V1.01 Firmware for Nex7 greatly improves shadows

[agorabasta]

My only gripe with the Nex7 noise performance was that the strongly lifted shadows could often get a pink/purple tint especially at higher ISO's.

Generally speaking, a pure purple shift in the shadows is quite normal for the ACR raw process for all bodies that don't apply any sophisticated manipulations to the raw black point in firmware. That's caused by stronger gain applied to the weaker red/blue data and can be cured with shadows tint adjustment under the 'camera calibration'.

The problem with the Nex7 was that shadows tint was not purple, it was more red-pink; so couldn't be corrected with that simple ACR adjustment. Moreover, that tint was at times not the same from shot to shot, it could differ greatly between camera restarts, etc. It looked like the black level cut-off was coded dynamically in the firmware, but that was simply not done right. Btw, Nex7 was not the only cam that suffered from that - I've seen the same with a55 and some much worse with NexC3. From the a77 raw files that I tried to analyse, it seems that cam shares the problem as well. On the other hand, the Nex5N never had that problem at all - and that is why some people still believe it to have better noise performance while in fact it's all about raw processing and converter limitations.

With the recent 1.01 firmware update for Nex7 it's clear that the problem is totally gone. So now the Nex7 shadows are always as good or better than Nex5N, it's not just sometimes. So there may be some improvement for the a77 with the 1.06 firmware too...

[jeep1]

That is very interesting. I wondered also what exactly was meant by wide angle edge improvement? Is this the magenta edge cast that can be seen at certain angles and lighting conditions? Although that mainly applies to third party lenses so can't really imagine Sony being that helpful.

[agorabasta]

The Sony's own SEL16/2.8 had terrible purple shift in the corners under bright natural light conditions.
Unfortunately, I can't test it now - under dim or artificial lighting it didn't show too much anyway.

[agorabasta]

Just tested SEL16 on pale blue autumn sky. The 'cornershift' correction is obviously applied directly to the raw data, rather than being merely coded into the metadata.
Below is the result using an old version of DCRAW (Ufraw 0.18) which naturally ignores all the correction metadata.

[edrice]

Well, this is definitely good news! I just now got around to doing the update so haven't had a chance to test it in daylight but will do it soon. Thanks for all your work!

Ed

[artington]

I have noticed a resolution improvement on the 18-55mm kit lens.

This photo, of beach huts at Hengistbury Head, was taken at 18mm. 1/500, F11, ISO 100.

The crops below are split into two posts because of the 3 image limitation

They show the original and approximate 100% crops at the centre, lower corners and left edge

[artington]

The last two

[agorabasta]

I have noticed a resolution improvement on the 18-55mm kit lens.

You've got CA in the very middle of the frame - means the active OSS element was strongly out of balance.
You simply can't make any res comparisons under such conditions. The OSS must be off when evaluating resolution, especially the OSS as bad as found in the 18-55 and the same in the 55-210 too.

[artington]

I have noticed a resolution improvement on the 18-55mm kit lens.

You've got CA in the very middle of the frame - means the active OSS element was strongly out of balance.
You simply can't make any res comparisons under such conditions. The OSS must be off when evaluating resolution, especially the OSS as bad as found in the 18-55 and the same in the 55-210 too.

You are right about the CA - but why would the OSS have been on when this as taken at 1/500 at 18mm?

[agorabasta]

You are right about the CA - but why would the OSS have been on when this as taken at 1/500 at 18mm?

If the OSS is not switched off in the menu, it stays on. And the active group is slowly floating while you have the shutter button half-pressed. And it may float very far from the optically optimal position.

On the other hand, if the OSS in 'off' position parks the active group off-centre, the problem would be present even in that 'off' position.
But I have never seen any lens doing so, and I've seen too many copies of native E-mount lenses.

[David Kilpatrick]

Sigma's OS in many lenses does park the lens off-centre when power is off, centres it when power is on, and floats the group when OS is enabled. It requires active control to be centered. I do not know what OSS in NEX needs - whether the lens must be powered to hold a centered position, or defaults to it without needing 'effort' to keep it there.

David

[agorabasta]

I think the Sigma OS implementation is among the very best. I have never seen it introducing any distortions in the frame centre, even if the lens is not top of the line EX or something.
And their active centring may be a good idea too, as the active group's central position with the OS off may stay non-reliant on any mechanical lock.

[agorabasta]

Just found a nasty glitch in the Nex7 flash behaviour. The rear-sync mode is not simply 'rear', it's 'rear-slow' - the cam measures for the ambient only, then fires a flash in the end of exposure. So no normal rear-sync. And it doesn't matter if the internal flash is used or the F43.
And, btw, the normal pre-flash is still fired, flash exposure is OK, it's the shutter speed that's measured for the ambient light alone.
And I don't know if the problem was there with the old firmware.