A biomorphic digital image sensor
read more
Citations
Neuromorphic Silicon Neuron Circuits
Event-based Vision: A Survey
A Survey of Neuromorphic Computing and Neural Networks in Hardware.
On Spike-Timing-Dependent-Plasticity, Memristive Devices, and Building a Self-Learning Visual Cortex
STDP and STDP variations with memristors for spiking neuromorphic learning systems.
References
CMOS Circuit Design, Layout, and Simulation
Analog VLSI and Neural Systems
Foundations of vision
CMOS image sensors: electronic camera-on-a-chip
Eye, brain, and vision
Related Papers (5)
Frequently Asked Questions (18)
Q2. What is the main source of power consumption for the current-feedback event generator?
Short-circuit current at the event generator’s input inverter is the main source of power consumption because the input slew rate is low.
Q3. What are the contributions in "A biomorphic digital image sensor" ?
In this paper, an Arbitrated Address Event Imager ( AEE ) was designed and fabricated in a 0.6-μm CMOS process.
Q4. What is the power consumption of the current-feedback circuit?
Because the energy consumption is independent of the input slew rate in their event generator, the current-feedback circuit guarantees constant power consumption per cycle.
Q5. How does the arbitration process reduce jitter?
In addition, reducing FPN will also decrease temporal jitter since the arbitration process minimizes collisions by synchronizing pixels in a row and distributes the row access.
Q6. How can the arbitrator reduce temporal jitter?
by increasing the bandwidth of the arbitration and/or reducing the nominal spike rate per light intensity, temporal jitter due to arbitration and collisions during readout can be reduced.
Q7. What is the way to limit the power consumption of an inverter?
To limit power consumption, a starved inverter can be used, where the output current is limited by a current source to a few nanoamperes.
Q8. What is the effect of power-supply noise on the switching voltages for the individual event?
Power-supply noise can strongly influence the switching voltages for the individual event generators and will be observed as jitter in the interspike interval.
Q9. What is the effect of the row-pipelining algorithm on the image?
The row-pipelining algorithm can impose some synchrony between pixels in the same row, provided they are exposed to the same light intensity.
Q10. How many updates per second can be achieved?
The imager provides a very large dynamic range of 120 dB in uniform bright illumination and when no lower bound is placed on the update rate per pixel, a low power consumption of 3.4 mW in normal indoor lighting and is capable of a maximum of 8.3K updates per second per pixel under local bright illumination.
Q11. How can The authorreduce the power consumption of an event generator?
In order to decrease the power consumption of the event generator, it is necessary to increase its gain, at least in the vicinity of the threshold.
Q12. What is the output current of the capacitive-feedback event generator?
In the capacitive-feedback event generator design, a fraction of the output current is fed back ( is the series capacitance of , , and is the load capacitance).
Q13. What is the simplest way to trade dynamic range for a pixel update rate?
Depending on the application and the light intensity falling onto the sensor, imaging can always trade dynamic range for pixel update rate.
Q14. Why are the integration, readout, and reset cycles executed asynchronously?
A. Imager Statistics and Light SensitivityBecause of the output-on-demand nature of the proposed imager, the integration, readout, and reset cycles are executedmainly asynchronously.
Q15. What is the way to reduce the jitter in the interspike interval?
thedrag on the power supply is likely to be much larger (as much as 10 times larger from simulations), which will further exacerbate the problem, resulting in larger temporal jitter.
Q16. What is the power consumption of an inverter as an event generator?
the power consumption of the inverter as the event generator is about four to five orders of magnitude greater than that of a minimum-size inverter in a digital circuit.
Q17. What is the difference between the frequency-modulated sequence of addresses?
As a result, the system represents light intensity on a pixel as a frequency-modulated sequence of addresses, where the time interval between identical addresses (pixels) is inversely proportional to the intensity.
Q18. Why is the FPN value worse than other CMOS imagers?
This value is worse than other CMOS imagers, primarily because FPN reduction steps, such as correlated double sampling (CDS), cannot be easily performed on time-domain phototransduction.