D3200 Announced

[eteam]

Anyway, worth noting on the spec-side, the biggest difference between the D3200 and the D800 is that the former only has a 12-bit A/D converter and the latter a 14-bit. Not sure how much this will matter once the files are encoded though

Considering that the output format in both cases is 8-bit (per colour channel) 1920x1080, derived from 12-bit (or 14-bit) (per channel) output of a 6x or 9x resolution sensor, it stands to reason that any differences in the video output of D3200 vs. D800 are likely *not* due to differences in the number of data bits issued from their respective A/D converters. In other words, the data set used to generate the video output massively greater (in both tonal and spatial resolution) than the derived result -- for either camera. That's merely my opinion, I've made no effort toward technical forensic analysis.

(although using a D800 with an external capture box via HDMI might see a significant gain in this area whereas this is impossible with the D3200 - HDMI is not 'clean' and limited to 720p max only).

Are you sure that the D3200 is limited to 720P HDMI output? The user manual suggests otherwise. I do not have a D3200 for personally verifying D3200 features, but any such limitation would be... unexpected .... and noteworthy.

-- Bob Elkind

[Samuel H]

the ADC lives in a linear world, the output format has seen a log transformation
if the D3200 has 12-bit A/D converter, it can only give you 12 stops of DR out of the sensor (then you may earn a little more with the image downsampling)

which makes me think that sensor has 14-bit A/D converters, just like the Sony A77, A65, NEX-7: it seems to be the same IMX096 sensor, with very similar per-pixel DR (i.e. before downsampling) to the D800
http://www.dxomark.com/index.php/Cam.../(brand3)/Sony

but I aggree that for video mode in-camera image processing will have a much bigger impact than this ADC thing: the D7000 also has similar per-pixel DR to the D800 (http://www.dxomark.com/index.php/Cam...(brand3)/Nikon), yet in-camera image processing is a lot worse and the final result is nowhere near as nice as with the D800


(edit: to see per-pixel DR on those links, click measurements, dynamic range, screen; the "screen" measurement is without downsampling, the "print" measurement is after downsampling to a common size of 8mpix; comparing those you can also see how the spatial resolution that eteam mentioned helps the D800 rise above the D7000, even if tonal resolution was initially the same; this is all for RAW stills mode, though: the downsampling method in video mode doesn't work so well)

[eteam]

Aside from the incidental point on which we are agreeing (12-bit vs. 14-bit A/D output bits does not directly translate to differences in video output quality), here are a few nits on which our fellow readers might nibble...

if the D3200 has 12-bit A/D converter, it can only give you 12 stops of DR out of the sensor (then you may earn a little more with the image downsampling)

1. Do not assume that the A/D readout from the sensor is linear.

2. Do not assume that limiting .NEF (RAW) files to 12-bit data format is evidence that the A/Ds are "only" 12-bit.

3. In a world of H.264 (compressed) 8-bit video, what do you think is the net impact of (hypothetical) 24-bit data rounded to 8-bit data vs. (hypothetical) 27-bit data rounded to 8-bit data (before compression) ?

... the D7000 also has similar per-pixel DR to the D800, yet in-camera image processing is a lot worse and the final result is nowhere near as nice as with the D800

You are suggesting that the DXO "quality" measurements do not necessarily translate (directly) to IQ from the camera. Many would agree with you on this point.

... you can also see how the spatial resolution that eteam mentioned helps the D800 rise above the D7000, even if tonal resolution was initially the same...

In the context of video quality there are other, more important differences between the D800 and D7000 than spatial resolution (pixel count). Remember that Star Wars III was "filmed" with Sony HDC-F950 1920x1080 digital cameras -- only 2 Mpixels! You could easily convince yourself that sensor resolution greater than 1920x1080 could (and would) only complicate (impede) the generation of high-quality video.

It's important to keep specifications in their proper place. I would value Philip Bloom's assessment of video quality well above any inferences made by either the camera datasheet or DXO sensor measurements.

-- Bob Elkind

[Samuel H]

fair enough - and also: I didn't know there were non-linear A/D converters... interesting

[eteam]

I didn't know there were non-linear A/D converters... interesting

You don't need non-linear A/D converters, you just need the correct algorithm to render properly scaled, linear RGB from the 'raw' A/D converter output from the sensels.

You were drawing inferences from the 'RAW' file bit depths of the D3200 and D800 (or perhaps to the camera sensors themselves?) to their respective video output. I am pointing out for you and the larger audience that these inferences are probably not very useful.

As an electronics design engineer, in this comparison I would say (simply) the proof is in the pudding. The amount of data processing which takes place between the sensor and the output video files is extensive, to say the least.

-- Bob Elkind

[Samuel H]

no, I wasn't talking about RAW file bit depths, but about ADC bit depth: Whitecat said "the former only has a 12-bit A/D converter and the latter a 14-bit"
if that's so, my understanding was that the D3200 could not produce more than 12 stops of DR out of the sensor (the 12-bit A/D converter would be a bottleneck)
but maybe that's wrong, I have very limited knowledge at that depth

[Samuel H]

Ok, so the canon T4i/650D is here, and it seems expensive (35% more so than the D3200) and a bit underwhelming (no idea on aliasing/moire yet, but no mention of all-I or IPB, and bitrate numbers that match the old H.264 codec, don't bode well), so my interest on the D3200 is renewed... Any owner around here with some time to take the set of RAW stills that would allow me to measure DR and check that Flaat-for-the-D800 works as intended on the D3200 too? And any further comments on aliasing/moire? (compared to the old APS-C Canons, 7D-60D-600D-550D). Also, I see the focus aid button is on the left side, can it be remapped like on the d800?

edit: I've been looking for samples on vimeo and it seems that aliasing and moire on near-horizontal detail are horrible... I still hold some hope that it's because people are shooting 720p or with high sharpness in-camera, but...

[Whitecat]

Are you sure that the D3200 is limited to 720P HDMI output? The user manual suggests otherwise. I do not have a D3200 for personally verifying D3200 features, but any such limitation would be... unexpected .... and noteworthy.

Sorry - it will also do 1080i (but not p). It's not 'clean' though.

Just another couple quick notes about my time shooting with the D3200 - there is no way that LCD screen is accurate, despite what Nikon was claiming about the similar cast observed with the D800. The rushes looked way too red and way over saturated. Was able to pull it all back, thankfully, but I'd definitely want to heftily tweak the picture profiles before starting next time.

For the record, the D5100 seemed much better in this department (colour accuracy on the screen).

Also, I typed earlier that I thought the D3200 exhibited slightly better control of the rolling shutter than the D5100, but I'm withdrawing this comment, after reviewing stuff in post I think it's pretty much equally bad on both cameras.