very happy with this unit
very happy with this unit
It is not equal in performance, as a Toyota Auris is not a Ferrari.Originally Posted by Jim Andrada
It is equal in results in most of the usual cases, as both cars above have same performance in a traffic jam. You have to go to Indy speedway to make a difference as for normal roads you can violate traffic law the same with both. Are you to print 2400 dots per inch? a Lightjet prints 300...
It is easy to proof that for the above enlargements there is no difference in MTF of USAF 1951 target, nor it cannot be separated by human eye. And mood it will depend the most on human operator that processed tha image in order to reduce DR to the output medium, if we display all the excessive DR of a neg to paper we'll get something flat.
Here there is the difference: What Ansel was doing whith hands for Moonrise prints. Now this is done with curves and Under/Over expose tools. For this reason a professional scan looks better, becasue scanist do that all day long, like Graniere did it for Salgado.
Then it comes the thing, a FM2 it's a "worse" camera than an F5, but you can take better shots with it, more difficult, but deppends more on the operator.
If I had a Pro scanning service bussiness I'd buy a Hassy X1 or X5, $10k or 25k it's nothing for a bussiness. With the X1 it's faster and perfect, you don't mind it if comes a Velvia with 3.8D shadows there.
For a private usage, I can bet I can obtain 100% of the performance needed with a V750, and equal mood in most of the cases, for usual enlargements. Exeption 135 and underexposed Velvia.
It is more about confort, holders etc than real performance, what counts: the scanits (and I'm not a good one... but there are...)
Oh... this is a matter that passionates me, I'm a hard core defender of movie film stock, "Taliban" supporter of JJ Abrams, Mendes, NOLAN, TARANTINO, SPIELBERG !!!! I see every one of their pictures shot in film just to delight technical excellence from Vision 3, I cry like a baby when I see every well tecnically executed scene, this is orgasmic...
What can do Alexa in front of a IMAX film camera? Nothing !!! just it can help making preshots to prepare real shooting. Alexas were stored when JJ arrived with Dan, the stars, the tech crew and the Vision 3 65mm cans. This was to shot a $2 Billion box office release, plus $3Bn in merchandise. Can't they afford Alexas ? Yes, they can, very useful for preshots !! Only Vision 3 can capture all Daisy Ridley beauty, not the Alexas, by no means, and there is a technical demonstration for it.
But Rogue One (Dec 2016) "A star wars story" is being shot now in Alexa 65, really a good camera. One day Alexa will emulate Vision 3 perfecty, not for now.
Star Wars 8 (Dec 2017) is being shot now in Vision 3.
That's the answer: They do not keep DR, as they they don't keep it from an Alexa.
DCI (https://en.wikipedia.org/wiki/Digita...ma_Initiatives) recommends 48 cd/m2 projection performance, a shame. They project shadows but no light, much inferior that old good projection film. TVs have 300cd/m2 even 1000. Also Alexa has near same DR than V3.
Vision 3 is used by Disney/Lucasfilm because some advantages:
1) Stars look pretty with V3, look Daisy Ridley !! Take an Alexa and try... if you want I can explain it, it takes a bit.
2) Uncompressed 64k IQ raws, Alexas are 4k or 8k, Disney will sell Star Wars 8 in 2040 with the broadcasting quality that digital crap can have by then. (Note I use "crap" ironically : )
3) Smooth highlight roll-off, beter volume depiction than ends in viewer's immersion, the the viewer becomes a witness, beyond a viewer.
No, not irrelevant at all. Just I showed that HDR blending of images was used as an universal digital tool, it is used in a consumer DSLR and in a high end Alexa, this was the point, and for this is also useful with flatbeds and Drums.
But one thing.. not very useful for BW if silver was developed until 2.5 densities,a common situation, then the silicon works perfectly alone, With Velvia underexposed and deep shadow... then please use Multi-exposure, because silicon's DR do not reach required performance.
Thanks, for discussing this with me, lets remove personal involment and let's go to technical discussion, I'm the first that I want to learn and to know if I'm mistaken, with science and reality proofs.
This is subjective, let's separate things. So why ?
Because tonal depiction? microcontrast? image ressolution? grain depiction... ? Or we don't know why?
I'd like concrete claims to discuss it. All of we are used in the technical analysis of images, we know the source of every blur component, the sironar's S aberrations, in corners, in center, the film MTF, the shake, the silver solvent, the enlarger vs contact, etc, etc. So let's go to the technical points as we do when comparing a N to an S glass, as "Bob" do
may we? or we are going to be tied to the "I like this more" argument ?
tonal depiction, microcontrast, image ressolution, grain depiction, please let's later discuss in those terms
Scanning is a very deep science, nobody scans perfectly.
For example: http://petapixel.com/2015/02/21/a-pr...ith-photoshop/
did you know that with this technique one can match drum ressolution performance with a bare V750 ? (DMax apart, applied to BW film, not velvia)
And that it takes less effort than a wet mounting if PS automated?
Not at all, I master Kernel IOPL0 programming, since DDK to WDK, but I'm not to make a driver for V750, I just hook the process before the digital chain. Modifiying illumination, and making 2 scans, each with silverfast multi-exposure, to end blending both by an utility program that I'm just started writting. This is appart of our discussion of a standard V750 vs other.
I been using DDK and WDK, for RT I use to hook IRQ0, with 8254 throwing 200KHz to IRQ0 and passing control to Win task scheduler only 100Hz, or what set in Multimedia timers with timeBeginPeriod, https://msdn.microsoft.com/en-us/lib...(v=vs.85).aspx
Also I hook IRQs from a driver to timestamp events, also I triger flash/camera from a RS232 Pin and use performance counters to timestamp, thread is working in a single core (affinity mask) as each core can have a different count. I've some lead under skin with these things, perhaps you have more...
I've some 1million hand written C++ lines in my back, (hand written, not compiler counter) most in machine vision from scratch or with OpenCV, 90% of time using Point Grey cameras. I do not writte bugs to go fast. This is my tech background. Ask me about any C++ compicated topic and I'll answer it in seconds, polymorphism, casting, MFC, machine vision algorithms, how every OpenCV function works.... anything you want to check I know... I don't know all, of course.
You said multi-exposure was not useful for V750, I replied it to show that HDR image blending is a powerful digital tool to expand sensor DR, in a cheap DSLR it blends images of 2 different captures, with the Alexa it is the same shot passed by 2 amplifiers in parallel working at different ISO, I pointed the Alexa to show that this tool is universal in digital image enhancing, from a consumer DSLR to an Alexa, including Flatbeds and Drums. Of couse it is not necessary for V750 with 2.5D developed BW film, as silicon sensor performs well. It is useful if you bring TX to 3.0D and want the most detail in lights, and very useful for velvia/provia if deep shadows, even the V750 will not perform well there, no Flatbed will perform well... X5 will perform, not your flatbed...
fixed exposure is not hardware, but software, "Epson Scan" soft don't do it, Silverfast do it, they have a Patent.
I said another thing, I've dismantled a V750 and I replaced the cold cathode illuminator by a LED illuminator that has 4x the power, for the moment I only made preliminar tests, it heats up, film and box. When it finished I'll make an scan with the regular illumination power (with silverfast multi-exposure) and another one also with SF M-E. Then I'll have 2 images I will autoalign and blend. This is because I've Velvia 8x10 sheets to shot, and 5x7 frozen Ektachrome. Also I've personal project to depict with 8x10 Velvia some landscapes I love, and the modified V750 is part of the project, to not depend on drums. We'll see if it will work ok, I'm in it.
I was not using % as enlargement ratio but as surface ratio (crop vs image surface) in order you can calculate how big the print has to be to show that detail, this is the hammer hits made by the flinstones, 6 centuries ago.
It is a general frame of a 50m high church, if V750 can depict (see crop) every flintstone hammer hit on every stone, at what print size image quality from the V750 is ot enough ???
Please, take calculator and tell me your opinion...
If you scale full image with the same crop magnification the full image will have 5m high, make the calculation. If you look it at reading distance... yes you'll see that the Creo performed a bit better, not very much. But if you look the monster 5m high print from say 3m it is physically impossible you notice it, and there are not people with "sight score" 400, people has 80 or 140, not 400. And for a 2m print size absolutely no difference from V750 to a Creo or an X5.
Look, for TMax 100, if your print size is 3x4cm (in a newspaper) you are not to notice if it comes from a 135 or a 8x10" negative, or a ULF camera. Just the same.
At 1200DPI it cannot be noticed if it is scanned overglass or at the optimal heigth, because circle of added confussion is much smaller than 1/1200 inch, I calculated it to be 1/2000 of an inch, I've pending to measure it well with a USAF 1951 target and MTF, as I plan to scan velvia 8x10 overglass and I want to know exacty what impact has overglass, and know if I have to make a custom holder with side clamps to get the neg tight in place or not.
1200 dpi is a resolution than V750 works exactly the same than a creo or an X5, no absolute difference, impossible. At 3200dpi, you will notice it. At 8000dpi the X5 rocks, but anyway film MTF is the main limitation at 8000, if there was no shake or subject moviment at 1/15
Have you tried it? Take the 8x10 and scan it V750 and Creo at 1200 dpi, then put it in PS layers, auto-align, then substract images, then move curves a bit, contrast etc to match, this until you see a near all black after substraction. Then compare crops.
(The reference image in the autoaling has advantge, the autoalign-substraction is find contrast-cruve settings than makes images match in tone gradation, this to discard ICC issues)
I bet you'll see absolute nothing of nothig: Nothing.
If you scan at 3200dpi you'll notice something, but at 4m height, perhaps.
Make the test... !!!
BTW, nice shot, superb, I like it a lot.
Are we sure Mustafa has not came back as someone else. holy shit I have to leave this discussion, hope it turns out well.
Of course...
I'm roman catholic, only extremist with film : )
I just want to turn it interesting, going to real facts, and throwing science light on it.
Then go out and make some photos. I don't get how you can live in a gorgeous country and waste your time writing 1 million lines of c code in order to get a slight improvement from a consumer flatbed scanner while you could be out making art.
I really think you're lost, Pere. Your message is clear: you like tinkering with your V700, someone else prefers a Creo and I myself stick with my 4990 until it dies on me (scanning 8x10 off the glass - sue me). It's different for everyone. You don't have to sway them to enjoy doing what you do.
Having said that, this is a forum dedicated to photography. If you want to discuss the intricacies of driver programming or the physics of machine vision, I'm sure there are great places for that. This is probably not it.
I'm new to 8x10 and I'm preparing for it, all DIY, from developing to scaning the complicated Velvia, with E-6 in the middle. I'm almost ready, for the moment I've some spare time to discuss and to test my points of view aganist better photographers than me.
I'm not lost at all, my message is clear but it is not what you say.
What I say is that a 1200 dpi scan of a BW 8x10" cannot be distinguished if it was made by a V750 or a Creo or a X5, because at this resolution all those machines perform exactly the same as they are IT8 calibrated, and they are not at their optical limit by far. And also say that a Creo do not make much a difference to a V750 when dealing with 3.8 Velvia densities, but X5 will do.
I don't want to discuss about M.V. and drivers, just replied Ryan I've also a background about digital imaging to discuss in depth about digital chain.
This is about photography: Serious Hybrid Process. Nothing like darkroom, but today cultural difussion is throught the net, this leads to digitalization.
This discussion is important, for me included, if one can and want afford both an expensive scanner and a Sironar-S collection, including macro, there is no problem. But many people can have to choose between a new scanner and a glass or film, then it is very important to know in what way and in what situations a more expensive scanner will improve the result.
My statement is that for LF and BW a V750 performs like a Creo and like a X5, specially if we are not to print beyond 3m size and viewing the monster at reading distance.
And that there is no tonal difference at all.
I'd like to go beyond "better result", and discuss with technical arguments if there are any differences (LF / BW) in tones, sharpness, grain depiction, microcontrast... and at what precise print size it is seen... just the information it will be useful to take decisions.
And one option is to stay with a V750 that will perform exactly the same than a drum for LF-BW, and keep the money to hire drum service for 3.8D Velvia-Provia that cannot be done on any flatbed, Creo included, if deep shadows there.
Others say that Creo will render a better image for LF-BW.
Then I ask... How it can be this if we all scan LF at 1200 or 2400 dpi as much, at 2.5D, 3.0 exceptionally?
At those requirements there is no difference, I state. Let's be serious about what is performance and what is hype.
Last edited by Pere Casals; 26-Jul-2016 at 11:24.
Pere, the Epsons cannot meet their claimed Dmax spec in the real world. I have a Stouffer step wedge that has densities ranging from 0.05 to 3.03. In a one-pass linear scan, my Epson V850 does not show any differentiation above step 10, which is 1.37 density. Applying the most extreme curve the software is capable of, applying multi-exposure, and applying every other trick you can use to recover highlight detail, the V850 still cannot differentiate between step 21 (3.03 density) and step 20 (2.89) density. This means the Dmax is below 3.03 (not 4.0 like Epson claims). Of course, applying all those functions to get up to ~3.0 Dmax results in a file that's completely unusable! To produce a file that can actually be edited and produce a smooth range of tones, the scanner seems to have a usable Dmax somewhere around 1.5 and, even then, it's definitely shouldering off and losing separation between steps at that point.
Ha Bob, i'm a gluten for punishment so im posting once more then im done too!
Pere, I honestly have never seen anyone say so much but really say so little. You joined this forum this month. These other photographers have been shooting film possibly longer than I have been alive. I would look at there advice a bit more before you dismiss them. As far as my post, you just conveniently skip over where you have contradicted yourself and start spilling out more information that has no relevance. You just said at 1200 dpi there is no difference between a $600 v700 and a $20k x5. So your saying the same optics are used in both? Or are you really trying to tell me I can't tell the difference because of DPI or lines resolved? By that logic there should not be any detail difference between lens on the camera. I mean we cant see that fine detail, oh but wait... we can... optics matter. None of what your saying is going to change the fact that the lens on a v700 is just not as good as a creo. Thats what this topic was about.
The WDK is for hooking hardware into windows not the reverse. Drivers are written in C or mostly ASM not C++ thou obviously if you know C++, C is easy but you keep quoting C++. Kernel programming has nothing to do with any of this. IOPL0 is not even a thing, IOPL is a flag on an x86 system and I think there is something in the STDIO for linux with it? Neither has relevance here. The ADC obviously happens on the scanner so you would have to desolder, dump the firmware, rewrite it likely using ASM with no docs to add most of what your talking about. There are a number of issues you have to solve like the init calibration is done pre-scan everytime, changing the lamp would throw everything off. You would have to override that as well in the firmware. Among other things. Lets say none of that was true. And you could access it all in the drivers. Where is it your getting the source code? If you decompile it your working in ASM not C++ and how are you expecting to compile it again or are you saying you can modify raw machine code? Even if you did something like DLL injection you would still need to decomplile the drivers and work in ASM to know what to hook saying it was even there. And you would have nothing but pointers and flags with no names to modify. Dumping anything to a RS232 connection (there is no pin) would do nothing since thats a serial bus. Using IRQ0 or just the PIO is really kind of archaic and really low level. timeBeginPeriod() is part of a timing routine thats the correct way to time something, ok... why you would use IRQ0 lowlevel then call a high level windows function... Not that any of that matters because you said you were going to hook into it before the digital chain... so your going to hook into an analog chain as if that is a thing... *thumbs up* I'm sure everyone here is looking forward to seeing this Frankenstein v750 you have made (or are making, as you have said its both done and not done). That can out perform an x7 just using multiple exposures.
What were all wondering is why are even discussing all these. This thread was about the quality difference between a v700 and a IQSmart2. What does any of what your saying have to do with that? You dont even have an IQSmart2 to compare. Your basing everything off scanner examples that the IQSmart2 wins but saying if you hack a v750 thats some how going to make the lens and scanning method on a v700 better than a IQSmart2. A scanner you dont even have to try?
Thank you everyone else for all the input. If some crazy person reading this is selling an IQSmart2 I will be looking for one in the fall! And im done!
Ryan Mills
IQSmart2 not better at all than a V700 for LF and BW
Hello Ryan, let's go...
IQSmart2 is better than a V700 for some jobs, to scan large format BW sheets at 1200 or 2400 dpi it will perform exactly the same than a V700, because both have optical performance in excess and excess capability for 3.0D BW densities, with no difference in results. In smal prints the same, in moster prints beyond 4m it may be some difference looking at reading distance.
No difference in BW tonality.
For 135/120 may be a little edge for the IQSmart2.
For Velvia 3.8D deep shadows none will perform, a drum will
Yes, I state this. V750 has a optical performance of 2400 to 3200 dpi this may agree is 1200 to 1600 LPPMM high contrast MTF at extintion with a USAF 1951 target, you scan 9600, but there is no more optical information there.
A X5 we may agree it delivers 4800 to 6400 dpi true performance, this would be 2400 to 3200 LPPMM.
If put it in a bitmap of 1200 dpi the V750 information will be reduced to 1200 dpi, (pixel binning, it is called). With the X5 the IQ also will be reduced to 1200 dpi
The quality of the 1200 dpi from V750 will be the same of the X5, because we degraded in both cases form higher quality.
An example, imagine you take 2 shots of 24Mpix, one from D610 and from a D3200. The D610 image has 22 mega perceptual pixels and the D3300 shot has 12, if we resize the image to 1536x1024 will have indistinguishable images. Do you understand this?
Want to talk about DDK etc, not of my interest but if it's your's, let's go:
You can make drivers hat are not to control hardware at all, for example I open ports for a User mode application from a Driver. Every win proces has a IOPM table that specifies what port numbers, classic example if Porttalk http://retired.beyondlogic.org/porttalk/porttalk.htm, I still use source code from this. Note these are not TCP ports, but mapped registers of the PC hardware, for example to program 8254 or 8259 chips, that today are firmware implemented in chipset.
IOPL is Input Output Priority Level. Since Intel protected mode (386+), it It occupies bits 12 and 13 in the FLAGS register. It can be 0,1,2,3. Windows uses only leves 0 and 3. Processor only can move from a level to the other by executing an ISR. All user level apps run in IOPL 3, drivers can run in IOPL 0 o 3, in case level is 0 this is a kernel driver and a special instruction set is available, while in IOPL 0 software can directly access all resurces, including physical memory, in IOPL 3 you can only acces virtual memory, at the end via HAL.
You are wrong I told of may skills as a programmer, I do not use this in my V750 modification projed, read again my message...
COM 1 is linked to the 3F8 register, by writting in that register you can modify manually the state HI/LO of each pin, making what is known by bit banging, also you can progran the UART and let it manage the comms, thats the usual way. In machine vision I use bit banging to signal camera trigger because there I avoid glue logic and I know when I take the shot with uS precission.
Of the V750 I've only modified the illumination, nothing about drivers or hardware, read again above.
You should know, I'm just answering you.
IQSmart2 is better than a V700 for some jobs, to scan large format BW sheets at 1200 or 2400 dpi it will perform exactly the same than a V700, because both have optical performance in excess and excess capability for 3.0D BW densities, with no difference in results. In smal prints the same, in moster prints beyond 4m it may be some difference looking at reading distance.
No difference in BW tonality.
For 135/120 may be a little edge for the IQSmart2.
For Velvia 3.8D deep shadows none will perform, a drum will
I'm also done, if you cannot argue anything beyond kernel programming...
Also thank you everyone else !!! best regards !!
******************************************
But remember, hype and performance... not the same
Posting Permissions
Bookmarks