1) Probe Based Data Storage (This work is in collaboration with Nanodynamic systems lab, University of Minnesota)

Research

Objective: This project aims to develop a probe based data storage system which operates by encoding information as topographic profiles on a polymer medium and can achieve significantly higher aerial densities as compared to conventional data storage systems.

Physical Model: Probe based high density data storage devices employ a cantilever beam that is supported at one end and has a sharp tip at another end as a means to determine the topography of the media on which information is stored. The information on the media is encoded in terms of topographic profiles. A raised topographic profile is considered a high bit and a lowered topographic profile is considered a low bit. There are various means of measuring the cantilever deflection. In the standard atomic force microscope setup, which has formed the basis of probe based data storage, the cantilever deflection is measured by a beam-bounce method where a laser is incident on the back of the cantilever surface and the laser is reflected from the cantilever surface into a split photodiode. The photodiode collects the incident laser energy and provides a measure of the cantilever deflection (See Figure 1).

Figure 1: AFM Setup

Dynamic Mode AFM: In the dynamic mode operation, the cantilever is forced sinusoidally using a dither piezo. The oscillating cantilever gently taps the medium and thus the lateral forces are reduced which decreases the media wear. Using cantilever probes that have high quality factors leads to high resolution, since the effect of a topographic change on the medium on the oscillating cantilever lasts much longer (approximately Q cantilever oscillation cycles, where each cycle is 1/f0 seconds long and Q and f0 is the quality factor and the resonant frequency of the cantilever respectively). (See Figure 2)

Figure 2: Dynamic Mode AFM

Cantilever-observer Architecture: As mentioned earlier, the high Q cantiever gives high signal to noise ratio and less tip-media wear but it is limited in bandwidth. It can be argued that the quality factor of the cantilever can be reduced to fasten the decay of the transient resulting in higher bandwidth. However, low Q operation results in lower resolution and higher forces. This tradeoff between resolution and bandwidth can be well tackled by using the cantilever-observer framework shown in Figure 3.The cantilever-observer framework decouples the bandwidth from Q allowing one to use a high Q system giving better resolution and high bandwidth at the same time. It gives the flexibility to shorten the impulse response of the channel and cancels the effect of the dither at the output.

Figure 3: Cantileve-observer Architecture

Channel Model: The modeling of the read channel of the probe based data storage as a communication channel is not so straightforward as the model should predict essential experimental features and should remain tractable for data storage purposes. The discretized read channel is shown in Figure 4.

Figure 4: Discretized read channel model for probe based data storage

Thresholding Detectors: With proposed read channel model, various thresholding detectors like locally most powerful (LMP), generalized likelihood ratio (GLRT) and Bayes detector can be developed. It can be shown that cantilever deflection and innovation output signal dont provide hit detection but thresholding does provide hit detection (See Figure 5).

As it is shown in Figure 5, hits can be clearly detected by using the threshold detectors. For different tip media interaction, receiver operating characteristics (ROC) are shown in Figure 6.

(a)

(b)

(c)

Figure 5: (a) Cantilever Deflection signal (b) Innovation signal (c) Output of LMP detector

From Figure 6, it is observed LMP detector outperforms all the detectors in terms of complexity and probability of error. When hits are spaced very closely in the order of cantilever frequency, thresholding detectors will not be able to detect hits as significant intersymbol interference (ISI) will be present in innovation signal. In this case, advanced channel modeling and viterbi detection is used.

Figure 6: ROC graphs for (a) 1.3 nm tip media interaction (b) 1.5 nm tip media interaction (c) 1.7 nm tip media interaction

(a)

(b)

(c)

Markovian Modeling of tip-media interaction: Markovian modeling is chosen for modeling the nonlinearity block in read channel model. It should be noted here that read channel is not a discrete memoryless channel (DMC) as nonlinearity block has a memory corresponsing to cantilever system response (See Figure 7).

Maximum Likelihood Sequence detector: The bit detection problem can be posed as a problem of finding maximum likelihood sequence detection. After making assumptions and solving the problem, it turns out that problem can be solved by using dynamic programming in particular Viterbi decoding.

Figure 7: Markovian Modeling of tip-media interaction

Simulation Results: The first resonant frequency of the cantilever f0 = 63.15 KHz, quality factor Q = 206, the number of hits in high bit duration is equal to 13 i.e. q = 13, the value of forcing amplitude equal to 24 nm, tip-media separation is 28 nm, discretized thermal and measurement noise variance are 0.1 and 0.001 respectively. With these thermal and measurement noise values, a Kalman observer is designed. In Figure 8, BER for various detectors for different SNR are shown. Comparison of various detectors for simulation data. The Bayes curve is not visible in the graph as it coincides with the LMP curve.

Figure 8: Comparison of various detector BER for different SNR

Experimental Setup: In experiments, a cantilever with resonant frequency f0 = 71.78 KHz and quality factor Q = 67.55 is oscillated near its resonant frequency. A freshly cleaved mica sheet is placed on top of a high bandwidth piezo. This piezo can position the media (mica sheet) in z-direction with respect to cantilever tip. A random sequence of bits is generated through a FPGA board and applied to the z-piezo. High level is equivalent to 1000 mV and represents bit ‘1’ and low level is 0 V and represents bit ‘0’ thus creating a pseudo media profile of 6 nm height. The bit width can be changed using FPGA controller. Thus we can change the bit width from 350 µs till 60 µs. The tip is engaged with the media at a single point and its instantaneous amplitude in response to its interaction with z piezo is monitored. The controller gain is kept sufficiently low such that the operation is effectively in open loop. The gain is sufficient to cancel piezo drift and maintain a certain level of tip-media interaction. An observer is implemented in another FPGA board which is based on the plant’s free air model and takes plant dither and deflection signals as its input and gives out innovation signal at the output. The innovation signal is used to detect bits by comparing various bit detection algorithms. The experiments were performed on Multimode AFM, from Veeco Instruments. Considering a bit width of 40 nm and the amount of time for scanning of 60 µs gives a tip velocity equal to 2/3 × 10-3 m/sec. The total scan size is 100 micron which means the cantilever will take 0.15 seconds to complete one full scan. Read scan speed for this operation is 6.66 HZ. The read scan speed for different bit width can be found in a similar manner. With this setup, experiments are performed for different bit widths varying from 60 µs to 300 µs (See Figure 9).

Figure 9: Experimental Setup for probe based data storage

Experimental Results: Innovation signal obtained from experiments is used to find BER for different detectors. In Figure 10, it can be seen that Viterbi decoding outperforms all the detectors and gives minimum BER. Details about this work can be found here.

Figure 10: BER for Viterbi, LMP and GLRT for different bit widths.

Related Publications

1. Naveen Kumar, Pranav Agarwal, Aditya Ramamoorthy and Murti V. Salapaka, "Maximum likelihood sequence detector for dynamic mode high density probe storage ", IEEE Trans. on Communications, 2009 (to appear).

2. Naveen Kumar, Pranav Agarwal, Aditya Ramamoorthy and Murti V. Salapaka, "Maximum likelihood sequence detector for dynamic mode high density probe storage ", IEEE Globecom 2009 (to appear).

3. Naveen Kumar, Pranav Agarwal, Aditya Ramamoorthy and Murti V. Salapaka, "Channel modeling and detector design for dynamic mode high density probe storage", 42nd Annual Conf. on Information Sciences and Systems (CISS), 2008

2) Transient force-atomic force microscopy (TF-AFM): a new interrogation method (This work is in collaboration with Nanodynamic systems lab, University of Minnesota)

TF-AFM imaging was invented in the NanoDynamics Systems Lab. Prior to this method of imaging most imaging methods used steady-state signals like amplitude and phase for imaging purposes. These signals and the related imaging techniques are fundamentally limited by the quality factor (Q) of the cantilever. TF-AFM interrogation bandwidth is independent of the Q of the cantilever and thus the bandwidth can be improved by a factor in the order of Q.
The TF-AFM method employs a model of the cantilever. The model of the cantilever is determined by standard frequency response methods. This model of the cantilever is built into a Field Programmable Gate Array (FPGA) card. The main concept is to compare the output as produced by the FPGA card that implements the model and the actual cantilever output. The notion utilized is that any mismatch between the FPGA output and the actual cantilever output is derived due to fact that the sample affects the actual cantilever but not the FPGA implemented model. This observer based method can detect the sharp transitions in the sample features.
The images shown below show that the TF-AFM based interrogation of DNA produces a highly resolved image of DNA when other imaging signals like height, phase and amplitude provide a poor image of the DNA. The DNA sample is a Lammda DNA that is approximately 2 nanometer in height. The lateral scan size is 2 micrometers. The scan rate was 12 Hz. The images were obatined in real-time in the sense that the image can be viewed as the scan proceeds. The experiment was done on a Veeco Multimode AFM, with one of the auxiliary imaging signals being the output of the FPGA board. The scanner used was the J-scanner.

Objective: To develop an algorithm for imaging which can image at a very good resolution for very high scan rates as compared to conventional imaging techniques. (Details are omitted here as we are going to patent this new imaging approach by end of this year.)

3) Multiuser scheduling and resource allocation

As the wireless access to the Internet becomes increasingly popular, the downlink (from the base station to the mobile users) may have to transport more traffic. However, supporting multimedia traffic, such as voice, video, and data, and making efficient use of the radio resource are very challenging tasks for the downlink OFDM wireless communication systems. This is due to the following: 1) the scarce radio resource and the limited base station transmission power; 2) the time-variant channel conditions resulting from the fading and user mobility; and 3) the diverse quality of service (QoS) requirements of multimedia users, in terms of delay, delay variance, throughput, and bit error rate. One of the promising approaches to efficiently support multimedia traffic in downlink OFDM systems is to employ a resource management scheme at the link layer which can dynamically allocate bandwidth to mobile users in accordance with the variation of traffic load and channel conditions. We have proposed a new algorithm which combines downlink resource management and power and subcarrier allocation scheme for packet transmission are combined in OFDM systems. Simulation results show that system throughput is increased using our algorithm. It also gives zero average and maximum transmission delay with increase in target transmission power. More detail about this work can be found here [1] , [2].

Return to top