PR Newswire: SiOnyx in collaboration with the Army Research Office (ARO), has successfully demonstrated pixel-scale detectors with room temperature Detectivity (D*) exceeding 1x10(14) Jones. This number basically says that the noise equivalent power is 1.0e-14 Watt for a detector having 1 sq. cm area and 1 Hz bandwidth.
At 1um wavelength the energy of photon is 1.24 eV or roughly 2e-19 Joules. So SiOnyx figure means that SNR=1 is achieved at about 50K photons/s flux in 1 sq. cm area detector at a speed limited by 1 Hz BW filter. Nice number.
This is said to represent a full 10x improvement over traditional silicon detectors. This paves the way for SiOnyx to participate in new sponsored research programs with the Army Night Vision and Electronic Sensors Directorate (NVESD) and Defense Advanced Research Projects Agency (DARPA).
this will kill the image intensifier tube !!!
ReplyDeleteThe NEI of an InGaAs based detector is about 3-4e+9 ph/s/cm2. 50e+3 ph/s/cm2 is 4-5 orders of magnitudes improvement compared to InGaAs SWIR sensor. It's hard to believe !
ReplyDeleteThe newest unit of energy is a "jones", as in I'm "Jonesing" for some electronic transition.
ReplyDeleteI don't understand something...
ReplyDeleteThe sensor gets an SNR = 1 with 50K ph/sec and an integration time of 1sec.
If we assume a 10um pixel, a 1 sq. cm. detector has 1M pixels.
50K ph / 1M pixels = 1/20th of a photon per pixel.
how exactly does one get an SNR of 1 this way? The shot noise alone will kill.
The D* number relates to a one-pixel detector. Detectivity D* Jones unit is in fact cm*Hz^0.5/W.
ReplyDeleteBasically, it means that each sq. cm of area generates 50K equivalent photons of noise. Depending on what type of noise it is, it somehow scales with the detector area, most probably proportional to sq. root of the area. Then 10um pixel would generate 50 eq. ph of noise and would need 50 photons to achieve SNR=1.
And 1um pixel would generate ~17 eq. photons of noise, if SiOnyx scales it down to that size. All these numbers assume 1 Hz BW of the noise measurement.
@ "And 1um pixel would generate ~17 eq. photons of noise"
ReplyDeleteMy bad, from 10um to 1um area scales 100x, so noise scales 10x. Thus, 1um pixel should generate 5 eq. photons of noise.
50K electrons for SNR=1 on 1cm2 surface. This means that the SNR on each 10um pixel is only 1/1000. At this performance, we have very roughly 60e noise per second on the 10umx10um detector. Not bad !
ReplyDeleteHow do you get the measured number? (Do you use a 1 cm^2 detector?) And, does the statistic actually scale linearly with pitch? (50K e- @ 10K um to 50 e- @ 10 um to 5 e- @ 1 um)
ReplyDeleteMy gut feeling is that a large-pixel measurement won't tell the whole story for small pixels.
Of course the measurement on a large photodiode is not the same as that of a small photodiode. The characteristics of a photodiode can be seen roughly as the combination of the area junction and the sideway junction. A small photodiode doesn't have the same sideway and area ratio as that of a large photodiode. That is why you have to measure large photodiode of different shapes in order to get the sideway and area characteristics. With these, you can have more precise prediction on a small photodiode.
ReplyDeleteBesides the measurement of a large photodiode can be falsed by some hot points inside. When you measure small photodiodes, the hot points have much less probability to fall inside a single photodiode.
Hope this is helpful for you ...
Yes, thanks.
ReplyDeleteYangNI, I'm not so sure about your InGaAs comparison. Reported InGaAs dark currents are about 2 nA/cm^2 so the shot noise would be much lower than your NEI number.
ReplyDeleteAlso, InGaAs detectors have over 1e14 Jones detectivity which is the same order of magnitude reported by SiOnyx.
Has anyone seen a reported dark current for black silicon?
Currently, it seems to me InGaAs is better in the SWIR but given that black-silicon is relatively immature it could overtake InGaAs as development continues.
It depends on the operation temperature ... I talked about 300K operation conditions. I think that most InGaAs FPAs have a D* less than 1e14.
ReplyDelete