Pixel Density (Thread 5)

The first four are linked below for reference:

+++++++++++++++++++++++++++++++++++++

First thread:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37482379

Second thread:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37532054

Third thread:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37548176

Fourth thread:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37559698

+++++++++++++++++++++++++++++++++++++

OK, the reason for the fifth thread is to respond to this post:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37572774

Seriously ... if that is all you've been saying then a hundred thousand+ words over the space of a week seems a tad excessive ...

Yes, it does seem excessive, doesn't it? Yet, people don't accept that more pixels result in more IQ (options) for any given sensor size, assuming equally efficient sensors , and/or don't believe that more pixels for a given sensor size do not reduce sensor efficiency as a general rule (true for CMOS, not necessarily true for CCD).

Of course, there are sensors with higher pixel densities that are less efficient than sensors with lower pixel densities, but the opposite is true as well. So, there is no correlation between pixel density and sensor efficiency (again, for CMOS, but possibly so for CCD).

The "hundred thousand+ words" are explaining and demonstrating the above. Lee Jay presented photos that demonstrated the positive effect of a higher pixel density:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37549285

where either the sensors had the same efficiency, or that the differences in efficiencies was not significant enough to be an issue in the photo he presented. You presented a demonstration where the differences in sensor efficiency did make a difference:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37559698

for the photos you presented, which I discussed here:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37569300

Ideally, this post should summarize the four threads, and be the end of it. However, like anything else, ideal situations rarely exist.

So increasing pixel density at lower densities generally improves resolution and image quality up to a point. At which point, any further increase causes an exponential increase in noise, decreasing image quality. Thank you geniuses.

Bokeh_freak said:

So increasing pixel density at lower densities generally improves resolution and image quality up to a point. At which point, any further increase causes an exponential increase in noise, decreasing image quality. Thank you geniuses.

Click to expand...

I can't remember where anyone got that result. Could you perhaps link a post or two?
--
Bob

Great Bustard said:

Ideally, this post should summarize the four threads, and be the end of it. However, like anything else, ideal situations rarely exist.

Click to expand...

A man can dream ...

I take issue with Lee Jay's example. The 1:1 big pixel image is cropped and scaled an enormous amount ... this is surely at the point of being a pathological case. And since it is widely quoted by the GoF, it does not meet the "widely applicable" standard for such use ...

That's one opinion at least.

Now ... moving on to BOBN's wonderfully arrogant response to this image:

Great Bustard said:

Great Bustard said:

Processing?
I'll do it myself properly. might need a new thread.

Click to expand...

Click to expand...

The answers are "no processing, just cropping" ... and define "properly" when there are only jpegs available on the site. Or was that just another case of unbridled ego getting the best of the writer?

Perhaps things would be more believable if we pit the latest full frame camera against the latest APS-C ... and let's give the FF cam a 2 stop advantage this time ...





Now that's pretty close ... but even so the D3s has better saturation and skin texture.

The numbers still are not quite adding up. So at the least I'd say that, without finding and using pathological conditions as with Lee Jay's hyper-crop or the really dark images GB or BOBN (I forget which) used, we are more or less unable to see the exact effects in the real world.

There is always more or less distance than we expect because everything is not equal out there.

But so far the FF images are actually better than they should be in every case ... and I'm thinking that those smaller pixels on the APS-C sensors are the culprits

--
I am but one opinion in a sea of opinions ... right?
http://kimletkeman.blogspot.com
http://letkeman.net/Photos

Bokeh_freak said:

So increasing pixel density at lower densities generally improves resolution and image quality up to a point. At which point, any further increase causes an exponential increase in noise, decreasing image quality. Thank you geniuses.

Click to expand...

That's not what was said. You're welcome, master of reading comprehension.

I'm sorry was there something that I said that was incorrect? Do we have any 100mp low light cameras out there?

Kim Letkeman said:

Great Bustard said:

Ideally, this post should summarize the four threads, and be the end of it. However, like anything else, ideal situations rarely exist.

Click to expand...

A man can dream ...

I take issue with Lee Jay's example. The 1:1 big pixel image is cropped and scaled an enormous amount ... this is surely at the point of being a pathological case. And since it is widely quoted by the GoF, it does not meet the "widely applicable" standard for such use ...

That's one opinion at least.

Now ... moving on to BOBN's wonderfully arrogant response to this image:

Kim Letkeman said:

Kim Letkeman said:

Processing?
I'll do it myself properly. might need a new thread.

Click to expand...

Click to expand...

The answers are "no processing, just cropping" ... and define "properly" when there are only jpegs available on the site. Or was that just another case of unbridled ego getting the best of the writer?

Perhaps things would be more believable if we pit the latest full frame camera against the latest APS-C ... and let's give the FF cam a 2 stop advantage this time ...





Now that's pretty close ... but even so the D3s has better saturation and skin texture.

The numbers still are not quite adding up. So at the least I'd say that, without finding and using pathological conditions as with Lee Jay's hyper-crop or the really dark images GB or BOBN (I forget which) used, we are more or less unable to see the exact effects in the real world.

There is always more or less distance than we expect because everything is not equal out there.

But so far the FF images are actually better than they should be in every case ... and I'm thinking that those smaller pixels on the APS-C sensors are the culprits

Click to expand...

My opinion on comparing jpgs is well known -- not useful in terms of comparing tech, but the best way to compare if you're a jpg shooter.

Moving right along, the D5000 ISO 1600 crop definitely looks cleaner than the D3s ISO 6400 crop to me, as one would expect.

Furthermore, the ISO 6400 D700 crop does look slightly cleaner and more detailed than the ISO 3200 D5000 crop, again, as expected, since the D700 would gather 0.2 stops more light for the same f-ratio and shutter speed, and a lens will resolve better on a FF DSLR than a crop DSLR (near the center, anyway, for the same f-ratio).

Thus, by taking advantage of the higher resolution, the D700 could extend it's lead over the D5000 by applying NR and normalizing the detail.

But, again, if we're talking jpgs, then I really make no predictions, because the variations in the jpg engines could be huge.

In any case, the advantage of FF is two-fold in this type of comparison:

1. FF puts over a stop more light on the sensor for any given f-ratio and shutter speed.
2. FF resolves more detail due to the larger sensor unless the crop camera has a lens that resolves better in proportion to the sensor diagonals

There is a third, but minor advantage for FF as well: sensors are often (but not always) a bit more efficient at higher ISOs. So, with the FF sensor using a stop higher ISO than the crop sensor, it may close the gap if it is less efficient, or expand the gap if it is more efficient.

Bokeh_freak said:

I'm sorry was there something that I said that was incorrect?

Click to expand...

Yes. Here it is:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37574866

So increasing pixel density at lower densities generally improves resolution and image quality up to a point. At which point, any further increase causes an exponential increase in noise, decreasing image quality.

So long as the sensor efficiency remains at least as good with the higher PD (pixel density, and, of course, same size) sensor, then the higher PD sensor will always have higher IQ.

In the situations when the additional noise at the higher level of detail is objectionable, we can apply NR to the photo from the higher PD sensor, and thus have either a slightly cleaner photo with more detail, or a slightly more detailed photo with the same noise, since NR is more efficient than binning.

Bokeh_freak said:

Do we have any 100mp low light cameras out there?

Click to expand...

Not that I'm aware of. But we do have the 5D2 (21 MP) and 5D (13 MP), and the 5D2 beats the 5D, hands down. Why? Because the 5D2 has a more efficient sensor. Would the 5D2 sensor have been more efficient still if it had the same 13 MP of the 5D? No.

Bokeh_freak said:

Thank you geniuses.

Click to expand...

You're welcome.

By your logic, why wouldn't the mfg increase it further than... To 40...60...100? I'm pretty sure they have the capability considering they have 16mp on small sensors.

Bokeh_freak said:

By your logic, why wouldn't the mfg increase it further than... To 40...60...100? I'm pretty sure they have the capability considering they have 16mp on small sensors.

Click to expand...

I don't have an answer for that, but I'm sure Bob has a good idea. So, the rest of this post is pure supposition on my part.

More pixels requires a greater data throughput. Thus, increasing the PD may result in a lower frame rate. It is possible that the D3 was made 12 MP to keep a high frame rate, as opposed to the D3x, which had a high pixel count.

Furthermore, the microlens tech may be the limiting factor. That is, a higher PD sensor may be a slam dunk, but the current microlens tech may not be up to snuff. Nikon has recently made some large advances in terms of microlenses.

In addition, perhaps a sensor with more pixels requires more power, and the additional power introduces more noise. Thus, they need to first reduce the power requirements of the sensor to where the EMI can be shielded.

Lastly, at 100%, photos will appear less detailed due to the higher enlargement they are being viewed at, in addition to the higher per-pixel (as opposed to per-image) noise. This will lead the uneducated public to believe that the IQ is lower. So, public perception requires the engineers to hold back.

As I said, all the above are pure supposition on my part. Bob is the one who is more qualified, by far, than I am to answer that question.

And, by the way, that is a good question.

The F550 arrives soon and many are hoping there will be a noticable improvement over previous models. I think you mentioned Fuji has a particularly good duel micro- lens set-up. Would you expect the 16mp BS I sensor and EXR pixel binning technique to break new ground or be no more than a small incremental step, at best?

Sorry if this is only loosely about PD. Thanks,

Nick

Great Bustard said:

Kim Letkeman said:

Great Bustard said:

Ideally, this post should summarize the four threads, and be the end of it. However, like anything else, ideal situations rarely exist.

Click to expand...

A man can dream ...

I take issue with Lee Jay's example. The 1:1 big pixel image is cropped and scaled an enormous amount ... this is surely at the point of being a pathological case. And since it is widely quoted by the GoF, it does not meet the "widely applicable" standard for such use ...

That's one opinion at least.

Now ... moving on to BOBN's wonderfully arrogant response to this image:

Great Bustard said:

Great Bustard said:

Processing?
I'll do it myself properly. might need a new thread.

Click to expand...

Click to expand...

The answers are "no processing, just cropping" ... and define "properly" when there are only jpegs available on the site. Or was that just another case of unbridled ego getting the best of the writer?

Perhaps things would be more believable if we pit the latest full frame camera against the latest APS-C ... and let's give the FF cam a 2 stop advantage this time ...





Now that's pretty close ... but even so the D3s has better saturation and skin texture.

The numbers still are not quite adding up. So at the least I'd say that, without finding and using pathological conditions as with Lee Jay's hyper-crop or the really dark images GB or BOBN (I forget which) used, we are more or less unable to see the exact effects in the real world.

There is always more or less distance than we expect because everything is not equal out there.

But so far the FF images are actually better than they should be in every case ... and I'm thinking that those smaller pixels on the APS-C sensors are the culprits

Click to expand...

My opinion on comparing jpgs is well known -- not useful in terms of comparing tech, but the best way to compare if you're a jpg shooter.

Moving right along, the D5000 ISO 1600 crop definitely looks cleaner than the D3s ISO 6400 crop to me, as one would expect.

Furthermore, the ISO 6400 D700 crop does look slightly cleaner and more detailed than the ISO 3200 D5000 crop, again, as expected, since the D700 would gather 0.2 stops more light for the same f-ratio and shutter speed, and a lens will resolve better on a FF DSLR than a crop DSLR (near the center, anyway, for the same f-ratio).

Thus, by taking advantage of the higher resolution, the D700 could extend it's lead over the D5000 by applying NR and normalizing the detail.

But, again, if we're talking jpgs, then I really make no predictions, because the variations in the jpg engines could be huge.

In any case, the advantage of FF is two-fold in this type of comparison:

1. FF puts over a stop more light on the sensor for any given f-ratio and shutter speed.
2. FF resolves more detail due to the larger sensor unless the crop camera has a lens that resolves better in proportion to the sensor diagonals

There is a third, but minor advantage for FF as well: sensors are often (but not always) a bit more efficient at higher ISOs. So, with the FF sensor using a stop higher ISO than the crop sensor, it may close the gap if it is less efficient, or expand the gap if it is more efficient.

Click to expand...

I am afraid that now that I see things in terms of variables, it appears that you seem to want to explain away differences in IQ by playing the variable card every time a camera with a lower PD appears to have a good as or better IQ than a camera with the higher PD.

There seems to be no exception to this from what I can gather

And you still seem to remain keen to add a variable, NR to the higher PD image as the way to prove it is does in fact have better IQ than the lower PD image.

In this thread you are concerned about the negative effect of jpegs and are calling into doubt the comparison of jpeg images because they process the RAW but then you want to use NR on one of the images and thats OK!

Without NR, from what I have seen so far, the very best that can be said about the comparisons I have seen is that the higher PD image might on a good day, get close to the low PD image.

You say that the reason that the high PD image it is not as good is because of variables like ISO efficiency or .2 f stops difference or resolving proportions to sensor diagonals or jpegs vs RAW and before that it was heat. Yet in the last thread you repeatedly claimed that camera variables were minimal to the point of insignificance for equivalence tests, when I raised the issue with you in response to your friends comments to Kim's last test, where they listed a whole bunch of different camera variables that invalidated his test.

If they are insignificant in equivalence tests, why are you listing so many now. If they are significant as you imply in this post then the points I made in my last thread were right and you are not controlling variables properly in your equivalence tests but if they are not significant which is what you told me in response to the criticism of Kim's test, then they don't explain the differences in IQ that we are seeing in these images.

Perhaps the lesson that users should draw from all this is that they are much better off using lower PD cameras because in the real world there is a plethora of variables between lower and higher PD cameras, but it turns out that those variables always favour the lower PD camera. Which is more or less the empirical evidence that most users have been finding in the real world.

duplicate posted in error

Wellington100 said:

I am afraid that now that I see things in terms of variables, it appears that you seem to want to explain away differences in IQ by playing the variable card every time a camera with a lower PD appears to have a good as or better IQ than a camera with the higher PD.

Click to expand...

Yeah, imagine that -- varaibles matter. For example, a low PD system with a sharp lens outresolving a high PD system with a dull lens. A low PD systems with a more effiicent sensor having lower noise than a high PD system with a less efficient sensor.

Obviously, I'm up to no good taking these variables into account.

Wellington100 said:

There seems to be no exception to this from what I can gather

And you still seem to remain keen to add a variable, NR to the higher PD image as the way to prove it is does in fact have better IQ than the lower PD image.

Click to expand...

Often, I don't want to apply NR, because I prefer the more detailed photo instead. Does the phrase "more IQ options" ring a bell?

Wellington100 said:

In this thread you are concerned about the negative effect of jpegs and are calling into doubt the comparison of jpeg images because they process the RAW but then you want to use NR on one of the images and thats OK!

Click to expand...

That's exactly correct. The amount of NR, if any, applied to a photo is an option that I, as a photographer, have control over. The PP applied by a jpg engine is outside my control.

If, as is all too often the case, the writers of the jpg engine decide that low noise is the holy grail, and all but kill every bit of detail in that quest, I end up with a "horrid smear" from a poorly written jpg engine.

Wellington100 said:

Without NR, from what I have seen so far, the very best that can be said about the comparisons I have seen is that the higher PD image might on a good day, get close to the low PD image.

Click to expand...

Then look harder. Look at a pic of a scene from a 5D and a 5D2 taken with the same settings and get back to me.

Wellington100 said:

You say that the reason that the high PD image it is not as good is because of variables like ISO efficiency or .2 f stops difference or resolving proportions to sensor diagonals or jpegs vs RAW and before that it was heat. Yet in the last thread you repeatedly claimed that camera variables were minimal to the point of insignificance for equivalence tests, when I raised the issue with you in response to your friends comments to Kim's last test, where they listed a whole bunch of different camera variables that invalidated his test.

Click to expand...

Here's the pic:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37549285

Not a long exposure (thermal noise doesn't apply), not too much in the way of shadows (read noise less an issue).

Was the higher PD sensor unfairly favored with a more efficient sensor? Is that relevant when the claim is that a higher PD sensor delivers a photo at least as good so long as the sensor is at least as efficient?

Wellington100 said:

If they are insignificant in equivalence tests, why are you listing so many now. If they are significant as you imply in this post then the points I made in my last thread were right and you are not controlling variables properly in your equivalence tests but if they are not significant which is what you told me in response to the criticism of Kim's test, then they don't explain the differences in IQ that we are seeing in these images.

Click to expand...

I take it that I wasn't "perfectly clear" with this post:

http://forums.dpreview.com/forums/read.asp?forum=1012&message=37572594

Wellington100 said:

Perhaps the lesson that users should draw from all this is that they are much better off using lower PD cameras because in the real world there is a plethora of variables between lower and higher PD cameras, but it turns out that those variables always favour the lower PD camera. Which is more or less the empirical evidence that most users have been finding in the real world.

Click to expand...

Ah. Got it. My 5D is better than the 5D2. The E3 is better than the E5. The 50D is better than the 7D. It makes so much sense now that you say it.

Kim Letkeman said:

Great Bustard wrote:
Now ... moving on to BOBN's wonderfully arrogant response to this image:

Click to expand...

I'm cool with that, I can take a put-down from Kim but plug on regardless, so that's what I'm doing here

Kim Letkeman said:

Kim Letkeman said:

Kim Letkeman said:

Processing?

Click to expand...

Click to expand...

Click to expand...

The answer is, ex camera JPEG, which really renders the whole demonstration valueless, since it is likely that the in camera JPEG processing will be very different

Kim Letkeman said:

Kim Letkeman said:

Kim Letkeman said:

I'll do it myself properly. might need a new thread.

Click to expand...

Click to expand...

Click to expand...

and that was my intention

Kim Letkeman said:

The answers are "no processing, just cropping" ... and define "properly" when there are only jpegs available on the site.

Click to expand...

but there are only jpegs of that image . On the other hand there are raws of the normal still life test scene (which at least is a real scene as opposed to the poster in Kim's demo)

Kim Letkeman said:

Or was that just another case of unbridled ego getting the best of the writer?

Click to expand...

Well, trying to think of another thing we could do to illustrate the point,

Kim Letkeman said:

Perhaps things would be more believable if we pit the latest full frame camera against the latest APS-C ... and let's give the FF cam a 2 stop advantage this time ...

Click to expand...

Or, we could do this. 8 different cameras lots of different pixel size, taken from the raw files, all processed the same way dcraw -6 -W (this time its got some gamma correction so not too dark for you, but should, hopefully be exactly the same for all, cropped and resampled to 8MP equivalent in cinepaint, 16 bit. Written out to JPEG 85% quality 0.06 smoothing, same for each. FF cameras at 6400, APS-C at 32 (to approximate equal light on the sensor, as per Steen's suggestion). IR's exposure is a little inconsistent, but I can't do much about that.

Anyhow, since small pixels are clearly inferior, it should be easy to order these by pixel size. Off you go.

































--
Bob

NIK11 said:

The F550 arrives soon and many are hoping there will be a noticable improvement over previous models. I think you mentioned Fuji has a particularly good duel micro- lens set-up. Would you expect the 16mp BS I sensor and EXR pixel binning technique to break new ground or be no more than a small incremental step, at best?

Sorry if this is only loosely about PD. Thanks,

Click to expand...

As I said in my post above, this question is best answered by Bob, who is much more familiar with the new tech than I am. I will follow his reponse with interest.

Could you please comment on my query - 8 lines up.

Thanks,

Nick

Great Bustard said:

Bokeh_freak said:

By your logic, why wouldn't the mfg increase it further than... To 40...60...100? I'm pretty sure they have the capability considering they have 16mp on small sensors.

Click to expand...

I don't have an answer for that, but I'm sure Bob has a good idea. So, the rest of this post is pure supposition on my part.

More pixels requires a greater data throughput. Thus, increasing the PD may result in a lower frame rate. It is possible that the D3 was made 12 MP to keep a high frame rate, as opposed to the D3x, which had a high pixel count.

Furthermore, the microlens tech may be the limiting factor. That is, a higher PD sensor may be a slam dunk, but the current microlens tech may not be up to snuff. Nikon has recently made some large advances in terms of microlenses.

In addition, perhaps a sensor with more pixels requires more power, and the additional power introduces more noise. Thus, they need to first reduce the power requirements of the sensor to where the EMI can be shielded.

Lastly, at 100%, photos will appear less detailed due to the higher enlargement they are being viewed at, in addition to the higher per-pixel (as opposed to per-image) noise. This will lead the uneducated public to believe that the IQ is lower. So, public perception requires the engineers to hold back.

As I said, all the above are pure supposition on my part. Bob is the one who is more qualified, by far, than I am to answer that question.

And, by the way, that is a good question.

Click to expand...

I think there are lots of reasons. I think the processing and memory limitation is one. Sensor pixel count is actually dragging behind Moore's law, which if everything is due to memory and processor, it should match, but I think the manufacturers are taking a price/profit advantage each generation.

then there are marketing reasons, for a start, if the 100MP cam was released now, where would the next update come from? Why would anyone buy a new camera? Also market perception. If the market expects 16MP as the next step, what would it make if suddenly offered 100MP? Would people actually buy, especially since there would be a chorus of apparently informed pundits saying how bad it would be.

Plus, i think many markets follow the pro markets, and pro's are quite conservative, they want advance but not risk. Sometimes its easy to get wrong. In the 2nd generation pro cameras Nikon decided 4.7MP was enough, which is what most pro's said at the time, Canon decided the market would take 8MP, and were right. What is interesting is amateur cameras at all levels from DSLR to P&S tend to trail the top end, but rarely get ahead.

One other reason is that they reach process limits for the fab line their working on. canon had kind of hit a block with the 50D/5DII until they moved to their new fab line - it's one reason fabless strategies are more agile.
So, in summary, a combination of all those reasons.
--
Bob

NIK11 said:

Could you please comment on my query - 8 lines up.

Click to expand...

Just did, like to take my challenge?

--
Bob

bobn2 said:

NIK11 said:

Could you please comment on my query - 8 lines up.

Click to expand...

Just did, like to take my challenge?

Click to expand...

I think we have crossed over. My above query is about the upcoming F550. Care to speculate?

Nick