CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD - page 16 - Creative Uses of CHDK - CHDK Forum
« previous next »
Pages: 1 ... 11 12 13 14 15 [16] 17 18 19 20 21 ... 71   Go Down

CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD

  • 704 Replies
  • 169031 Views
*

Offline casrap

  • **
  • 90
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #150 on: 10 / November / 2012, 06:21:31 »
Advertisements
Sorry, you've just lost me completly  :o. Your way of communicating and choice of words does not agree with my way of doing things and how i wish to be treated online, so i'll leave this thread alone in the future.
I'll just say this (once more): dark CURRENT is only depending on 2 things:
- temperature
- used sensor technology.
Dark current are the electrons that break free from the substrate (without being excited, nothing to do with flux) and end up in the pixel wells were they sit until read out or excite even more electrons before being readout. How you want to call it does not matter.
For the picture quality side for astrophotography dark current is the most important parameter, as it is what you see as background in your pictures when taking long exposures (after xx seconds the signal from the dark current is bigger than the readout noise).
Duh, don't want to post here..
byeeeeeeeeeeee

« Last Edit: 10 / November / 2012, 09:09:10 by casrap »
Logged
*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #151 on: 10 / November / 2012, 11:53:16 »
This is a worthwhile temporary diversion because I feel it is relevant.  The question again is "why is the S90 so good for a small-format CCD ?"

While what casrap said that "Dark current are the electrons that break free from the substrate (without being excited, nothing to do with flux) and end up in the pixel wells were they sit until read out" is very true, that is not the problematic issue in small cameras.  When taken in ideal theoretical isolation, what casrap said "after xx seconds the signal from the dark current is bigger than the readout noise" can be true under some conditions, that is, dark current noise can be the dominant pixel noise especially at higher CCD temperatures.  On-chip manufacturing variations in pixels when extreme result in the well-known "hot pixel" effect which appears as a severe non-uniform fixed pattern image noise.  Once you begin to cool the device (typically dark noise drops by a factor of 8 every 20C drop in temp) the other (very realistic and relatively strong) sources of noise (shot noise, 1/f noise, readout noise etc) come more and more into play.  One notorious noise source that is very difficult to control not mentioned anywhere is the circuit around the photoelectron-to-voltage converter (floating diffusion amplifier, FDA) on the CCD chip, whose typical outside-world interface is an open-drain transistor.  I have seen manufacturers of commercial (linear) CCD spectrometers, devices where dark performance is paramount, use a simple resistor as a current source there and wonder why their products are so noisy.  Dark performance immediately suffers just from the thermal resistor noise being injected into CDS input stage (which also has to be low noise).  But worse, because these are compact circuits, the presence of DC-DC converter power supply switching noise and logic switching noise to which the resistor is connected can never be completely suppressed.  Even very small PS voltages cause significant secondary effects in the CDS-ADC chain, such as beat frequency noise, that result in effects that dominate image noise especially at lower CCD temps.  In typical household camera use of course, this is not an issue.  But when you want to use a pocket camera for astronomy at long integration times with very low-light sources, this problem becomes important to handle.  In order to mitigate it as in high-end scientific astronomy applications, an *active* current source must be used on the FDA output drain, and the best way to do that is with discrete bipolar technology, a circuit made of many transistors.  Of course that introduces circuit bulk which is not an issue in scientific applications for Mt Palomar, but for a small camera where real estate is at a high premium, even using SMD transistors would make the circuit impractical.

So we know already from my early tests that dark performance is significantly better on the SX110 over the S50, and dark performance of the S90 is better than the SX110.  Here is the explanation.  Follow the yellow signal line (upper left, p8) on the IXUS (SX110-like): http://photo-parts.com.ua/parts/Datasheets/MB39C303Ixus%2065.pdf.  Here any existing PS voltage noise (VDD on CCD pin 16 & on Q2501) is funneled directly into the AD9923 CDS input pin A6.  Recall we cannot get rid of all switching noise on the PS rail.  Also, my interpretation of this circuit is that the CCD output uses an *on-board* pullup resistor.  Hence the output topology is not what one would call "scientific grade" because the user has no control over the output circuit.

What's the solution ? ... an active current source using bipolar transistors as described above.  This is major technological advance by Sony and is exactly what the S90 (and the like) implements as a single chip http://www.sony.net/Products/SC-HP/datasheet/90205/data/a6811921.pdf.  The chip sits between the S90 CCD and its CDS (ADDI9003, Analog Devices likely custom variant of the AD9923A).  They call this device a "buffer amplifier" which naturally has to have an ultra-low-noise-figure input, but, its true performance boost comes from the active current source: p2, follow ISF pin to the first current source on the left (where the user sets the desired current), and then the next current source just to the right that affects IN. 

Therein lies the magic of the S90: power supply noise get rejected by the 2nd active current source.  Lastly, the active current source makes conversion linear across the full-well photoelectron range of the pixel, making light conversion linearity in the S90 as best as it could be.  In conclusion, one easy way of interpreting all this is that in non-S90 cameras one never gets to see "real dark noise" like in Mt Palomar because it is masked by circuit noise.  This is especially true of the S50, because either the CDS is poor or not implemented at all.  The SX110 (and IXUS above) is a huge improvement over the S50 because of the excellent CDS topology (AD9923).  However, the S90 is hugely different from all predecessors: the CXA3791 (or similar) active current source and no on-CCD-chip pullup resistor gives this camera a "scientific grade" architecture.  Thus the S90 CCD has excellent potential for cooling.  Kudos to Sony, Analog Devices and Canon for this amazing feat of engineering.

Hence once we've achieved mechanical minimization and I have the CCD out on a TEC, only then will we know what its *true* dark performance might be (shot, 1/f, readout, dark current, etc).  That's when a very interesting phase will begin. 

« Last Edit: 10 / November / 2012, 12:17:19 by SticK »
Logged
*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #152 on: 10 / November / 2012, 12:50:51 »
@casrap
Relevant to your CMOSvsCCD photo http://chdk.setepontos.com/index.php?action=dlattach;topic=8801.0;attach=7364 you can't use the camera LCD to compare intensity differences because the camera LCD display routine fudges the real intensity of an image so it can been seen easily after taking a dark shot, and that is especially true of a long exposure dark shot.  Fudging will likely vary among different cameras, and more so between a CCD and CMOS camera.  You have to compare the RAW images.  Furthermore, to examine circuit noise performance (please try to understand with diligence my previous post) you have to set the camera at maximum ISO (variable gain amplifier gain, VGA gain).  Those can differ from camera to camera and you have to set them to be equal as you did here to be of any comparative value ... but ... a gain setting on one camera is not necessarily the same as on another // that's difficult to work out.  Furthermore, what you need to do is SNR as I explained in my first answer, that is, you need to have a reference light source and take a photo with both cameras.  Also as I explained, you have to compensate for the difference in CCD sizes between the two cameras, if different.

The apples-vs-spaghetti CMOS-vs-CCD difference is this: each pixel on a CMOS device has an active component (circuit) that uses up significant pixel sensor area real-estate.  In that sense alone, a CMOS device will never approach the light collection ability of a CCD.  Flux and light collection is very important to analyze, because noise in general increases as the sqrt(2)x(area) whereas a constant reference flux increases linearly with area, hence larger CCDs have better SNR capability than small ones for the same flux impinging on the pixel sensor area (photodiode).

So why not use a large format CCD like a DSLR APS-C for example, in your PFP telescope architecture ?  The answer is left as an exercise .. try to figure this one out.

edit:
VERY IMPORTANT: when doing any kind noise or SNR analysis in long exposures, you have to disable dark subtract, and that can only be done with CHDK.  As the sensors warm up, more with liveview enabled, hot pixels might begin to show.  Only then can you *begin* to have meaningful results.
« Last Edit: 10 / November / 2012, 13:23:05 by SticK »
Logged
*

Online reyalp

  • ******
  • 14080
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #153 on: 10 / November / 2012, 18:32:36 »
Quote from: SticK on 09 / November / 2012, 22:01:25
On the PowerShots I've tested, you can't set a 5 minute exposure, only 15s.  CHDK extends it 64
On some cameras (those with exp_drv_task implemented) cameras, CHDK enables exposures up to ~2048 seconds.
Logged
Don't forget what the H stands for.
*

Offline blackhole

  • *****
  • 937
  • A590IS 101b
    • Planetary astrophotography
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #154 on: 11 / November / 2012, 03:40:44 »
Quote
In that sense alone, a CMOS device will never approach the light collection ability of a CCD
I would not be just concurred with that statements.
CCD is superior to the cheap CMOS built-in Web camera, but compared to the BSI-CMOS sensor, a CCD sensor is inferior. The problem with the CCD sensor its construction, where the wiring is done above the photo diode, which reduces the possibility of collecting light. In low light conditions, the CCD is always working with a much larger gain than the BSI-CMOS sensor, which increases the noise. If you make the image with the same settings in low light conditions on the CCD and BSI-CMOS, the picture taken on the CCD sensor will always be with more noise due to the above reasons.
http://asia.cnet.com/whats-the-deal-with-backlit-sensors-62061250.htm
Logged
*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #155 on: 11 / November / 2012, 08:27:27 »
@blackhole
Quote "but compared to the BSI-CMOS sensor, a CCD sensor is inferior."
This is an interesting discussion.  Backside-illuminated sensor technology is not new.  Hamamatsu has been doing it for many years and spectrometers I use have BSI (linear) monochromatic CCD sensors.  The principle is very simple in that the final sensor once fabricated and is still wafer-borne, the wafer is then polished down to 20 microns thickness from the bottom.  That improves quantum efficiency of the sensor by about 40% compared to an FSI (frontside-illuminated) sensor of the same technology.  Hence a BSI CCD sensor can approach 90% quantum efficiency at mid-band: an application of theoretical physics // you can't get better than that.  So your statement should be refined to read: "but compared to the BSI-CMOS sensor, an FSI CCD sensor is inferior."  That would be true, all else being equal, there would be a small difference.  Thus a BSI CCD sensor still far outperforms a BSI CMOS sensor in low light applications.  In a BSI color sensor, the CFA is deposited on the bottom of the wafer.  The new revelation to me from your article is that the S90 in the table actually uses a BSI (image) sensor.  However, the sensor in the S90 is BSI-CCD, not BSI-CMOS // amazing technology.

The moral of the whole story: don't let anyone trick you into believing a CMOS sensor is better than a CCD sensor in low light applications.  Keep in mind this is a forum that makes efforts to identify and cut through marketing hype.
« Last Edit: 11 / November / 2012, 09:09:16 by SticK »
Logged
*

Offline Microfunguy

  • ****
  • 4037
    • StereoData Maker
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #156 on: 11 / November / 2012, 09:00:38 »
Get on with it, let us have some action on the hardware  :)
Logged
*

Offline SticK

  • *****
  • 779
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #157 on: 11 / November / 2012, 09:16:06 »
@Microfunguy
Cute!  .. I should add that I have been successful at accessing the IRIS PI.  I plan to access all of them before reassembly.
Logged
*

Offline ahull

  • *****
  • 634
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #158 on: 11 / November / 2012, 17:07:20 »
I had another google trawl in the hope of turning up a service manual for some of the newer models, with limited success.

I did find some loosely related hardware information, again its for some older ixus models, but I suspect it might still be of some interest in the current quest as it gives some indirect insight into the adjustment of the mechanism.

 http://tinyurl.com/apognbx  ;) Cheers waterwingz

There are also some more schematics at the end of the PDF, but they are not as detailed as the ones we saw before.
« Last Edit: 11 / November / 2012, 17:39:06 by ahull »
Logged
*

Offline waterwingz

  • ******
  • 12138
    • Waterwingz
Re: CHDKPTP: S90 Primary Focal Plane Configuration - hacking out the CCD
« Reply #159 on: 11 / November / 2012, 17:17:53 »
Logged
Ported :   A1200    SD940   G10    Powershot N    G16
Pages: 1 ... 11 12 13 14 15 [16] 17 18 19 20 21 ... 71   Go Up
« previous next »
 

Related Topics