Fuji X-Pro 1 surprising noise and Ha performance

[buzz]

I own a Fuji X-Pro 1 for normal photography and was thinking of which EOS to buy for casual astrophotography. I checked some of the forum entries in SGL and CN and went into DPReview and had a look at the noise performance of the various recommended EOS models - 1000/1100/50/500/450/550 etc. What struck me was the EOS1100 noise performance was pretty good, only bettered by the considerably more expensive EOS 5D MKIII.

I noticed that there is quite swelling community using the Fuji for IR photography and, with a 720nm filter it takes reasonable IR images, like the old Kodak High Speed IR film. It can see my remote control light and I figured it's red sensitivity might be quite extended. I took a simple static shot of Orion nebula which was very promising.

I then plugged in the X-Pro 1 into DPRreview's noise comparison chart. (RAW, no noise reduction for all) . The Fuji has the same noise at 1600 ISO as the Canon's had at 200. Now - either DPRreview have messed up somewhere or this is a potential hidden gem.

Granted I have to use a 10x live-view to focus and remote control is nothing more sophisticated than a long cable release (mechanical) - but this has got to be worth an experiment.

So, I'm going to buy a used 1100D, unmodded, and when the evenings roll in a bit more, I'll do some back to back's and post back here.

[Coco]

Please buzz! Get it mounted on that FLT98 and lets have a peep

[laser_jock99]

If you're interested in Fuji's IR cameras then the Fuji IS Pro is worth a look. When they got discontinued in 2011 they were sold off very cheaply (I got one for £350 from Clifton Cameras). Interestingly the noise level of these Fuji CCDs seems to be lower than similar Canon/Nikon offerings. I'm occasionally even tempted to shoot at ISO3200! You can see some of my Fuji IS Pro pics here: http://s48.photobuck...L?sort=3&page=1 The Fuji Hyper Utility V3 software lets you see Live View on PC and you can control the camera too. But I always use an intervalometer and the in-camera live view anyway as it is much simpler in the field. Here you go 200s of The Wall in NGC7000 Cygnus @ ISO3200

[buzz]

I'll keep an eye out for that iS fuji. Nothing shows up on google right now.

I have a refurbished 1100D for just £169 on its way to me and I'll first just do some calibrated dark frames between the two cameras at the same temperature and post the noise results here. Without decent skies - I think I can still predict the Ha sensitivity - since I have a 7nm Ha filter - I can simply take a few images of a incandescent light bulb through the filter and note the relative signal levels through Green and Ha.

[Coco]

I'll keep an eye out for that iS fuji. Nothing shows up on google right now.

I have a refurbished 1100D for just £169 on its way to me and I'll first just do some calibrated dark frames between the two cameras at the same temperature and post the noise results here. Without decent skies - I think I can still predict the Ha sensitivity - since I have a 7nm Ha filter - I can simply take a few images of a incandescent light bulb through the filter and note the relative signal levels through Green and Ha.

Any chance of seeing some stacked darks Chris?

[buzz]

It seems the Fuji is in another league on dark frames. The RAW results bear out the DPreview results. Very low colour noise and dark noise is not bad either.

Here is a link to my test results so far:

http://www.toinfinit...se_results.html

and a screen dump of the two stacked dark files next to one another, with the same screen stretch. It is so astonishing, I reckon the Fuji must be doing something to the RAW files. I'll have to do a proper star field test to confirm it is not swallowing stars and when the T-adaptor arrives, do a relative Ha/SII to green sensitivity analysis.

I just wish the Fuji had remote control and download.

[laser_jock99]

Interesting result. If they are true RAW files then any noise reduction algorithm should be off. Are you able to disable the Noise Reduction in the shooting menu? You can (and should for astro) with both my Fuji IS Pro and my Fuji S5 Pro.

[buzz]

Yes - I'm an experience photographer (ARPS) and have had more cameras than I care to admit to my wife in all formats. RAW should technically be just that - but I turned off all noise reduction (long exposure and high ISO) just as a precaution anyway (it should only change JPEG appearance). I forgot to mention in these tests that both were set to ISO 800 and during the exposure the camera rear LCD screen's were on, showing the bulb second count. I haven't done any analysis yet of whether during the 25 minutes of exposure, the sensor became noisier. I also will do some bias noise tests and come up with some real pictures as soon as I have the opportunity. In the RAW conversion, I used ACR at default settings (latest version) and just plain converted the images to TIFF.

I don't think the Fuji camera is fiddling the RAW - when I look at photographic images with long exposures, there is no sign of deliberate softening - also borne out by DPreview's high ISO images too, in comparison to others.

I have a set of Baader filters - I will stick a green filter on the end of the 2" to t-adaptor tube and photograph a light bulb at +2 stops. Without changing the exposure, I'll take another with Ha and SII and will note the relative pixel values. It will probably have to wait until the weekend.

[buzz]

I just realised that in my haste I posted the Maxim screen with a TIFF vs a FITS. It amounts to the same - this time I have stretched both previews between 0 and 4000 to show the noise

[dph1nm]

Remember that Canon adds a constant pedestal to its RAW output - 256 for my 1000D, but higher (512, 1024?) for 14bit cameras. No idea whether Fuji does this, nor whether your conversion routine takes this into account.

You also need to compare the noise in elections not ADU. So you need to know the gain at ISO800 for both cameras.

NigelM

[Coco]

I just realised that in my haste I posted the Maxim screen with a TIFF vs a FITS. It amounts to the same - this time I have stretched both previews between 0 and 4000 to show the noise

I just realised that in my haste I posted the Maxim screen with a TIFF vs a FITS. It amounts to the same - this time I have stretched both previews between 0 and 4000 to show the noise

Looks very promising... I wonder how the XE-1 would fair ? That little camera looks like a good replacement for my old GF1..

[laser_jock99]

It will be very interesting to see how this camera performs on real DSO work. Keep the tests coming.

[buzz]

Good feedback all, yes, I do need to work out the ADU/electrons - just done the bias test stack of 10 x 4000th's second - canon is averaging 500 and the fuji is averaging 50. Maxim Std deviation of bias is 30 on the Fuji and about 300 on the Canon.

I'm just trying to remember how to work out the gain. Nigel - can you remind me?

I'll do the colour sensitivity test tomorrow. I just need to dismantle my filter wheel first.

My friend sent me a Nikon D800 5-min dark RAW exposure. It was not noisy but very bright and uneven, with bars and gradients - looks like the Nikon is doing something to its RAW file to smooth it but cannot accommodate what looks like amp glow.

[buzz]

It has taken several attempts to defeat the camera processing with the narrow band test and I'm not convinced I have won. I eventually produced the comparison table below. It is not a clear cut outcome though. The Fuji measured pure green with the Baader filter but the Canon, with the same exposure and fixed white balance had half as much red too - hence the yellow tinge. The SII levels are the same but the Canon appears to have better Ha sensitivity. What I cannot tell is if the Canon is artificially boosting the red channel. The acid test will have to be a back to back real DSO comparison, which I hope to do in a few weeks time when I visit a dark field site.

The file is a small tiff file, so you can see the RGB values for yourself. ISO 800, 500th second, daylight fixed white balance

fuji canon reds.tif

[Coco]

So are we looking at low noise but fainter response to HA?

[dph1nm]

I'm just trying to remember how to work out the gain. Nigel - can you remind me?

There is a good article on measuring gain (and other things) here

NigelM

[buzz]

Nigel, thanks - I found a similar article and have set about working out the e/ADU for the Fuji. I just realised why the Fuji gave pure green in the TIFF file and the Canon did not - the Fuji does not have an anti-alias filter. Certainly the Fuji does not have such a step cut-off - the SII response is better ... it is just whether the Ha is better too. Next time a present any results, I realise I need to be more comprehensive to be of use to others.

[buzz]

OK folks, here are the results from my weekend test on the gain and noise performance of the Fuji. I followed the Craig Stark method and got a nice straight line for gain for a range of exposures (V vs. M) the slope was 0.58 e/ADU. (ISO800)

I went onto the noise test. No such luck. The master bias and 1-10min dark frames all had a mean value of 250, give or take a unit. In fact the trend was slightly the opposite - with the 10min figure having a slightly lower mean than the shorter dark frames.

(In these experiments, I had to do a Nebulosity batch convert of the Fuji raw files to FITS; The bayer pattern of the fuji is unconventional but it seemed to make no difference to the outcome of the ImageJ analysis.)

I would say that the Fuji is doing something clever so as to limit the average dark noise. The std deviation goes up with time, but the mean comes down!

So, like the reported Canon manipulations using the CCD periphery - you cannot scale darks on the Fuji. It all comes down to doing a back to back test. I will have to wait until my holiday in August.

[buzz]

I briefly imaged the Heart Nebula this week with the Canon and Fuji, using an IDAS filter and 300 second subs at ISO 800. I only managed 4 subs each before the clouds defeated me. The Fuji has definitely less noise and the faint nebulosity was better defined as a result.

I would say the Ha sensitivity was similar but since the Fuji has less noise, one can manipulate the signal to a better degree. I would also say, after my prior analysis with green, Ha and SII filters that the IR roll-off on the Fuji is not as abrupt.

I took a single 300s image of M31 whilst I was at it. The result was really smooth with good definition and bodes well for a longer experiment.

The principal drawback is the need to convert RAW files to TIFF in PS or Aperture and the lack of an external interface. Maybe Fuji will develop a USB link on its latest viewfinder-less version. For the occasion I need a wide field view, since I own the Fuji and my son has taken a shine to the Canon, I would be happy to continue to use the Fuji. I just need to buy a very long cable release!

[MooMoo]

Has anybody got information on the sensor and processor used in this camera?

I really am tired of Canon being the defacto astrophotography camera, I have one myself but with to escape the ever changing Canon mounts, unlike Nikon which is proud of not changing the mounts as often as Canon has.

What we need is an open source DSLR specification.

People have made open source GSM phones from Arduino's , I have these boards myself.

[buzz]

Well, I think all the camera manufacturers have changed mounts in one way or another many times over. It doesn't really make much difference.

I have to say the T-thread adaptor to my Fuji locks securely with no slop, which is something I cannot say about the EOS. What is a real shame is that the Fuji adaptor is really deep to maintain the 55mm spacing but also deep enough to have taken a filter wheel and OAG between it and the T-thread. I might buy a second one and see if I can get someone like SRB to modify it for me. I know some take great delight in modifying cameras and putting filters / coolers etc on them. There comes a point however when it may be more sensible, if not less satisfying, to just buy a used CCD camera.

[Gtech]

Sorry to drag up an old(ish) topic but I've also been keen to understand the characteriscs of the X-Pro 1 and X-E1 cameras.  So I thought I'd share my findings.

My analysis started as an exercise in determining the optimum exposure for metered sky glow in the field. I've since gathered a sizeable number of darks, flats and bias' to help determine what ISO/Expsoure settings would yield the best return for X-Trans astrophotography.

I adopted Craigs Starks well documented method for determining the following and used DCRAW 9.19 for RAF > TIFF conversion (since most proprietary Raw converters Adobe, C1 Silkypix do voodoo to the raws). All analysis and calcs were done with AstroArt5 and Excel :

First off, I noticed that the X-Tran sensor had similar ISO gain to my 5D2 and was much lower than the 1100D. Saturation was marginally higher than the 5D2. Calcualted dynamic range was also marginally better at lower ISOs. Read noise was also better than the 5D2.   So far so good!

There's no prizes for noticing that Fuji have applied some (Canonesqe magic) or black point dark current reference intensity scaling to long exposures, since the average levels decrease over time.

The conclusion I draw from the data is that ISO800 is the sweet spot for these two models while working with 2, 3 or 4 min exposure times.

The X-Trans sensor captures almost as much IR out of the box as my modded 1100D. While its 16MP - 4.8um pixel size and 15.6 x 23.6mm format yeilds a favourable 1"/pixel on my MN190 with a slightly wider FOV than a 1100D.

I guess the only real drawback is the lack of a Fuji SDK that could offer up future BYE or APT functionality for the X-E1.  Unlikely, although I've just sourced a wireless intervalometer (Phottix Aion) that will offer a big step forward in remote control which I will couple with an Eye-fi card for snapshot focussing.    

I'm now looking forward to some clear nights and some field trials.....

[dph1nm]

Are these numbers  in e- or ADU? The dark standard deviations seem VERY high. On my Canon 1000D the dark noise hardly gets above the read noise in 30secs.

NIgelM

[Gtech]

The average levels and SD values are in ADU. RON is in e.

[buzz]

I changed my X1-Pro for the X-T1 with the battery grip. The tiltable display is very useful.  I bought a JJC intervalometer from ebay which has a removable short lead and replaced it with a standard 1m 2.5mm jack to jack.  The only challenge left is focusing.  I tried the wifi remote idea on the iPad but there is no way to zoom in or do the kind of live view you get with EOS.

I tried the bahtinov mask again and repeated manual focusing trials and noted the step position on the focuser on each attempt. I was getting within 10 steps (40 microns) each time.