Hacker News Reader

Comment by 07d046 | Hacker News Reader

I can't imagine spending $6000 on a computer for editing photos and then saying "It also means I don't plan to buy and meticulously use a color calibration device". My computer might not be half as fast, but at least I know my blues are the shade of blue I think they are.

View on HN

Fair point :) I should reconsider. The UltraSharp line is factory calibrated but yea that doesn't mean much as it varies between display.

In what sense are they calibrated if things vary between displays?

I am under the impression they do some basic tests that don't entirely accommodate for uniformity of illumination and consistency of color across the display. In other words - I think maybe they only test one part of the display.

ICS profile with the screen’s unique profile baked into the EDID? Just spitballing.

Factory calibration is a thing in UltraSharp line, can't be compared to hardware calibration but works pretty well out-of-the-box.

http://www.tftcentral.co.uk/reviews/dell_up2718q.htm#hardwar...

They do a base calibration. Do a Google search for Dell Calibration report. It shows what they calibrate. Keep in mind, it's still a pretty loose calibration. I've never once gotten two of the same models that look the same out of the box.

This has driven me nuts more than once. I’ve very intentionally ordered two “matching” monitors at the same time, supposedly from the same lot, and they were vastly different. It may be an overreaction but since a number of years ago I’ve gone to just one big and nice monitor (even if it’s not technically as nice as the others) because those details get in my head so much.

As the saying goes: “comparison is the thief of joy” - so I remove the second thing to directly compare to.

It is factory calibrated for general desktop usage, but not necessarily for photograph reproduction.

Buy the cheapest Datacolor Spyder5(S5X100). The hardware is same between all the versions and only the software differs. Then get the open source DisplayCAL(https://displaycal.net/) that works with Datacolor sensors and other manufacturers. I calibrate my display once or twice a year. I also use it to calibrate my gaming computer and laptop.

Or you can do it manually the old way with a colour reference sheet. More laborious and requires good lighting (preferably daylight).

I use the Open Source Hardware colorhug and it works with DisplayCAL

  http://www.hughski.com/

If you are also printing than you need to spend more money on a calibrator that works with both screens and printers.

I can't tell you how huge the difference can be even on high end monitors. I had a friend that swore all Macs were calibrated and he got mad at me when I laughed. After we did the calibration he realized how wrong he was.

Also buy this: SpyderCHECKR 24 and use it as the first picture at any new site. Color correction within 10 seconds.

https://www.amazon.com/dp/B00LPS46TW/ref=wl_it_dp_o_pC_nS_tt...

I used to work at Datacolor, and I specifically worked on Spyder and SpyderCHECKR. It's nice to see that those get lots of love when it comes to color calibration :) but it still amuses me that they haven't figured out how to deal with the fact that their high-end Spyder unit sales are getting supplanted by the open source color calibration libraries..

Shhhh don't let them know. Funny thing is that Professional Shops and video companies are really starting to embrace these open source libraries.

So what is your thought on the Color Hug 2? http://www.hughski.com/colorhug2.html

I bought the first color hug and it is still serving me well but I really like the new color hug 2 especially the latency tool.

> Shhhh don't let them know. Funny thing is that Professional Shops and video companies are really starting to embrace these open source libraries.

Oh, I haven't worked there for a few years, but someone I know already brought it to the attention of upper management(pointing out the fact that Amazon reviews for Spyder products literally tell people to use the Open Source libs) and they've basically just ignored it as a whole. Their products are cool, and there were a few great people, but the company itself was... I won't really go further into it. Alas, I haven't been involved in imaging in quite sometime so I can't really comment on Color Hug :/

Not only between displays, but also between lighting conditions in the environment where the display is used. In principle, you should recalibrate every time you switch bulbs – even more important any time the monitor moves to a different room!

My understanding is that calibrating the monitor only takes the light emitted from the monitor into account, so the environment doesn't matter for calibration. That said, the way we perceive colour varies depending on environmental lighting, so if you want to be particular you should use D50 or D65 lights in the room, especially if you're printing.

It depends on the calibration device you are using - some also take ambient light measurements to compensate.

Well, it's calibrated to a specific delta-E so the amount of variation is limited - in practice, unless you're in a room with controlled lighting and are doing professional work (neither of which seem to be the case) calibration doesn't really matter.

(I went with an LG 32UD99 w/ some similar criteria of decent content creation performance and some gaming - it was reasonably priced, performs well enough, and more importantly, the thin bezels and lack of logos is quite schmexy - will have a hard time going back to anything branded. The LG comes w/ Freesync but I have since swapped out my Vega 64 for a 1080Ti, and as you mentioned, HDR is pretty pointless atm on PCs so ¯\_(ツ)_/¯. LG showed a slightly improved 32UK950 at CES.)

Anyway nice build writeup, didn't spot anything too out of wack on the build-side and the inclusion of the CL timing chart was a nice thing since I've often seen people get confused/mess that up.

As a fellow long-time Mac user forced to migrate (mostly in Linux, but Windows for VR and LR) I liked the part of the writeup on environment/keyboard tweaks. One thing I didn't see was about privacy. Windows 10 is very invasive. I used O&O ShutUp10[1] to go through all that stuff but there are a lot of tools [2] and other considerations [3].

Besides WSL, I've still found Cygwin to be indispensable as there's still system-level stuff that doesn't work outside the WSL (interacting w/ PowerShell scripts and such).

[1] https://www.oo-software.com/en/shutup10

[2] https://www.ghacks.net/2015/08/14/comparison-of-windows-10-p...

[3] https://senk9.wordpress.com/checklists/windows-10-privacy-ch...

> My computer might not be half as fast, but at least I know my blues are the shade of blue I think they are.

I've wondered if one could do color calibration using any ordinary digital camera and some clever software, but don't know enough about how color calibration actually works on computer to figure out if it would actually work.

The idea is that you would have some reference photos of assorted physical objects of known colors. The physical objects are chosen to be items that are commonly found around the house or office, or are cheaply and easily obtained.

You pick such a physical object, place it next to your monitor, and tell the software which object you are using.

The software then displays several images of that item on the monitor. One is the reference photo, and the others are tweaked versions of that reference photo that tweak the colors.

You then take a photo with your camera, showing the images on the monitor and the object itself. You upload that to the software.

The software then compares the object in the photo to the images on the monitor, and figures out which best matches. From this, it should be able to deduce some information about how accurately the monitor is displaying color, and what adjustments could be made to improve it.

Note that this does not depend on the camera being accurately calibrated. It just depends on the camera being consistent, and having enough color resolution to see the tweaks that the software applies to the images on the monitor.

Note also that this should work for more than just calibrating monitors. It should also be able to figure out a color profile for your camera along the way. Then it should be able to print some test images, have you take photos of them, and from those and the camera color profile figure out a color profile for your printer.

Interesting idea, but I think there's a variable you're missing, which is the light by which the physical object is illuminated by. The colour temperature and spectral power distribution of the light makes a difference. If you used a custom white balance in your camera you should be able to get the colour temperature close to right (depending on how neutral the camera is), but to calibrate colours you'll need a high quality light source (real daylight is quite good, but varies a lot). At this point, a cheap calibrator is probably easier and far more accurate.

Color calibration is a chain, you need to know if your camera is calibrated too, so you need at least standardized targets, finding random objects of known color that cover the whole color space sufficiently is not trival, only then you could use your calibrated camera to measure a displayed target.

But I wouldn't cut corners there and just get a calibration device for about $100 plus a standard color calibration target. Certainly more sensible investment, then the difference between a 2k and 6k rig.

Having purchased the i1Display 2, the Huey Pro and then the Spyder 4 Elite these devices have a pretty finite lifespan of a couple years at most.

If the manufacturer doesn't discontinue software updates for your latest OS the sensors in the devices themselves seem to use fairly cheap plastic which is susceptible to colouration over time causing the calibration's to drift quite significantly.

It might be more cost effective to just have someone calibrate your equipment for you rather than buying another plastic baubel with questionable longevity.

Serious question: If you're not selling work to be printed or resold as stock, how important is calibration these days?

Given the proliferation of the screen as a photo consumption device, and considering how just about every screen out there has slightly different properties (color temp & gamut, brightness and contrast, OLED vs LCD), what will calibration necessarily accomplish?

Calibration puts you (hopefully) in the middle of the distribution of errors, so your viewer is more likely to see pretty much what you intended.

I publish a lot of black and white images in part to further reduce the uncertainty of the end-viewer's calibration. Can't get a perceived color cast if there's no color.

>> Can't get a perceived color cast if there's no color.

But you can still have other problems depending on the dynamic range of your photos.

I often edit my photos so that stuff in shadows are not perceptible. The problem is that I edit on my laptop LCD screen, but on my phone OLED at default settings, I see clearly everything that was supposed to be hidden in the shadows.

> Calibration puts you (hopefully) in the middle of the distribution of errors

That would be an interesting research topic! Perhaps the average color of the un-calibrated displays is not the same as the one of a calibrated display.

For example most TV are put in some very colorful mode from factory and most user don't bother to switch it. The same is true for old Samsung mobile phones which had a lot of popping colors. You could switch them to more realistic tones but nobody ever did this.

I just use f.lux e mess all my calibration :-)

Another way to mess it is to buy a glass with "digital lenses" that will filter the blue spectrum.

there's also this gem: "I'm not a professional photographer. I'm merely a hobbyist."

o_O

I'm pretty sure a professional would rather buy a dual-socket Xeon based machine (say 2x E5-1650 v4) that has 4x the memory performance for closer to half the price.

A lot of professional photographers don't know much about computers and CPUs. Remember that "professional photographer" crosses a very broad spectrum of photographers, since the bar is basically "I charge for my work".

For better or worse, many professional photographers often rely on the advice of friends and salespeople for "something fast enough for what I need to do".

Why should that tip off your BS detector? There's a long, long, long, long history of amateurs lavishing far far more attention on their rigs than pros ever could. It became a trope in Renaissance literature.

Basically, people get rich doing other things and then dive into their hobbies buying all top-of-the-line gear. Whereas the pro started with nothing and learned the basics, so they are able to get much much better results from substandard gear.

Holds across all creative disciplines.

Why so negative? Some people spend a lot more than that on what I'd classify as less important stuff.

I'm not in a position to travel the world with high end photo gear. (I just recently replaced my 10 y.o. desktop pc case)

...but I am in a position to ve happy together with thise who can.

(And I also have rewarding and less expensive hobbies like learning to fix my stuff and coding side projects :-)

For $6000 he could have got something like:

CPU: 2 x Intel Xeon E5-2637v4, 3.5GHz (4-Core, dual threaded, HT, 15MB Cache, 135W) 14nm

RAM: 128GB (8 x 16GB DDR4-2400 ECC Registered 2R 1.2V DIMMs)

Graphics: NVIDIA Quadro P4000 GPU, 8GB GDDR5, 105W, Single-Width, PCIe 3.0 x16,

Seagate 1TB Exos 7E2 HDD (6 Gb/s, 7.2K RPM, 128MB Cache, 512n) 3.5-in SATA

just a quick build i did over at Silicon Mechanics

There's no real benefit of that much RAM in Lightroom (Premiere Pro is a different story). I also don't need spinning disk storage as I have a 12TB NAS (which is 3 years old, I can probably replace all the disks for 10TB models) for when I archive my shots, so faster M.2 SSDs were a priority.

The dual CPU setup doesn't help Lightroom either -- it's not terribly efficient for multiple cores so I can't imagine it would do any better for dual CPUs. I'm trying to find the adobe faq on the topic I thought I've read about that.

Quadro has some merit if I wanted a 10-bit workflow since I already have a 10-bit panel.. but I did kinda want to dabble in some gaming and VR with this build as a side-benefit so I went with a GTX card.

I'm clueless when it comes to graphics cards. Yeah, in your case, you might really want to just massively overclock a few cores. :)

I work with huge 3d images my self 600x600x600 pixels/voxels, can become quite demanding when calculating warps, etc.

Running latest lightroom on a 1950x w/1080ti card and CPU % doesn't go beyond 20%. Just unfortunate all this computing power is wasted.

Oddly, Lightroom does not parallelize across multiple photos in a job. Try cutting your photos into smaller batches and launching multiple batches in parallel.

You would figure that's an obvious and simple feature to offer, at least as an option.

> There's no real benefit of that much RAM in Lightroom

This really depends on what you are editing. 13mp snaps from your ancient DSLR? Yeah, no point. 52mp snaps from your digital medium-format camera? You'll probably want more than you'd think. 4x5 or 8x10 negative scans? Open the bloodgates.

(and remember that by default Lightroom does not parallelize across multiple photos within a batch job, so multiply your measurements by 4x or whatever if you're going to be launching multiple batch jobs at once to fully occupy a high-end processor.)

128 GB is overkill, but there's a solid justification for at least 32 GB or 64 GB for power-user situations.

That dual CPU setup gets you 4x the memory performance though... Which is likely your main bottleneck.

They included a $1499 monitor in the list as well, along with a $400 VESA arm mount, and hundreds of more dollars in accessories.

Overall, with your Silicon Mechanics build they'd probably be closer to a $8000 build. On top of that, a P4000 is pretty much a beefed up GTX1060, which gets smoked by a GTX1080TI if one is going for GPU performance. They're not really comparable builds.

I missed the monitors. Fair point.

Reply on news.ycombinator.com