5d II HTP : tell me the truth...

<p>Hi!</p>

<p>I am currently traveling a 5d II that I acquired just before I left home.<br>

I have shot mainly at 200 iso, with HTP activated, thinking that it wont affect RAW images, as it is<br>

normally with this kind of gadgets, such as Auto Lightning and the likes.<br>

Today I was in need of some lower iso setting for the first time, and I have noticed that it<br>

was not possible to do so with HTP activated, which makes looking further and discovering<br>

that it HTP was therefore affecting raw images.<br>

Then I came across some web forums when the problem is discussed, saying that HTP was<br>

bringing noise in the shadows...<br>

However, my (very) superficial tests (I do not carry my computer, nor have access to any raw<br>

converter) don't show me a noise difference, at least in jpeg.<br>

Could you please tell me the truth about it? does it bring noise (or even worse, banding) in the shadows or not?<br>

Thank you!</p>

 

<p>Yes, In my experience it does increase the noise at any iso setting , shadows or not.</p>

<p>The question is not really "does it" but "can it?" </p>

<p>You can always introduce a set of conditions to force a behavior that you want to (or don't want to) see. </p>

<p>In other words: your mileage may vary.</p>

<p>I actually don't use it, but all of the literature does say it increases shadow noise. The reason for the iso 200 is technical (it allows less light to hit the sensor that iso100) so allows more detail to be captured in highlights. I am not sure it actually affects the raw, but may rather tag it to be processed a certain way by DPP--not sure if other processors will recognize it or not. In reading about it, it almost sounds like it might just underexpose a stop or so and then push the exposure back up--which generally will increase the noise in the shadows. Anyway, with the noise reduction software out there, including ACR's new algorythm for that, I doubt it would have a major impact on your image quality. I shoot my dsmkIII at 1600 a lot and after processing in DxO, and to a lesser extent in ACR, the noise is almost totally eliminated even in a 40x60 print.</p>

<p>So, the question might really be do you really need it? My experience is that between the lcd on the 5dII, or the dsIII, and the historgram I don't have high value blowout except where I allow it. In those cases, htp wouldn't help in any case as it is generally not a situation of a stop or so, but more like many.</p>

<p>Well, use HTP when you need it and turn it off when you don't... HTP is very useful in many situations but neutral or detrimental in many others. I use it only when I really want more detail in the highlights than in the shadows, and only with the 200 ISO setting.</p>

<p>Whether it adds noise in the shadows it is debatable and very difficult to qunatify under typical conditions becasue HTP or not, the sensor has so much tonal capacity and if the photographed scene exceeds that value, something's going to give and you might get more shadow noise because of underxposure or blown highlights becasue of overexposure. At the end of the day you are the judge so process and print a typical picture with and without HTP and (honestly...) compare the two. Just make sure that you make a good paper print of the size you'd normally do because at 100% on a monitor one can see many invisible things, including the crap from your 72 DPI monitor...</p>

<p>I am still not sure exactly what HTP does. I understand it in outline. When HTP is on, the shot is exposed at one stop slower ISO than you have set (which is why you can't use HTP below a set ISO value of 200), and in all but highlight areas the value reported from each photosite is then doubled to get you back to where you expected to be (except that this results in a noisier image than you would get without HTP, and with one bit less depth up to the highlight threshold). But above that threshold, the doubling process is rolled off, so that at photosites where the intensity is between the threshold and saturation, contrast is reduced. If you look at the DPReview review of the 60D, there is a graph showing the behaviour. What I don't know is exactly at what stage this process happens. If the raw image is created just as it would be at one stop lower ISO and the values are then remapped, then the same effect could be achieved by the user just by exposing with one stop slower ISO and applying a modified tone curve during post-processing, although it might be rather a fiddly business. But if the raw image is created directly by changing the way that the charge on each photosite is mapped to a digital value in the raw image, it might not be possible to reproduce the exact effect during post-processing. Either way, it's pretty clear that the raw file is actually modified when HTP is used, whereas all the other settings (apart, of course, from the ISO setting itself) are simply recorded as recipes that are applied when the raw image is converted to a viewable image, and can be changed if desired.</p>

<p>I agree with Michael that HTP can be useful but should be used only when needed. In addition to its value for high-luminance highlights (where each of R, G, and B is close to saturation) it is also valuable when just one channel is near saturation. Commonly this is the red channel. Although current-generation sensors are better than earlier ones, it is still possible to see a curious effect with some red fabric dyes that reflect red light so strongly that the red channel is saturated at what in luminance terms is a normal exposure level, with the green and blue channels well below saturation. With such fabrics, the blue channel is often the next strongest, and so the red-blue balance shifts towards blue, turning a bright red into a sort of pimiento red. In my botanical work I often see a red-green effect when I am photographing some of the intensely orange South African daisies. In this case the shift is from orange towards yellow, for the same reason. HTP helps to control it.</p>

<p>The probelm with HTP is that switching it on and off is extremely inconvenient. It is deeply buried among the Custom Functions. Even using My Menu, you can only register CF groups, not individual CFs, so the number of steps required to change HTP is quite substantial. It ought to be possible to do so from the Quick Control Screen; there's no obvious reason why it should be a problem for Canon to implement this.</p>

<p>I have to say that I cannot vouch for what HTP actually does to photographs either. As I understand it - and I could be incorrect - the general idea is that it allocates more data points to describing luminosity at the bright end of the scale and allocates fewer to the lower end. This might, or so I presume, allow you to retain more detail in the bright areas of the scene though it may not (or, again, so I presume) have any effect on the point at which bright subjects blow out.</p>

<p>Again going by my assumptions rather than testing, it would follow that either that values at the dark end of the luminosity scale might get lost or that there would be fewer data points available to describe them, with the theoretical result that differences in values (whether part of the signal or the noise) might become larger, perhaps creating more obvious noise in shadows and/or banding.</p>

<p>I do know that by comparison to film digital capture responds differently at the extremes of bright and dark tones. At the bright end, I see two things. One, there is a sort of "hard stop" at the point where the image blows out and essentially all detail is lost - with film (exceptions noted) there is more or less a roll-off curve at this end of the tonal scale. Second, and related, values just below the blown out point seem to be not quite enough different from the maximum value, and it can be difficult to retain highlight details. (In post I often deal with this in highlight areas with some photoshop techniques involving curves constrained to the brightest values.)</p>

<p>So, after all that blabbering... I don't use HTP, and I don't think I need to. In those situations where it might have an effect, the data are still there (I think!) in the raw file and I don't have any problem getting the image I want in post.</p>

<p>Dan</p>

<p>If you use HTP at ISO200 you will see little increase in noise. However at ISO 800 lower mids and shadows start looking like grains of friggen sand. At 1600 they look like dad burn pebbles. I would never use HTP as a default but might use it at ISO 200 for an emergency sunny day landscape.</p>

<p>Thank you all for your kind answer!<br>

I shoot raw + small jpeg. Jpegs are really just used to have some pics to be sent to my friends and familiy<br>

easily, from any computer I come across when traveling. All my serious work is made from Raw.<br>

Now I really think that Canon could have warned users that HTP was affecting raw as well.<br>

I think it's crazy that you have to search the web to get know this.<br>

I am looking forward, a bit anxiously, to seeing how DXO 6 will manage this.<br>

Have a nice w.-end<br>

Chris</p>

<p>Christa:<br>

Nowhere in the camera manual Canon writes that HTP is a cure-all panaceum and that HTP should be left on all the time.</p>

<p>The fact that all RAW files are processed is given (HTP or not) so if you have a notion that RAW is something "straight off the sensor" you are wrong: the in-camera software does a lot of work interpreting the data obtained from individual photocells/pixels and that's why improvements/changes in the camera's firmware can dramatically affect/improve the quality of RAW files without any hardware changes. The fact that HTP will remap the RAW file is actually AFAIK the only way (at least to date) to achieve the desired result, i.e. effectively underexpose the highlights in relation to the shadows to prevent "blown highlights" and lack of shadow detail (if the entire picture was exposed for highlight detail shadows would be underexposed as well, resulting in even less detail and more noise.) I don't know the internal architecture of Canon hardware and software well enough to judge whether the actual file remapping is really needed - or simple tagging would suffice - but I gather that, at least for now, there is no other way. Maybe in v. next...</p>

<p>And, even with the best tools, imitating the HTP in post is very, very difficult, bordering the impossible. Yes, you can bring up some shadow detail and change the density of the highlights but you will pay dearly in noise and tonality so why not use the HTP functionality when you are likely to benefit from it (as in the proverbial black tux/white wedding dress situation) regardless of any real of perceived noise artifacts, because you are likely to get way more artifacts while trying to remedy the situation in post...</p>

<blockquote>

<p>I think it's crazy that you have to search the web to get know this.</p>

 

</blockquote>

<p>Canon - and Nikon, and Sony, and Pentax, and the rest of 'em, <em>all </em>keep the workings of the "selling points" secret.</p>

 

<blockquote>

<p>I am looking forward, a bit anxiously, to seeing how DXO 6 will manage this.</p>

</blockquote>

<p>Not sure what you're saying here: DxO will not be able to see the metadata tag that tells DPP to HTP process the file: all it will see in a c. 1 stop underexposed file that you will have to fix yourself.</p>

 

<blockquote>

<p>if the entire picture was exposed for highlight detail shadows would be underexposed as well, resulting in even less detail and more noise.</p>

</blockquote>

<p>Michael, if HTP in the 5D Mk II is anything like HTP in the 40D, then the entire file is underexposed by about a stop - just look at an HTP'd RAW file in something than DPP and this much is clear - and then "pushed" by the HTP process (nothing much more fancy than a curve, as far as I can tell) back to "normal" exposure, which is precisely why the shadow noise increases.</p>

<p>I actually find it easy to mimic HTP myself, and frankly can do a better job of balancing the shadows and highlights, because I can apply <em>per file</em> adjustments, rather than the in-camera, one-size-fits-all solution HTP provides.</p>

<p><em>And under no circumstances would I try to manage highlights by first underexposing the file by a stop... </em></p>

<p>Hmmm...I don't see a stop underexposure with HTP on 40D, 1Ds3, 1D3 and 1D4 (I'm not familiar with 5D2 other than for video.) What I see OTOH is way more highlight and unaffected shadow detail with HTP (in the proverbial traditional wedding situation, there is detail on a black tux, seams et all, and detail on the white wedding dress, and no perceptable noise @ ISO 200 with 1Ds3.) One caveat: HTP requires good exposure (gray card here I come...) so the best results are obtained under controlled conditions, say, with the same light setup for wedding formals.</p>

<p>Again, personal likes and dislakes aside, HTP IMO is a valuable feature when used judiciously.</p>

<blockquote>

<p><em>And under no circumstances would I try to manage highlights by first underexposing the file by a stop... </em><br>

<em> </em></p>

</blockquote>

<p>If one or more of the R, G, B channels is saturated (at the 14-bit level to which raw data is recorded after A/D conversion on current Canon bodies), what alternative is there to reducing exposure to (just) desaturate all channels and then manipulating the tone curve?</p>

<p>Whatever the exact workings of HTP (see my earlier post), this is in effect what it is doing for you, by one stop (which may, of course, be too much or too little), and in the range below the highlight threshold it is then compensating by doubling the value. I recall that someone actually unpicked an HTP RAW file to investigate this, and found that all the values below the threshold were even. This means that, regardless of effects on noise, you are losing one bit of colour depth precision below the highlight threshold. However, a reduction from 14 bits to 13 bits is unlikely to be obvious, even under the demanding conditions of wedding photography. Since displays are mostly 8-bit (or even 6-bit on laptops) apart from a few very high end 10-bit displays, and most printer drivers are 8-bit, you would probably need to be using a 16-bit printer driver to have much chance of seeing any difference. </p>