMT competitors. On average, the MSW2-1001ELGA has an IIP3 of 10dB higher than its GaAs pHEMT competitors. This is in addition to all the previously stated performance advantages.
Alternative Switch Technologies
To this point, we have only considered transistor switch technologies. In this limited space, SOI is uniformly superior to competing technologies. There are non-transistor switches as well, which operate very differently and each have their place.
Micro-electromechanical (MEMs) Switches
A transistor switch is a purely electrical device in the sense that nothing physical is moving. A MEMs switch consists of an actual physical contact that is made and broken with each switching cycle, just as you would expect from the switch diagram. This has some significant benefits including:
- DC and RF are switched together
- Linearity and power handling are very high
- Insertion loss can be low
There are also fundamental disadvantages:
- Isolation may not be as good
- Slower than purely electronic switches
- Requires high switching voltage
- Lifetime limited by number of switching cycles
- Lower operating frequencies
From these points, it should be clear that there are applications where MEMs are more appropriate and applications where a transistor switch is appropriate.
PIN Diode Switches
Another type of purely electrical, solid-state switch is the PIN diode switch. There are two major issues with these switches:
- While transistors are natively three port devices, PIN diodes are only two port devices. This means some kind of diplexing circuit is required to separate the control circuitry and the signal, typically a bias tee with an inductor and blocking cap.
- They require an active current to turn on and off, so the current must be controlled and dissipate power.
PIN diodes have an advantage in that they can handle significantly higher power levels than SOI, however. In some applications that don’t require broadband, low frequency operation PIN diodes may be preferable. This is especially true at very high frequencies or in very niche, low volume applications.
Why not SOI?
Given these benefits, it would seem all switches on the market would be built using SOI instead of GaAs pHEMT. Nonetheless, several companies have released GaAs pHEMT switches in the recent past. Marki also considered, and dismissed, the idea of making GaAs pHEMT switches several times. Why?
The reason is that silicon development requires its own, separate development ecosystem. There are very few resources that can be used for both GaAs and silicon development. Production of SOI devices requires separate suppliers for chips, packaging materials, packaging services and design software. While the end product can be tested and used exactly the same as GaAs products, the design and production is completely different. It requires development of a completely new team.
Prior to Marki, only high volume communications companies were able to support an SOI capability. Marki made the jump, we have the capability, so what now?
The ability to make SOI switches means that you should expect more, superior and differentiated switch products for our core customers. Different throw counts, power handling, linearity and other spec tradeoffs will be available to our customer base. More variety will be available for smaller volume customers.
The ability to design in silicon means the ability to add intelligence to any product. We have a lot of ideas on how to use this intelligence, but we are just getting started. When combined with Marki’s legendary packaging prowess, the possibilities are endless.