Ota Qe

The optical testbench focal plane uses Hamamatsu S1337 photodiodes to measure the absolute photon flux from monochromatic flatfields. These hermetic ceramic packages behave well under vacuum and at cold temperatures. We carefully calibrate them as a function of wavelength and temperature against the usual NIST-calibrated photodiodes. Figure 3 shows representative QE values for OTAs which are thinned to 45 micron (no substrate bias applied) and OTAs which are thinned to 75 micron (-20 V substrate bias applied to fully deplete the CCDs to reduce charge diffusion).

400 600 800 1000 Wavelength (nm)

Figure 3. The quantum efficiency of 45 ^m thick OTAs is the same as 45 ^m thick CCID20s, tested by many people with many different setups. The QE of the 75 ^m thick OTA is distinctly higher in the red, in accordance with its increased opacity to red photons.

400 600 800 1000 Wavelength (nm)

Figure 3. The quantum efficiency of 45 ^m thick OTAs is the same as 45 ^m thick CCID20s, tested by many people with many different setups. The QE of the 75 ^m thick OTA is distinctly higher in the red, in accordance with its increased opacity to red photons.

The QE we measure from 45 micron thick OTAs is virtually identical to that of CCID20 devices whose QE has been determined by us and others with completely different apparatus. However, the 75 micron thick OTA has a markedly improved red response beginning at about 750 nm. By wavelengths of 1 micron, the 75 micron device has a 50% higher QE than the 45 micron thick OTA, as expected from the greater absorption depth for IR photons.

4.5 Substrate Bias

Figure 4 shows how the various components of an OTA cell are isolated. The result is that it is possible to run parallel clocking voltages below the ground of the NMOS transistors making up the logic without current flow between them. We therefore get better parallel clocking and transfer from pixels into the serial register. By applying a negative potential to the substrate connection, we bring the surface boron implant to a negative potential with respect to the buried channel where the electrons collect. This enhanced electric field fully depletes the silicon and decreases the charge diffusion.

OG REF CCD PEF

OG REF CCD PEF

Figure 4. There are three distinct reference potentials in the deep depletion OTA: the ground for the pixels and amplifier, the ground for the logic, and a substrate connection. Use of n-implants adjacent to the p-type implants for the channel stops create back-biased pn junctions which prevent current flow along the surface, and a high potential on the scupper connection depletes the bulk and prevents current there as well.

Figure 4. There are three distinct reference potentials in the deep depletion OTA: the ground for the pixels and amplifier, the ground for the logic, and a substrate connection. Use of n-implants adjacent to the p-type implants for the channel stops create back-biased pn junctions which prevent current flow along the surface, and a high potential on the scupper connection depletes the bulk and prevents current there as well.

This feature permits us to trade off red QE against charge diffusion. The very red QE is essentially proportional to the thickness of the devices, whereas the charge diffusion is also roughly proportional to thickness and inversely proportional to the square root of substrate potential. Fe55 x-ray images immediately show that the substrate bias has a big effect. The number of single-pixel events increases dramatically as the substrate bias is increased.

We have found that x-rays are a very good quantitative measure of the extent of charge diffusion, and are much easier to provide than extremely small optical spots. We normally examine all x-ray events, decide which are K-alpha events, normalize them, and then assemble a distribution of pixel values found in the event. There are some very characteristic features visible in Fig. 5.

Figure 5. The histogram of pixel values around x-ray events can be matched in detail with a model which depends only on the surface charge diffusion length. The values are normalized to the full 1620 e- of single pixel events.

Obviously there is a spike at 1, which is the single pixel events - all the charge found in a single pixel. The slope change at 0.7 corresponds to x-rays which convert at the surface of the device, in the center of a pixel. This discontinuity in the volume available for making events of a given fraction causes the discontinuity in slope. We find that we can fit the observed distribution accurately using a simple model which has an exponential absorption of x-rays, a constant electric field causing a diffusion size which scales as the square root of the height above the backside where photons are absorbed. The one free parameter of these models is the charge diffusion at the surface.

When we are confident that we understand the detailed distribution of x-ray pixel values, we can construct a simpler statistic which can be converted to physical charge diffusion length. The statistic, called "qdiff, is defined as the sum of the 8 adjacent pixels divided by the highest pixel of an x-ray event (and is therefore 0 for no diffusion at all and 8 for infinite diffusion.) We process x-ray images by identifying events, determine those which sum to K-alpha events (to avoid multiple x-ray photon events), stack the events by centering on the brightest pixel, and create the qdiff statistic. Application of the model above permits us to calculate a theoretical qdiff for arbitrary surface rms diffusion and we can thereby convert qdiff to surface rms in units of pixels and then microns.

As a function of substrate bias we find that the charge diffusion length scale gets smaller, as illustrated in Fig. 6. We obtain the same diffusion for different sized pixels, not surprisingly, and much greater diffusion at a given substrate bias for the 75 micron thick device compared to the 45 micron thick device. We regard a diffusion of 3.2 micron (rms) to be the limit of what we would like to have, since this corresponds to 0.2 arcsec FWHM degradation of images. We can achieve this amount of diffusion with the 45 micron devices and no substrate bias or application of a -40 V bias to the 75 micron devices. (Note the diverging diffusion for the 75 micron device as the substrate bias approaches 0; without a bias, this device is not fully depleted.)

Figure 6. The charge diffusion can be measured from the x-ray histogram or qdiff statistic.

We get identical results from devices with the same thickness but different pixel size. The diffusion depends on the thickness of the OTAs as well as the substrate bias voltage.

Figure 6. The charge diffusion can be measured from the x-ray histogram or qdiff statistic.

We get identical results from devices with the same thickness but different pixel size. The diffusion depends on the thickness of the OTAs as well as the substrate bias voltage.

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