LS 4000 - blooming and density range

[pellicle]

Hello

I have just got my 4000 back from Nikon in Finland where (among other things) it was meant to be cleaned. I'm new to this scanner so have no prior knowledge of what its able to do. So I thought I'd pop my Stouffer 21 step stepwedge into the FH-3 and give it a whirl.

I must say I'm surprised at the results (which are here on my blog).

I don't know if anyone has done any 'testing' of this nature before with one but the results are clearly inferior to my Epson 4990 flatbed for density range and blooming.

I'm just seeking opinions, as I've already contacted Nikon service and sent them an email and that link. However the service tech doesn't get back to work till monday ... so till then I thought I'd ask here.

Thanks

 

[pellicle]

Hi

I've dismantled my LS-4000 myself (since its the weekend) and sure enough the optics were filthy. Careful cleaning of the mirrored surface has improved things to a great extent, but still things are not 'ideal'

perhaps its as good as it gets ... but I'm inclined to try better shading inside the scanner

 

[Marco B]

One thing that intrigues me is the fact that your Epson 4990 translates deepest black in values of about 15,15,15, and your LS4000 sets them at 0,0,0 looking at your graphs...

Actually, except for this, the graphs themselves don't look that much different in response, both hit the ground at about step wedge step 17. Maybe, like you say, this is just as good as it get's with the DMax performance of the LS4000, and you are just left with a much better focusing and sharper scan result (that still would make your buy worth it).

Actually, it slightly reminds me of the blown out highlights (255,255,255) I used to get on my Canon 9950F before I hit the "Execute-(self-)calibration" button and learned something new... this was a few years back at a time I didn't know anything about digital photography.

 

[Marco B]

Well, come to think of it, isn't the non-linear response of the LS4000 also maybe a possible indication that it fails to execute it's self-calibration?

If this support page by Nikon Canada is right, the LS-4000 is also supposed to perform a self-calibration on start-up.

 

[pellicle]

One thing that intrigues me is the fact that your Epson 4990 translates deepest black in values of about 15,15,15, and your LS4000 sets them at 0,0,0 looking at your graphs...

I perhaps didn't as agressively set the absolute dark points as much in the scanner software ....

Here is the data for the 3200

step# linear linear std %error
1 233 1.42 0.60944206
2 187 2.48 1.326203209
3 149 3.14 2.10738255
4 120 3.68 3.066666667
5 97 4.08 4.206185567
6 79 4.04 5.113924051
7 64 3.53 5.515625
8 51 3.44 6.745098039
9 42 2.93 6.976190476
10 34 2.81 8.264705882
11 29 2.75 9.482758621
12 24 2.29 9.541666667
13 20 2.11 10.55
14 17 1.73 10.17647059
15 14 1.59 11.35714286
16 12 1.51 12.58333333
17 11 1.48 13.45454545
18 10 1.47 14.7
19 9 1.47 16.33333333
20 8 1.46 18.25
21 8 1.45 18.125

and for the Nikon

step# linear (median) analgo gain 0.8 log % error
1 249 2.396199347 0.962329055
2 219 2.340444115 1.068695943
3 170 2.230448921 1.312028777
4 137 2.136720567 1.559650049
5 107 2.029383778 1.896620353
6 88 1.944482672 2.2096394
7 72 1.857332496 2.579628467
8 61 1.785329835 2.926770221
9 51 1.707570176 3.348176816
10 42 1.62324929 3.864879263
11 33 1.51851394 4.601557394
12 26 1.414973348 5.442205185
13 19 1.278753601 6.73028211
14 9 0.954242509 10.60269455
15 5 0.698970004 13.97940009
16 3 0.477121255 15.90404182
17 2 0.301029996 15.05149978
18 1 0 0
19 1 0 0
20 1 0 0
21 1 0 0

Actually, except for this, the graphs themselves don't look that much different in response, both hit the ground at about step wedge step 17. Maybe, like you say, this is just as good as it get's with the DMax performance of the LS4000, and you just are just left with a much better focusing and sharper scan result (that still would make your buy worth it)

that's how I'm feeling at the moment. Certainly for 35mm the extra precision if focus is well well worth it! But for my 120 and 4x5 stuff the Epson is acceptable for most things (and Drum scans are available for the rest).

 

[Marco B]

What about the "self-calibration" feature of the LS4000, does it go through such a phase at start-up? The non-linearity as you graphed keeps slightly bugging me as well...

 

[Marco B]

By the way, here's a short discussion of the Analog Gain feature you used, I did not yet know what it did, since my Canon 9950F doesn't have it.

http://photo.net/digital-darkroom-forum/007Qy9

It seems you might be able to set lamp-brightness as well, as opposed to using Analog Gain?? Maybe increasing that, might allow for a bit more shadow depth, although you should be careful not to blow out highlights.

Quote from the Photo.NET thread:

"It is important that you realise that the Analog Gain control adjusts the scanner exposure time NOT lamp brightness as is implied in the Nikon manual. Also note that adjusting the Master channel is to be preferred over the individual Red, Green and Blue channels. The scale is measured in Photographic EV terms and so each 0.333 increment is theoretically equal to 1/3 stop photographic. Personally I find altering the Master channel by much more than 0.33 to be exceedingly damaging of either highlight or shadow detail, so be warned this tool is NOT the ultimate "get out of jail card" for poor exposure that some would have you believe."

Oh... and mind the possible implication of the 0.8 analog gain you set, see the remark above again, that suggests that going over maybe 0.33 might actually reduce or damage your shadow detail instead of improving it...

 

[pellicle]

Hi

What about the "self-calibration" feature of the LS4000, does it go through such a fase at start-up? The non-linearity as you graphed keeps slightly bugging me as well...

yes it does. The linearity kink is probably related to the analog gain. Let me repeat the test with no linear gain and I'll post that as an addendum to the end of the blog ... gimme an hour or so

 

[Marco B]

The linearity kink is probably related to the analog gain. Let me repeat the test with no linear gain and I'll post that as an addendum to the end of the blog ... gimme an hour or so

That might make some sense, also in the context of what the Photo.NET commentator wrote. By bending the curve, you will loose or possibly clip ultra deep shadow detail, but gain better separation in the "normal" deep shadows, with RGB values spaced further apart.

 

[pellicle]

addendum added

 

[pellicle]

That might make some sense, also in the context of what the Photo.NET commentator wrote. By bending the curve, you will loose or possibly clip ultra deep shadow detail, but gain better separation in the "normal" deep shadows, with RGB values spaced further apart.

well it depends on what is 'normal' deep shadow. For slides I think its a bit too poor a performance as slide has a narrow range of scene brightness especially with shadows. With Negative this will be bad for skies and high intensity light areas as this part of a negative is usually pretty dense anyway. I notice that these and especially where the least amount of error appears appear to neatly fit within band of negative (but be careful of over exposure here).

 

[Marco B]

QUOTE from blog:
"seems that the linearity step at about 13 is still there"

Well, I have no experience with Nikon Coolscans at all. Maybe someone with experience with Coolscans can comment on your observations of non-linearity and if this is sign of an issue with the scanner (which seems likely), or any kind of error in post processing.

Another issue that comes to my mind, and that I haven't actually been pondering enough about before, is what happens to scan results if the light source grows old, as in a second hand years old much used scanner... might that effect your scanresult significantly and cause such non-linearity?

 

[pellicle]

Hi

QUOTE from blog:
"seems that the linearity step at about 13 is still there"

what happens to scan results if the light source grows old, as in a second hand years old much used scanner... might that effect your scanresult significantly and cause such non-linearity?

qood question. I'd expece that with cold cathode tubes it would be a significant issue. I can't say that I've heard of LED's having such issues. Its not impossitble, so it would be interesting to be able to test. Given they are semiconductors any drift in behaviour may be use related rather than time related, but as an ex-electronics tech (FM radio) and electronics engineer my first suspicion would be that such degradation would appear in other equipment making them quickly unusable. For example the amplification curve would no longer operate correctly and dynamics in amplifiers (which use transistors and diodes) would suffer.

I'll do some investigations and post on my blog if I find something.

 

[Marco B]

OK, I hadn't realized your Coolscan uses a LED light source. Don't know anything about ageing of those, except that what I have read about normal white LEDs is that they use phosphorescence to create white light, due to the inability to create a working "gap"(is that the correct term) that produces white light directly. Don't know what the stability of those phosphorescents are.

But maybe the Coolscan mixes up white light from different coloured LED's?

Well, considering your background, you undoubtedly know ten times more than I will ever do about these subjects...

 

[Marco B]

You might find this interesting, from Wikipedia (http://en.wikipedia.org/wiki/LED). So it does seem even LEDs suffer from a gradual loss in light output, and not sudden failure...

Lifetime and failure
Main article: List of LED failure modes

Solid state devices such as LEDs are subject to very limited wear and tear if operated at low currents and at low temperatures. Many of the LEDs produced in the 1970s and 1980s are still in service today. Typical lifetimes quoted are 25000 to 100000 hours but heat and current settings can extend or shorten this time significantly. [25]

The most common symptom of LED (and diode laser) failure is the gradual lowering of light output and loss of efficiency. Sudden failures, although rare, can occur as well. Early red LEDs were notable for their short lifetime. With the development of high power LEDs the devices are subjected to higher junction temperatures and higher current densities than traditional devices. This causes stress on the material and may cause early light output degradation. To quantitatively classify lifetime in a standardized manner it has been suggested to use the terms L75 and L50 which is the time it will take a given LED to reach 75% and 50% light output respectively.[26]