Sensor Spectral Quantum Efficiency

Started 2 weeks ago | Discussions thread
OP Bernard Delley Senior Member • Posts: 1,208
Re: OK ...

Iliah Borg wrote:

Bernard Delley wrote:

Iliah Borg wrote:

Alexey.Danilchenko wrote:

Bernard Delley wrote:

The one putting thins in perspective with monochromator performance, I have already been given. If a monochromator were used, its spec sheet would do instead. It is on the readers judgement/comment, if 'my' setup compares reasonably.

Nobody uses specs of monochromator to take out its effect on a signal. Instead you assemble a system (light source if any, integrating sphere on output or optical fiber) fix all the components and calibrate the fixed system function (response) with a known signal - calibrated lamp (FEL) or using various models of tungsten halogen lamps. Then you can account for the signal changes in the measurements. This absolute calibrations only work if everything is fixed, if you change attachment or position of say fiber optic taking signal out of monochromator, you have to recalibrate.

You also keep asking for constructiveness from all people commenting here yet a lot of information about spectral measurements and calibration of various kind for instruments involved are out there - Google them.

What is not readily available is how to use a film camera to calibrate the projection of grating because the setup is quite complex, unless the grating is mounted on a stepper stage compensated for backlash and slack through an encoder. This way one can use a photodiode to measure just the centre, thus it will take 30+ shots with a digital camera to get a single measurement.

If I would do these measurements professionally, I would of course use a monochromator and integrating sphere setup, calibrated photospetrometer etc, for the flexibility of the setup. What I wanted to say about the monochromator, say with common specs about 1nm wavelength accuracy and slit settable resolution to better than say 10 nm, it would be more like checking specs than calibrating for the wavelength dependence. Calibrations for the electron number and the photon number at lambda would still have to be done.

With my setup shown in the scheme post, I clearly need to do the 'monochromator' specification for wavelength reading accuracy and resolution. -- I showed that in my post "Calibration 1: wavelength" . No stepper is needed for that ! It can go in a 'single' shot. In practice I do extra shots, also for each laser.

I have learned some lessons on how to do electron calibration (camera gain) to better than 1%. It is not as simple as one may think first. The papers showing bigger scatter on camera response have not mastered this.

The major issue is with the calibration of the photon flux. Clear limitations of my measurements are that I am using a photospectrometer as calibration source with the following limitations. 1) its ambient mode calibration has no specs. I just guess what it might be. !! 2) it has an inconvenient aperture. There is no handle at its integration time so its low light sensitivity limit cannot be pushed down. 3) is about the definition of ANGULAR sensitivity, which is a HARD ISSUE in any setup. I see it as the main reason why photometric calibrations are skirted if they are not absolutely needed. However, the challenge for QE really is knowing the photon flux or photon number in the exposure.

In my setup scheme, illumination of the sensor is through a lens of 160mm FL on the DX format. With the grating, the aperture ratio is about f/5.6. This is a realistic illumination scenario for camera pixels. On the blue and red ends of the spectrum, this optics gives a skew angle of incidence up to 4.3 degrees off normal incidence. With the approximate Lambert characteristic of the i1Studio ambient mode, this should remain well under 1% reduction of sensitivity as compared to normal incidence.

If one places a Nikon F mount camera body without lens at the port of an integrating sphere, one has effectively an aperture ratio between f/1.1 - f/1.2 for the central pixels. The pixels likely have a narrowed angular sensitivity distribution, not cosine. So calibrations are needed, if QE relevant for longer lenses with near normal incidence is of interest. The ANGULAR sensitivity issues can easily amount to more than 10% differences in the QE result unless this is carefully studied and calibrated. This angular study goes further than I wish to go with my present little project.

Lets see the proof of the pudding, lets see the results from your setup.

Our setup is to produce colour transforms to convert device RGB to XYZ, and the results make customers happy. But I thought the title of the thread is not "show me your spectral data, and I'll show you mine". When I will be asking for an advice on our setup, or seek solutions for any inaccuracies resulting from it (currently I have none of the sort, only happy users and customers), I will start a thread posting calibration and measurement data, drawings, photos, anything that may help my peers to suggest improvements or to determine the source of the problems.

OK, the most important is to keep the paying customers happy !

I must assume now that you have such spectral response data for many camera models. Even if your purpose does not require the challenging absolute calibration, It would be very interesting for me, and probably some readers of this thread, to see your spectra for one of the camera models I showed, to judge the level of (dis)-agreement between the approaches on the result. I do no intend to claim any superiority of my cheap approach.

Post (hide subjects) Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark MMy threads
Color scheme? Blue / Yellow