"Biggest Advance in Camera Design"

[Wayne Morellini]

Wayne, I do not see how Open Hardware camera can be cheaper than similar performance proprietary one. Designing for hackability imposes additional design requirements, and currently developer of the proprietary system has more "freedom" than one working on the open design. Proprietary system developer is not limited by the development tools (we only use freebie FPGA tools so our users can recompile bitstream without spending extra thousands $$ in addition to the camera cost), they can sign NDA with the chip vendors, can use closed interface standards using the documents, not collecting reversed-engineered pieces of the information here and there, as we have to to be able to use modern sensor standards. To say nothing about closed products manufacturers are routinely infringing on each other patents, as Marc Levoy wrote when working on their Frankencamera.
On the other hand, there are applications where Freedom is appreciated, where users are eager to pay extra. Not a huge market, but it exists. And I do not believe in "poor man's Red".

Andrey

Sure enough Andrey, they usually can't be done cheaper, especially with a FPGA based design. But manufactures dumb down designs and put a lot of markup on them. Red achieved its killing by marking down in a custom design for massive volume, for a cinema camera). However, it is possible to do a lot cheaper on retail with a design spread across markets to get enough volume on consumer components already manufactured in millions of units. Are they ever going to natch Red, maybe not, but they can be innovatively the camera for everybody else. So I am encouraged.

However this is how they can be cheaper for manufacturers. A platform based on their design would allow a small manufacturer to quickly enter the market without the cost of designing one themselves. If the core can be transferred over to asic, then the better, or if a hardware neutral soft design could be had to work on mass components. I am eyeing the Snapdragon 800 at the moment, wondering what could run on it, GoPro/cineform is said to be eyeing mobile components too, and twice as powerful+ multi-arm systems are coming.

As for closed bad aspects, individual companies have to apply for permission. and a standardized individual license can be negotiated that companies can use to apply, even a group package can be arranged. There are many ways to skin the cat, but in the end we are likely going to see cheap Chinese manufacturers. Once off and running, of popular, after a few years, chip manufacturers might like to do a custom part that all the manufacturers can buy. I was looking into a soft programmable FPGA replacement chip that aims to be manufactured in the billions for my old designs. The 32 bit version might be more suitable, and there is still a massive market for that.


Thanks


Wayne.

 

[Andrey Filippov]

But manufactures dumb down designs and put a lot of markup on them.

Wayne, I do not think so - competition is strong and the market is not monopolized. So you are just competing with established players. And the strongest advantage for cheap hardware is just production volume. Yes, small and inexpensive boards like Arduino have sufficient production volume, but how can you build as many cameras as Red and make them even cheaper - I do not see.

However, it is possible to do a lot cheaper on retail with a design spread across markets to get enough volume on consumer components already manufactured in millions of units.

Yes, you can make alternative firmware for Canon, as people do. Or "free up" Red cameras by developing alternative code for their inexpensive, volume-manufactured products. But developing a camera from scratch - I do not see any "millions" here.

As for closed bad aspects, individual companies have to apply for permission. and a standardized individual license can be negotiated that companies can use to apply, even a group package can be arranged. There are many ways to skin the cat,

You can apply, but you will not get. Or get only for yourself and that will render the end product to violate the free software/hardware licenses. Or you'll make it just a closed product where "open" will be used just for marketing purposes - Microsoft loves this word too.

Andrey

 

[Will Keir]

Make a film about it!

It's funny that some people seem to think that to make a great camera you plug a great sensor into some kind of recording mechanism and you're done. One only has to look at the bother Aaton got into with their Dalsa sourced sensor to know that there's vastly more to it than that. I'm not about to go into details, but there's things you only learn about cameras when you "do them for real" and there's gotchas you never believed would occur until you're faced with dealing with them, especially as dynamic range, sensor size and frame rates increase.

Graeme

 

[Wayne Morellini]

Wayne, I do not think so - competition is strong and the market is not monopolized. So you are just competing with established players. And the strongest advantage for cheap hardware is just production volume. Yes, small and inexpensive boards like Arduino have sufficient production volume, but how can you build as many cameras as Red and make them even cheaper - I do not see.

Easily. I don't have much time to describe every facet, I am wanted on a forum far away, and it will take a number of days work.

We are talking video camera market, they might call it a cinema camera, but it is just a fancy video camera. They under deliver and markup all the time. Ambarella was one of the few real competitors and Santo, Kodak etc pumped out outstanding for the price devices, only steps away from uncompressed record ability. At a fraction of the cost they delivered, less a few cheap to implement buttons and features. I've looked at a $799 still interchangeable lens camera reduced to $299 today, and it has worse video capability then the 5 year old Sanyo HD 2000, or other various cheaper new ones. No credible reason for it, just bloody mindedness.

Now, take a step back, which two companies control most if the sensor market?

Taking a reference board on mass produced items and using it, drastically reduces your cost and all the mass produced components are cheaper. I will tell you the truth, several years ago I looked on the oem section of one of those cheap for the quality $200-300 ambarella based
HD cameras, and they had OEM cameras making capability starting at a thousand units, a thousand, it is all done reference designs. If such manufacturers had a quality reference design and parts list, then yes, they could do cheap cameras as just one of there lines.


>Yes, you can make alternative firmware for Canon, as people do. Or "free up" Red cameras by developing alternative code for their inexpensive, volume-manufactured products. But developing a camera from scratch - I do not see any "millions" here.

Hmm, you are not considering I am not talking about from scratch, but taking a reference board obviously with the mass produced parts. and "adjusting" the code for that at least. You get major cost reduction, because your part becomes another tablet/phone/camera based on the design with a few differing parts. Now you can sell the product as a tablet/phone and/or a consumer camera with Hugh quality mode and pro/cinema mode (I don't care so much for high end cinema camera as much as something that gives basic cinema results and good pro video). As a quality product with something extra, over 90% of your sales should be consumers not using it in cinema/pro video mode, bring the volume up and price down. If somebody had the money today they could start work on this. These root OEM manufacturers produce hundreds of product lines for themselves and under contract, and have camera divisions. Up to now, they have not had a cheap cinema/pro video reference platform to choose from to do better cameras. The day of FPGA is drawing to an end, as you can get more and more powerful mass produced cheap parts capable of the same work at cheap power. It seems they will become 4k pro capable in a big enough form factor before process technology maxes out benefits of decrease size, it is close, but a tablet should be able to accommodate more easily.


>You can apply, but you will not get. Or get only for yourself and that will render the end product to violate the free software/hardware licenses. Or you'll make it just a closed product where "open" will be used just for marketing purposes - Microsoft loves this word too.

Not really. What is not covered by another agreement but their own work can be free and open, and the parts owned elsewhere not be in that case, just a compromise but no conflict that their own work can be open. If you modularize the design the sensor interface, most things, can be swapped around to support different interfaces/sensors/codecs etc, including a set of usably free ones. Ways to skin a cat, if it is not working one way then find another.

With all standards groups they can arrange group licenses under a standard, and for whoever wants to sign individually. So you could go to bluray, JPEG group etc and eventually gain access to a standardised agreement so to speak, which was in every bodies financial interests to fast track licensing, use and popularity to make more money. So I did say it could be done eventually, to give them time to prove the value of the idea, years, when companies start to see the profit potential.. So then they can negotiate streamlined access to licensing for companies wanting to use the design. Standard reference/standards business.


Thanks

Wayne.

 

[Wayne Morellini]

Make a film about it!

.

 

[Andrey Filippov]

Easily. I don't have much time to describe every facet, I am wanted on a forum far away, and it will take a number of days work.

Wayne, I'm still having hard time understanding how to put your vision in the practical plane. All these reference designs are surely used for information by most hardware developers for both proprietary and open designs, as part of the all documentation you need to process. Kodak was really great in keeping their documentation open - something that I do not see from others. But anyway - those reference designs I would say are less than 20% of the design input data and are applicable to cheap Chinese manufacturers of proprietary stuff also.

Integrated circuits, resistors and capacitors, screws and nuts, lenses are also mass-produced items, and all the camera manufacturers can make and do make use of them. So were is your competitive advantage?

Not really. What is not covered by another agreement but their own work can be free and open, and the parts owned elsewhere not be in that case, just a compromise but no conflict that their own work can be open. If you modularize the design the sensor interface, most things, can be swapped around to support different interfaces/sensors/codecs etc, including a set of usably free ones. Ways to skin a cat, if it is not working one way then find another.

With all standards groups they can arrange group licenses under a standard, and for whoever wants to sign individually. So you could go to bluray, JPEG group etc and eventually gain access to a standardised agreement so to speak, which was in every bodies financial interests to fast track licensing, use and popularity to make more money. So I did say it could be done eventually, to give them time to prove the value of the idea, years, when companies start to see the profit potential.. So then they can negotiate streamlined access to licensing for companies wanting to use the design. Standard reference/standards business.

Wayne, maybe I'm mistaken, but I got an impression that Apertus plans to use MIPI and sensor documentation under NDA and so keep it closed from their users. Haven't heard they negotiated licenses compatible with the open designs with the owners, so I believe it is difficult or impossible. Otherwise they would not go that path. And if they already signed NDA - there is no easy way out of those agreements.

But I will be really happy to see I'm wrong and Apertus will make those restricted standards documentation available for free software developers.

Andrey

 

[Jake Wilganowski]

Update on the cam and first images:
http://nofilmschool.com/2013/11/fir...35mm-camera-prototype-axiom-alpha/#more-70100

 

[Wayne Morellini]

Wayne, I'm still having hard time understanding how to put your vision in the practical plane. All these reference designs are surely used for information by most hardware developers for both proprietary and open designs, as part of the all documentation you need to process. Kodak was really great in keeping their documentation open - something that I do not see from others. But anyway - those reference designs I would say are less than 20% of the design input data and are applicable to cheap Chinese manufacturers of proprietary stuff also.

We are speaking about general examples, another such example is graphics cards, were the design is mainly the reference design, even 100% for a good example, as manufacturers gave up the expense of developing their own. As for cheap camera manufacturers, the case maybe different, the software features selected different, but you could just 100% follow a reference design otherwise.

Not to get confused I am talking about what is possible, and separately what us possible under the axiom vision. Both are vitally important, the general what is, is important for axiom to be aware of.

>Integrated circuits, resistors and capacitors, screws and nuts, lenses are also mass-produced items, and all the camera manufacturers can make and do make use of them. So were is your competitive advantage?

Simple, a complex circuit mass produced can replace expensive low volume custom parts at a fraction of the cost. It might not be an advantage relative to anybody else doing it, but a firm is no longer at a disadvantage of competing on so many costly custom/low volume parts. So, your question is a bit confusing, but in context of a firm using costly parts it is a more competitive way to do it.


>Wayne, maybe I'm mistaken, but I got an impression that Apertus plans to use MIPI and sensor documentation under NDA and so keep it closed from their users. Haven't heard they negotiated licenses compatible with the open designs with the owners, so I believe it is difficult or impossible. Otherwise they would not go that path. And if they already signed NDA - there is no easy way out of those agreements.

As I said, what they could do, or could have done, in this case. At the moment the agreement stands, but to make a industry around their design they 'could' negotiate a seperate agreement in future for eventual manufactures (the manufacturers themselves should get together with them to do this eventually). Alternatively, they could keep that section of the camera a sealed entity, and you get to change everything else. It is like saying that an open source design is not open source because we don't have access to the internal priority circuit design of the Arm processor chip, of course it is still full open source in reference to our use of the part. So the bits that are open source remain so, and the bits that are not are not. However, it doesn't really stop anybody else from doing an open source version for other sensors and interfaces. If you want to get picky about it, you would have to go to antitrust laws about restriction of trade preventing use of a 'standard' interface. A standard being alleged not to be an interface that is restricted to exclusive use by a sole party or group. To conclude something can't be done is to not find a solution and stop it being done.

I'm sure you yourself Andrey, like many successful people, had lots of people, and a few significant ones, tell you what you were trying to do was not practical, unlikely, impossible, a dream or whatever other dismissive thing people say, but you like many, succeeded and did it.

>But I will be really happy to see I'm wrong and Apertus will make those restricted standards documentation available for free software developers.

What is relevant is that we have a full speed full featured interface to sensors available to the rest of the system, it does not matter what the details are that can remain hidden. All we should see is the data/control black boxed interface on the other side. As long as it is simple fully passing functionality and thin, there is no restrictions to performance and functionality, even if such data has to be repackaged as a subset commands and data of all possible commands and data generically in the passing interface to hide priority data and commands.

 

[Andrey Filippov]

What is relevant is that we have a full speed full featured interface to sensors available to the rest of the system, it does not matter what the details are that can remain hidden. All we should see is the data/control black boxed interface on the other side. As long as it is simple fully passing functionality and thin, there is no restrictions to performance and functionality, even if such data has to be repackaged as a subset commands and data of all possible commands and data generically in the passing interface to hide priority data and commands.

Wayne - I'm sorry. I read the text several times but was not able to completely understand it. Let's take just the last statement- about the sensor interface.
Sensor interface is an important part of any camera, especially if you plan others to be able to add their own boards with different sensors. If it is based on closed proprietary standards AND you signed NDA (so you can not disclose it), your users/developers can not work with it. Users who are not more or less developers themsales usually do not care much about freedom of the device. Unless they also turn to the Dark Side, pay significant member fee and get the same documenttaion. So your "open" and "moduular" system turns out to be rather useless without that important info, you can not attach anything to what you called a "black box" without information about the sensor interface.

 

[Robert Ruffo New]

Easily. I don't have much time to describe every facet, I am wanted on a forum far away, and it will take a number of days work.

We are talking video camera market, they might call it a cinema camera, but it is just a fancy video camera. They under deliver and markup all the time. Ambarella was one of the few real competitors and Santo, Kodak etc pumped out outstanding for the price devices, only steps away from uncompressed record ability. At a fraction of the cost they delivered, less a few cheap to implement buttons and features. I've looked at a $799 still interchangeable lens camera reduced to $299 today, and it has worse video capability then the 5 year old Sanyo HD 2000, or other various cheaper new ones. No credible reason for it, just bloody mindedness.

Now, take a step back, which two companies control most if the sensor market?

Taking a reference board on mass produced items and using it, drastically reduces your cost and all the mass produced components are cheaper. I will tell you the truth, several years ago I looked on the oem section of one of those cheap for the quality $200-300 ambarella based
HD cameras, and they had OEM cameras making capability starting at a thousand units, a thousand, it is all done reference designs. If such manufacturers had a quality reference design and parts list, then yes, they could do cheap cameras as just one of there lines.


>Yes, you can make alternative firmware for Canon, as people do. Or "free up" Red cameras by developing alternative code for their inexpensive, volume-manufactured products. But developing a camera from scratch - I do not see any "millions" here.

Hmm, you are not considering I am not talking about from scratch, but taking a reference board obviously with the mass produced parts. and "adjusting" the code for that at least. You get major cost reduction, because your part becomes another tablet/phone/camera based on the design with a few differing parts. Now you can sell the product as a tablet/phone and/or a consumer camera with Hugh quality mode and pro/cinema mode (I don't care so much for high end cinema camera as much as something that gives basic cinema results and good pro video). As a quality product with something extra, over 90% of your sales should be consumers not using it in cinema/pro video mode, bring the volume up and price down. If somebody had the money today they could start work on this. These root OEM manufacturers produce hundreds of product lines for themselves and under contract, and have camera divisions. Up to now, they have not had a cheap cinema/pro video reference platform to choose from to do better cameras. The day of FPGA is drawing to an end, as you can get more and more powerful mass produced cheap parts capable of the same work at cheap power. It seems they will become 4k pro capable in a big enough form factor before process technology maxes out benefits of decrease size, it is close, but a tablet should be able to accommodate more easily.


>You can apply, but you will not get. Or get only for yourself and that will render the end product to violate the free software/hardware licenses. Or you'll make it just a closed product where "open" will be used just for marketing purposes - Microsoft loves this word too.

Not really. What is not covered by another agreement but their own work can be free and open, and the parts owned elsewhere not be in that case, just a compromise but no conflict that their own work can be open. If you modularize the design the sensor interface, most things, can be swapped around to support different interfaces/sensors/codecs etc, including a set of usably free ones. Ways to skin a cat, if it is not working one way then find another.

With all standards groups they can arrange group licenses under a standard, and for whoever wants to sign individually. So you could go to bluray, JPEG group etc and eventually gain access to a standardised agreement so to speak, which was in every bodies financial interests to fast track licensing, use and popularity to make more money. So I did say it could be done eventually, to give them time to prove the value of the idea, years, when companies start to see the profit potential.. So then they can negotiate streamlined access to licensing for companies wanting to use the design. Standard reference/standards business.


Thanks

Wayne.

There are many markets, ad they don't all want the same cameras, nor work the same way. There is high end, low end and the middle in the pro world of anything, then there is ads, music videos, features and television, fiction and non-fiction. Even within non-fiction a show about architecture might use an Epic, because buildings aren't going anywhere, whereas a show that follows a bicycle race might prefer an ENG EX1 rig with a powered zoom built-in . Each of those markets have different needs and challenges, and thus would prefer a different camera. For example the RAW workflow of Reds is awesome for some, an annoyance for others. Consumers want something else altogether, mainly light weight and ease of use, not really a lot of control/power if it makes thing complicated in any way. (And easy vs powerful/control are in many cases a choice that must be made, you can't have your cake and eat it too).

There is no mass market as you imagine, that would be happy with just one camera, all the time. This is only imaginary.

Here's a quote from Jim: "Everybody asks me how to become a millionaire. I tell them first become a billionaire, then start a camera company."

 

[Gavin Greenwalt]

Open source only works when the people who use it can contribute to it. IT software works great as open source because you see a need and you can implement it. APIs and Python Modules etc work great because the people who are writing it would be writing it anyway.

This needs a photographer and a camera assistant to debug a FPGA chip. Good luck with that.

 

[Andrey Filippov]

Open source only works when the people who use it can contribute to it.

Gavin, you are exactly right describing the OSS model, but it is not the only way how it can work with the hardware. We run OH company for 12 years and so far never got any contributed hardware design. Only in the camera I'm working on right now we plan to use Verilog module made by others and released under GNU/GPL.

There is other side of the Free/Open designs - there is a market (not huge, but it exists) where users need it and so are willing to pay extra for this freedom.

Andrey

 

[Jens Rosengren]

This needs a photographer and a camera assistant to debug a FPGA chip. Good luck with that.

Haha true. I've had both good and bad open source experiences, things like Wordpress work fantastically. I've seen other open source projects become crap when developers decided they didn't get enough appreciation and left the projects, ending it all.

Looking forward to the result, hopefully it will affect digital cinema in a positive, noticeable way!
Hopefully it will be able to bend the way things work, try to get the most out of hardware like magic lantern did with dslrs!

 

[Wayne Morellini]

Wayne - I'm sorry. I read the text several times but was not able to completely understand it. Let's take just the last statement- about the sensor interface.
Sensor interface is an important part of any camera, especially if you plan others to be able to add their own boards with different sensors. If it is based on closed proprietary standards AND you signed NDA (so you can not disclose it), your users/developers can not work with it. Users who are not more or less developers themsales usually do not care much about freedom of the device. Unless they also turn to the Dark Side, pay significant member fee and; get the same documenttaion. So your "open" and "moduular" system turns out to be rather useless without that important info, you can not attach anything to what you called a "black box" without information about the sensor interface.

The problem is that you are looking at it from a different direction to what I was meaning. I am still a bit off, have just sat up. I will have to cut these conversations short, as I have to devo;te the better part of my days to other things. I am talking about not knowing the interface details, for it all to be black boxed and only knowing a axiom specific interface to the black box, to hide the ip of the real interface, and also the option of picking any interface that is not closed to use with sensors that use those. This of course means not just anybody can add a sensor, unless it uses an non closed interface instead. Now of course, only one group needs the license/nda to do the black box.

The user is interested in openness to plug in new functionality modules and sensors. Manufacturers are interested in using and modifying if desired the design. If they want a new sensor then the ones supported by the black box they can get agreements to implement the changes themselves. On the end though, they may go yo the sensor manufacturer and show them potential volumes from the industry and ask them to adjust the? black box design. Axiom only has to specify the black box physical, em, thermal and interface for the. A effective solution to a tricky issue.

Again Andrey, this is about effective solutions, what can be done, not what they are current lay doing.

The rest I've answered above.

 

[Andrey Filippov]

I will have to cut these conversations short, as I have to devo;te the better part of my days to other things.

Wayne, let's do it. I fail to understand anything at all.

 

[Wayne Morellini]

There are many markets, ad they don't all want the same cameras, nor work the same way. There is high end, low end and the middle in the pro world of anything, then there is ads, music videos, features and television, fiction and non-fiction. Even within non-fiction a show about architecture might use an Epic, because buildings aren't going anywhere, whereas a show that follows a bicycle race might prefer an ENG EX1 rig with a powered zoom built-in . Each of those markets have different needs and challenges, and thus would prefer a different camera. For example the RAW workflow of Reds is awesome for some, an annoyance for others. Consumers want something else altogether, mainly light weight and ease of use, not really a lot of control/power if it makes thing complicated in any way. (And easy vs powerful/control are in many cases a choice that must be made, you can't have your cake and eat it too).

There is no mass market as you imagine, that would be happy with just one camera, all the time. This is only imaginary.

Here's a quote from Jim: "Everybody asks me how to become a millionaire. I tell them first become a billionaire, then start a camera company."

It just scrapped my reply again when I pressed submit on the inline quote form. I will have to restart.

Hi, Robert, welcome. I'm wrapping up long discussions here, as it is now known that I've been invited by Apertus to contribute ideas over at their forums a while ago, and this discussion has been stopping me from starting in the last week. I need to devote more of my day to do things. I'm not here trying to undermine or dictate what they are trying to do, but advocating possible effective solutions to objections here. It is easier to knick than to build effective solutions.

I've answered everything you mentioned here and elsewhere (except for the difference between being a billionaire and becoming a millionaire? Less than one cent. So neither does Jim saying it cost him lots and lots of money prove that it will cost lots and lots of money (he did do it in a most expensive way) or that becoming a million air costing less than a cent prove it will cost less than a cent). The truth is different things can be done in different ways more cheaply. For what Red did it was a good plan as long as they got the numbers they did.

As I have written before and have readied a post for on Apertus. Most forms of shooting can be supported through different software selectable operational/control modes and modules on the one camera. As for mass market, you might be referring to my sperate idea from Apertus, for a camera based on commodity parts that is sold as a point and shot stills,video or tablet camera, but has professional performance modes of operation, to be a professionals pocket camera that any professional would love to have, but that can also be mounted, extended (modules etc) and put in shooting case of any form factor to share it the style of shooting. The sales for a quality product can be in the millions, mostly to normal users, and to most professionals as a personal camera. The price, very cheap. This is the type of camera I tried to sell Jim on before they eventually went to the Scarlet fixed. There is no disputing, sub $500 internal 4k cameras with adequate sensor have been out for a while, just no uncompressed recording path. Nikon j1 was one.

 

[Wayne Morellini]

Thanks Andrey. My thoughts are fairly compressed and flowing. Unless people match the tempo and flow they have trouble understanding. I've noticed around here that the engineers thoughts tend to jump around skipping through the flow. It confused me for a while, as there was usually nothing there for them to fade out on (one of the things that happens with not so smart people who cant understand, rather than skipping to knowledge). My mind can skip to massive amounts of knowledge when reading or writing, but self orders into logical flows around the flow I'm reading These flows can be ordered multi-flows. Eventual solutions to difficult problems are seldom about what we know, but often about what can be done.

Re-edit: Hmm, I might need to clarify here. Not so smart people will fade out missing pieces of text, and past that link things wongly to their "experience" knowledge, even extrapolate wrongly from that, so we are nearly all guilty of that at times. A smart person though might read but link to variable information that is jumping through their head that relates to something, as a localised conclusion, skipping pieces of the flow around and through it, but not actually analysing what is being said consistently and thoroughly, the flow so to speak, the jumble of conclusions produces wrong results. This is hit and miss and can miss 90% of what is meant and produce 90% more arguments.