[jeffkrol]

I've had my K20D for a little over a week now and thought I'd chime in with another sample. I'm using Dead Pixel Test to provide the counts. Took three shots from both my K10D and K20D at ISO 400, 1/30 sec, 2 sec mirror, dial @ M, with lens cap on. K20D - Slow speed NR = Auto, High-ISO = Off. K10D - Noise reduction = Off.

3 images from K10D ==> count of 6 hot pixels on each
3 images from K20D ==> count of 190, 196, 195 hot pixels respectively

These are using the default value of 60 for the luminance threshold. Is this the most accurate value for this test?

Just posting as an additional point of information. Didn't sync all settings (just those noted above) across cameras so if there are other settings that affect the numbers, please let me know as I'm curious.

As a sensor test, that methodology should be fine. What I'm curious about is the k20 @ 100 or 200 iso and if the numbers are close to the 400 numbers.
BTW: I you translate those numbers into percents you get approx:
.00006% failure vs .00131%.....

[reddream]

I've had my K20D for a little over a week now and thought I'd chime in with another sample. I'm using Dead Pixel Test to provide the counts. Took three shots from both my K10D and K20D at ISO 400, 1/30 sec, 2 sec mirror, dial @ M, with lens cap on. K20D - Slow speed NR = Auto, High-ISO = Off. K10D - Noise reduction = Off.

3 images from K10D ==> count of 6 hot pixels on each
3 images from K20D ==> count of 190, 196, 195 hot pixels respectively

These are using the default value of 60 for the luminance threshold. Is this a reasonable value for this test?

Just posting as an additional point of information. Didn't sync all settings (just those noted above) across cameras so if there are other settings that affect the numbers, please let me know as I'm curious.

One additional set - same parameters as above except drive mode to single instead of 2 sec mirror

3 images from K10D ==> count of 6 hot pixels on each
3 images from K20D ==> count of 2, 2, 1 hot pixels respectively

My Result:

1. ISO 200, 1/60, 2 sec mirror, M, lens cap on, High ISO Noise Reduction weak
13 pixels when threshold = 60
397 pixels when threshold = 20
628 pixels when threshold = 17
2064 pixels when threshold = 9 (still visible)

2. Same setting except 2 sec timer
0 pixels when threshold > 17
1 pixels when threshold = 17
108 pixels when threshold = 9

[rfortson]

I tried the method posted over at DPR and I could only see two hot pixels. I didn't shoot a very good subject (a fabric), which made it hard to count the pixels. However, I didn't see any large number.

I've used the hot pixel mapping before (just to try it out - no indication of a problem). I wonder if that prevented me from seeing the potential hot pixels?

Anyway, seems like my K20D works as advertised. I'm still as pleased as can be with it.

[reddream]

I tried the method posted over at DPR and I could only see two hot pixels. I didn't shoot a very good subject (a fabric), which made it hard to count the pixels. However, I didn't see any large number.

I've used the hot pixel mapping before (just to try it out - no indication of a problem). I wonder if that prevented me from seeing the potential hot pixels?

Anyway, seems like my K20D works as advertised. I'm still as pleased as can be with it.

I did a pixel mapping also before my test and got hundreds of visible pixels. I think you are lucky.

Attached Images

 

[reddream]

I did a pixel mapping also before my test and got hundreds of visible pixels. I think you are lucky.

I just viewed this post on my desktop computer (with dell LCD) and couldn't see any hot pixels in my attached crop and I had to increase some degree of brightness to make them visible. Then I switched back to my old IBM T40 laptop, all those hot pixels are obvious. So using a tool such as dead pixel test could help to reach a fair conclusion.

[jeffkrol]

I did a pixel mapping also before my test and got hundreds of visible pixels. I think you are lucky.

seems to be in the pipeline to be solved:
One hour ago i visited PENTAX European Headquarter in Hamburg and talked with the german chief technican Mr. P. about the hotpixel effects.
He said that it is not a hardware problem - it is a firmware related question. Since yesterday evening there is a firmware-fix ready and we implemented it on my K20D - IT WORKS !
This update will be implemented in the next firmware version - it will be available after quality testing phase in the near future.
So they worked fast and fixed it - it´s NOT a hardware question - and everybody can update the K20D´s via normal sd card update process.
K20D / HOTPIXEL Problem fixed: Pentax SLR Talk Forum: Digital Photography Review

[falconeye]

I'm using Dead Pixel Test to provide the counts.

Thank You for the link.

I applied it to the dark 700x700 dark area crop which I provided in the other thread:
https://www.pentaxforums.com/forums/pentax-news-rumors/26955-officially-emerg...tml#post252551

Using it, I see that the noise floor and dark area exposition is centered around 30 and stops at 53. Some of the noise may be hot pixels, actually

My confirmed hot pixels are from 56 to 140 (13 altogether), with an average of 76 and a median of 69.

If I exclude two hot pixels with 116 and 140 luminance (may be static hot pixels which I maybe could map out), I get hot pixels from 56 to 75 (11 altogether), with an average of 66 which is within the median.

11 random hot pixels for the crop are 330 for the image.


If I assume that random hot pixels build up proportionally to exposure time (2s), I would get a typical luminance of 1 at 1/30s and random hot pixels would completely disappear in the noise floor. Probably same for ISO.

[RiceHigh]

New Super Aggressive Hot Pixel Ruling Out Algorithm?

seems to be in the pipeline to be solved:
One hour ago i visited PENTAX European Headquarter in Hamburg and talked with the german chief technican Mr. P. about the hotpixel effects.
He said that it is not a hardware problem - it is a firmware related question. Since yesterday evening there is a firmware-fix ready and we implemented it on my K20D - IT WORKS !
This update will be implemented in the next firmware version - it will be available after quality testing phase in the near future.
So they worked fast and fixed it - it´s NOT a hardware question - and everybody can update the K20D´s via normal sd card update process.
K20D / HOTPIXEL Problem fixed: Pentax SLR Talk Forum: Digital Photography Review

[Edvinas]

New Super Aggressive Hot Pixel Ruling Out Algorithm?

If it solves the problem - who cares?

[rfortson]

I did a pixel mapping also before my test and got hundreds of visible pixels. I think you are lucky.

Oh well, I'll take it. For some reason mine doesn't exhibit the same problem as the others posted here. I'm glad, and also glad there appears to be a firmware fix for others not so lucky.

I'm back to using my K20D with no worries.

[reddream]

New Super Aggressive Hot Pixel Ruling Out Algorithm?

No. I don't think Pentax could work out a new algorithm in couple of days by firmware. In my opinion, they just forgot to subtract dead/defective pixels in some scenarios (2s timer, remote 3s, etc.).

[falconeye]

For some reason mine doesn't exhibit the same problem as the others posted here.

I think it is quite important to know whether the problem is universal or not.

Over at German forum DFN (the friendly person reporting back from Pentax Hamburg is a member) there was a report about "no problem with mine" as well. The claim was withdrawn later on (he used ISO800 where NR already kicked in).

So, your report is now the only one of not having it. Can you please upload a darkframe shot, or report the hot pixel count in a darkframe shot using the tool reported here (StarZen Technologies, Inc; DataFlex tools and books), threshold level 20?

Settings: lens cap on, 2s shutter delay, 1/30s or shorter, ISO400, JPEG high quality, NR off, Long Exposure NR AUTO.


Thank You very much. It would be great news if not all units exhibit the same phenomenon.

[falconeye]

No. I don't think Pentax could work out a new algorithm in couple of days by firmware.

Agreed.

The person who visited Mr. P of Pentax Hamburg meanwhile added a detail: According to Mr. P, the fix will be applied to a subpart of the firmware which is Samsung sensor specific, so really low level. Whatever it means.

[rfortson]

I think it is quite important to know whether the problem is universal or not.

Over at German forum DFN (the friendly person reporting back from Pentax Hamburg is a member) there was a report about "no problem with mine" as well. The claim was withdrawn later on (he used ISO800 where NR already kicked in).

So, your report is now the only one of not having it. Can you please upload a darkframe shot, or report the hot pixel count in a darkframe shot using the tool reported here (StarZen Technologies, Inc; DataFlex tools and books), threshold level 20?

Settings: lens cap on, 2s shutter delay, 1/30s or shorter, ISO400, JPEG high quality, NR off, Long Exposure NR AUTO.


Thank You very much. It would be great news if not all units exhibit the same phenomenon.

I did copy the somewhat less stringent techniques in the DPR post (ISO 400, Tv mode, 2 sec mirror lockup, High ISO NR turned off, 1/15s, LE NR Auto) and only saw two hot pixels. To be honest, I usually wouldn't even use those settings (the 2sec mirror lockup being the exception - I definitely use that), so pushing it to even more settings that aren't practical don't make sense to me. Plus, every camera that I've ever used where I took a shot with the lens cap on showed hot pixels. My solution to that was to quit taking shots with the lens cap on.

[falconeye]

To be honest, I usually wouldn't even use those settings

I see your point and I think along the same lines. The whole issue may be of no practical relevance. But this we don't know yet for sure. We just hope so.

Anyway, we are already turning a bigger wheel here. Remember the FDIV bug in some Intel processors? It had no relevance to 99.999% of customers. Still, Intel had to recall them all and spent big money.

Meanwhile, I have started a detailed study and found that up to 10000 pixels can be affected in an image (they seem to just follow their own noise pattern curve belonging to some higher ISO; so they are hard to detect visually as 99% of them are pretty dim).

If you didn't do a lens cap test, I think it is permissable to assume that your unit wouldn't be the single exception known to date. Probably, your shot didn't contain enough dark area to spot the hot pixels. Most of them are pretty dim actually...