I want to determine the Dynamic Range, Signal to Noise Ratio etc of a pixel. I have simulated the pixel in Cadence Virtuoso.
Is it possible to analyze Dynamic Range, Noise level from Virtuoso? If yes, then how?
If No, then can Verilog or another software can help the purpose?
First, you need to measure the input referred noise current. in the calculator build the following function: integrate(VN2) up to 1.5BW. then take the square root of this integration. then divide it by the mid band gain. This is the input referred noise current Iin,n. for BER of 10-12 use Q=7 so the photo current sensitivity Isen =7*Iin,n. for optical sensitivity Psen =10log(Isen/R) where R is the photodiode responsivity.
I hope it is clear.
Dear Mr. Minhaz , Adding to what ha been said by Dr. Mohammed,
The dynamic range of any linear sytem or subsystem is the ratio of its maximum input to its minimum input for the output signal to be free of distortion and distinguishable from noise. The lower limit is set by noise generated in the system or the device and the the upper limit is set by the nonlinear distortion of the input signal. The equivalent input noise level is calculated as dr mohammed described. However the maximum value limit is determined by calculating or measuring the transfer curve and determining the one dB compression point or the knee point between the linear and the saturation curve of the transfer curve.
I hope that i shed some light on the answer.
Minhaz
Take a look at the two references below. The first reference does a lot of architectural comparison with key metrics like dynamic range, SNR etc. using MATLAB/SIMULINK. Can the same exercise be done with Cadence? The answer is not straight forward. You can certainly do noise analysis to figure out input referred noise (as mentioned by Mohammed Atef) with Cadence. An analytical example is given in the second reference. But for dynamic range you will need to know the highest input signal that can be processed by your system without severe distortion. This upper limit is defined by well capacity or a more complex function for a sigma delta based architecture. However, this particular metric can be easily modelled numerically (see first reference).
So long story short, SPICE simulators are good for transient behavior and noise modeling. Use a high level abstraction for determining Dynamic Range, SNR, flux etc. MATLAB/SIMULINK can do this very effectively. I have not seen a Verilog example.
1. S. Kavusi, and A. El Gamal, "Quantitative study of high-dynamic
range SD-based focal plane array architectures," Proceedings of the SPIE
Electronic Imaging, vol. 5406, Apr. 2004;
2. N. Bluzer, R. Stehlik, "Buffered Direct Injection of Photocurrents into Charge-Coupled Devices," IEEE Journal of Solid-State Circuits, vol. sc-13, no. 1, pp. 86-92, Feb. 1978.
Hello Minhaz
the dynamic range DR of a pixel is the ratio between the maximum light signal that can be processed and the minimum which is constrained by the noise level. This means, the Dynamic Range i [dB] is given by:
DR = 20.log ( fullwell capacity [e-} / readout noise [e-])
both values given in electron numbers.The FW full well capacity is the maximum number of electrons which the pixel can process. Both values are pixel size dependent.
Following to this, the maximum pixel dynamics corresponds to the SNR of light, which can
be expressed:
SNR = 20. log (FW / sqt(FW)) in [dB].
So your calculation should go as: estimate from given SNR value in dB the FW fullwell value
and then from the needed dynamic range DR in dB in the 1st equation. From this you can
conclude to noise level, in single electron [e-] units.
Comming to the sizing of the pixel:
Equation for the pixel capacitance is given as C = ( FW . q ) / Vrange
where q is the elementar charge constant 1.6021...×10−19 C and the
Vrange your given voltage range for the pixel in [V].
Following to this calculation the needed size of your pixel is estimated from the diode capacitance value [fF/um²] and the sidewall value [fF/um]. This way you get your pixel size.
The equation Unoise= sqrt (k.T/C) gives you the noise voltage value for your
pixel size (capacitance). Convert it to single electron units to compare it with
the noise value in electron units you got from the 1st equation constraining the DR.
The noise value should be less than in 1st equation to fullfill your DR specification.
Best regards
Richard
- Badges
- Science topic
- Similar topics
- Computer Science and Engineering
- Artificial Intelligence