ROICs

CMOS technology allows complex input amplifier circuits to be designed to temporally integrate the detector signal. Figure 9 depicts schematic examples of the three most widely used, and simplest, input circuits—the source follower per detector (SFD), the capacitive transimpedance amplifier (CTIA), and the direct injection (DI).

Figure 9. Circuit schematics for the three most commonly used input circuits for CMOS ROICs: (a) the SFD, (b) the CTIA, and (c) the DI.

Table 2 provides a description of the advantages and disadvantages of each circuit. The SFD is most commonly used in large-format hybrid astronomy arrays as well as commercial monolithic CMOS cameras. It is simple, has low noise, low power, and low FET glow. The CTIA is more complex and higher power but is extremely linear. The DI circuit is used in higher-flux situations and has the same low power and FET glow as an SFD. The gain (volts out / charge in) of the input circuits are determined by charge integration capacitors: for an SFD, it is the detector capacitance; for CTIA and DI, charge is integrated on ROIC capacitors. ROIC capacitors are determined by the layout of the ROIC and can be varied according to the application.

4.3 Noise Reduction Techniques for CMOS Hybrids

There are three general types of noise in CMOS hybrids: temporal, fixed pattern, and random telegraph signal (RTS). Temporal noise may be divided into white or uncorrelated noise and 1/f noise. This noise may be from the detector and/or ROIC, depending on the design. Fixed pattern noise is caused by residual nonuniformity after calibration, which in turn may be caused by 1/f noise or drift in key system parameters such as focal plane temperature. RTS is randomly occurring charge trapping/detrapping events and is highly dependent on CMOS process and design parameters.

White temporal noise has usually been the dominant type of noise in CMOS devices used astronomy. A common technique to reduce this noise is to sample each detector element multiple times during a long integration time, as illustrated in Fig. 10.

There are two common methods of analyzing multiply sampled data. One is Fowler sampling, where the first N samples and the last N samples are each averaged and the two averages are then subtracted. Noise is reduced by root N if sample-to-sample noise is uncorrelated. This technique also performs a CDS function, reducing noise introduced by resetting the integration capacitor (kTC and other noise sources). The second technique is to perform a least squares fit of all samples to a straight line or polynomial.

Table 2. Comparison of Attributes of the Three Most Common Input Circuits.

Circuit

Advantages

Disadvantages

Comments

Source follower

Simple

Gain fixed by detector

Most common

per detector (SFD),

Low noise

and ROIC input

circuit in IR

.. ."self-integrator"

Low FET glow Low power

capacitance Detector bias changes during integration Some nonlinearity

astronomy

Capacitance

Very linear

More complex circuit

Very high gains

transimpedance

Gain determined by ROIC

FET glow

demonstrated

amplifier (CTIA)

design (Cfb) Detector bias remains constant

Higher power

Direct injection (DI)

Large well capacity

Poor performance at

Standard circuit

Gain determined by ROIC

low flux

for high flux

design (Cint)

Detector bias remains constant

Low FET glow

Low power

Time

Figure 10. Illustration of multiple sampling of a detector during a long integration. The sawtooth line represents the voltage on the output of a detector input circuit as a function of time. The upwardly rising ramp represents the integration of photocurrent; the sudden downward voltage is the reset that begins the next integration. Arrows show regularly spaced points in time where the integrated signal is sampled.

Figure 11 compares the signal-to-noise ratio (SNR) for Fowler sampling with the sample-up-the-ramp (SUTR) noise reduction technique for the theoretical case where all samples are uncorrelated. This example uses a total of 64 samples scaled so that Fowler-1 (CDS) SNR is set to 1. The SNR

of the Fowler sampling is a function of sample pairs with 32 being the maximum in this case. A peak in SNR where the number of Fowler pairs is one third of the total number of samples (take the last third of the samples and subtract from the first third while ignoring the middle third) and is a general result. The Fowler SNR peak is 6% below the SUTR peak. Detector dark current and photocurrent are integrated, and therefore later samples are correlated to earlier ones. In the limit of background-limited performance (BLIP), Fowler-1 (simple CDS) yields the best SNR, which is 9% better than SUTR [4]. Noise as low as 2 electrons rms has been achieved with cryogenic (30 K) CMOS ROICs in a Fowler-16 mode [5].

Figure 11. Example of SNR improvement with multiple uncorrelated samples. SNR as a function of Fowler sample pairs is shown in comparison to SUTR. In this example, there are a total of 64 samples spaced over equal time intervals. SNR is scaled for SNR=1 for Fowler-1 (CDS). The peak in Fowler sampling occurs when the number of Fowler pairs is one third of the total number of samples and the middle third of the samples are ignored. This peak is 6% below the SUTR SNR value.

Number of Fowler Pairs

Figure 11. Example of SNR improvement with multiple uncorrelated samples. SNR as a function of Fowler sample pairs is shown in comparison to SUTR. In this example, there are a total of 64 samples spaced over equal time intervals. SNR is scaled for SNR=1 for Fowler-1 (CDS). The peak in Fowler sampling occurs when the number of Fowler pairs is one third of the total number of samples and the middle third of the samples are ignored. This peak is 6% below the SUTR SNR value.

Was this article helpful?

0 0
Telescopes Mastery

Telescopes Mastery

Through this ebook, you are going to learn what you will need to know all about the telescopes that can provide a fun and rewarding hobby for you and your family!

Get My Free Ebook


Post a comment