Earthshine blog

We had a method to measure exposure times – but the method depended on an LED inside the telescope system, and it appears that the system scattered light which interfered with exposures. So we had it turned off. From data collected before the device was turned off it is possible to look at the relationship between requested and supposedly actual exposure times.
For requested exposure times below 10 ms it appears we did not get anything useful. For requested exposure times above that there is a bias which decreases with the increasing duration requested. It is simple enough to fit a regression line to the above, it is:

exptime_actual=-0.0045021465+exptime_requested*1.0004696

and use this relationship to correct all exposure times. We have tried this.

On the left, results without correction of exposure time. To the right with.

We have extracted B and V magnitudes from the set of ‘good images’. We have converted to magnitudes using Chris’ NGC6633 and M22 calibrations. For B and V images taken within ½ and hour of each other we have calculated the mean B and V values for the BS and then B-V. We plot these values against lunar phase as well as the time of day, and we see that there is a phase- and time-of-day-dependence in B-V. When we compare B-V from images where the exposure times have NOT been bias-corrected with B-V from images where the bias HAS been corrected we realize that the problems are much worse for the bias corrected images, suggesting that the bias-correction is not valid.

Hmmm.

That in turns suggests that the method for measuring exposure time is faulty. As far as we know the method was based on passing IR-light through the shutter aperture to a detector and then calculating the opening time of the exposure from data collected. We are not sure if this was based on a timing mechanism or a flux-collected method.

From our own images of a constant light source we ought to test the stability of the shutter. From the lab we have Ahmad’s laser images, and from MLO we have many stellar images, Jupiter images and also hohlraum images.

A webcam on the PXI now points to a thermometer display showing the tmperature somewhere on the telescope and somewhere in the dome. When the heaters are turned on the telescope readout cycles between about 16 degrees C and more than 60 C.

We presume that the temperature is telling us something about an area near one of the two shutters. Since the shutter shows the sticking behaviour through these temperature cycles the sticking problem is either unrelated to the temperature of the shutter or the temperature we can see is not from an area close to the shutter inside the telescope.

We think that the sticking shutter is the one inside the telescope because the system that reads the shutter time also indicates failure whenever a picture is ‘bias like’, which would not be the case if the failing shutter was the front barrel one.

After speaking to Ahmad yesterday about the shutter performance I am not
sure we fully understand it. If I understood Ahmad correctly there will
be shutter dropouts now and then due to the temperature-related issue
that causes the shutter simply to not open. Then there is the issue that
sometimes the shutter opening time is measured close to zero without
actually being close to zero – an issue due to an occasional ‘spike’ in
the signal from the shutter which causes a timing circuit to start and
stop very quickly.

So these two modes can lead to images that have a measured exposure time
of close to 0 – about half actually are close to zero and the rest are
OK but the measured exposure time is no good.

Can the measured exposure time be trusted for other situations?
Apparently it depends on how short the requested exposure time is.

I took about 200 images of the moon from our database and extracted
requested and measured exposure times as well as the total counts in
each image. I then considered the standard deviation of the fluxes
(=counts/time) for both the measured and the requested exposure time. I did this for increasing exposure times by raising a lower limit on the
times considered.
The black line is for measured
exposure times and the red is for the flux based on the requested
exposure time. For exposure times below 10 ms the standard deviation of
the flux is smallest if you use the requested exposure time. For
exposure times longer than 10 ms the standard deviation is smallest if
you use the measured exposure time.

The result is the above graph. We only have data for exposure times up to 25ms – this can be extended later.

There is a clear change in behavior at 10 ms . this is also reflected in this

lab report by Ahmad and Rodrigo.

It would seem that only very shaky
results can be had for exposure times below 10 ms. That is, if you
really need the absolute flux in an image do not expose under 10 ms.
CoAdd mode will still work, of course, since the whole image is affected
in the same way by shutter variability and we do not need accurate inter-image fluxes for
this mode to work.

The issues here do have an impact on the ability to work with the
non-simultaneous modes, and some careful analysis of the measurement
error is needed – e.g. by using a constant-flux lamp in the dome or just the inside of
the dome or the clear sky during shorter intervals.

As the lamp is now accessible we can test this!

The Canon shutters we tested – see here and here, seems better than the Uniblitz shutter we have – at a fraction of the cost. For very few dollars we could have had a substantially more reliable shutter.

We have come across another problem with the shutter: Not only is it variable when it shouldn’t be – i.e. we have complete frame dropouts and we have a spread in observed counts from sources that are constant – stars at short exposure time excluded here due to scintillation – but we also now have examples of FITS file headers that show the measured to exposure time to be 0.000000000000001s when it is patently not – for instance, the image is fine and shows a well-exposed Moon.

This limits our possibility to use the ‘not simultaneous’ modes of observing – i.e. BBSO and modified BBSO.

Chris and Peter observed the bright star Markab on the night of September 30 2011. We noticed that changing the size of the Front Iris — from 30 mm setting to the 40 mm setting had no effect on the amount of light coming from the stars. In both positions we were getting peak counts in the star of around 56000, and total of 246000 in a 10×10 box centered on the star.

We were expecting the flux to increase in going to the 40 mm iris from the 30 mm iris by a factor of (40./30.)^2 = 1.8.

Are we misinterpreting what this iris does – or did it not move? We will ask Dave about this cosmic mystery!

On night of Septembter 10, 2011, 28 images on the almost full moon were obtained at 0.015 second exposures. We compute the mean level of these images. They are dominated by the full moon, the sky flux is negligible, and the moon is well centered.

The mean count rate for these frames is 5546.7 counts / 0.015 sec and the standard deviation of the mean counts over all 28 frames is 8.4 counts – if this standard deviation is all due to shutter timing error for these frames, the timing error is ~1.4%. On these short timescales the shutter timing is at least this accurate.

Longer sequences with bias frames in between the lunar exposures would be good. None of the frames dropped out, but we know that this can happen occasionally, with as much as 20% of the frames dropping out.

The shutter is a bit of a dissapointment! It fails 30-50% of the time, and there is considerable variability, and bias, in the exposure durations. Only for 1s exposures is the scatter less than 0.1% and the bias negligible. We have heaters on the shutter, but apparently not enough. Low temperatures (‘below 5C’) are supposedly not acceptable – yet we see failure at 9-10 C.

We also see ‘dragging’ of the image. [Image needed here!] This used to be due to the shutter staying open after the image started reading out. We see a variation on this: For the overexposed images of various stars blooming occurred in th evertical direction – well, ‘dragging’ of these spurs occurred in the image. This is hardly anything to do with photons arriving on the image during readout but is some issue related to electrons and their transfer from column to column – perhaps suggesting that the previous ‘dragging’ also is related to saturated or highly charge pixels.

Here is the link to the Uniblitz home page – we are thinking of using the LS6 shutter: http://www.uniblitz.com/products/shutters_detail.asp?s=3&ss=6

This is the Canon EOS 350D parts catalog. The shutter unit is part CG2-1603

Added in 2013: I posted a query online about how to drive this shutter, but … [link]

Here are the results of some experiments I made with a Canon 350D to see how accurate its shutter works.