Mike Engles

Lives in United Kingdom United Kingdom
Joined on May 24, 2006

Comments

Total: 57, showing: 1 – 20
« First‹ Previous123Next ›Last »
In reply to:

SteB: The first thing I noticed was that in the A7 III pixel shift images there is clearly visible lateral chromatic aberrations which are absent from the A7 II images. This either indicates the use of a different lens, or that the A7 II had some sort of image processing to remove CA, which itself would have influenced colour rendering.

Yes, the eyes of all the subjects show a curious colour fringing, not visible in all the other examples

Link | Posted on Nov 13, 2017 at 09:33 UTC
On article Cinematic 4K footage shot with the Apple iPhone X (308 comments in total)

As someone who does not do video, just stills with a 5D3, I can only say that the film does look really very good, sharp, well exposed and pretty convincing. I don't own a Apple or anything Apple and never will
MDE

Link | Posted on Nov 10, 2017 at 12:08 UTC as 35th comment | 1 reply
In reply to:

pro photo 2011: This is film SLR camera nostalgia going overboard.

There are plenty of secondhand film SLR cameras on eBay if you feel like shooting film sometimes. But let me assure you that you won't be doing that too often. Film is just too troublesome. And remember, ISO is not adjustable in a film camera. You'd have to finish that roll before you can change to another film with different sensitivity!

Changing bags can get you out of trouble if you as I used to do loaded your own film into spools. If on the rare occasion the film is over wound and parts company from the spool, a changing bag is a saviour, allowing you to reattach the film to the spool, put it back in the cassette and then rewind. This once happened to me in Whitby and I did not have a bag. Luckily I knew that the eminent British Photographer had a studio and museum in the town and I was allowed to use his darkroom to reattach and rewind my film.

Link | Posted on Nov 8, 2017 at 22:40 UTC
In reply to:

pro photo 2011: This is film SLR camera nostalgia going overboard.

There are plenty of secondhand film SLR cameras on eBay if you feel like shooting film sometimes. But let me assure you that you won't be doing that too often. Film is just too troublesome. And remember, ISO is not adjustable in a film camera. You'd have to finish that roll before you can change to another film with different sensitivity!

Perhaps it will also increase sales of 'changing bags'
If you don't know what a changing bag is? It is a black bag large enough to take a camera, which has two sleeves, that gather around the wrists of your hands. You put camera and film backs plus a blanking plate, if they make one into to the back, then your hands and do the necessary.

Link | Posted on Nov 8, 2017 at 10:54 UTC
In reply to:

pro photo 2011: This is film SLR camera nostalgia going overboard.

There are plenty of secondhand film SLR cameras on eBay if you feel like shooting film sometimes. But let me assure you that you won't be doing that too often. Film is just too troublesome. And remember, ISO is not adjustable in a film camera. You'd have to finish that roll before you can change to another film with different sensitivity!

Hello
Yes film is fiddly, but it is not a great deal more fiddly to change films between rolls. You just need to work out how to sync the start of a film reliably and then wind on the requisite number of frames plus one. Nikons had very precise winding.
MDE

Link | Posted on Nov 7, 2017 at 19:55 UTC
In reply to:

M Jesper: Still using LR5 and CS6 on the latest version of Windows. I don't think LR6 will just 'break entirely' any time soon.

Hello
It might not be the OS that breaks Lightroom, but the Adobe servers that underpin the licencing and codes

Link | Posted on Oct 27, 2017 at 14:20 UTC

The latest update on offer is 6.13, so what is this final 6.13 update coming later in the year?

Link | Posted on Oct 27, 2017 at 08:50 UTC as 27th comment | 1 reply
In reply to:

Marty4650: Interesting stuff.... but still not worth $10 per month for the rest of my life.

They have to do better than this if they want me on their subscription plan.

Interestingly Exposure X3 does not do 'AUTO' and I cannot find a RGB info.
It is an interesting programme and looks very Lightroom like

Link | Posted on Oct 22, 2017 at 13:50 UTC
In reply to:

stratman1976: Okay Adobe, now fix the sluggish LR performance already!

Best thing is to buy Adobe shares and the dividends will pay for the subscriptions

Link | Posted on Sep 20, 2017 at 19:48 UTC
In reply to:

Richard Murdey: As noted, you get fine steps at the highest brightness the ADC range accepts, and 1/2 fewer intervals for each EV below that.

Even at 14 bits though, very few digital steps available for very low brightness signals. The first three EV are digitized at 2, 4 and 8 steps, and 16 steps is probably a bare minimum to acceptably resolve smooth tonal variations.

Starting from the high end, you likely keep the top two EV as highlight protection, then pack your actual image in the 5-6 eV range that follows, so the bottom range of your data is recorded in 1/64 or 1/128 intervals at 14 bit, or 1/4 that at 12 bit. So everything is fine. But with high DR scenes using ADL or similar DR enhancement, or when pushing underexposed areas manually, 12 bit gets ugly fast.

(aside: are we sure the amplifier used to record the sensor output is linear and not logarithmic? If nonlinear all this discussion goes out the window...)

It brightens them, but relatively less so than the highlights which can be torn apart if there are not enough bits. Imagine a balloon. with evenly spaced dots from the knot end to the apex. inflate the balloon and knot it. All the dots spacings are changed and get raised, but the highlights are particularly affected and gaps can appear in 8 bit files. High bits fill in the gaps. If you do not have enough bits and then try and use curves to bring out the tones in a sunset, posterisation is extremely likely. This is the problem with try to process 8 bit jpegs. If you really want to do extreme processing in RAW, ideally 16 or 18 bits are necessary so that enough bits are allocated to the higher tones. They way round this is to apply the gamma to the RAW analogue data before the A/D. That way all tones get the same number of accuracy. If they did that only 12 bits would be needed. It has to be said that no manufacturer does that.

Link | Posted on Sep 4, 2017 at 10:51 UTC
In reply to:

Richard Murdey: As noted, you get fine steps at the highest brightness the ADC range accepts, and 1/2 fewer intervals for each EV below that.

Even at 14 bits though, very few digital steps available for very low brightness signals. The first three EV are digitized at 2, 4 and 8 steps, and 16 steps is probably a bare minimum to acceptably resolve smooth tonal variations.

Starting from the high end, you likely keep the top two EV as highlight protection, then pack your actual image in the 5-6 eV range that follows, so the bottom range of your data is recorded in 1/64 or 1/128 intervals at 14 bit, or 1/4 that at 12 bit. So everything is fine. But with high DR scenes using ADL or similar DR enhancement, or when pushing underexposed areas manually, 12 bit gets ugly fast.

(aside: are we sure the amplifier used to record the sensor output is linear and not logarithmic? If nonlinear all this discussion goes out the window...)

You really need to think about what inverse gamma 2.2 does to the file, where in essence the distribution of lows and highs reverse. Your camera or RAW editor will always apply that curve, otherwise to you do not see very much.

Link | Posted on Sep 3, 2017 at 09:06 UTC
In reply to:

G Sciorio: This is why Micro 4/3 is kinda stuck. Unless there's some major advancement that is exclusive to smaller sensors, 'the little system that could' will become 'the little system that could not'.

Noise in sensors is due to heat. Cool the sensor, less heat.
You might have less heat if the camera were white instead of black, less heat absorbed. You could have small fans, or some heatpipe arrangement with fins or a mini refrigeration box or to really make a difference, liquid nitrogen cooling!

Link | Posted on Sep 2, 2017 at 12:38 UTC
In reply to:

MrBrightSide: This is troubling. If 12-bit only has room for roughly 12 stops and 14=14, does that mean we will be stuck at 12 or 14 stops forever unless sensor manufacturers come up with a 24- or 36-bit sensor?

Well as far as I know Quantum effects are probabilistic and those are not digital as you can have fractions. If they were truly digital, why have a ADC converter as they would already be in the digital domain.
You are very correct in mentioning noise as this masks the dectectabilty of low photon counts and the effect that since we cannot YET detect these, there in no correlation with photon numbers and sensor output. So at the very low end non linearity is apparent but not necessarily so. Once a critical mass of photons are detected, to over come the noise threshold, there is I think a direct correlation between number of photons and the sensor output and from then on the sensor is linear, till it is overwhelmed.

Link | Posted on Sep 2, 2017 at 11:53 UTC
In reply to:

MrBrightSide: This is troubling. If 12-bit only has room for roughly 12 stops and 14=14, does that mean we will be stuck at 12 or 14 stops forever unless sensor manufacturers come up with a 24- or 36-bit sensor?

Well as far as I know Quantum effects are probabilistic and those are not digital as you can have fractions. If they were truly digital, whyhave a ADC converter as they would ready be in the digital domain.

Link | Posted on Sep 2, 2017 at 11:12 UTC
In reply to:

Richard Murdey: Isn't the point though that for 12 bit, the lower light ranges are encoded at very poor resolution, say from 8 to 16, i.e. 8 steps for 1 EV change in light intensity, whereas for 14 bit that same range of light intensity would be encoded from (using your diagram above) using the range from 32-64, i.e. 32 steps or 4x the "smoothness" of the gradation?

That's why 12 bit looks so often more pixellated on very gradual color transitions.

That always seemed the reasonable explanation to me.

What needs to the mentioned is that an inverse gamma curve of about 2.2 is applied to the RAW data in your camera to view the image and also in your RAW converter. We have to have this because human vision is non linear.This has the effect of totally stretching the mid tones and highlights. The data belonging to those areas can be so stretched, that especially in the sky tones, posterisation will occur in an 8 bit file'if you do not have high enough bit depth. Think of an balloon and mark dots evenly with pen along its length. Now blow up the balloon, knot it and look at the dot spacing. At the knot end the spacing will be pretty good, but as you go up to the top, the spacing will be very uneven. High bits will fill in the gaps. The problem is that after the inverse gamma is applied the top end does not have enough bits and the bottom has too many. The way to get round this is to apply inverse gamma before the ADC to the raw data readout from the sensor. Analogue TV worked like this

Link | Posted on Sep 2, 2017 at 09:56 UTC
In reply to:

MrBrightSide: This is troubling. If 12-bit only has room for roughly 12 stops and 14=14, does that mean we will be stuck at 12 or 14 stops forever unless sensor manufacturers come up with a 24- or 36-bit sensor?

Sensors are linear and analogue, so can produce an infinite number of levels and tones as is natural light and only limited by the inherent total noise in the sensor and readout circuits. Active cooling will really reduce sensor noise, perhaps soon we will have a small refrigeration plant to carry around for active sensor cooling!

The conversion to digital can be any number of bits, but would then have very large files and difficult to process. The trick is to have enough bits to cover the usable information you have. A high number of bits do allow very fine variations in tones to be recorded rising to near the infinite in analogue capture. High bits also allow one to be pretty sloppy about exposure as we also have a lot of spare information. So if we have underexposed by 5 stops, to preserve highlights, we can apply a linear gain to shift the whole spectrum We can also apply a non linear gain (curves) to favour parts of the range,to open up the tones of a beautiful sunset.

Link | Posted on Sep 2, 2017 at 09:36 UTC
In reply to:

Dodi73: Interesting article on a math point of view. Unfortunately it's a bit too high for me,and for this reason I'd like to ask you (maybe it's obvious for you, not for me) why camera RAWs always have 14 bits as a maximum limit (is it a PHYSICAL, i.e. sensor can't be designed to get more, or PRACTICAL, that's is useless to go beyond 14 bits?) Will we ever have 16,18,20 (or even 32) bits cameras or not ?
I beforehand apologize for the dumb question, but this is something I have never understood and I sense this is the right place to ask for. Thanks in advance

It all depends on the basic noise of the sensor.
If you could cool your sensor with liquid nitrogen, the base noise floor would be much lower and that is another way of improving the shadows. Heat equals noise and worse noise floor. The dynamic range would be increased, but the fineness of the tonality would be the same in a 12 bit system. A 20 bit system, would give a immense dynamic range and the gradation of tones would also be immense. The whole thing would be very forgiving of any carelessness as there would be a huge amount of redundant data, but also huge files to process. There has to be a trade off.

Link | Posted on Sep 1, 2017 at 22:17 UTC
In reply to:

mpgxsvcd: Something the article doesn’t mention is that if the scene you are taking a picture of doesn’t have 12 stops of dynamic range in it then it really doesn’t matter at all if your bit depth is 12 or 14 bits.

I see a lot of people take a simple landscape picture and say it has lots of dynamic range. However, when you look at the histogram of it neither side is anywhere near the edges of the histogram.

Yes the more bits you have the more forgiving a system is. You have redundant data and you can then 'range' the data to enhance particular areas, without adversely increasing noise if the correction is linear (gain) or non linear (curves)
If you have a high bit depth you can also expand the mid tones and highlights,to better depict the varied tones of a sunset, without creating gaps or spikes (posterisation) in those tones. The fewer the bits you have to play with, the greater the precision you have to achieve in capturing the data in the first place. That is what the great photographers did with their zone system

Link | Posted on Sep 1, 2017 at 22:04 UTC
In reply to:

craigvolpe: Probably a dumb question, but if linear response creates excess information in highlights, why don't they make sensors or image formats that map light logarithmically?

They did in the analogueTV days as the camera and the tv acted in a non linear way. If that happened in a digital camera and the analogue data from the sensor was gamma encoded before the ADC, one would only need 8 bit for correct tonal separation, but the signal to noise would be pretty bad especiall if we had underexposed and increased the gain(exposure). 14 bit encoding would shift the noise floor way down, so we would have much more leeway in exposure accuracy at capture and we would gain a lot of fine resolution in all the tones of the picture. Remember the sensor captures an infinite number of tones, gamma encoding before the ADC, will still give an infinite number of tones and having optimised the sensor output to match human vision, 12 or 14 bit information, would give us the best of all worlds.

Link | Posted on Sep 1, 2017 at 20:54 UTC
In reply to:

sirhawkeye64: Curious as to how well this works? I have probably some 7000 slides I'd like to get scanned somehow (no joke, I really do have around 7k slides). This probably wouldn't be the best solution though, but interesting and good for someone who has maybe 100 or so. Not so great if you have more than perhaps a few hundred. And unfortunately, this is a device that you'd probably scan your slides, and then have very little use for it afterwards. Unless it was good enough for you to start charging people to scan their old slides.

Hello
That is what I once thought, I only did about 400.
MDE

Link | Posted on Aug 25, 2017 at 19:58 UTC
Total: 57, showing: 1 – 20
« First‹ Previous123Next ›Last »