The classification of OCC technologies has been given.
Being a part of non-flicker modulation series, with the vast usage of rolling shutter cameras in the camera market, the use of rolling shutter cameras in OCC systems takes advantage of the high sampling rate of the rolling shutter mechanism, which sequentially exposes pixel lines to the incoming light. The Rx sampling rate of these systems, which reflects the pixel sampling rate of the rolling shutter, is much higher than the Tx optical clock rate. In principle, the frame rate of a high-speed camera Rx must be higher than the Tx optical clock rate to satisfy the Nyquist rate.
Some modulation schemes for rolling shutter OCC systems submitted to TG7m include camera-OOK (C-OOK) from Kookmin University (see this previous post), Pulse Width/Position Modulation (PWM/PPM, or MPM mode) from Panasonic, rolling shutter-FSK (RS-FSK) from National Taiwan University, and the camera-M-FSK series (CM-FSK) from Kookmin University (see this previous post). Related works discuss OOK based line coding developed by PureLiFi [see here] and M-FSK developed by Carnegie Mellon University [see here].
Rolling shutter OCC – A Classification
The highlighted works can be grouped into two types: time modulation series (OOK schemes and PWM/PPM), and frequency modulation series (M-FSK schemes, and RS-FSK ). The modulations listed by TG7m were designed to be compatible with a wide range of image sensors, whereas those presented in the related works did not take into account the compatibility considerations. In particular, C-OOK enables the detection of packets that are missing because of a sharp decline in the Rx frame rate. A pair of asynchronous bits (Ab) allows for the detection of two-thirds of these missing packets. Two other modulation schemes, RS-FSK and CM-FSK, are compatible with cameras that have widely different frame rates, readout-times, inter-frame gap times, and pixel sampling rates. The typical bandwidth of smartphone cameras is also supported. Moreover, PWM/PPM is designed to support bandwidths of ≥10 kHz, which facilitates greater data rates.
All of these rolling-shutter-based OCC systems share a problem that results from the trade-off between the data rate and the communication distance; this trade-off means that the amount of data available is proportional to the resolution of the light source on the image. A detailed analysis of the performance shall be discussed thoroughly in the following section. From our perspective, these modulation schemes are most suitable for indoor scenarios or for smartphone users located in nearby buildings because the distance is usually no more than 10 m and the camera receiver is equipped with a pedestrian’s smartphone. Researchers have shown that the rolling shutter OCC technology is to deliver accurate indoor localization and navigation, or some other promising services for inside buildings where GPS cannot reach.
Data Rate – Distance Tradeoff
Given the operation of the rolling shutter camera, the actual number of samples (pixel rows) acquired from the captured image of the light source at distance d, Nrow(d), is calculated as
where w is the image width (in case the rolling axis is along the width of the image sensor), L is the normalized length (diameter) of the light source along the width of the image sensor, d is the distance between the light source Tx and the camera Rx, and FOV is the field of view of the camera.
The actual number of samples acquired from an image, Nrow(d), determines the amount of information that an image can capture. Two different approaches to demodulation, frequency domain and time domain based demodulations are described as follows.
1) Performance Limitation of Frequency Demodulation
The sampling rate of the image sensor, which is the frequency at which a row of pixels is sampled, must be at least twice the rate of the highest signal frequency, according to Nyquist’s theorem. This relationship is described by the following equation:
where fmax is the maximum resolvable frequency, fNyquist is the Nyquist frequency, and fs is the sampling rate of the image sensor (i.e., the sampling rate of lines of pixel).
The frequency resolution (df) is dictated by the acquisition time and given as
where T is the acquisition time (of a rolling image), Nrow is the number of samples (i.e., pixel lines) acquired throughout the diameter of the light source along the rolling direction of the image sensor.
From those two above equations, the condition for the camera Rx to differentiate a frequency is that the size of the light source on the captured image is large enough, namely,
Accordingly, the requirement of the light source size on the captured image leads to the upper-limited communication distance, which can be computed as follows
Thus, the maximum distance of transmission is proportional to the frequency resolution and the size (diameter) of the light source.
2) Performance Limit of Temporal Demodulation
Unlike FSK, wherein a single frequency symbol is demodulated per image, OOK or Packet PWM/PPM systems demodulate the entire packet including multiple symbols per image. Thus, the size of the data sub-packet shall be short enough to be captured entirely by an image in the case where data fusion technique is not applied. The conditional size of the light source on the captured image is expressed as
where R(DS) is the data sub-packet rate, and Tsub_packet is the length in time of the data sub-packet.
Consequently, the communication distance is limited as a function of the data sub-packet size as follows:
Moreover, besides the distance limitation which is linearly proportional to the data sub-packet length as shown in (14), the temporal rolling shutter decoder shall require the number of samples to clearly differentiate between the on and off states of the captured waveform in the time domain. A minimum of four samples per clock is suggested to identify whether the transmitted state during the clock time is on or off. The condition is expressed as,
where N (B/sub_packet) is the number of binary clocks per data sub-packet.
Fig. 3 presents the simulated bit error rate against the pixel Eb/No performance of the OCC systems employing OOK, VPPM, and FSK with various dimming levels. We applied pulse width control for VPPM and FSK dimming; however, we applied the amplitude control for OOK dimming as it is suitable for an OCC system. Undeniably, Binary FSK (BFSK) shows the best performance because it requires the lowest pixel Eb/No to achieve the desired BER. VPPM outperforms under dimming. VPPM dimmed at 10% provided a better performance than OOK dimmed at 25%. This is because the amplitude dimming control is applied for OOK to maintain the link rate. Fortunately, our experiment of pixel intensity along with the calculation of pixel Eb/No always guarantee the pixel Eb/No at greater than 30dB for the light-of-sight link at a predetermined indoor distance such as 10 m, ensuring that the OCC systems perform well under low dimming.
Fig. 4 displays the bit rate against distance performances of OOK, VPPM, and FSK modulation schemes. Temporal modulation series (including VPPM and OOK) provide higher bit rates at the same distance compared to frequency modulation series (such as RS-FSK and CM-FSK). However, for the communication distance, the OOK scheme cannot extend the communication as much as FSK can. Above Equations describe the calculation of the maximum ranges of all those schemes. The trade-off between the distance covered and the data rate suggests the selection of a proper modulation for the planned system.
An improvement over Data Frame Format
To extend the maximum distance of communication, a fusion technique has been proposed in which the sub-packet is merged from two adjacent images using asynchronous bits (Ab). This sub-packet is decodable if every image captures no less than half of the data sub-packet size. Hence, the size of the image allowed is two times smaller than that of the typical data sub-packet:
Consequently, the communication distance is limited as a function of the data sub-packet size with Ab as follows:
Within the maximum communication distance, the decoder shall also require the number of samples to clearly differentiate between the on and off states of the OOK or the VPPM signal in the time domain as expressed the previous equation.
Fig. 16(a) describes how data fusion is applied to collect and recover data from different images. Each data symbol shall be transmitted along with its clock information, allowing the camera decoder to group together images belonging to a data symbol. Each image provides a part of the data sub-packet; however, adjacent images are fused into a complete data subpacket. The fusion technique is beneficial in terms of distance because at a further distance the decoder operates well even if a part of the data recovered from the captured image is incomplete.
In Fig. 6, we present bit rate as a function of distance performances of the OOK modulation schemes. Our comparisons are performed against conventional OOK modulation formats, i.e., the data subpacket shall comprise a preamble and a payload. The fusion technique is implemented by inserting a single asynchronous bit at the start and end of each payload. OOK modulations using the fusion technique significantly outperform conventional OOK modulations. Especially, at the same required data rate, the maximum distance is markedly enhanced with the fusion. Similarly, at the same distance needed, the amount of actual information data is increased for each image (i.e., the subpacket payload after the overhead part is removed) and the fusion technique gains twice the data rate achieved by the typical method.
The distance-data rate tradeoff is problematic for OCC system designers implementing rolling shutter camera.
PHY IV modes within IEEE 802.15.7m result from the best efforts of the technical contributors. The customization of PPDU for OCC without overhead subfields has its reason.
We have quickly gone through Rolling-shutter-OCC modes, the next post will discuss Screen OCC modes (i.e., Screen-to-Camera Communication).