There is a very good, lengthly review of the Z7 by Adam Wilt at the Pro Video
Coalition site:
http://provideocoalition.com/index.php/awilt/story/review_hvr_z7_hvr_s270/

Tom Valens
Tamalpais Productions
Forest Knolls, CA

Thanks Tom for posting the link to the review. Interesting read. It's great to finally read a meaty review where someone has really put the unit through some major testing.

Interesting to read that the HDV codec seems to hold up better in the native progressive modes.

If others find new reviews, please pass on the info.

Thanks again . . .

Ian

Boy, Adam sure is thorough, isn't he?

I was interested to see that increased gain results in decreased resolution. It certainly sounds like Sony is using a pixel-averaging process to reduce gain noise.

I was also interested to see that Adam did not report any of the focus or lens issues other users have reported. This and Marshall's experience would seem to suggest that these are the factors of a few lemons (or perhaps damaged) cameras and is not indicative of the whole model.

Overally, a very good review, and my confidence is restored.

Super review and makes me glad I went for the Z7

Super review and makes me glad I went for the Z7

I agree. I finally feel good about my Z7 purchase. I am working on some wedding footage that I took last week and promise to post a short clip, but what I see so far is amazing and exactly what I had hoped for.

I did get a chance to chat with Dave Williams at the IVA in Chicago. For the most part he likes his Z7 as well, but does not think the auto focus is the best in low light and has been shooting manual focus in those situations.

Bruce

Interesting to read that the HDV codec seems to hold up better in the native progressive modes.

Although MPEG-2 has the option to encode fields separately (when there is motion) -- 1080i HDV does not make use of this option. Both fields are tossed into a buffer and encoded together. So the more motion -- the more computation is wasted on handling the motion. Progressive avoids this problem -- which is why the codec issues with 1080i HDV have never affected 720p HDV.


Note -- Adam's statement is not true: "[Note that marketing materials for the V1 referred to its sensors as “960x1080 (effective)”, but the specs for the camera show them as 1440x810, same as on the Z7 & S270, although in a 1/4” size instead of 1/3”.]"

All 3ClearVid chips are 960x1080 with 1,036,800 pixels. The EIP interpolator chip requires the 1080-rows. The EIP creates a full 1920x1080 image. Sony does not use pixel-shifting which Adam correctly points out does not work as well as Panasonic claims.

Here's a pix from the Z7 brochure.

Note -- Adam's statement is not true: "[Note that marketing materials for the V1 referred to its sensors as “960x1080 (effective)”, but the specs for the camera show them as 1440x810, same as on the Z7 & S270, although in a 1/4” size instead of 1/3”.]"

Steve, kindly point me to the incorrect part of my statement.

The marketing materials for the V1 use the same image you show from the marketing materials for the Z7, and state that the sensors are 960x1080 (yet 960x1080 implies a pixel aspect ratio of 2:1 (as on the Z1), while the schematic shows square [1:1] photosites and square interpolated pixels).

The operating manuals for both cameras state that the sensors have 1440x810 pixels, on page 127 in the V1 manual; page 125 in the Z7 manual. (This implies a pixel aspect ratio or photosite aspect ratio of 1:1, and better matches the isomorphic appearance of the zone plates--960x1080 would imply half the resolution in one direction compared to the orthogonal direction, and that's not what we see on these cameras, though we do on the Z1.)

So which part of my statement do you dispute: that the marketing materials say 960x1080, or that the specifications list 1440x810?

Steve, kindly point me to the incorrect part of my statement.

Sony's NYC V1 intro presentation shows a V rez of 1080. Here is a pix from the NYC presentation.

Because Sony sent me a prototype V1 in the summer before it was released, I had many conversations with them. I've always used the number 1080x960. And, Sony never had a problem with the 960x1080 number. In fact, even when my questions were passed to Japan -- with that number -- it was never corrected.

Given Sony's 960x1080 number and the fact the chips were capturing progressive frames, 1080 rows was an ideal number of rows to pass to the interpolator. I worked with Sony on this issue -- and again no one in the USA or Japan raised a flag.

However, like you I was concerned about the pix of the photosites. Perhaps wrongly, I assumed the artist drew a "representation" that didn't show them having the correct pixel aspect-ratio that would space them out in a 16:9 ratio whilst also having them in a diamond arrangement.

The marketing materials -- both USA and UK -- show the chips V rez to be 1080 for both cameras. So, we have a choice to believe Sony has made a multiyear error in their marketing material -- and the manual writers got the numbers correct -- or Sony's entire marketing organization blew it.

Later this week I'll be talking with Bob and Juan about other public statements they've made -- so I'll certainly ask.

I am the first to admit that I don't know what the *actual* number of active photosites is on the chips in either the V1 or the Z7; all I know is that one set of numbers is given in the marketing materials, and another, different set in the specs in the manuals. I was simply presenting the stated numbers, as stated, and describing the contexts in which they were stated.

Either or both sets of numbers may be correct depending on the manner of counting and the interpretation of the diagonal arrays. If you look at the NYC graphic, it's really 960 wiggly dual-photosite columns by 1080 staggered single-photosite rows; it's a weird way to count 'em up, and could just as easily be said to be 1920x540, or 1920x1080, depending on your frame of reference.

But, having said that, how do you get 1440x810? Given that there's clearly multiple-photosite-per-pixel processing going on, or at least a non-one-to-one-mapping of photosites to pixels (equal ISO ratings in interlaced and progressive modes; reduced resolution as gain is boosted) is 1440x810 the actual sensor resolution, and 1920x1080 an effective post-processing figure based on photosite staggering (a 50% boost, like green-channel pixel shift but without the disadvantages)? Or is 1440x810 a measure of degradation from an actual 1920x1080, again due to staggering (a 33% loss, due to active area overlap)? I can make handwaving arguments either way. If I look at this stuff too long, I'm reminded of Larry Niven's "Singularities Make Me Nervous"; call this one "Diagonal Photosites Make Me Nervous" [grin].

If you can get clarification from Sony on which set(s) of numbers to believe, and when to believe 'em--and how the two different sets of numbers came about--please let us all know.

If I look at this stuff too long, I'm reminded of Larry Niven's "Singularities Make Me Nervous"; call this one "Diagonal Photosites Make Me Nervous" [grin].

I totally agree with your feeling. I read multiple explanations of the diamond arrangement and became totally confused. Finally, I wrote down what made the most sense to me -- and sent that to Sony for any corrections. Frankly, I never was sure when I sent technical writing to Sony USA to send to Japan for "comments" -- that the lack of any correction meant I was right.

By the way, I never noticed the 1440x810 because the prototype came with no manual.

I couldn't reach either one. Will try again next week.

If you look at the NYC graphic, it's really 960 wiggly dual-photosite columns by 1080 staggered single-photosite rows; it's a weird way to count 'em up, and could just as easily be said to be 1920x540, or 1920x1080, depending on your frame of reference.

But, having said that, how do you get 1440x810?
An interesting discussion. It took me a little while to see what was meant by the above, but now it reminds me of those puzzles where the same image can either show a convex or concave image - depending how your brain snaps on to it!

If you count a column as "tip to tip", I believe you'll see there to be 1920 discrete (interlocking) columns with 540 pixels in each. Count a row as "tip to tip" and the perception will be 1080 rows with 960 pixels in each. Either way, the total number of actual pixels is 1,036,800, with corresponds with what Sony state in their marketing - "approx 1,037,000 pixels". This is true for square pixels, explaining Adams zone plate observations.

So, where does "1440x810" come from? My suspicion is that it's an "effective" resolution figure after processing, and probably fairly arbitrary, certainly not representing actual physical photosites.

Such a pattern would make sense for relative ease of processing in a way that 1440x810 wouldn't - especially if 1080i is to be derived. Getting that would be very clumsy from 810 physical rows, wouldn't it?

810 is the number you get when you derive a 16:9 frame from 1550 pixels across. It is a theoretical number that isn't related to the actual resolution of the Z7.

The CMOS sensors in the Z7 and s270 are actually 960x1080. By a combination of offsetting and interpolation, the horizontal resolution is increased to 1440 pixels across. For all practical purposes, this is real resolution. The interpolation is only used when the image is so strongly one of the primary colors (RGB) that the offset sensors don't capture the extra resolution. In reality, very few things are pure primary colors so this is hardly an issue, and when it is, the interpolation works well.

Most HDV cameras use sensors that are only 960x540. The 960x1080 sensors of the Z7 and s270 give it an edge when shooting progressive footage.

810 is the number you get when you derive a 16:9 frame from 1550 pixels across. It is a theoretical number that isn't related to the actual resolution of the Z7.
I believe you meant 1440, not 1550? But yes, 810 is either 1.5x 540, or 3/4 of 1080 - same as 1440 is to 1920 and 960.

I believe the sensor can be seen as 960x1080........ or 1920x540, depending how you count. (Maybe easier to just say (as Sony do) that it's 1,036,800 pixels!?) Hence it seems logical that the *effective* resolution is equivalent to a non diagonal type of halfway between those max and min figures - 1440x810. But that is a theoretical number, agreed.

If you count a column as "tip to tip", I believe you'll see there to be 1920 discrete (interlocking) columns with 540 pixels in each. Count a row as "tip to tip" and the perception will be 1080 rows with 960 pixels in each. Either way, the total number of actual pixels is 1,036,800, with corresponds with what Sony state in their marketing - "approx 1,037,000 pixels". This is true for square pixels, explaining Adams zone plate observations.

That's how I finally counted it -- using the zig-zag since there are no strictly H or V photosites.

Informative and interesting. Thanks for sharing the info Tom it shed some light on the Z7 for me.

Very interesting test. I would also like to see it compared to the competion, however. Has anyone found a comparison test of the Z7 to the Canon XH-A1? I googled a lot, but can't find any...

Time for a new mega shootout: Z7 vs. Xh-A1 vs. JVC 200 with improvements in chip performance (recording in 720 60p) vs. Panasonic's new 150 AVCHD @ 21 mbps.

Get out yer guns, hombres!

My answer from Sony -- "depends on how you count."

I've written back and said the problem is given the NYC pictures from Sony -- I can't find a way to count 1440x810.

However, I'm inclined to think Adam was correct as 1440x810 is a perfect 16:9 pixel ratio. It makes sense that a 16:9 chip would have this physical arrangement.

However, I can think of two ways the photosites could be placed.

1) in linear rows of 1440 photosite. 810 rows. Each photosite tilted 45-degrees. This is the way some Sony pix show things to be.

2) Every other row and column offset by 1/2 site. Now there are 405 rows of 720 pixels. Count in a zig-zag manner and you get 1440x810. Each photosite would NOT be tilted 45-degrees. It is the arrangement that creates diamonds. This is way other Sony pictures show things to be.

BUT:

From either arrangement, I can't see a way of COUNTING that gets 960x1080!!!!!

Clues MIGHT be that if every third site was dropped counting horizontally the result would be 960. And, if you somehow could increase the vertical count by 1/3 you would get 1080.

However, other than the factor 1/3 is involved in both -- the third that must be lost is not equal to the third that must be added. And, 1/3 seems like a "non video" number.

SO:

Who can count 1440x810 and get 960x1080?

OR:

Assuming I was correct and the chip has 960x1080 -- Who can count and get 1440x810?

My assumption was that the PIXEL-aspect was 1.33 so the layout fit a 16:9 chip. I explain the 1.33 as about 2/3 of each photosite contains transistors and 1/3 is photosensitive. Thus, the Sony chip's site would be .44 active and .90 inactive -- which is a 30% fill-ratio.