# Re: Diffraction limit for APS-C DLSRs

Discussion in 'Digital Photography' started by Kennedy McEwen, Jan 4, 2011.

1. ### Kennedy McEwenGuest

In article <>, Alfred Molon
<> writes
>At what aperture does diffraction start limiting significantly the
>resolution with the current brand of APS-C DLRs (with 12, 16 and 18MP)?
>
>Is F11 still fine, can you go up to F16 or F22?

Work it out yourself.
Radius of the Airy Disc is about 1.22 x L x f/#, where L is the
wavelength of light, say 0.55um for the middle of the visible band.

The closest that two points can be resolved by the lens is when the edge
of one disc is just at the centre of the other, which means the points
are separated by one Airy Disc radius, and you need two pixels in that
distance to be able to resolve the separation.
so 2p = 1.22 x L x f/#
f/# = 1.64 x p / L

For 18Mp APS-C the pixels are about 4.2um pitch, so the *maximum* useful

So f/16 won't resolve any more than f/11 in practice.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 4, 2011

2. ### Kennedy McEwenGuest

In article <>, Alfred Molon
<> writes
>In article <>, Kennedy McEwen
>says...
>> So f/16 won't resolve any more than f/11 in practice.

>
>Thanks. So roughly one should go beyond F11, unless perhaps DOF is
>needed, which however comes at the expense of available resolution.
>

I hope that was a typo - "shouldn't go beyond f/11 unless..." ;-)

>One more question, what effect does the diffraction have - is it a low-
>pass filter?

Diffraction is a low pass "ramp" filter - the contrast reduces fairly
linearly with spatial frequency gradually tapering off towards the
cut-off frequency of 1/(L x f/#) above which there is zero contrast.

>If so, sharpening could solve (partly) the problem at the
>expense of SNR.

Sharpening can partially compensate, particularly where the MTF is still
significantly greater then zero, but obviously can't recover contrast
when diffraction has attenuated it to zero in the first place - at and
above the diffraction limit.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 4, 2011

3. ### PeabodyGuest

Kennedy McEwen says...

> For 18Mp APS-C the pixels are about 4.2um pitch, so the
> *maximum* useful f/# is about f/12.5.

> So f/16 won't resolve any more than f/11 in practice.

Just to be sure I understand... So, an APS-C camera of five
years ago with, say, 8mp resolution would be less subject to
diffraction limiting, although of course it has less
actually be a usable non-limiting setting?

I'm just looking for reasons to keep my XT.

Peabody, Jan 5, 2011
4. ### Peter IrwinGuest

Peabody <> wrote:
> Kennedy McEwen says...
>
> > For 18Mp APS-C the pixels are about 4.2um pitch, so the
> > *maximum* useful f/# is about f/12.5.

>
> > So f/16 won't resolve any more than f/11 in practice.

>
> Just to be sure I understand... So, an APS-C camera of five
> years ago with, say, 8mp resolution would be less subject to
> diffraction limiting, although of course it has less
> resolution to start with. But for such a camera, f/22 might
> actually be a usable non-limiting setting?

The difference between 8 and 18 megapixels should only be a hair
over one stop. 18 megapixels is only 1.5 times the resolution of
8 megapixels. (I say only 1.5 times, but then again 6x4.5cm film
is only about 1.5 times a linear improvement on 24x36mm film.)

> I'm just looking for reasons to keep my XT.

If your camera is working for what you want to do with it, then
that should be enough reason to keep it.

Peter.
--

Peter Irwin, Jan 5, 2011
5. ### Kennedy McEwenGuest

In article <>, Peabody
<> writes
>Kennedy McEwen says...
>
> > For 18Mp APS-C the pixels are about 4.2um pitch, so the
> > *maximum* useful f/# is about f/12.5.

>
> > So f/16 won't resolve any more than f/11 in practice.

>
>Just to be sure I understand... So, an APS-C camera of five
>years ago with, say, 8mp resolution would be less subject to
>diffraction limiting, although of course it has less
>actually be a usable non-limiting setting?
>

Pretty much, although it won't be as far as f/22 though f/16 would be
worth using. In general, the lower the resolution of the rest of the
system, the higher an f/# you can use before diffraction becomes the
dominant resolution limitation. Also, don't forget that this is the
maximum f/#, beyond which diffraction actually degrades the camera
resolution. Some diffraction blur will be visible a stop or so lower
although, as already mentioned, this can be partially compensated by
sharpening at the expense of increased noise.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 5, 2011
6. ### Kennedy McEwenGuest

In article <>, Pete D
<> writes
>
>
>"Alfred Molon" <> wrote in message
>news:...
>> In article <>, Pete D
>> says...
>>>
>>> Totally impractical for the range of lens sizes I use.

>>
>> Why? You just buy the filters in the size of your largest lens and one
>> (step up) adapter for each (smaller) lens size you have. Adapters cost a
>> few Euro or dollar each.

>So how many will I need to go from my 77mm down to my 49mm?

One. ;-)
I have one that came with my Zuiko 18mm f/3.5, but its not for sale.
http://www.mir.com.my/rb/photography/hardwares/classics/olympusom1n2/shar
ed/zuiko/htmls/18mm.htm
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 7, 2011
7. ### nospamGuest

In article <>, Pete D
<> wrote:

> >> Totally impractical for the range of lens sizes I use.

> >
> > Why? You just buy the filters in the size of your largest lens and one
> > (step up) adapter for each (smaller) lens size you have. Adapters cost a
> > few Euro or dollar each.

>
> So how many will I need to go from my 77mm down to my 49mm?

one ring, although you might be better off with two if you have lenses
between those two sizes.

nospam, Jan 8, 2011
8. ### Kennedy McEwenGuest

In article <080120110811089174%>, nospam
<> writes
>In article <>, Pete D
><> wrote:
>
>> >> Totally impractical for the range of lens sizes I use.
>> >
>> > Why? You just buy the filters in the size of your largest lens and one
>> > (step up) adapter for each (smaller) lens size you have. Adapters cost a
>> > few Euro or dollar each.

>>
>> So how many will I need to go from my 77mm down to my 49mm?

>
>one ring, although you might be better off with two if you have lenses
>between those two sizes.

You can buy a whole set for a few dollars/euro/pounds.
http://cgi.ebay.com/120643691309
http://cgi.ebay.com/150540750157
http://cgi.ebay.com/330396934335
http://cgi.ebay.com/370468290744
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 8, 2011
9. ### Kennedy McEwenGuest

In article <>, Pete D
<> writes
>
>
>"Kennedy McEwen" <> wrote in message
>news:...
>> In article <>, Pete D
>><> writes
>>>
>>>
>>>"Alfred Molon" <> wrote in message
>>>news:...
>>>> In article <>, Pete D
>>>> says...
>>>>>
>>>>> Totally impractical for the range of lens sizes I use.
>>>>
>>>> Why? You just buy the filters in the size of your largest lens and one
>>>> (step up) adapter for each (smaller) lens size you have. Adapters cost a
>>>> few Euro or dollar each.
>>>So how many will I need to go from my 77mm down to my 49mm?

>>
>> One. ;-)
>> I have one that came with my Zuiko 18mm f/3.5, but its not for sale.
>> http://www.mir.com.my/rb/photography/hardwares/classics/olympusom1n2/shar
>> ed/zuiko/htmls/18mm.htm

>
>
>What a monster.

What a brilliant, high performance lens though. I use it on my Canon 5D
regularly as it still out-performs any equivalent.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 10, 2011
10. ### John TurcoGuest

Kennedy McEwen wrote:
>
> In article <>, Alfred Molon
> <> writes
> >In article <>, Kennedy McEwen
> >says...
> >> So f/16 won't resolve any more than f/11 in practice.

> >
> >Thanks. So roughly one should go beyond F11, unless perhaps DOF is
> >needed, which however comes at the expense of available resolution.
> >

> I hope that was a typo - "shouldn't go beyond f/11 unless..." ;-)
>
> >One more question, what effect does the diffraction have - is it a low-
> >pass filter?

>
> Diffraction is a low pass "ramp" filter - the contrast reduces fairly
> linearly with spatial frequency gradually tapering off towards the
> cut-off frequency of 1/(L x f/#) above which there is zero contrast.
>
> >If so, sharpening could solve (partly) the problem at the
> >expense of SNR.

>
> Sharpening can partially compensate, particularly where the MTF is still
> significantly greater then zero, but obviously can't recover contrast
> when diffraction has attenuated it to zero in the first place - at and
> above the diffraction limit.

Kennedy:

times are >earlier< than messages you're responding to!

--
Cordially,
John Turco <>

Marie's Musings <http://fairiesandtails.blogspot.com>

John Turco, Jan 12, 2011
11. ### Kennedy McEwenGuest

In article <>, John Turco
<> writes
>Kennedy:
>
>times are >earlier< than messages you're responding to!
>
>

Interesting! That post does seem to have a time stamp before Alfred's
post it replied to, however it is the only example I have noticed on a
brief check of a few recent samples. Can you point me to any others?

My clock is OK at the moment, however it does resync to my ISP server
every hour - I guess that the time server could have been out on that
previous post, but unlikely. Also, I believe that the time I sent the
message was close to the time stamp on it. Perhaps Alfred's post that
it was in reply to was time stamped later than when it was actually
sent.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 12, 2011
12. ### David J TaylorGuest

> Interesting! That post does seem to have a time stamp before Alfred's
> post it replied to, however it is the only example I have noticed on a
> brief check of a few recent samples. Can you point me to any others?
>
> My clock is OK at the moment, however it does resync to my ISP server
> every hour - I guess that the time server could have been out on that
> previous post, but unlikely. Also, I believe that the time I sent the
> message was close to the time stamp on it. Perhaps Alfred's post that
> it was in reply to was time stamped later than when it was actually
> sent.
> --
> Kennedy

You could try NTP - it's available for many operating systems, including
Windows:

http://www.satsignal.eu/ntp/setup.html

Better than syncing every hour.

Cheers,
David

David J Taylor, Jan 12, 2011
13. ### Kennedy McEwenGuest

In article <igjrd5\$9q1\$-september.org>, David J Taylor
<> writes
>> Interesting! That post does seem to have a time stamp before
>>Alfred's post it replied to, however it is the only example I have
>>noticed on a brief check of a few recent samples. Can you point me
>>to any others?
>>
>> My clock is OK at the moment, however it does resync to my ISP server
>>every hour - I guess that the time server could have been out on that
>>previous post, but unlikely. Also, I believe that the time I sent the
>>message was close to the time stamp on it. Perhaps Alfred's post that
>>it was in reply to was time stamped later than when it was actually
>>sent.
>> -- Kennedy

>
>You could try NTP - it's available for many operating systems,
>including Windows:
>
> http://www.satsignal.eu/ntp/setup.html
>
>Better than syncing every hour.
>

Doh! What do you think NTP does?

As it is, NTP is built into the Mail and News software I use. The
reason it "syncs" every hour is because that is what I have configured
it to do, which is more than adequate and certainly good enough to catch
changes to and from daylight savings time.

Further, having looked at Alfred's previous message it has the following
NNTP-Posting-Date: Tue, 04 Jan 2011 10:56:07 -0600
From: Alfred Molon <>
Newsgroups: rec.photo.digital
Subject: Re: Diffraction limit for APS-C DLSRs
Date: Tue, 4 Jan 2011 20:56:04 +0100
...
X-Received-Date: Tue, 04 Jan 2011 16:56:08 UTC

From the NNTP Posting Date, we see it was received by Alfred's News
Server at 16:56:07 GMT and, from the X Received Date, received by my
News Server at 16:56:08 GMT, one second later. Yet Alfred's computer
time stamped the post as 19:56:04 GMT, some 3hrs later than it was

In fact, the headers of all of Alfred's posts in this thread up to that
of Tue, 11 Jan 2011 06:47:37 GMT show this 3hr discrepancy!

As I suggested earlier, that is precisely where the discrepancy has
arisen: at some stage Alfred's computer gained at least 3hrs, but has
now been corrected. Any response, not only my response, to such posts
within 3hrs would appear to be out of sequence.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 12, 2011
14. ### David J TaylorGuest

"Kennedy McEwen" <> wrote in message
news:...
[]
>>Better than syncing every hour.
>>

> Doh! What do you think NTP does?

NTP does not "sync every hour", it has an adaptive poll interval between
about 1 minute and 20 minutes according to the system requirements, and it
alters the clock rate to provide better accuracy than simply stepping the
clock.

> As it is, NTP is built into the Mail and News software I use. The
> reason it "syncs" every hour is because that is what I have configured
> it to do, which is more than adequate and certainly good enough to catch
> changes to and from daylight savings time.

It's very unlikely that your software includes full NTP - it certainly
doesn't if it has a sync interval setting. More likely your software has
SNTP - simple SNTP.

[]
> As I suggested earlier, that is precisely where the discrepancy has
> arisen: at some stage Alfred's computer gained at least 3hrs, but has
> now been corrected. Any response, not only my response, to such posts
> within 3hrs would appear to be out of sequence.
> --
> Kennedy

I'm glad that you've resolved the issue.

[Posted at 09:20 UTC]

Cheers,
David

David J Taylor, Jan 13, 2011
15. ### Kennedy McEwenGuest

In article <igmg41\$hkf\$-september.org>, David J Taylor
<> writes
>
>"Kennedy McEwen" <> wrote in message
>news:...
>[]
>>>Better than syncing every hour.
>>>

>> Doh! What do you think NTP does?

>
>NTP does not "sync every hour", it has an adaptive poll interval
>between about 1 minute and 20 minutes according to the system
>requirements, and it alters the clock rate to provide better accuracy
>than simply stepping the clock.

It is set to one hour intervals because, as I pointed out to you, that
is how *I* have it configured and, since my clock wasn't out in the
first place, it couldn't "provide better accuracy".

>I'm glad that you've resolved the issue.
>

I didn't have an issue to resolve, however I have identified that any
issue lay elsewhere.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 13, 2011
16. ### David J TaylorGuest

> It is set to one hour intervals because, as I pointed out to you, that
> is how *I* have it configured and, since my clock wasn't out in the
> first place, it couldn't "provide better accuracy".

[]
> --
> Kennedy

To clarify, that's SNTP you are using, as NTP automatically sets the best
poll interval, and does so dynamically. Other features of the NTP
algorithms provide better accuracy.

Cheers,
David

David J Taylor, Jan 14, 2011
17. ### Kennedy McEwenGuest

In article <igp43e\$2f0\$-september.org>, David J Taylor
<> writes
>> It is set to one hour intervals because, as I pointed out to you,
>>that is how *I* have it configured and, since my clock wasn't out in
>>the first place, it couldn't "provide better accuracy".

>[]
>> -- Kennedy

>
>To clarify, that's SNTP you are using, as NTP automatically sets the
>best poll interval, and does so dynamically. Other features of the NTP
>algorithms provide better accuracy.
>

To clarify, you can't make something that wasn't out more accurate.
I suggest you direct your advertising campaign to those who need it.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 14, 2011
18. ### David J TaylorGuest

> To clarify, you can't make something that wasn't out more accurate.
> I suggest you direct your advertising campaign to those who need it.
> --
> Kennedy

using the full NTP, but the simplified SNTP.

Cheers,
David

David J Taylor, Jan 14, 2011
19. ### Kennedy McEwenGuest

In article <igqa9c\$qh3\$-september.org>, David J Taylor
<> writes
>> To clarify, you can't make something that wasn't out more accurate.
>> I suggest you direct your advertising campaign to those who need it.
>> -- Kennedy

>
>using the full NTP, but the simplified SNTP.
>

And a correction in your assumptions David. My PC clock wasn't and
isn't at fault so adding unnecessary crap certainly isn't going to help.
Go shove your software where its needed.
--
Kennedy
Yes, Socrates himself is particularly missed;
A lovely little thinker, but a bugger when he's pissed.
Python Philosophers (replace 'nospam' with 'kennedym' when replying)

Kennedy McEwen, Jan 15, 2011
20. ### David J TaylorGuest

> And a correction in your assumptions David. My PC clock wasn't and
> isn't at fault so adding unnecessary crap certainly isn't going to help.
> Go shove your software where its needed.
> --
> Kennedy

Why so impolite? I never said your PC clock was at fault and NTP is not
"my" software, but it is widely used to keep PCs and servers in sync.
That's all. No need to get upset about it. It sounds as if I caught you