FS7: 5 reasons the URSA Mini is better than the Sony FS7

[Aaveum]

Heresy mode: On

I see no reason why the images wouldn't be considerably better in every way. Not least because the FS7 is running a sensor designed years if not a decade ago

Heresy mode: Off

This doesn't play a as large role as it would imply. The industry standard and benchmark, the ARRI Alexa, runs on an over 7 years old sensor, which actually goes back to the D20 sensor released in 2005. Since the current sensor is an evolution of that we could say it's a decade old design. The F35, also mentioned in this thread, produces one the best images and runs on a sensor which is at least 13 years old. The Panavision Genesis, which uses the same sensor was released in 2004.

The reason for this is we have long reached and passed the point of diminishing returns.

 

[starcentral]

I can open the files before any debayering is done and see the raw greyscale channels. I can either open them in Nuke or there's a nice package called RawDigger, this can do it as well. This allows me to see the 4 channels (RGGB) and see the actual recorded values.

When you open up various DNGs you can scoot across the values. The FS700 & FS7 are always even values, meaning that the smallest bit is always a 0, never a 1. This means the value is 11 bits with a padded 0. By comparison Bolex and the other BMC cameras will show odd and even values, a full 12 bit range. In the case of the FS7 slow motion, it's padded two bits, so 10 bits in 12.

The BMC cameras also include a 1D LUT, a linearisation table. Normally a DNG stores a linear range (FS700, FS7, Bolex), which shows a doubling of light, so 12 bits means 12 stops unless there's some subtle curve on it but that's pretty the real world case. With the BMC cameras, including the Ursa Mini, there's a 1D table which means the data in the file can be more log like and the DNG reader will apply the transform and that's why a BMC file can hold more than 12 stops and also distribute those stops better.

BMC have always done DNG raws very well. Issues in the past have been more about their colour science. Resolve has always been the glue that knocks that into shape.

There's a whole level of what RGB data is stored, whether it's in a colorspace or not (it never used to be) i have no idea with the Ursa.

Wow that is some cool stuff you are doing - hacking video files! Impressed. Any idea why some clips are 11 bit and some are 12? As I understand it there are three RAW modes or something on the URSA.. could that be why? also are you looking at 4.6k created files, or perhaps are a few from the 4.0k camera? Maybe there is a some difference there. Great sleuth work nonetheless.

Dennis before you think about swapping out an in demand F55 for an Ursa 4.6k mini I would wait for feedback from those shooting in the field. I have more questions about the practical aspects of working with it than the image quality in raw.

No it's not like that. I would keep the F55 or downgrade to F5/FS7 as the workhorse for most of the work I do. Then the F35 or 4.6K would come out for lower budget projects, indie stuff, or just ones that I want more of that oomph on.

I'd feel naked without a "proper" camera like FS7/5/55 and only having the URSA. As I said a few threads ago the URSA is not better than the FS7 when you look at features, benefits, reliability, etc… I only like the 4.6k so far because of the image.

Heresy mode: On

I see no reason why the images wouldn't be considerably better in every way. Not least because the FS7 is running a sensor designed years if not a decade ago

Heresy mode: Off

But the FS7 is an all round workhorse and has 'pretty good' across everything it does.

But Dennis, this may be the camera for you. I've always maintained that RAW is where i think you'll see your mojo. The Ursa and Resolve will go together perfectly, so there's a known workflow for you. If you can handle the data rates then i think it's well worth looking at. IMHO nice as the F35 is, i suspect the post workflow will be a tough cookie.

I'm interested too, but will wait everything out until i need to get something else, and let Red, KineMax and Panasonics into the wild and see what the feedback is like.

My concern with the Ursa is sensitivity, what's the high iso like? I do shoot FS700 at 6400 and above some times. If i can't achieve the same exposure on the Ursa then that's more difficult.

Exactly. The FS7 can't be beat for an all around camera and in a straight up 4.6k vs FS7 I see no reason to go for the 4.6k unless the characteristics of the image and RAW is what you need. Even that will come with headaches for some.

It's the same as many who've incorporated a FS7 and a RED into their kit. Maybe they don't want to lose the "RED" jobs but if the FS7 was really capable of making the same images as RED then it would have happened by now. I do feel like XAVC 10-bit is over rated. It's fantastic for a lot of things but maybe 10-bit would have been better for a 10-12 stop camera where each stop was getting more data. Now, the data allocation just gets diluted… so again as I've said before paper specs are one thing but go take the images through post and see for yourself what's what. In post my BMPCC beats my F55 XAVC hands down. We're talking a $995 camera vs $35,000. Add the Sony RAW module and ok well things change drastically but all in all my opinion now is that 10-bit is not enough for the F series and maybe making it worse is the XAVC compression. SR in 10-bit on the other hand is almost like a different world. Codec, bit depth, all this makes your head spin.

About ISO, I'd be hard pressed to shoot over 3200 ISO on my F55. 6400 ISO max but then it's grain surgery later on with NR in post. For narrative work where I would use lights 800 ISO is all I need and the URSA 4.6k seems good at 800-1600 ISO so far from what I've seen. But we definitely need some more field tests.

RED has gotten too expensive for what it is. They've gone ahead and protected their higher end camera's (and rightfully so) by adding a lot of compression on the new lower cost RED models. For that reason I'm out. As for KineMax, Panasonic, well they just never really tickled my fancy and I guess as a pocket cam owner since December 2015 the little thing has blown me away. I know Blackmagic makes a lot of people angry with their announcements and delivery but I think the products are really well built and designed, and these guys have a huge background in color and signal processing in general.

What an interesting time it is and I'm wondering what NAB will bring.

Wasn't this sensor designed years ago as well? It's the camera that is late not the sensor

do we have any findings from the pro res recording and how good/bad it is?

what is the audio like? Are the preamps good enough? Does phantom power work properly?

Screen good enough for pulling focus? Decent monitoring lut implementation? How much flex is in the mount compared to the Fs7 or the Red Raven?

I know there's no internal ND and that the camera is very sensitive to IR contamination like the last generation of cameras were. How does it play with IRNDs? Is it neutral with the use of Firecrests?

These are all good and legit questions. For me personally I don't shoot any camera and use it's audio for anything more than scratch track so I'm not concerned about preamps, number of inputs, etc.. The F55 can accept AES 4 channels of audio digital from a mixer which is cool, but again never used it and don't want to be tethered anyway. A 5" screen even in HD might be hard to pull focus from but I've done it off 7" with only 1280x800 screens without any problems, and the 1st AC's I've worked with. Maybe the 4.6k non-mini with 7" screen is better way to go? As for internal ND I rarely use internal ones and either screw on 82mm filters with an assortment of adapter rings, or if using CP.2's then run a matte box and filter trays. Personally I hate matte box and filters, etc.. I'm quite lazy.

My only question is when is the 4.6k shipping?

 

[Ben Scott]

Now isn't it?

 

[paulcurtis]

This doesn't play a as large role as it would imply. The industry standard and benchmark, the ARRI Alexa, runs on an over 7 years old sensor, which actually goes back to the D20 sensor released in 2005. Since the current sensor is an evolution of that we could say it's a decade old design. The F35, also mentioned in this thread, produces one the best images and runs on a sensor which is at least 13 years old. The Panavision Genesis, which uses the same sensor was released in 2004.

The reason for this is we have long reached and passed the point of diminishing returns.

I'd agree partly. Sensor design is about trade offs, what is more important at any one time. For example sensitivity is often traded for colour separation. If the dyes on the CFA are wide spectrum they let more light in, but that light isn't as pure as it could be. See contamination issues with Red at the start. Or even the IR issues we see today. IMHO these are all do with that balancing act.

I think it's the CFA evolution that's perhaps of more interest.

The Varicam is interesting because of this and that dual ISO nature of the sensor. That really is AFAIK quite different.

And the Ursa Mini, well it will be interesting to see what BMD have really done with it. With 2 green photo sites on the sensor for each pixel i do wonder whether people are experimenting with making them different (they're not in the Ursa).

When Sony did their sensors they were about the resolution. Arri was about the colour. But the market has changed now and whilst there may be nothing super different in hardware, those evolutionary tweaks do add up to a difference.

So in summary, the core of sensor tech is somewhat locked to physical properties. But what gets built around this does appear to change and improve.

cheers
Paul

 

[Ben Scott]

Yup a good example is the Dragon sensor being matched to new boards/ASICs in the Weapon and now having less noise and better compression, and faster data pipes/media meaning it can do simultaneous Redcode/Pro Res.

I guess different areas of tech move at vastly different paces.

 

[paulcurtis]

Wow that is some cool stuff you are doing - hacking video files! Impressed. Any idea why some clips are 11 bit and some are 12? As I understand it there are three RAW modes or something on the URSA.. could that be why? also are you looking at 4.6k created files, or perhaps are a few from the 4.0k camera? Maybe there is a some difference there. Great sleuth work nonetheless.

I think hacking is a bit strong, it's not that difficult! But i do like to know where the image is coming from! You can learn a lot by looking at the steps an images goes through to a final image. I did my own poor mans debayer in Nuke for a laugh (well, okay it's not really that much of a laugh) but it did teach me a lot of colour science.

When you see the image in that form, you can quantify differences.

You could well be right about when i saw 11 vs 12 bit, perhaps that's down to different modes on the Ursa Mini. All the files i've looked at are 4.6K that's the only one i'm interested in.

Exactly. The FS7 can't be beat for an all around camera and in a straight up 4.6k vs FS7 I see no reason to go for the 4.6k unless the characteristics of the image and RAW is what you need. Even that will come with headaches for some.

One thing that comes to mind is that with that extra space you can go back to film days and actually look outside a 4K frame through the viewfinder. This would save quite a few booms in shot and also make for better operating choices. I really find the extra resolution appealing, more so for cropping, stabilisation and generally getting a great 4K crop.

About ISO, I'd be hard pressed to shoot over 3200 ISO on my F55. 6400 ISO max but then it's grain surgery later on with NR in post. For narrative work where I would use lights 800 ISO is all I need and the URSA 4.6k seems good at 800-1600 ISO so far from what I've seen. But we definitely need some more field tests.

I would like to see 6400. I know there have been times where i've had no choice. With RAW it's a bit better in terms of pulling detail out. I had some RAW images that were really pushing it, i was really happy how much i could extract from them.

Does the Mini even go up to 6400?

cheers
Paul

 

[starcentral]

I know a lot of people in post and 9.5 times out of 10 when they shoot RED in RAW they immediately transcode to ProRes HQ and that becomes the file for editing, grading, even VFX work believe it or not. Yeah you need storage but it's cheap price to pay for more image quality if you look at it that way.

I'll check my library but I might have 6400 ISO somewhere. If not I can shoot some quickly one night.

Not sure if mini 4.6k goes that high but you could just push it in post and it'd sort of be the same thing.

 

[starcentral]

This is the thing. We don’t know yet. You can only take through post what is been shot and posted and that is not enough to judge yet. So it looks like you are also getting caught up in comparing specs. Because till I see some stress tests I don’t have enough footage to make a decision. So what would be helping the decision is the specs.

First of all not being 4K doesn’t matter. Second, just now you realize the F35 is larger and needs bigger support? The battle tank part is what I like. I miss this with my FS7. I miss that high end build of the big boys instead of the little toy build. So I’m fine with that! I never used an Ursa Mini but the Ursa is not on pair with the build of the big boys.

I agree as long as you are not talking about the F35. But I already said the Ursa looks better than the FS7. Also looks better than the F5 and F55. But the F35 has a look of its own. It’s not Sony for some reason.

Well for me I can take some frames that someone shot without any lighting or major effort and the stress test would be putting that image through post and seeing what I can do to it, how it looks in the end, or what kind of treatment it may or may not need. i.e. film emulation LUTs, or other post "tricks", etc..

But yes we need more innocent non staged shots and that's how you can judge it, but better to shoot them youself indeed so you know for sure.

I did realize the F35 was larger before when I finally made up my mind to take the plunge, but in the back of mind I had reservations, and some close DP friends were also chiming in about size, cost to repair, external deck, tethering, etc..

cheers,

 

[Ben Scott]

I know a lot of people in post and 9.5 times out of 10 when they shoot RED in RAW they immediately transcode to ProRes HQ and that becomes the file for editing, grading, even VFX work believe it or not. Yeah you need storage but it's cheap price to pay for more image quality if you look at it that way.

I'll check my library but I might have 6400 ISO somewhere. If not I can shoot some quickly one night.

Not sure if mini 4.6k goes that high but you could just push it in post and it'd sort of be the same thing.

You could say that of absolutely any raw workflow to be honest.

My experience is you either cut and grade native redcode raw, or you do a 'onelight' pass grade in Redcine-X (grading to a redgamma) and hand that .RMD sidecar over with the footage and they'll switch to RedLogFilm before transcoding so they have a starting point primary grade log file.

 

[Aaveum]

I'd agree partly. Sensor design is about trade offs, what is more important at any one time. For example sensitivity is often traded for colour separation. If the dyes on the CFA are wide spectrum they let more light in, but that light isn't as pure as it could be. See contamination issues with Red at the start. Or even the IR issues we see today. IMHO these are all do with that balancing act.

I think it's the CFA evolution that's perhaps of more interest.

The Varicam is interesting because of this and that dual ISO nature of the sensor. That really is AFAIK quite different.

And the Ursa Mini, well it will be interesting to see what BMD have really done with it. With 2 green photo sites on the sensor for each pixel i do wonder whether people are experimenting with making them different (they're not in the Ursa).

When Sony did their sensors they were about the resolution. Arri was about the colour. But the market has changed now and whilst there may be nothing super different in hardware, those evolutionary tweaks do add up to a difference.

So in summary, the core of sensor tech is somewhat locked to physical properties. But what gets built around this does appear to change and improve.

cheers
Paul

You said you partially agree but then went a bit off topic. The point is, the current industry standard is a 7 years old sensor. That questions the point of putting any weight on the old "new is always better" concept. This was my point.

 

[Ben Scott]

All sensors are a couple of years old at least by the time they hit the street.

The Alexa sensor hit an absolute sweet spot and sweet time in the market when the majority of people (content creators AND consumers) were not even vaguely interested in resolutions above 2k except for a few key players and for archiving. I wonder if Arri can repeat the feat when it comes time to introduce a new sensor which will obviously have to be 4K at the very least, which can't be miles away. Certainly if the Alexa sensor was released today it most certainly wouldn't last 7 years.

I assume the Arri65 is just a larger slice of the Alexa wafer.

 

[starcentral]

yeah I agree with Aaveum.

Sensors themselves are not what's holding back the class. Personally I'd take a sensor in a plastic electronic project box so long as someone could develop a really good OLPF, CFA, and color science for it - and I'd run it tethered to a 7Q to record 4K and be a very happy camper. I don't care about XLR jacks, ND filters, and a dozen other features buried in a menu somewhere.

 

[paulcurtis]

You said you partially agree but then went a bit off topic. The point is, the current industry standard is a 7 years old sensor. That questions the point of putting any weight on the old "new is always better" concept. This was my point.

Me? Off topic? Surely not...

All i meant was that in the case of the Ursa Sensor design started, what a few years ago? Production hitting now.

In the case of the F5 it was released 2012? And the design of that sensor probably started a few years before that.

So the Ursa sensor is perhaps 6 years ahead of the sony? I can't say when the design of it started, but certainly BM tweaks to it are a lot more recent.

And my argument is that in 6 years there is quite a lot of change, perhaps not with the silicon itself but with the rest of it. So to my mind the Ursa sensor could be considered a generation newer? I'm sure there are Sony sensors in R&D right now that would be pretty amazing.

cheers
Paul

 

[basspig]

In regards to audio I'm not so sure any of them should/ could be classed as great .......other than for guide tracks including Red, Alexa, so Mini Ursa is in good company.

Exception is Sony IMO. pretty rock solid audio in their better cameras showing their broadcast origins.

XDCam EX line is the only camera I trust for DAT quality audio and flat 20-20,000Hz preamps.
The FS7 is quieter, but that 7.4dB rolloff by 20Hz disqualifies it for 2-channel budget concert shoots. That said, the line inputs are stellar, if you want to drag around and worry about powering an external high grade preamp.

 

[Ben Scott]

OMOD XLR for Red is going to be absolutely class-leading.

 

[DLD]

... in the case of the Ursa Sensor design started, what a few years ago? Production hitting now...

Fairchild listed their "off the shelf" sensor in November, 2014 (I posted the link on the Kinefinity thread). One has to assume a couple of years of work prior to that, although a lot of R&D can obviously be general or concurrent. BMD then has had a year and a half to customize the chip for their product (they claim to have designed it but it's unlikely to have been more than "ordered to spec").

And the 4.6 Mini is shipping but there one better consult a local retailer.

 

[starcentral]

Does anyone know anything about possible version 1 and version 2 sensors on the 4k mini?

I was reading about the black hole sun issue on them. Some claim it was a "version 1" sensor thing only.

How to tell if a camera has a v1 or v2? (not that I'd consider the 4k over the 4.6k anyway but genuinely curious)

 

[NorBro]

I'm sure this was mentioned somewhere here, but in the BM video that addresses the public, Grant said that the delay was because of the failure to implement the global shutter (to their standards). So they could have most likely shipped this camera sooner (like in July 2015).

That OMOD XLR looks fresh.

 

[NorBro]

Does anyone know anything about possible version 1 and version 2 sensors on the 4k mini?

I was reading about the black hole sun issue on them. Some claim it was a "version 1" sensor thing only.

How to tell if a camera has a v1 or v2? (not that I'd consider the 4k over the 4.6k anyway but genuinely curious)

I was actually part of this conversation recently, and I claimed (along with dozens of others) for the past year that it was a V1 vs. V2 thing, but John Brawley recently said it wasn't.

It's a fact that the URSA Mini 4Ks have a V2 sensor in them and I - along with many others - have not seen the black sun spot in this camera (where my V1 URSA had it horribly), but there is no official statement from BM that the sun spot is fixed in the 4K Mini (like there was for the Pocket via firmware).

 

[starcentral]

man sound like a future firmware update may very well indeed un-lock global shutter. After all, the Fairchild based super35mm sensor they are using supports it. :thumbsup: