Thoughts on my new computing hardware

[Roger Baylon]

Bob--Thanks for suggesting that I save the profile. I had no idea that I could, but --thanks to your hint--it was easy to figure out and I have done it now!!!

One comment -- my new Beelink i5 8th generation mini-computer has worked flawlessly under ACP all night long for two nights. But the purchase was scary, because the system arrived with no documentation for setup. Fortunately, the company provides instead a super competent IT professional from Uraguay who walked me through every step! What if I they had assigned me a less competent tech support?
One question: My new Beeklink provides 4k display support but only 10 bit color depth. Is 10 bit enough for ACP imaging acquisition but maybe not enough for good processing with PixInsight? Or is 10bit enough for everything since I am very much at the amateur level.
Roger

[Bob Denny]

I can't speak to processing with Pixinsight, but I doubt that 10 bits color will be sufficient. ACP's acquisition process would not be adversely affected.

[Roger Baylon]

Hi Bob, You are very helpful - as always.
As to 10 bit color depth on my new Beelink mini- computer, Apple's new 6K Pro XDR display that sells for $4,999 claims 10 bit color depth. See info below.
So maybe 10 bits is ok for BOTH imaging and for processing in pixinsight?
Roger

https://www.apple.com/pro-display-xdr/specs/

XDR (Extreme Dynamic Range)

• Brightness: 1000 nits sustained (full screen), 1600 nits peak¹
• Contrast ratio: 1,000,000:1
• Color: P3 wide color gamut, 10-bit depth for 1.073 billion colors

[Eric Dose]

10 bit integers are definitely not good enough for serious image processing, in any program. One must retain many more digits during intermediate computations than are needed in the end product. The human eye's "bit depth", which Apple is catering to, has little or nothing to do with it.

In any case, floating point representations are usually best for processing workflows that have lots of steps.

[Roger Baylon]

Eric,
I appreciate your comments. I suspect that I am going down a rabbit hole (whatever that means) with this particular quest for clarity.
My new mini-computer has an Intel i5 processor which handles 64 bits for computations. So Bob Denny seems to thin that is fine for imaging. So far so good.

Since I spent 6 months trying with only partial success to get my macbook air with parallels run ACP and Maxim DL, I don't want to go down another dead end. The good news is that my mini-computer has showing encouraging potential for imaging, which was the main reason for buying it. But it would be nice if it could do imaging processing too.

By color bit depth, I mean the number of bits that are available to display the various shades of grey. For example, 8 bits can display 256 shades or levels. 10 bits can display 1024 levels.
15-16 bits of color depth is called high color. 24 bit seems to be called true color.

I have just found this on the PixInsight website, which is the software I am experimenting with:
A minimum color depth of 16 bits per pixel is required, but for serious work a true-color, 24-bit video mode is a must.

I am definitely getting confused--and must now pay the price for not having seriously studied computer science.
Warmly, Roger

[Dick Berg]

Roger, (Eric,)

I've been biting my tongue, but I can't help commenting.... The ten bits Apple is talking about is 10 bits per channel, or in other words 30-bit per pixel video. 2^30 = 1.073 billiion. Does the Beelink "10-bit color" also mean ten bits per channel? Not much documentation on their website to tell. Even if you had the usual kind of $200 ViewSonic monitor, 10-bit (total) color would look terrible.

[Eric Dose]

Dick: Yes, you beat me to the response.

Roger: if you're talking about color images, you absolutely must distinguish between bits-total vs. bits-per-channel. 24-bit color is 8 bits per channel (as RGB), barely adequate for final presentation (esp. for high dynamic range images as astrophotos usually are). Again, 8 or 10 bits *per channel* will prove utterly inadequate for intermediate processing. Just use floating point numbers for your image processing. PixInsight always does, if memory serves.

What numerical format and depth you use in the final image rendering is a completely different subject. 8 bits per color channel has been standard monitor presentation for some years, as Dick notes. But 8 bits per channel won't suffice for very high pixel counts (because the human eye is very, very good at detecting edge artifacts in smooth gradients, which is why dithering was a popular "remedy" in digital imaging's early days). And 8 bits per channel sure won't be enough to represent astrophotos, very marginal even if HDR processing is in play.

As for Apple's "10 bit", I assume with Dick that it probably means 30 bits total (RGB).

If you see "16-bit color", that almost always means a subset of the colors available in "24-bit color". If you see "32-bit color", it almost always means 8 bits per channel, the 4th channel being "alpha" (opacity), often useful for image processing (layering), but not for final presentation where (by convention) alpha is exactly one.

Color and image processing don't have to be a rabbit hole. It can be learned. But first, the terminology must be mastered, else all is confusion. I hope our comments help a "bit".

[Roger Baylon]

Dear Eric, Dear Dick,
I appreciate your help. It all makes sense. I have to give you a short reply because I am afraid of being timed out in the process of getting my thoughts together.
It seems to me that you both must be right: 10-bit color depth must be per channel. So 30-bits total.
In that case:
- I can safely use this mini-computer for image acquisition with ACP and Maxim. It seems to work well. If I remember correctly, I think that Dick suggested months ago that I consider a mini-computer.
- I may be able to use this mini-computer on a temporary basis, as I develop my skills. Also, my monitor may be limited to 8bits so I could be monitor-limited rather than computer-limited. Finally, I am not sure that my Apple Macbook Air that I was planning to use for processing is better than my mini-computer.
- In the meantime, I will be searching for a monitor that is good for PixInsight. I am hoping that I can get one for $500 to $600. Am I way below the mark? Does anyone have a suggestion for a good monitor? Ideally about 25-27 Inches?
Warmly, Roger

[Eric Dose]

Whatever monitor you get, remember that calibrating the color balance, black and white points, and gamma is what really makes the monitor perform for image processing. Photoshop has some routines to help you make the adjustments, PixInsight might also.

And don't go too very cheap with the graphics card, although they're all pretty good now compared to the bad old days.

[Niklas Storck]

Hi Roger

I happened to see this post and feel there is a bit confusion about bits and what is going on in computers and image handling in the answers.

I think You is talking about the bits in the LUT table (Look Up Table) in the computer screen. In that case 10 bit is quite good. My excellent Eizo ColorEdge CX270 that can display 99% of Adobe RGB has a 10-bit LUT. They are professional photographic monitors costing about 2000$ each. The LUT choses from 16 bits color information from the graphic card and select appropriate colors for the chosen settings, in the screen that is. I am not aware of a monitor with bigger LUT. Even a reference monitor costing 20 000 $ has 10-bit LUT.

There is a lot of bits and bytes in image handling.

To start we have the cameras, that usually have 14 to 16 bits per pixel. If it is a color camera there is 14 to 16 bits per pixel for each red, Green and Blue primary colors. My old Nikon D800 had 14-bit but the software in the camera changed that to 12-bit per RGB if I saved raw format. Saving jpeg is only 8-bit. Usually, the astronomical CCD or CMOS cameras export images in 16 bits per pixel regardless of the internal resolution.

The software can choose to change the bits in the image when importing it. I think PixInsight uses 32-bit floating-point precision during the computing. That is reasonable because there is an increasing risk for rounding errors when using 16-bit integer math. There is a lot of complexity in image handling with different kind of noise, but that is another and longer story.

When saving the pictures after integration of the subs you take, it is good practice to save a 32- or 64-bit image before stretching and final image handling.

Finally, when you want do show your wonderful pictures on a website it is probably best to export a 8-bit jpeg (in sRGB) because it is likely to look best on all the screens out in the world.

Your computers 64-bit memory channels can handle all the formats you ask it to handle, depending on the software. The 10-bit screen LUT does not interfere in any way to what the internal work in the computer does.

Good luck!
Niklas Storck