A pinboard by
Pengyu Zhang

Postdoc Researcher, Stanford University


Design and implement systems that can compute and operate with microwatts of power budget.

The wide deployment of Internet-of-Thing (IoT) devices is limited by two fundamental factors: battery energy density and wireless radio power consumption. The first limiting factor, battery energy density, has only improved by three times over the past twenty-two years. As a result, an IoT device can only continuously operate for several days on a coin cell battery. One example of this is Fitbit where you have to charge the wristband once every several days. The second limiting factor is the wireless radio power consumption. When we turn on a wireless radio for data transmission, the radio itself consumes tens to hundreds of miliwatts power and becomes the power bottleneck of an IoT device.

To address this problem, we propose a fundamental shift on how an IoT device should communicates its data. Instead of generating the wireless signal and transmitting it out, we propose using backscatter communication for connecting IoT devices. In backscatter communication, a device takes an ambient wireless signal as its excitation signal and reflects it back to a receiver. During the signal reflection, the device modifies the backscattered signal to embed its own information. Since signal reflection only consumes microwatts of power, backscatter communication is very efficient. Despite its efficiency, we actually do not see the wide deployment of backscatter-based IoT devices mainly because we do not have backscatter readers, such as RFID readers, around us to communicate with backscatter-based devices. The overhead of deploying such RFID reader-based infrastructure is so high such that we do not see the wide deployment of backscatter-based IoT systems.

To address this challenge, we propose a system that can leverage commodity radios, such as WiFi and Bluetooth, for backscatter communication. In this system, a backscatter radio can talk to commodity WiFi and Bluetooth radios without any hardware modification on these radios. The key observation here is that a signal transmitted by a commodity radio, either WiFi or Bluetooth, is generated using a codeword from a codebook. If a backscatter radio can transform a codeword in the excitation signal into another codeword in the backscattered signal, then we have the opportunity of using commodity radios for receiving the backscattered signal and decoding the IoT device information. Such transformation can work because both the excitation codeword and the backscattered codeword stay within the same codebook.


Monitoring Quantitative Ultrasound Parameter Changes in a Cell Pellet Model of Cell Starvation.

Abstract: Although it has previously been shown that the spectral analysis of ultrasound backscatter data is sensitive to the cellular changes caused by apoptosis, the sensitivity of spectral analysis to oncosis or ischemic cell death had not previously been studied. Whereas many anticancer treatments induce apoptosis, others induce cell starvation, or oncosis. HT-29 colon adenocarcinoma cells were formed into pellets and covered in phosphate-buffered saline at room temperature for 56 h. The pellets were imaged every 8 h with high-frequency (55 MHz) ultrasound and the raw radio-frequency data processed. The lack of nutrients available to the cells induced cell death due to oncosis. The attenuation slope, speed of sound, spectral slope, and midband fit were estimated at each of the eight time points to identify changes as the cells died due to starvation. The spectral slope decreased monotonically over the 56 h, whereas the attenuation slope showed an increase between 1 and 48 h, followed by a slight decrease between 48 and 56 h. The midband fit did not vary over time. The speed of sound increased from 1514 to 1532 m/s over the first 24 h, after which time it plateaued. These in vitro results indicate different trends in ultrasound parameter changes from those of in vitro apoptotic cells, suggesting that these different methods of cell death can be identified not only by morphological markers, but also by specific ultrasound signatures.

Pub.: 22 Jun '17, Pinned: 28 Jun '17