Canon may be expensive but...

[CR Backup Admin]

Users are now reporting this thread because of the personal accusations.  Calling another names is not professional and degrades the forum for everyone.

Please agree to disagree and move on, arguing back and forth solves nothing.

[sanj]

Come on!

[Neutral]

Was reading this thread and could not help but putting some comments.
Let's look at the root of the discussed HTP issue 
What is the blown highlights means for the sensor? 
Really blown highlights ??
This means that photodiode of the photocell got saturated and any increase in number of captured photons cannot increase current via this photodiode .
Anybody is going to argue about that ?
Then what could be the remedy for this?
Answer is simple - reduce the number of photons.
How this could be done?
Use EC to  -1  and then later in processing  change gamma in tone curve to raise mid-tones and shadows  to the level which would be produced if camera  was used with EC=0.
This could be done manually or camera could do that automatically using HTP mode but reporting that all was done with EC=0 , actually cheating camera user.
This all simple physics - anyone could see response curve of the photodiode and understand what it is all about.
It does not matter how you name what is the happening but essence of the process does not change with this.
If photodiode is saturated no further on chip electrical post processing could recover highlights - they already blown.Full stop here.
This is ABC of electical engineering.

[neuroanatomist]

@ Neutral - Thanks, and of course, that's absolutely true if the blown highlights are due to saturation of wells with photons prior to application of the user-selected amount of gain (but note that some gain is applied even at ISO 100, since the 'base ISO' of most current sensors is actually a bit less than 100).

But highlights can also be blown by less-than-full photon wells being subjected to too much analog gain as they are read out. That can happen at higher ISO settings (which are often needed for action-stopping shutter speeds or narrower apertures chosen for sufficiently deep DoF at handholding shutter speeds).  In that case, simply reducing the gain (i.e. ISO) by one stop will preserve one stop of highlights, and that's what HTP does at ISO 400, for example.  No change to aperture or shutter speed, so no change in the amount of light hitting the sensor, but rather, a (clandestine) reduction in the analog gain.

[TheSuede]

Actually, what Neutral wrote isn't really true.
A well constructed sensor isn't even close to saturating the charge well when the raw ADU reaches maximum, at base ISO. Constructing a sensor that way would mean some serious linearity problems.

The reason why sensors are made to behave like this is that the charge well capacitance behaves like any other capacitance, the potential difference has to be above a certain threshold for the transfer from cell to well to be linear. If you map the ADU/photometric exposure relationship at base ISO you'll see a slowly decaying rate as you approach blown values in the raw file. This rate of decay can tell you how much bigger than the maximum allowed translated digital value the well really is, and it's usually bigger by about a factor of 2.
So no, no modern sensor "blows to white" due to oversaturation in the charge well. The point where that happens is more than 1Ev past blown white in the raw file at the lowest analog gain ISO ("real" base ISO).

The sentence or meaning of "blown white" always relates to the base data you're looking at, no matter what it is. In a jpg, something blown is pegged at 255. In a raw file something blown is pegged at some value close to 15-16,000 depending on ISO and what channel you're looking at. And no, that value isn't 16383 as it "should be" in a 14-bit file, most cameras don't "fill" the raw file - Canon usually use a real raw value range ADU of either 1024-15,000 or 2048-15,000.

[Neutral]

@neuroanatomist & @TheSuede :

I think you are both correct but there are two aspects here.
 Hopefully modern sensors are designed in a way to prevent photocell saturation in normal conditions and at base ISO saturation point of photocell (or point just below) should match upper level of input voltage range of correspondent  ADC. This is required to obtain maximum possible DR from the sensor at base ISO.

 With lower light input (so less input photons) higher digital ISO should increase analog circuit gain so less photocell output should be amplified to the same upper input voltage of ADC.
This also explain why sensor DR is reducing 3db with ISO going up one step (for properly designed up do date modern sensors - e.g. from Sony you see that starting from base ISO).
 So for HTP at ISO higher than base ISO negative exposure compensation could be done just by reducing analog circuit gain.
But at base ISO this might not work - we still would need to reduce number of photons by reducing exposure time or using ND filters to prevent photocell saturation.
 Unfortunately there is no freely available information for sensor cell full path gain distribution (from photocell up to ADC) to see actual sensor performance in this respect (for different sensors). This would be interesting to see - actually this is one of the most important parts of the system design.

[neuroanatomist]

So for HTP at ISO higher than base ISO negative exposure compensation could be done just by reducing analog circuit gain.
But at base ISO this might not work...

In fact, that is precisely the reason that with the HTP function enabled, the user cannot set the camera to ISO 100 (although technically, that's not really base ISO, which is in the ISO 70-80 range).  When HTP is enabled, ISO 200 is the lowest user-selectable ISO, so the camera is able to reduce analog gain by one stop at any user-selected ISO setting (the expansion ISOs where digital gain is applied post-ADC are also unavailable).

[dlleno]

If I may make an observation of the way my  mind naturally imposes a context to the HTP discussion:

this forum  is about Canon cameras and the way they work.  With few exceptions the reason people come here is to exchange information about Canon cameras, present and future to be sure.   Therefore, when I read the expression "the mechanism of HTP" the primary context is, quite naturally and automatically, "what happens inside my camera when I enable HTP".   It means "how are the principles of headroom and ISO gain applied inside my camera".  It means "what effect does enabling HTP have on the way my camera behaves".

So with the behavior of Canon Cameras paramount to the discussion, I don't see why it is so difficult to think of Canons implimentation of HTP in  these terms:   

Set your Canon camera up to take a picture at ISO 400. 
pause. 
apply cyanoacrylate to the ISO dial -- just not to the electronic gain structure inside the camera .
take the picture
pause again.  look at the dial.  whew it is still set to 400.
enable HTP
look at the ISO dial again.  yep, the cyanoacrylate is still there.  ISO still set to 400
take another picture.  how did the camera behave differently under the influence of the HTP setting?

THAT, to me, is the "mechanism of HTP".  Reverting back to the general application of base ISO headroom, the special case of ISO 100, the benefits or characteristics of non-Canon cameras or sensors -- none of that helps describe what Canon cameras do under the vast majority of use cases when HTP is enabled. 

Furthermore, a discussion on the differences between Canon and Sony sensors (and the supporting electronics), as regards their respective behaviors at high levels of light (near saturation) and how to advantage each sensor to the given situation is engaging to be sure ,but doesn't help us understand how Canon Cameras behave differently when HTP is enabled, compared to the way they behave when HTP is not enabled.

[neuroanatomist]

Quite the price, the Canon counterpart is priced better true.

My heart sinks. Canon may price their next 800 to match this...

Back to the actual topic of this thread...

If/when Canon releases an updated 800mm 5/f.6L IS, it may very well be in the same price range as the Nikon 800/5.6, if not higher. 

Consider the MkII versions of the 500/4 and 600/4 lenses.  Look at the selling price of the MkI versions at the time the MkII's were announced, add in the price increase that Canon applied during the long delay between announcement and availability, and you see that the MkII lenses were a 45-50% increase in price.  Based on the current 800/5.6 price, that gives a range of $19-20K for an 800/5.6 II.

And don't expect Canon to throw a TC into the deal...   

[wickidwombat]

Quite the price, the Canon counterpart is priced better true.

My heart sinks. Canon may price their next 800 to match this...

Back to the actual topic of this thread...

If/when Canon releases an updated 800mm 5/f.6L IS, it may very well be in the same price range as the Nikon 800/5.6, if not higher. 

Consider the MkII versions of the 500/4 and 600/4 lenses.  Look at the selling price of the MkI versions at the time the MkII's were announced, add in the price increase that Canon applied during the long delay between announcement and availability, and you see that the MkII lenses were a 45-50% increase in price.  Based on the current 800/5.6 price, that gives a range of $19-20K for an 800/5.6 II.

And don't expect Canon to throw a TC into the deal...   

I bet higher, and made from titanium and carbon fibre...