Part two: Panasonic Lumix DMC-GH4 / Sony Alpha 7S Comparative Review

The Gh4 has 4x less the area the Sony a7s does because one is a full frame camera, and the OTHER (GH4) is a APS-C , and if you bought the sony for stills you definitely bought the wrong camera because id say its more of a video camera because of the 12mp because bigger pixels is better for video so idk why the spatial resolution would make it better?. Like he said Apples & Oranges

On the surface, I'd think the same thing. Since giving the gh4 a second glance, I'm comparing it to the a7S from a mostly video taking perspective. They both have interesting positives and negatives and are one-of-a-kinds in their own respects.

i shoot nikon and fuji and they are not even trying in realm of video. I needed something better months ago but keep putting it off while I wait for something ideal. The a7S with internal compressed 4K recording would be ideal. I'm sure something like that will arrive next year but am not sure I want keep dealing with the footage out of a d800 and hero 3 until then.

I rented the GH4 for a week and took some nice video. I could not get a reasonable still out of the camera. Not even one I would use for anything. I used the 14mm pancake and a couple of Nikon AIS lenses. Back when I had a GF2, I consistently got better results. Not sure what is up with that.

Low light performance was an issue for video. I had to shoot wide open for 70% of my Friday shoot. Here is the resulting video:

https://www.youtube.com/watch?v=t2hZU_-E4nY

I shot all in the GH4 and edited in final cut pro. Please excuse the jerky camera movements. I have a proper tripod head on the way along with the a7S I ordered. :)

At the end of the day, I need the low light ability and I also want something I can take great stills with. Overall though, the GH4 is a more polished video taking machine. It just lost out in a few areas that are important to me.

At least mine is logical while your statement that, "an infinitesimally small sensor would need no light to produce a perfectly fine image," is totally illogical.

First, "infinitesimally small" is still small and therefore requires some light, duh!!!!! For "no light" to be applicable, it would have to be no sensor area to follow my logic!!!! Duh!!!!!!!!

Stop trying to justify your position and spend more time first understanding what I'm saying. You're acting irrational because you're getting emotional by taking it personally.

Your statement that "Therefore, the same captured scene will be 'sampled' using many more photons, which means less noise." requires there to be comparatively more light per square area in order to have "more" light striking a larger sensor since that same larger sensor invariably requires MORE light. Just saying "many more photons" is not only an incorrect generalization but also patently false.

You're wrong, dude, keep trying.

um... read the first post, it was a simple comment from which you started the s*^t talk with "reductio ad absurdum" so please blame yourself for your reply.

What you finally pointed out: "if the number and size of pixels..." is my mistake. I meant to say just that "if the size of pixels..." So I admit my error. And had you pointed it out in your first comment, I would have gladly admitted it. Your first comment didn't mention that and went on a rant about an irrational relationship between smaller sensor and no light, and so I didn't read it again thinking there was in fact a mistake. Thank you for pointing it out, but stating a smart ass comment doesn't help the conversation.

My point still holds true when you omit that extra word. Please read it again without the word "number" and if you still feel it's illogical, let me know. Like I said: "It's only true if the pixel pitch is lower (meaning the pixel density is greater.)"

Qyuote: "My statement: "an infinitesimally small sensor would need no light to produce a perfectly fine image" follows from your logic, so I'm glad you agree it's completely illogical."

You could not have such a small sensor, because the concentrated light would burn through the sensor. Hence your example neglects what is going on.

If the small sensor could work as well as the big sensor, quality would be defined by the lens. As m43 improves, the lens quality increasingly becomes the issue. And quality can be depth of field, lens size and weight too ...

What he is getting at is NOT invalid. Think about photons. The same lens will project the same number of photons through it at the same focus and aperture settings, no matter what size the sensor is. Now . . . it matters what the image circle is. One lens that is 150mm might make a huge image circle at f8 (such as the 150mm lens for my 4x5 large format camera), while another lens, such as the Sigma 150mm f2.8 macro will make a much smaller image circle at f8. Do you think that if you put a full-frame sensor behind both lenses that you will get the exact same amount of exposure with the same shutter speed? NO the light projected by the large format lens gets "spread" out, rather than focused on a smaller area . . . or am I mistaken? If I am indeed mistaken, then how can the Metabones Speed Boosters work? http://www.metabones.com/assets/a/stories/Speed%20Booster%20White%20Paper.pdf

Nothing personal either but saying "reductio ad absurdum" doesn't lead me to a typo.

On the topic of "normalized," for an image on a monitor or laptop screen, one must take into account the resolution of the screen. On a 4k monitor, one doesn't need any more than an 8mp image to equal the screen's resolution. If the image is enlarged for any reason other than aesthetics, then it's not normalized anymore and it's about pixel level detail.

Let's get one things straight, of course a larger sensor will always yield a better image, but that talk doesn't stop at 35mm equivalence, as you well know. I was talking about generalized statements concerning more and less light, and in that context, sensor noise is directly related to pixel pitch (and of course tech).

From a tech view, noise is the key and comparing an 8mp image (m43) on a 4k monitor with a ff 32mp image, there is no difference with all having the same fstop, shutter, and pixel pitch. The extra pixels can't be resolved.

JamesD28, one must take into account technology differences too. That said, my point is about the fundamental relationship with noise per square area. That boils down to pixel pitch (and technology, of course). I'm NOT saying a larger sensor is the same. The principle of noise is about the size of the pixel receptor and the consequential amount of light IT receives, not the amount going through the lens. A bigger sensor needs more light proportional to it's larger area.

Tradeoffs with a smaller sensor have some benefits as long as we realize that as things get bigger, they also become more unweildy, that's why medium format isn't included in this discussion.

M43 can theoretically have lenses 2 stops faster for the loss of 2 stops in greater noise at high iso, given an equal sized lens (key!). But for this to be realized, M43 must always be 1/4 # of pixels to ff.

This is also about relative needs to relative tools. M43 is a great street shooter, ff for detail. etc...

I think these discussions oftentimes devolve into an apples vs oranges conflict (no computer pun intended). If the whole point is to have a bigger sensor, one would haul around a medium or large format camera on a tripod. Most don't. If the whole point was to shoot discreet images with reasonable quality, m43 tends to have an advantage, but at the pixel level, it's about physics principles, and then apples to apples. Stepping up, at what point is there too few pixels? Maybe 8-12mp. Too many, over 40mp?? seems to make file management unweildy. So each person has a sweet spot for their needs.

I spoke of pixel relationship because the original comment was generalized about light amount... which can be very deceiving. The editors should know better than to use generalized terms without context. I was just "calling" them on that... I get the generalized concept of more is better, but out of context it means little to each person. and for some, it's actually NOT better...

hmmm, noise IS increased amplitude of signal to account for reduced light when aperture and shutter speed are insufficient. noise has everything to do with pixel pitch. smaller pixel lens or "window" means less light, so once the need to boost amplitude arises, so does noise go up. It's actually the same in audio.

With all others equal, including pixel pitch, noise will only go up if amplitude is increased (increasing signal sensitivity to lost light levels). Period. that is what noise is (in daylight amplitude isn't raised so little difference). If you have more pixels, you have more resolution. Therefore, this greater resolution is what you are talking about.

In general with daylight, unless the pixel pitch is very small (ie, smartphone camera), there is no noticeable difference between m43 and ff. but m43 typically has too many pixels for it's size meaning light gathering capacity or pixel pitch is adversely affected by raising amplitude while in ff, it's not until high iso.

You keep tangenting while factually, a sensor with the exact structure that's 1/4 size (or 4x the size, my original statement) has same noise with the only difference being resolution/more pixels (this is a simple fact).

I said it because the author made a generalized statement about light, and saying that is not accurate without context. DPReview has stated over and over what I have said in context... this is nothing new.

Arguing that more light makes a bigger sensor better, is well, generalizing to an extreme. Each time I point out that pixel vs. pixel, if everything is the same, will result in same noise, you tangent back into the generalized statement that normalization justifies that more light is just better because it's a bigger sensor.

I really don't want to argue with you, but you simply won't accept a factual statement and tangent back to your normalization focus.

"Pixel-for-pixel everything is the same, but you're ignoring the fact that the larger sensor has captured more picture."

What exactly is "more picture?" It's higher resolution. More pixels per square area. A larger sensor has to capture more light, and that in and of itself doesn't constitute higher image quality. The reason the aperture is a larger diameter is it must be larger to cover the larger sensor with an image circle sufficient for the task.

Stating that "more light" means better is deceiving because it implies that a larger sensor that NEEDS more light means it's just better. That's too generalized without articulating that that larger sensor has the advantage of either keeping the pixel pitch greater or placing more pixels per square area in comparison to a smaller sensor.

It's simply NOT accurate to say the aperture is bigger therefore it's better, and that's what you are alluding to without clarification.

Richard, I have a hard time understanding this aspect of equivalence.

What I don't get is that the amount of light projected on the APS-C area of the 35mm sensor is the same amount of light projected on the actual APS-C sensor.

That being the case if you crop the image produced by the 35mm sensor, does it have the same amount of noise as the image produced by the actual APS-C sized sensor, assuming similar technology, pixel pitch and using the exact same lens?

Or put another way if you used the same tech and pixel pitch in the D7100 and the D800e (both released early 2012) the same 50 ƒ/1.8G, would an image taken on a D810 in DX mode have as much noise as the D7100 in 1.3x mode?

If I look at the D5200, the D7000, the D800 and the 5D2 on the studio comparison tool all images RAW ISO 100 (and magnify my MacBook view iva CMD-shift-'+') I see very little difference in noise, with noticeable but not great differences in brightness and a bit more aliasing in the DX (esp the 5200).

As was said further up, if you just use the APS-C crop from a full frame sensor (with the conditions you suggest), then yes, those two image should be identical.

However, if you use the whole of the full frame sensor then you'll have more light-collecting area (and hence lower noise when you consider the entire image).

Richard Butler, if you use the "whole of the full frame sensor" then you would have more resolution with the same pixel noise level, with required more light to produce a larger image circle to cover that ff sensor.

If you mean having a higher resolution image gives the appearance there is less noise, fine, but that's all. But if your monitor can't resolve that difference than there is no difference unless you blow the image up, crop it, or print it to a larger size.

Nah it's a 4/3 camera and you are comparing to FF. It's over priced.

"From youtube comparisons..." Remember that Youtube also has a denoising algorithm so any video posted on Youtube will have less noise. Also, depending on which video you saw, it might be a 4k-rescaled-to-1080p, which gives it an advantage over straight-1080p A7s footage.

The GH4 holds less detail in the shadows, there's no debate there. The A7s' slog-2 has a greater dynamic range that allows greater post production. So yes, A7s is actually a better "pro" video solution if you're going to drag the videos in DaVinci Resolve afterward. The A7s is not a low-light photography tool only, though I agree the GH4 is mostly for video. Don't see too many people touting that camera's stills quality.

@marc petzold
Two professional videographers also recently reviewed both the GH4 & A7S
http://philipbloom.net/2014/08/06/a7svideoreview/
http://www.eoshd.com/content/13197/sony-a7s-x70-pinewood-studios

@ Dr_Jon
"I’m surprised at how nice the quality is from APS-C (Super 35mm) mode on the A7S. I expected it to be a lot softer! There’s no signs of significant moire or aliasing either."
http://www.eoshd.com/content/13105/good-news-super-35mm-mode-sony-a7s-dramatically-reduces-rolling-shutter

Thanks for the explanation

Richard, you have used a different TIME and therefore there is a different EXPOSURE RESULT and that is why you cannot use the pictures in the comparison tool and of that because the result is very misleading when you compare!

If you're talking about the night scene with the Space Needle - those shots were taken on different days, but at around the same hour. In other words, something like 4-5 hours after sunset and 4-5 hours before sunrise. So, right around the darkest part of the night here in Seattle.

There really shouldn't be any appreciable difference in brightness levels of the scene. Save for lights in buildings that went on/off, & perhaps some slight atmospheric differences that may have changed the black level of the sky. But you can explore the scene and find streets/buildings that would not have experienced any appreciable changes in brightness due to being shot on different days.

Focal plane exposures were matched so any differences in exposure would be down to T-stop, which we don't believe to be too different between the lenses used.

If you're asking why we didn't shoot them at the same time side-by-side - that raises a point I always try to make about the importance of designing controlled tests vs. doing one-off side-by-sides. No one has the resources to test every combination of cameras side-by-side. A controlled lab setup, though, allows you to compare cameras on different days, months, years even. As our studio scene does. And we're working on designing a new low light (& daylight) lab scene that emulates the real-world spread of tones/brightnesses. In the meantime, this night scene worked quite well - it allowed us to talk about ISO performance across shadows, midtones, & highlights.

In fact, the differences in ISO performance between cameras match up with SNR data we've acquired for these cameras. That's a nice lab to real-world correlation that validates lab testing.

The irony is, with lab testing we'll still get people saying that we should go shoot it in the real world. You just can't win sometimes.

Richard, i can repeat myself but that i think that is not necessary.
Look at the ISO 1600 JPEG there you a whole different picture e.g. people on the left etc.etc. So you have a different light because the TIME is different,it is not a controlled environment! So that is why it is not use full for a comparison tool.
Again very misleading because you compare apples and oranges.A intelligent photograph MUST know this.

At Richard and Rishi,

You use a lot of words to explain things they at base wrong and again misleading.

Most people likes a more contrasts picture and there is a tendency to overexposure and most people do not like either.
Other people like more detail etc.etc.

So a compare tool you offer must be a neutral offer otherwise wait until you can give us the right tool it is now so subjective and you get misleading information

Except that the ISO performance pretty much matches up almost perfectly with what DxO's full SNR data predicts. Read page 2 of my original article introducing this scene/test here:

http://www.dpreview.com/articles/4613822764/high-iso-compared-sony-a7s-vs-a7r-vs-canon-eos-5d-iii/2

Results match up with what sensor size (and resulting shot noise contributions) and read noise measurements of these cameras would predict.

Don't look at parts of the scene that are mildly different b/c people are walking on the street or building lights have been turned on/off. There's enough consistency throughout the scene for you to draw the relevant comparisons.

Like I said before: you just can't win. Ultimately, we found these results to be quite helpful in demonstrating the points we wanted to make. And they match up with SNR data from lab tests.

If you can prove we're showing something that disagrees with what the SNR data and/or math would predict, feel free to come back with something concrete.

Are you going to have a section on auto-focus speed and continuous auto focus with the 35-100m lens? Or even the -4EV focus claim?
The GH4 finally is very competitive with most DSLRs in many ways, but my fear these things will be glossed over in this "video comparison" review.

@Lab D. We haven't got a 35-100mm lens but I will be looking at the continuous shooting of the GH4. This isn't a video comparison, it's just that a lot of what sets these cameras apart are their video features.

OK, thanks. Make sure you use one of the new Panasonic lenses (borrow a 35-100mm if you can). As with most systems some older lenses affect focus speed.

Thanks Richard.

I understand that and you're not expected to introduce another variable for EVERY camera review.
I'm saying it's justified for this PARTICULAR CASE as we are probably facing the current two best Hybrid cameras. It's very important to know how much you can stretch the ISO in video, because it's proven to be a very different matter from photos and with these two video centric cameras it's pretty much relevant.

Even if you would just publish a static page with selected crops it would already be a boon.

Thanks again.

You are fussy indeed. ;)
I take my GH2 to ISO800 regularly on venues and most of the videos hold up quite well, even without NeatVideo.
I can't imagine the GH24 being worse.
But I would really like to see how far can it go with acceptable video IQ.

try http://www.steves-digicams.com/ then

https://www.youtube.com/watch?v=GeixtYS-P3s

I hope Panasonic blinks soon. I just realized when I add lenses it is almost out of my price range too. Maybe a good competitor will show up next month too. Fuji? Sony? Canon?