CineD Super Impressed with the V-Raptor

[Jon Pais]

 

[Nick Vera]

The fact that it has slightly more DR than the Sony Venice 2 8.6k sensor was astonishing. On top of the frame rate options, flexible post, and size of the camera for travel, the RED team really did a fantastic job with the camera sensor. So hat's off to the team. Mind you some years ago, I would not have ever thought of shooting on RED again. But they did a fantastic job on this generation's sensor. Glad to have invested in one!

 

[Ryan Chernin]

Doesn’t this generation Alev/Arri sensor incorporate a form of temporal noise reduction at the sensor level? I’m wondering if that’s partially giving it a DR advantage over the Raptor when using imatest.

 

[Phil Holland]

Doesn’t this generation Alev/Arri sensor incorporate a form of temporal noise reduction at the sensor level? I’m wondering if that’s partially giving it a DR advantage over the Raptor when using imatest.

Correct. Their ISO weighting/calibration is also different. ARRI is using a dual gain architecture and normalizing the noise levels of the two different signals then applying what they now call a Texture to re-grain.

Though my measurements are going to produce different results than CineD's testing methodology, their method does produce consistent disparity metrics similar to mine. Meaning what they found between Raptor and the V2 or Raptor and the A35 is about right in terms of stops disparity even though our "count" is different.

But I will underline that there is more that should be looked at.

 

[Mike P.]

Doesn’t this generation Alev/Arri sensor incorporate a form of temporal noise reduction at the sensor level? I’m wondering if that’s partially giving it a DR advantage over the Raptor when using imatest.

I don't think so. No amount of post noise reduction gives REDs as much usable/colour accurate/colour consistent latitude as Arris (as seen in the raptor noise reduced -6 examples). A better way to look at it is that it's more akin to a per-frame "black shade" done on-the-fly than post noise reduction (doubt it's more destructive than a black shade, and even if it were, you certainly can't see it in the resulting images). Regardless, whatever Arri is doing is much (much) better than having internal/always-on highlight reconstructed stops whose most tangible benefit (whether intentional or not) seems to be artificially inflating Xyla chart numbers by two whole stops (because the chart is b&w, you/IMATEST can't see how unusable they are).

That gaudy yellow of the forehead/face/neck when they go only +2 happens frequently even when you expose the skin tone lower than 60% (if there's punchy/contrasty lighting, the high-side light isn't as direct/flat/soft, or the skin complexion has more sheen/grease). I've seen it often in even tame conditions on RED and nowhere else; even older/lower DR cameras (I'm talking 15 year old 5Dmk2... 11 stops, but +/- 5.5) don't crap out so badly when it rolls off into the extreme highs. It also makes exposing less bulletproof/standardized. "Brighter" but monochromatic parts of the frame that have a higher overall IRE than the 60%+2 (like the light grey/white chips of the chip chart which are actually higher on the scope) are still "captured", while the "lower" high side skin tone is clipped grossly. So if you're *not* using the stoplights you'd never know it's gone to poop and while owner/ops view all the raw tools on their camera's monitor, false colour, histogram, and video village (and/or Ops/DPs not familiar with stoplights) typically don't.

Though my measurements are going to produce different results than CineD's testing methodology, their method does produce consistent disparity metrics similar to mine. Meaning what they found between Raptor and the V2 or Raptor and the A35 is about right in terms of stops disparity even though our "count" is different.

Interesting. That's the first time I've actually heard you admit as much. So being that they have a +1 for ALF and +3 for A35, are you saying that RED's are 16+, and by that measure ALF is actually 17+ and A35 is 19+, or are you saying that Arri's advertised 15+ for LF and 17+ for 35 are accurate, but Monstro/Raptor are actually only 14+?

 

[Phil Holland]

Interesting.

It's more than a number rating when it comes down to it, but I'll do what I can to explain what the differences are.

At ISO 800 Arri is "about" +2-ish in highlights due to the Dual Gain/Dual Readout with the caveat of temporal NR and all that compared to V-Raptor at the same ISO. I mention the NR as that will have an impact on IMATEST readings, which is one reason why manufacturers don't go by that metric. If you test the whole ISO range of every camera IMATEST can be off by about 3 stops even, which is wild.

A35/Raptor difference at 800 ISO. From 18% gray you have about 9+ in highlights on ARRI Alexa 35. RED V-Raptor is 7+ or 8 exactly depending on how you interpret results at ISO 800.

TCDR. V-Raptor is 17+ Total Captured Dynamic Range. Monstro similar. Venice 2 is slightly less, but not by much, it is about a stop though.

Both V-Raptor and A35 see up to the 19 stop wedge on a Xyla pretty cleanly and there's a bit more than that. Nobody would say 19 usable for either camera, but you certainly would be using that at higher ISO ratings. But the weighting is more of the tale with that camera to where they put the stops. ARRI goes towards the highlights. Similarly to where RED favors shadows with their weighting/calibration. Shadows aren't wonderful at higher ISOs on the A35, but not bad. Raptor does pretty well in that category.

Venice 2 is Dual ISO however. 800 and 3200. Monstro shows one particular strength in having interchangeable OLPFs and you can get a rather clean image with the Low Light Optimized OLPF. RED moved away from that concept with Raptor for an OLPF that is more bulletproof optically when it comes to artifacts in tricky situations. Which I certainly approve of. But with a less aggressive color filter in there we'd likely have native ISO 1600-3200 ISO, but prone to dots for sure.

My more based take is Raptor, Monstro, Venice 2, and Alexa 35 are bloody close in Dynamic Range. But you'll want to know how to get the most of out of them if that's your goal. I've matched all of them and the side by sides mostly reveal subtle differences in image processing, color science, and that DR weighting. Noise/Texture is an interesting comparison too as you can truly see the impact of image processing or not.

The rundown on this is more of an interesting tale when you consider price point, format size, and frame rates. Also physical body weight/size is fascinating. When I see Monstro's sell for under $20K I'm a bit blown away compared to cameras available today. Raptor could be 1.5X or 2X the price and still compete as only the A35 "slightly" outperforms it in some categories (certainly not all), but please don't tell RED that. I also don't think that's what the general market RED aims at would want either. To that point the cost of Monstro back in 2017 seems to make a lot of sense considering it's longevity compared to everything new. I personally wonder what another $80K RED body could do, but that's not something that is happening currently.

To date I have tested the A35 twice (pre-pro and production), V2 twice, and 5 V-Raptors over the past year. All providing consistent results. Was worried that the XL had more jazz, but it produced the same as the smaller V-Raptor.

 

[Leonardo Newman]

Thank you, Phil, looking forward to your results if you decide to release them. I saw the xyla on Scott's live stream. I do agree that at the very least the testing seems consistent so you can still compare based on the consistency even if the numbers may not necessarily be right. I'm just confused as to why they are trying to say that the mini LF has a stop more DR when their own numbers and CVP's video dispute that very claim. They are trying to knock down points for highlight recovery but again, their own latitude test and CVPs latitude test show that whatever raptor loses in the highlights compared to LF, it makes up for in the shadows, at the very least they should be tied and if decided to really take the highlight recovery into account then the results push raptor over the edge. Again their own database now puts raptor in 2nd only behind Alexa 35, which like you said is using NR in camera, and if you gonna take away stops for highlight recovery, shouldn't you then also take away stops for NR? But yet they still report their results as saying raptor is a stop less than Alexa LF because LF doesn't have highlight recovery.

In any event, Mike I think what your saying about reds clipping quickly is laughable and very far from the truth. I can recover 5 stops over mid grey at iso 800, in most circumstances the only way your clipping your sensor is exposing for dark talent pointing directly at windows with no light on the talent, or over exposing. Any good DP even without knowing how to get the most out Red in particular would not be clipping any shots the way your claiming. It’s is very much misguided hyperbole that you wrote. Yes at 800 arri has more highlight range, but Red has more shadow range, they very much even out, CVP and even cined basically confirms raptor having equal too or more dr than LF depending on how you interpret their data. As it relates to highlight recovery, I have some example images that show off just how useful highlight recovery actually is. I don’t get why people are trying to hate on it. Also, red desaturates as it clips, (which you know, considering I saw you mention it before on here) I’ve never seen my cam do anything but turn grey at clipping, I posted an over exposure test in the Reduser (official) Facebook group, and the color doesn’t shift as you over expose, and it doesn’t become saturated as you clip, c300m3 and Sony Venice exhibit this behavior, but red surely doesn’t. So I’m confused as to you saying that is your experience. I’ve even done further testing where you can bring back some of the color in those grayed out details.

 

[Jon Pais]

As an aside, whereas ARRI used to desaturate color in the mid-tones and highlights, with saturation basically stopping increasing beyond middle gray, now, looking toward the future with HDR, the new REVEAL color science allows for bright, saturated highlights. You can always reduce saturation in post if you want to emulate the film look and ARRI is looking into putting something like that into their Alexa 35 for those that want it.

 

[Tom Gleeson]

I thought CineD look at the Raptor was very good but as they discussed themselves it was only a look at exposure latitude or dynamic range. Going back only 10 years DR was a big issue and the achilles heel of digital capture but as all the top end cameras move to ever increasing performance it seems a little less important. I would have loved if CineD could have looked at colour and gamut reproduction but I know only too well the vast amount of work that goes into doing credible testing and comparisons.

 

[Jon Pais]

Not sure if it's got any relevance to the conversation, but when it comes to HDR, well over 2/3 of the dynamic range of current display technology is situated in the shadows. On a PQ reference display of 1,000 cd/m2 (e.g. Sony BVM-HX310), you've only got a touch over five stops above 18% middle grey (26 cd/m2) before reaching peak luminance, whereas there are well over ten stops (sorry, I have trouble counting backwards!) until you reach .0005 cd/m2.

 

[Leonardo Newman]

Not sure if it's got any relevance to the conversation, but when it comes to HDR, well over 2/3 of the dynamic range of current display technology is situated in the shadows. On a PQ reference display of 1,000 cd/m2 (e.g. Sony BVM-HX310), you've only got a touch over five stops above 18% middle grey (26 cd/m2) before reaching peak luminance, whereas there are well over ten stops (sorry, I have trouble counting backwards!) until you reach .0005 cd/m2.

indeed, the meat of the image is still found below 100 nits, but the added color gamut and expanded contrast of HDR still makes images look nicer overall. It’s such a beautiful thing.