A more realistic impression of the Sony A7S low light performance at ISO 12,800

[hmcindie]

 

The C100/C300 low light test is irrelevant and incompetently handled, fwiw. Whoever performed it approached it wrong. Both cameras should look the same first of all, secondly there's no stopping down to compensate for the ISO push.

 

C100 will look "cleaner" than C300 because AVCHD will apply more NR and thus decrease noise (and details). C300 has 50mbps 4:2:2 mpeg2 and it will keep noise grain better and not smooth it.

 

If those two are your only complaints and a test becomes immediately "irrelevant"...that's just weird.

[Policar]

C100 will look "cleaner" than C300 because AVCHD will apply more NR and thus decrease noise (and details). C300 has 50mbps 4:2:2 mpeg2 and it will keep noise grain better and not smooth it.

 

If those two are your only complaints and a test becomes immediately "irrelevant"...that's just weird.

 

 

The C100 should have a slightly different noise texture, but it shouldn't be a different color temperature and exposure at apparently equal settings so right off the bat it's obvious that whoever did this is doing something wrong.

 

When you're examining low light performance by testing ISO, you need to stop down or change the shutter speed to compensate. If you expose everything much too brightly it gives an unrepresentative result...

 

In the test posted, they're boosting the ISO, but not stopping down or decreasing light levels to keep the exposure consistent. So the image won't look noisier… just brighter. The test is totally useless for this reason. If I point a C300 at a light bulb at ISO 80,000 of course it will be noiseless, it will be pure white… and it's that but a little less egregious. To compare ISO meaningfully you need to compensate and shoot at an even exposure, by stopping down, adjusting shutter speed, adjusting light levels etc… Look at virtually every other ISO comparison ever made.

 

This "test" is just… incredibly stupid and whoever performed it doesn't have a basic grasp of exposure. Not trying to be rude, but if you don't know what you're doing you might use this as a reference, and it's useless as one. The C300 is only this clean at extreme ISOs when extremely overexposed and, likewise, only noisy at low ISOs when underexposed. A series of shots exposed at key at different ISOs would have been useful, and that is how every other worthwhile ISO test I've seen has been handled. (I've seen this mistake repeated, but it doesn't make this any better or more useful.)

[noone]

check out "yak films"  filming on the A7S preproduction model

it looks very nice, sometimes clean and then not so clean (due to different sensor modes?, as they say in the description they have used some vintage nikon glass on it) we can also see the famous rolling shutter but then also some choppy motion (no ND filters I guess or I don't know) 

what you think guys?

'>

 

  

[Michael Ma]

This just in: http://goo.gl/6pXSyN

dpreview's iso tests of the Sony A7S

 

Something just occurred to me after looking at this.  We've been looking at a straight RAW to JPEG conversion the whole time.  Maybe even the black and shadows raised?.  Of course there will be noise at ISO12800 in RAW.  You can get the 5D Mark III to look noisy at ISO 800 if it's dark enough in RAW.

 

Looking how well the in-body noise removal looks for JPEG, I expect the video to perform as good as the smoked fish video, especially resampled to 1080p.

[aim120]

In that dpreview comparison the A7s has the cleanest jpeg image compared to any other camera,So i agree the video quality should be really good like those fish video.

 

I think Andrew rushed this article.

[Wulf]

Looks very nice. Sony is back! If possible, I would like to avoide shooting in raw... :-)

For several years during the DSLR-boom, there was no need for real progress in sensor technology - customers were happy with all this plastic bodies and the money flows.

Two years advantage is a long time nowadays, so I'm not that surprised that Sony came back when compared with the DSLRs of the last generation.

At least Sony has done it once before - they invented the D3 sensor and rescued Nikon btw.

[Kolen Cheung]

A quick question. Does the A7S has any 30 minutes limit for video for the US version?

Thanks!

[varicam]

A quick question. Does the A7S has any 30 minutes limit for video for the US version?

Thanks!

 

From what I heard, it does have a 30min limit. 

[Imaginate]

From what I heard, it does have a 30min limit. 

Seems like this is case as per Matthew Allard's review on Newshooter.

 

This is a real deal breaker for me and one of the MANY reasons why I'm going to keep my GH4.  I shoot mostly doc stuff with lots of long interviews and live events.  Also the dynamic range doesn't look that impressive at all. The 8 bit footage out of cam looks very DSLRish and not much better than a 5D(granted this will probably change a bit with 4k and 4:2:2). I wasn't that impressed with the Den Lennie footage either and this review just confirms it. The 10-bit image out of the GH4 looks like it's going to kick the crap out of most images from this camera. It might be a little too early to make this judgement without seeing more 4k footage of both cameras, but right now the GH4 footy looks better straight out of camera and so far @ 4k downscaled into 1080 on a recorder(yes I think the TEST clip of the little girls looks better than the Den Lennie stuff). Don't get me wrong, the low light is nice, but personally I don't want to see everything when it's dark. Kinda defeats the purpose of light and dark to me. 

[Imaginate]

I also forgot to add that I'm not that worried about the full frame aesthetic either. It's very nice to be able to put some nice L lens on this, but I think with the upcoming EF to m 4/3 metabones will close the gap.  

[hmcindie]

It seems you just came to the conclusion "10 bit from GH4 is better!" without actually seeing anything that confirms it.

 

Everything I've seen makes the A7s looks considerably better in every single category except 4k internally. Rolling shutter is still undetermined but Matthew Allard did say it's comparable to a 5d which would be usable for me.

 

Having a 30 minute "dealbreaker" is also funny as I work with people who do documentaries with a 7d still, heh.

[Jacek]

The 8 bit footage out of cam looks very DSLRish and not much better than a 5D(granted this will probably change a bit with 4k and 4:2:2). I wasn't that impressed with the Den Lennie footage either and this review just confirms it. The 10-bit image out of the GH4 looks like it's going to kick the crap out of most images from this camera.

 

Don't know if it means something, but according to dxomark, GH4 tonal range is max 8.7bit @ ISO100 and <8bit @ ISO > 400.

[jcs]

Last night I was shooting RED with Zeiss Super Speeds (@T1.3) in a very dark environment which we could not light (and had to wrap when it got too dark): the A7S would have been very helpful.

 

The GH4 works great in good light and for low light Neat Video cleans it up very well. The GH4 noise is mostly fine monochromatic and effectively works like dither to reduce banding.

 

5D3 14-bit RAW looks great and can cut between Alexa/Amira and RED Dragon (some of the most popular higher end cameras).

 

A popular cellphone camera shot this Bentley commercial:

[QuickHitRecord]

[Wulf]

Let's hope it was not really a prof colour gradist... :-)

[Michael Ma]

Video lit by a single firefly.   :D

 

That should be the new litmus test for shooting in low light.