Sony PXW-Z190 (1/3") and PXW-Z280 (1/2") at NAB

[Cliff Totten]

Question for beam splitter, prism and optical block experts;

THE BASICS:

As we know, light enters the front of the optical block and is split and mirrored into three directions. It ends up hitting three image sensors with red, green or blue optical filters. The three RGB images are combined on top of each other and processed together to form a full color image.

PROS:

- High resolving power because each RGB sensor is 100% full 4k resolution. (As opposed to a heavily reduced RGGB resolution from Bayer pattern de-mosiacing. Each Bayer-pattern channel only having 25%, 25% and 50% of 4k resolution)

- Three sensors produce more accurate colors and edges. On the other hand, single Bayer RGGB patterns require color binning or averaging from adjacent pixels to assemble it's image. Three sensors produce simple and highly accurate image assembly.


CONS:

- Incoming light is divided and the prism only allows 33% of that light to each of the three sensors. So, in low light situations, each sensor needs high amounts or pre-amplification to reach a nominal output level. (Eventually ends up as 0db for the user to adjust higher as needed)

- Note; With a Bayer-pattern layout, RGGB photosites are each receiving 100% of the incoming light on a single sensor.


THE NUMBERS:

I dunno for sure but does anybody find anything wrong with these statements? (I dont pretend to be an expert on this.) Each RGB sensor is only getting 33% of the incoming light from the lens. In daylight situations, this is perfectly fine as there is more light than you need. But in indoor situations or night time where light could be scarce, these sensors potentially will need high amounts of early stage amplification to get to a decent level at 0db. I'm guessing noise reduction is going to play a BIG, key role in these cameras.

The EX1r and other 1080 cameras had 1/2 inch sensors with only 2 megapixels. These 4k sensors are 4 times more dense so we are looking at much smaller photosites. The 1/3 inch model will be even smaller.

I have high hopes for these two models but how do people feel about the light gathering ability of these cameras? (Not the noise reduction power....the light gathering ability. Actual "Light" and "gain" are two totally different things.)

Obviously nobody knows for sure yet as there is very little material to watch out there on these guys now. So we can only wait and just talk physics and science for now.

Your thoughts?

[Dave Sperling]

Cliff -
One thing I think you're overlooking is how the Bayer pattern gets its R,G&B
Remember, pixels are basically B&W receptors,
So in a Bayer sensor, each individual pixel basically has a colored filter over it to limit it to 'seeing' that color of light.
As you know, filters cut down on light transmission. So the amount of light received by each pixel in a bayer sensor is not 100%, but is reduced by that color's filter factor.

[Cliff Totten]

You are certainly right. All image sensors see RGB optically filtered light and produce a voltage reading. This amounts to a "grey-scale" voltage value for each photosite on the sensor on the read cycle.

Yes, all RGB filters intoduce resistance to light. Both, in Bayer and three sensor arrays too. This light loss is "built-in" to the idea and is unavoidable for every sensor.

Im specifically questioning the prism's light dividing roll itself. Its job is to take the lens projection and divide that into three images and present each one to it's sensors at the end of the block.

This light split-up takes 100% of the lens' projection and gives 33% of that light to each RGB sensor. This is a "burden" that a three sensor configuration must suffer.

Single Bayer sensors dont need to alter the lens light and just take it all exactly as the lens projects it. 100% light.

Anybody have more to add. Im curious just how much loss a prism/beam splitter gives. Is it worse that 1/3 light transmission?

Possibly so?

P.s......Do these optical blocks divide FULL SPECTRUM light into three full spectrum images and allow the sensor to RGB filter it. Or, does the prism take in full spectrum light and divide it into red, blue and green light and then beam that to each sensor?

[Christopher Young]

Cliff ~

One cannot argue that because there are three sensors that each sensor receives only 33.333% of the incoming light. It is the cumulative value of the three sensors output that has to be measured to compare it to the output value of a single sensor.

In a three chip design each chip receives the entire image. Full-bandwidth R, G, and B signals making color processing much simpler. Each sensor can be optimized for the color it will 'see.' For example by chemical filtering to alter its spectral sensitivity. Adjusting the color balance of the overall signal is a simple matter of adjusting the gains and black levels of the three chips relative to one another.

In our hypothetical 4K three chip camera 100% of the red wavelength lands on the red chip because it is split out from the 'white' light coming in through the lens, therefore 4096 worth of red sensor photosites get evenly illuminated in red light. Subsequently 100% of the blue and green wavelengths arrive at their respective 4096 photosites with very little color wavelength cross contamination due to the efficient wavelength filtering used. From the this point each sensors gain and time constant are adjusted in real time to ensure that the clock out of the RGB signals are in time with each other and that their respective levels are correct. Importantly remember that collectively the combination of the three sensors represents 100% of the white light that came through the lens, bar prism internal reflective and refractive losses which in modern three chip cameras is within 1-2%

Now with the single chip sensor 100% 'white' light arrives at the sensor. The difference here is that with a traditional 2:1 Bayer sensor of 4096 photsites only the '2' which is 2048 photosites are going to be capturing the green wavelength. That's only 50% of the green wavelength within the white light coming through the lens. The other 50% of the green wavelength is going to be landing on photsites that are not filtered and tuned to receive green light. When it comes to the '1' of the 2:1 Bayer ratio that means only 25% of the 4096 photsites will be receptive to red and blue wavelengths respectively.

Therefore with the Bayer sensor approach the missing color data at each pixel is then interpolated, or guessed at, by various algorithms that try to use image statistics to make up for the fact that all three color planes are under-sampled, in violation of the Nyquist sampling theorem.

To sum up fairly simply:

Resolution

The mosaic filter approach lowers the spatial resolution available by roughly 30% compared to a monochrome sensor or 3 sensor approaches in which all of the photons are used for the image.

Sensitivity

With a single chip color cameras, not all of the pixels are sensitive to all colors so not all of the light contributes to sensitivity as it does with 3-chip cameras. The loss of sensitivity in a single chip 2:1 Bayer sensor due to its 50/25/25 light collecting sensitivity across the RGB photosites is considerably greater when compared to the light loss through a prism that has a transmission efficiency of around about 98-99% and that is delivering let's say minimum 98% of the total RG and B wavelengths to all three full raster RG and B sensors. Very little illumination value per wavelength lost in other words.

I have to qualify this last statement by saying that if a single and three sensor cameras had the same size sensors the single chipper would be left for dead compared to its three chip cousin. In the case of a single 4096 x 2160 Bayer sensor we have 8,847,360 photosites with the drawbacks and limitations of the mosaic design. In the case of full raster 4k three sensor cameras we have 26,542,080 photsites. All of which are being utilized to the max to capture their respective R, G and B photons. Generally speaking though three chippers have smaller sensors and which means smaller photosites which equates to them being less sensitive. Despite that the efficiencies of full raster three chippers compared to existing 2:1 Bayer mosaic sensors size for size puts the tree chipper way ahead for the time being.

As time progresses single chip sensors are becoming much more efficient and 'promises' of performance such as the likes of the Foveon sensor will doubt narrow the differences between single and three chip efficiencies. Price and efficiency will win out in the long run. It almost always does. A single Bayer type sensor design and a three RGB split prism images below.

Chris Young

[Cliff Totten]

THANK YOU CHRISTOPHER!!

The prism allows for 100% (almost) Red, Green and blue light transmission! It's that efficient at breaking out light frequencies?

Three sensor installations don't need RGB optical filters over each sensor. They are essentially "naked" grey-scale sensors. (with probably only optical low pass filters on top)

It's a shame that the Z450 is only a single Bayer sensor design. I suspect that a 4k optical block for 2/3 sensors will arrive soon. I can only expect that a new Z450 replacement with 3 sensor design will smoke the current Z450. (it would certainly make sense)

I wonder how far up they could go. A block for three 1 inch? (maybe that is pushing it too far on the camera size?)

This is some fantastic stuff you posted here, exactly what I was hoping we could talk about here. This does a great job showing exactly how these prisms work!

I really hope this gets allot of views here as this is a great post.

CT

[Christopher Young]

Cliff

Sony already have a 2/3" three chip CMOS full raster UHD camera, the HDC-4300. It also can do HD out at the same time which is what goes to air here. We have had them being used here on the league football, Australian Open tennis etc since 2016. The 4300 is also capable of 4 x overcrank for slo-mo replay. Being UHD capable they also have different prism coatings for the wider color gamut of 2020 for live HDR broadcasting. They are only available as a studio/OB digital triax camera not a camcorder. It would be a very expensive camcorder no doubt if Sony decided to implement this tech into a camcorder. From memory the camera heads were about $40-50K plus the various software HFR packages which were about $18K per package.

https://pro.sony/en_CA/products/4k-a...iable-hdc-4300

Sony also have the S35 single sensor HDC-4800 that is capable of up to 8 x overcrank in UHD and 16 x in HD.

https://pro.sony/en_CA/products/4k-a...stems/hdc-4800

It's this hi-end broadcast technology that is trickling down to the likes of the Z280 for the likes of us mere mortals.

Chris Young

[Paul Anderegg]

Chris, while watching season 2 of The Grand Tour (Top Gear 2.0) on Amazon Prime, I noticed all the studio cameras are shoulder style Sony rigs...I couldn't tell if they were the 4300's or Z450's.

And I would really love to see some effort put into 1080p/720p HD only cameras for news...your local news market will not be 4K anytime within the life expectancy of any new camera bought in the next few years. Imagine what the sensitivity and s/n ratio would be on the Z280 if it had only 2MP with the same tech applied. I am fairly sure the f12/13@2000 63db rating is for HD mode...Sony also seems to be joining the (high/low sensitivity) mode game, where thy give some ratings at 6db (high sens) and others at 0db (low sens). That is how they rated the Z190, at least they were honest about it.

Paul

*my secret dream, shooting for a 720p60 TV station, is for giant 1MP 720 sensors that can see in the dark! :-D

[Paul Anderegg]

I found this interesting...just aired new episode of 60 Minutes on CBS...saw one of the cameras in the background...an old Panasonic HDX900 720p DVCPRO HD TAPE camera! :-D

Paul

https://www.bhphotovideo.com/c/produ...2855/KBID/3801

[Christopher Young]

[QUOTE=Paul Anderegg;1944049]Chris, while watching season 2 of The Grand Tour (Top Gear 2.0) on Netflix, I noticed all the studio cameras are shoulder style Sony rigs...I couldn't tell if they were the 4300's or Z450's./QUOTE]

In studio they are the 4300's. Outside United Broadcast Facilities from the Netherlands who won the Grand Tour contract supplied all the location kit for season 2. Below is some of the kit used:

Cameras:

Arri Amira, Alexa
GoPro HERO4 (Minicams)
Panasonic GH4 (Car Interiors)
Panasonic GH5 (Drone)
Phantom Flex (slow motion scenes)
Sony HDC-4300 (Studio)

Ben Joiner was the DP on the show.
http://www.benjoiner.com/blog/the-grand-tour-season-2

Nice story on the GH4's
The Grand Tour ? The Small Camera Story ? ?HOME

The post production rundown:

Primarily ProRes
Color Color (ACES)
Aspect Ratio 1.78 : 1 (UHD)
Sound Mix Dolby Digital

Rushes were done on set by the DITs and were DNxHD36

Interesting Editors interview of just what is involved!
https://www.provideocoalition.com/aotc-grandtour

Chris Young

[Cliff Totten]

I'm curious to see how the Z190's three 1/3 sensor performance is compared to Sony single 1 inch-type sensor.

Anybody want to guess what the photosite size comparison is?

I suspect that Z190 sensors use the same stacked BSI, gapless architecture. So,...I'm guessing that three 1/3'rds at 8+ megapixels will have slightly larger photosites than the single 20mp 1inch type sensor? (especially after you consider the 16x9 crop in the 1 inch-type)...I'm not sure at all though.

Also,...we can only guess but does anybody want to to toss up ideas on dynamic range? I'll guess at 10-11 in SLog. I would be shocked if the Z190 had more than 11.

I see it has up to +18db gain and then a "turbo" mode.

I have not seen any published signal to noise ratio number on the Z190.

CT

[Ron Evans]

Maybe its the same sensor that is in the AX53.

[Paul Anderegg]

The PXW-X180 was 60db, but had some very obvious fixed patterns at 18db due to noise reduction. We can assume it is at least 60db given the specs on the Z280. I think the Z190 is probably about the same sensitivity as the X180, f9@2000 with 60db s/n ratio. The specs say it is f11@2000, but at "high sensitivity mode", which as we know, is basically +6db of gain, so the old F9@2000 of the X180 in normal mode is likely retained.

Paul

[Harry Pallenberg]

Any ideas when we'll see more info / footage on these cameras? Doug?

[Doug Jensen]

I've uploaded a bunch of my test footage from the Z280.

Sony PXW-Z280 test footage Part 1 on Vimeo

[Arild Pedersen]

Incompatible batteries. My Z280V only works with the original BP-U30. When I plug in a BP-U60 from BatteryChampion, I get "Battery Error please change battery".