ab announced an evaluation kit recently for its NanoEMS MEMS compass device, which can be built entirely in the metal interconnect layers of a CMOS wafer. This new way of incorporating MEMS fabrication into the standard CMOS process promises to revolutionise the inclusion of MEMS in mass produced consumer devices since it will allow low cost and high volume production.
The incumbent approach to creating MEMS does have some drawbacks. Each type of device – be it an accelerometer, compass or medical device – requires a custom process, and is therefore produced in relatively low volumes. MEMS structures are built on top of CMOS wafers, by depositing extra material in various post-processing steps requiring dedicated equipment. The lack of a consistent process on standard equipment means it can be tricky to capitalise on volumes of scale, while expensive and non-standard packaging only makes the situation worse.
Barcelona-based Boalab has invented a process capable of constructing MEMS inside the interconnection layers of a standard CMOS wafer, which it says will enable low cost, high volume production of all types of MEMS devices (figure 1). Josep Montanyà i Silvestre, CTO of Baolab, describes the company’s process as essentially a very simple post-processing etching step in the BEOL (back end of line) CMOS process, in which mask-defined MEMS structures in the metal layer are released by etching the inter-metal dielectric away.
Figure 1. Baolab’s NanoEMS process promises a quicker and cheaper way of constructing MEMS devices by embedding them into the metal interconnect layer on top of a standard CMOS wafer.
“There are 2 steps,” he says. “First the wafers go into the etching machine for 30 minutes. These machines can take around 35 wafers at a time, so the cost per wafer is very low. Second, the release etch holes [in the passivation layer] need to be sealed, which is done with a standard sputtering step.”
Apart from the final sputtering step, the MEMS structure uses existing material within the metal layers normally present in a CMOS wafer. Standard chip packaging techniques can be used. Importantly, all the steps can be done on standard equipment in any CMOS fab, differentiating Baolab’s process from other offerings.
The company’s IP is mainly in the area of how to build MEMS structures out of aluminium, which is not an ideal material, and in the design of the structures so that the inter metal dielectric can easily be removed. Feature sizes are an order of magnitude smaller than conventional MEMS, so Baolab calls the process ‘NanoEMS’ (nanoscale MEMS).
Constructed in this way, MEMS can interact directly with active circuitry and the company sees this as the way that MEMS will be integrated with system-on-chips in the future.
Beyond standard accelerometers and other MEMS sensors, Baolab is researching many alternative types of device that are either completely new, or are being given new potential by the low cost, high volume process that NanoEMS enables. Here’s a summary of some of the most interesting device types that will be enabled by this technology in the future.
One area Baolab is looking into as an application for its NanoEMS technology is vibrating antennas. A vibrating antenna is essentially an antenna that is not static, but rather deforms its shape in a periodical way.
“Using this technology it’s possible to escape from the limitations of classical antenna theory, including size limitations,” Silvestre says.
Vibrating antennas can be used to create antennas with extremely directional properties (‘superdirective’). This is desirable for communications since the antenna can effectively point to the basestation or mobile phone it wants to communicate with, which according to Silvestre, means spatial multiplexing schemes are feasible based on physical location. Spatial multiplexing in turn has the potential to rapidly increase bandwidth of network nodes.
“Consider the ‘internet of things’,” he says. “It’s nice to have nodes everywhere, but where will the channel bandwidth come from?”
Silvestre cited wireless communication between chips on a PCB as another application for the vibrating antenna. Having a MEMS directional antenna on chip could mean bypassing traditional communication lines and open up exciting new possibilities.
Integrating inductors onto silicon has always held problems since the spiral feature shapes required occupy a lot of area. It’s also difficult to couple the different loops of the spiral, which have to be constructed on different layers. Baolab has come up with a solution:
“Metals bend and we use this to our advantage,” Silvestre reveals. “Metal features [when the material underneath is etched] will automatically drop down.”
This intentional bending of metal features can be used to couple the loops in a spiral, creating an inductor. According to Silvestre, this technique holds possibilities for creating on-chip transformers with arbitrary winding ratios, something which is not currently possible. Wide metal tracks mean Ohmic losses are minimised.
Thermo-magnetic RF switches and antennas
An extension of the current technology, this future application of NanoEMS would introduce an additional layer on the CMOS – nickel.
“At low temperatures, nickel is a bad conductor of RF signals,” Silvestre explains. “If you heat it beyond 358°C [the Curie temperature], it becomes a magnetic material and is a good RF conductor. MEMS is a great way to implement this - 358°C sounds like a lot, but a MEMS element would be so tiny that its volume would be very small,” he points out, adding that using in the order of 0.1mW to heat the element would be acceptable; MEMS makes this achievable.
This principle can be used to create devices such as RF switches and antennas, and as a way to configure RF without having to use unreliable moving parts, it certainly holds a lot of potential.
Figure 2. The 3D NanoCompass takes advantage of Baolab’s NanoEMS technology to build a Lorentz-force magnetic sensor inside the interconnect layer.
The device types mentioned above are still in research and development at Baolab and commercial devices are therefore some way into the future. Today though, the company has evaluation kits available for its first product, a NanoEMS compass device called 3D NanoCompass (see figure 2). This compass chip contains integrated MEMS, analogue and digital circuits including self-calibrating magnetic sensors, ADCs and a configurable digital interface (figure 3).
Figure 3. A block diagram of Baolab’s 3D NanoCompass chip reveals MEMS, analogue blocks and digital circuitry integrated monolithically.
The 3-axis Lorentz force magnetic sensor outputs magnetic field strength for the X, Y and Z axes, which can be processed by an external processor to obtain the magnetic heading. The device features a heading accuracy of +/-5°. The sensor’s range is -1.2 to 1.2mT and its resolution is 0.3μT. Power consumption for 10 measurements per second is 2mW and standby current is 1μA. The digital control bus is configurable as either SPI Slave or I2C (standard or fast mode). The device is available in two package styles; a 3mm x 3mm x 0.9mm DFN and a 2mm x 2mm x 0.75mm BGA.
Silvestre concluded by saying that Baolab is strictly a product company and will not license its secrets any time soon. Following the compass device, the company plans to release a gyroscope product, and will continue working in the motion control space as it ‘offers a lot of options’.
Ultimately the technology’s main selling point is its price; the devices in Baolab’s near future can all be implemented with existing MEMS technology, but the ease of building these devices in a standard CMOS process points towards extensive economies of scale, something the awkward MEMS process technologies of today have failed to fully realise.