AA filter weakened, dying

[RichA]

The AA filter is likely to still be used, but could be minimized in
other cameras than the E-M5.

From Dpreview poster: (from French magazine)

L'Olympus OM-D E-M5 est l'appareil 4:3 qui, à ce jour, produit la
meilleure qualité d'image. Des questions sont donc rapidement
survenues concernant son capteur. S'agit-il d'un nouveau capteur
Olympus ou du 16 MP de Panasonic ? S'il n'y a pas eu de réponse
officielle d'Olympus, les excellents journalistes du magazine français
Chasseur d'images ont trouvé la réponse. Il s'agit bien du même
capteur que le Panasonic GX1 mais Olympus:

"a allégé le passe-bas en calant sa F de coupure un poil au-delà de la
F de Nyquist, alors que les Pana sont calés un poil en-dessous. Du
coup, il y a certes un très léger risque de moiré (c'est ce qui m'a
mis sur la piste, du reste, et j'ai eu la confirmation par un Ingé de
Pana par la suite), mais, du coup, le piqué natif est extrêmement
élevé (surtout avec les bons, très bons objectifs Oly). Cela permet de
ne pas trop accentuer, et même de lisser un peu activement à 1600 iso
en gardant du détail sous le pied. Réglage ultra pointu et fin,
nécessitant du doigté, mais résultats assez bluffants je trouve: le
meilleur 4:3 à 1600 ISO, fingers in ze nose !"

Propos de Ronan Loaëc sur le forum Chassimages.

Globally : it IS the Panny G3/GX-1 sensor, but Oly used a way weaker
AA filfter. The result is a risk of moire, but extremely sharp and
detailed files, allowing further smoothing for cleaner high ISO files.
They also claim they had confirmation from a Panny R&D tech.

I though it'd be of interest to many of you techy guys...

Marla.

crudely translated:

The Olympus OM-D-E device M5 is 4:3 which, to date, produced the best
image quality. Issues are occurring rapidly on its sensor. Is this a
new sensor 16 MP Olympus or Panasonic? If there has been no official
response from Olympus, the great journalists of the French magazine
Chasseur d'images have found the answer. It's the same sensor as the
Panasonic Olympus but GX1:

"alleviated the low-pass cutoff F wedging its a bit beyond the Nyquist
F, while the Pana are wedged underneath a coat. So, there is certainly
a very slight risk of moiré (c is what put me on the track, the rest,
and I got a confirmation Ingé Pana later), but, suddenly, the native
sharpness is extremely high (especially with the good, very Oly right
goals). This avoids creating too much, and even a bit of smoothing to
1600 iso actively keeping the details under foot. adjustment and ultra
sharp end, requiring careful handling, but I find quite impressive
results: the best 4:3 to 1600 ISO, fingers in ze nose! "

[David J Taylor]

"RichA" <rande...@gmail.com> wrote in message
news:51dc668a-6acd-4e8d...@b26g2000vbt.googlegroups.com...

> The AA filter is likely to still be used, but could be minimized in
> other cameras than the E-M5.

"AA filter weakened, dying ...."

As the sensor resolution approaches and surpasses that of the lenses,
there is less and less need for an anti-alias filter. The recent Nikon
D800 confirms this trend. No surprise. Oversampling with a weaker AA
filter works well in audio, as you may know, and in that application there
is no lens MTF to reduce the high spatial frequencies before they reach
the filter & sensor.

David

[RichA]

On May 29, 9:53 am, "David J Taylor" <david-
tay...@blueyonder.co.uk.invalid> wrote:
> "RichA" <rander3...@gmail.com> wrote in message

From a purely esthetic standpoint, using the D800E and not the D800
makes images look (at large sizes) like you have better lenses.
Arbitrarily, I can liken it to going from a $500 lens to a $1000 lens
or :(again, subjective) stopping the 17-55mm lens down from f2.8 to
f5.6.. I wish I had the two bodies to really run through their paces,
but I think that as evidence grows, the D800E will sell in greater
numbers than perhaps Nikon anticipated, next to the D800.

[Andrew Haley]

David J Taylor <david-...@blueyonder.co.uk.invalid> wrote:
>
> "RichA" <rande...@gmail.com> wrote in message
> news:51dc668a-6acd-4e8d...@b26g2000vbt.googlegroups.com...
>> The AA filter is likely to still be used, but could be minimized in
>> other cameras than the E-M5.
>
> "AA filter weakened, dying ...."
>
> As the sensor resolution approaches and surpasses that of the
> lenses, there is less and less need for an anti-alias filter. The
> recent Nikon D800 confirms this trend.

Erm, no. That'll be the D800e.

> No surprise. Oversampling with a weaker AA filter works well in
> audio, as you may know, and in that application there is no lens MTF
> to reduce the high spatial frequencies before they reach the filter
> & sensor.

But there may well be an analog filter before the audio DAC: even if
its modulator is running at 6.144 MHz (i.e. 128x oversampling) you
still don't want signals above fs/2 getting in there. The data sheet
I'm looking at (for the AD1871, fairly typical) shows an LRC input
filter. Maybe when we have 128x oversampling camera sensors there
will be some logic to all this "who needs an AA filter?" stuff.

Andrew.

[David J Taylor]

"Andrew Haley" <andr...@littlepinkcloud.invalid> wrote in message
news:wOidnR0I3OcoxVvS...@supernews.com...

> David J Taylor <david-...@blueyonder.co.uk.invalid> wrote:
>>
>> "RichA" <rande...@gmail.com> wrote in message
>> news:51dc668a-6acd-4e8d...@b26g2000vbt.googlegroups.com...
>>> The AA filter is likely to still be used, but could be minimized in
>>> other cameras than the E-M5.
>>
>> "AA filter weakened, dying ...."
>>
>> As the sensor resolution approaches and surpasses that of the
>> lenses, there is less and less need for an anti-alias filter. The
>> recent Nikon D800 confirms this trend.
>
> Erm, no. That'll be the D800e.

It's both - having the ability to compare with and without the AA filter.

>> No surprise. Oversampling with a weaker AA filter works well in
>> audio, as you may know, and in that application there is no lens MTF
>> to reduce the high spatial frequencies before they reach the filter
>> & sensor.
>
> But there may well be an analog filter before the audio DAC: even if
> its modulator is running at 6.144 MHz (i.e. 128x oversampling) you
> still don't want signals above fs/2 getting in there. The data sheet
> I'm looking at (for the AD1871, fairly typical) shows an LRC input
> filter. Maybe when we have 128x oversampling camera sensors there
> will be some logic to all this "who needs an AA filter?" stuff.
>
> Andrew.

Yes, there is a filter, of course, but if there are no frequencies which
may alias, there is nothing to filter. At the resolution level now seen
in the best cameras, there may be so little energy coming from the lens at
these high spatial frequencies in "typical" scenes that the need for extra
AA filtering is minimise - the lens does a good enough job on its own. Of
course, you will be able to find certain subjects where aliasing /does/
occur.

Will we still need AA filters with 50 Mp or 75 Mp sensors?

David

[Floyd L. Davidson]

"David J Taylor" <david-...@blueyonder.co.uk.invalid> wrote:
>"Andrew Haley" <andr...@littlepinkcloud.invalid> wrote in message
>news:wOidnR0I3OcoxVvS...@supernews.com...
>> David J Taylor <david-...@blueyonder.co.uk.invalid> wrote:
>>>
>>> "RichA" <rande...@gmail.com> wrote in message
>>> news:51dc668a-6acd-4e8d...@b26g2000vbt.googlegroups.com...
>>>> The AA filter is likely to still be used, but could be minimized in
>>>> other cameras than the E-M5.
>>>
>>> "AA filter weakened, dying ...."
>>>
>>> As the sensor resolution approaches and surpasses that of the
>>> lenses, there is less and less need for an anti-alias filter. The
>>> recent Nikon D800 confirms this trend.
>>
>> Erm, no. That'll be the D800e.
>
>It's both - having the ability to compare with and without the AA filter.

It's neither, and the suggestion the two of you are entertaining is
hilarious. The difference between a D800 and a D800E is very very small
compared to the sensor resolution required to eliminate the need for
an Anti-Aliasing filter with today's modern lenses. These are 36 MP
cameras and it will take upward of 150-200 MP to obsolete the AA filter.

>>> No surprise. Oversampling with a weaker AA filter works well in
>>> audio, as you may know, and in that application there is no lens MTF
>>> to reduce the high spatial frequencies before they reach the filter
>>> & sensor.
>>
>> But there may well be an analog filter before the audio DAC: even if
>> its modulator is running at 6.144 MHz (i.e. 128x oversampling) you
>> still don't want signals above fs/2 getting in there. The data sheet
>> I'm looking at (for the AD1871, fairly typical) shows an LRC input
>> filter. Maybe when we have 128x oversampling camera sensors there
>> will be some logic to all this "who needs an AA filter?" stuff.
>>
>> Andrew.
>
>Yes, there is a filter, of course, but if there are no frequencies which
>may alias, there is nothing to filter. At the resolution level now seen
>in the best cameras, there may be so little energy coming from the lens at
>these high spatial frequencies in "typical" scenes that the need for extra
>AA filtering is minimise - the lens does a good enough job on its own. Of
>course, you will be able to find certain subjects where aliasing /does/
>occur.
>
>Will we still need AA filters with 50 Mp or 75 Mp sensors?

Yes.

--
Floyd L. Davidson <http://www.apaflo.com/>
Ukpeagvik (Barrow, Alaska) fl...@apaflo.com

[Andrew Haley]

David J Taylor <david-...@blueyonder.co.uk.invalid> wrote:
>
> Yes, there is a filter, of course, but if there are no frequencies
> which may alias, there is nothing to filter. At the resolution
> level now seen in the best cameras, there may be so little energy
> coming from the lens at these high spatial frequencies in "typical"

> scenes that the need for extra AA filtering is minimised

... as is the need for a high-resolution camera ...

> - the lens does a good enough job on its own. Of course, you will
> be able to find certain subjects where aliasing /does/ occur.
>
> Will we still need AA filters with 50 Mp or 75 Mp sensors?

I think so. Lenses on such cameras will, hopefully be high-resolution
too, otherwise what's the point? There's no reason why they should
not be: for example, the Schneider Apo-Digitars already show what's
possible.

Andrew.