DIY Dual 8 Telecine HD Digitizer under $1000?

[granfer]

Roger gets to the core of the matter in his usual inimitable style, and by saying some of the things some of us were reluctant to say!


First off, to answer the question "Can a dual 8 digitizer be built for under $1000?" and then he adds ...
The answer CAN be yes but is best answered with a variant on the same question,"Can YOU build a dual 8 digitizer for under $1000?".

Actually, Thomas actually asked the question "With a Budget of $1000 is it possible to do the following?" (which is substantially Roger's question), but started the post with ... "I'm looking at building a Dual 8 Telecine HD Digitizer for under $1000". which read in context with the later question I interpreted (reasonably) as
"Can I build a Dual 8 Telecine HD Digitizer for under $1000"?

There is enough in the following paragraph to make a reasonable assessment of Thomas' current knowledge and expertise in the specific field (no disrespect intended Thomas, we have all previously made the same surmises and asked the same questions!), and Roger has from experience recognised the areas of concern and advised caution, because he knows that without the relevant knowledge and other practical attributes $1000 dollars would very likely be only a starting point for the machine Thomas envisages.

Here are a few of the items I believe would be required.
Projector- Variable speed and/or Ultras slow motion of 8 FPS Dual 8 Projector that historically is gentle and won't chew on film. (Ebay has tons of these)
Motor - It may be necessary to use a controlled stepping motor in order to control frame timing (I'm not sure what to look for or if there is a good projector that would already be suitable)
LED Lamp - Using combined RGB LED array to give a better white than the fake white LEDS at 100W-200W equivalence or about 1600 lumen - 3200 lumen
Controller - USB controller to control LED Gain levels and timings (can't find anything like this except cine2digits which make an obscure reference to it)
Power Regulator - Possibly additional power would be needed if the 2.5W from the USB port isn't enough to supply the controlled LED lights
Machine Vision Camera - 720P or higher (2MP preferred), ~25+ FPS, Global Shutter, controllable frame rates for timing purposes. (I think CCD is better but can a CMOS with global shutter work too or is it too grainy?)
Marco lens - Macro lens with c or cs mount than can fit inside of the original projector lens assembly
Mount - Adjustable camera mount to hold the machine vision camera in place with Projector

I, too, have built/modified a number of machines, and spent much more than $1000 doing it. But if I were starting from SCRATCH today, armed with nothing but my existing knowledge and experience, Thomas' Specification AND his $1000 ONLY, I still would be unlikely to achieve his aim.

Thank you, Roger, for putting it into perspective.

[retroman]

I can only agree to Roger.
building a machine for 1000 dollar, forget this.
the projector you need will be minimum 800 refurbished for regular8 and 500 for super-8 ( working machines ) , the camera starts at minimum 250 , the macro lens at 100 and so on....

so it is easier to buy a complete system which has been launched on the market

br

[MovieStuff]

Well, my goal wasn't to suggest that building a unit for $1000 was impossible but, rather, the limited budget doesn't leave any room for experimentation and failure. The more experience you have under your belt, the less you have to depend on trial and error.

[carllooper]

Granfer: I'm afraid this contributor seems to have lost the whole point of the thread (but then he often does). It's about COST not TIME! But then cost apparently is not of any consequence to him .....

Carl: "One really needs to know what one is aiming at. If it's to manufacture a competing product to existing integrated systems (such as the Retro8) that's a whole different ball game. But if the agenda is the creation of a one-off system of your own, which you can fine tune to your own particular needs, that's an entirely different story. Whether that can be done cheaper than a Retro8 or otherwise it is somewhat beside the point. You pay what you are willing to pay to achieve want you are wanting to achieve. "

The point being made here is that if you pay what you are willing to pay, to achieve what you what you are willing to achieve, then the cost relative to a Retro 8 becomes beside that point. Beside that point. Not anyone else's point. It is in this context (nobody else's) that I say it wouldn't matter what a Retro8 costs. All that would matter is how much money you have in the bank and how much of it you would want to spend or gamble.

The whole point of the topic is that COST is not "beside the point".

If that is the point of the topic, it's not my point.

He freely admits that his own machines have cost many times the Author's Budget,

Yes. That is constructive advice. One might think it's possible to build a machine for less than a 1000 and indeed, one might. But just as easily, if I'm any example, it can blow out. It all depends on what you consider is more important. If it were just cost then one can could just project the film on a wall and point a camera at it. There is always a trade-off going on, between cost and result. The alternative emphasis I'm putting on the debate is on the result, where the cost is a function of what result you are aiming at, and what knowledge you can acquire or already possess. Either way cost is not an independant variable or a fixed frame of reference. The cost can go up. But can just easily go down.

but offers no opinion as to whether someone with little or no previous experience can REALLY expect to achieve his aim within his very limited budget.

Its possible to do it for under 1000. But I have no experience in that. Thomas might very well be in a position to do that. Especially if you have the relevant knowledge - or can acquire it - which forums such as this can provide in more reflective moments. I know others who do have experience in building their own cheap setups. And indeed I'm working with someone at present on a new system that is expected to be built within just such constraint, ie. for $1000. Of course, it doesn't take into account the many years I've spent fucking around on more expensive setups. But I now own that fucking around. I can take credit for that. I can now make a particular kind of work that nobody else can, on my own custom setups. That's important to me, even if no-one else gives a shit. I myself don't give a shit.

To his credit, he does offer some salient facts regarding the obvious barren areas in the author's knowledge. And to those of us who are foolish enough to try and answer the actual question, he offers this ...

"To suggest (that) doing something yourself is somehow sad is, itself, very, very sad" - carllooper

Now I'm not sure whether that is supposed to be patronising, needling, taking the mickey or simply another example of the pseudo philosophical/arty outpourings we are used to ..... but rest assured, carlooper, that it affects me not. I am old enough and experienced enough to have met it many times before. Just, if you can, try being CONSTRUCTIVE in your answers to those seeking help.

Pseudo philosophy/art? So what would be real versions of such? Or is all philosophy/art to be regarded as "pseudo"? The quoted line takes exception to this line:

"Sadly, however, once we get an idea, we humans press on regardless ............" - Granfer

I don't find it sad that "humans press on regardless". What I find sad is the apparent need to rubbish those who do press on regardless. If Thomas pursues the task of building a film to video transfer system, against the advice he is given, that is his decision and it is not a sad one. It might be mistaken. But it's not sad. And indeed he might just as easily be not mistaken. He might very well put together a half decent system, for under a 1000. And that too will not be sad.

Before you accuse me of being DESTRUCTIVE in my advice rather than CONSTRUCTIVE, just try reading what I, and at least two others, actually said ... that we thought the answer to his question was NO.

There were no accusations of destructive criticism. The advice you've given is constructive. Very constructive. But characterising those who might ignore such advice, as "sad", is not advice. It is just conceit.

What else we wrote was aimed at indicating how he might achieve the aim of transferring the films WITHIN his budget.

That's good of you. Would you like a gold star?

Perhaps you would like to offer to do just that for him? Even Roger, brilliant engineer and innovator that he is, would likely say that there was no way that he could have built his first fully working Workprinter, from scratch, for the equivalent then of today's $1000 even with its understandably lower specification. I'm also sure that the development cost of his Retro 8 was many times that $1000, even with the benefit of his vast experience and existing facilities. (I have NO connection or financial interest in Moviestuff, nor do I own any of their machines: I have used them only as .an example of one of many people offering Commercially engineered Transfer Machines).

Fortunately it's a lot easier today then it is was when Roger started. Or indeed when I started. I'm quite keen on the idea of putting together systems cheaply. I don't want to spend more money than I have to. Now its easy to stay within budget if you really really want to. The only reason I went over budget is because my goal posts were (and remain) higher than any of my various budgets allow. If you lower the goal posts to what your budget can afford you'll be fine. The real trick is raising the goal post and reducing the cost. That's hard. But I'm sure it's achievable. I wouldn't want to suggest otherwise. But I'm not in any position to offer any real advice on that. And I don't. I've been quite open about that. That's why I mention my own experience, but at the same time putting a different spin on it - that a project doesn't have to be framed in terms of cost alone - even if everyone else does, or you yourself do. In reality everything I've done has been within budget anyway - for how can you spend money you don't have? I can't. Everything I've built has been built on my own money and what I could afford at the time and each has produced some sort of result, and each one better than the previous. The first one cost me no more than $800 and it gave me excellent results - better than anything I'd seen anywhere else at that time. But not as good as the goal post at which I myself was aiming. And so I reinvested more money. And more time. And more learning/thinking. Towards my own goal rather than any one else's goal. If that has ballooned to more than a 1000, or more than 2000, or more than 5000, that's because I didn't impose any limit other than what I could could afford each time. A natural limit rather than some arbitrary limit.

My advice to Thomas remains the same: No I do not imagine for one moment that you will achieve your aim for less than $1000. Don't let anyone who is not prepared to actually give you a straight YES or NO answer to your question lead you to commit yourself to indeterminate expenditure, if your limit is truly $1000.

Yes. If.

As for straight yes and no answers, they are the worst kind of answers. They are stupid. Not even you have given a straight yes/no answer. That is because you are not stupid.

[granfer]

:roll: :roll: :roll: :roll: No comment! The jury is out.

[carllooper]

Here's some figures.

Projector- Variable speed and/or Ultras slow motion of 8 FPS Dual 8 Projector that historically is gentle and won't chew on film. (Ebay has tons of these)
I used an Elmo K100 SM. I ended up buying 3 of these, ranging in price between 100 and 300, if my memory serves me correctly. But of course you only need one. They have a variable speed control, but I ended up using a stepper motor.
https://www.youtube.com/watch?v=mx8-vBzieKc

Motor - It may be necessary to use a controlled stepping motor in order to control frame timing (I'm not sure what to look for or if there is a good projector that would already be suitable)
Cost was about $15 for the first stepper I got, but it wasn't powerful enough. So I got the biggest one I could find for about $80, but that was overkill. So I spent $50 on a torque gauge to measure the torque and select a more appropriately sized motor (should I ever need the bigger one for another job). The usb controller for the stepper was $95. Power $8.

LED Lamp - Using combined RGB LED array to give a better white than the fake white LEDS at 100W-200W equivalence or about 1600 lumen - 3200 lumen
I didn't use an LED array on the Super8 tests I was doing. Just used a desk lamp and corrected in post. But for another project I purchased the components for an LED array. An 8 x 8 array emitting a total of about 64+ cd over a 90 degree angle (from which you can calculate lumens). Cost was $50 for all the LEDS.

Controller - USB controller to control LED Gain levels and timings (can't find anything like this except cine2digits which make an obscure reference to it)
The usb controller for the 8x8 LEDs was $125. Here's one:
http://www.phidgets.com/products.php?ca ... _id=1032_0
Check out all the other quite convenient components they make available, all usb controlled. The coolest thing for me is just the sdks thay have. In C++ and C# amongst others.

Power Regulator - Possibly additional power would be needed if the 2.5W from the USB port isn't enough to supply the controlled LED lights
Power for LED usb control was $8.

Machine Vision Camera - 720P or higher (2MP preferred), ~25+ FPS, Global Shutter, controllable frame rates for timing purposes. (I think CCD is better but can a CMOS with global shutter work too or is it too grainy?)
Varies in price depending on what you are after. A pointgrey cmos 4K cam goes for 800. My first tests were done using a 4K Canon DSLR (they have a C++ SDK) that I picked up for $500. I picked up a better one for $600. Eventually going to use the pointgrey cmos 4k - much faster than the Canon in terms of data transfer rates. And more easily programmable (have programmed them on otherwise commercial gigs and they're sdk is very pleasant). Plus they have a much smaller form factor so far more easily mountable. As previously mentioned a rolling shutter doesn't matter if the film is held stationary during capture. I'm currently exploring continuous moving capture using a rolling shutter camera on such - not because it's a good idea but simply because it's an interesting technical challenge! A friend of mine plays chess and does crossword puzzles for exactly the same reason.

Macro lens - Macro lens with c or cs mount than can fit inside of the original projector lens assembly
I found one at a garage sale for $20. But this can be otherwise quite expensive. Especially if you want good results. I was just lucky that a photographer down the road was retiring. We had a good chat about photography and how much it has changed over time. Found some bellows for $15. I eventually acquired a 3D printer for $1000 (!) and made a custom mount but prior to that just used cardboard and sticky tape. Worked fine.

Mount - Adjustable camera mount to hold the machine vision camera in place with Projector
Originally used G clamps but in subsequent versions used a manfrotto 3D head for $250. Also involved various metal plates and bolts etc (about $50). Have spent a lot more on mounts in subsequent work (xyz positioning plates etc). Mounts are important if you want precision results. Especially if going for high definition. In my original humungus over the top ridiculously ginormous definition tests (20K captures), just sneezing would induce vibrations in my setup, which blurred the high frequencies (miniscule details) I was otherwise interested in, despite heavy duty clamping and generous sandbagging. However that was due to the camera and projector being mated by the platform to which they were bolted. Better results became obtainable when I mounted the camera directly to the projector chasis. But the mount for that was even more difficult to make. Perhaps that's just me. I'm not the most skilled when it comes to bending the physical universe into place. My main area of expertise is software development. The rest is just an ongoing learning experience.


Carl

[Thomaseo1]

Thanks for your responses everyone all the feedback has been very useful for looking into this project. Just so everyone is clear I have yet to commit to buying anything as of yet and am so far at the stage of serious researching and exploring the possibilities of building my own Telecine machine. Once I feel that I have enough usable knowledge and feedback on areas I'm either guessing on or just unsure of I'll pull the trigger at that point that is unless other solution become apparently more viable.

For reference this is what I have worked with [youtube]http://youtu.be/P3Bq6_8ELuM[/youtube] end result [youtube]http://youtu.be/epNOfRCgWJI[/youtube] after working on this for about a few weeks I decided that the quality looked only slightly better than if it was recoded off a wall while manually holding the camera smeared with Vaseline(its that bad). The Actual film looks much better than this leading me to believe that a DIY cheap telecine will be by far much less time consuming, less than 2 week per film and give a bit better results than this junk ; even if my proposed telecine machine looks like slightly better crap than this I will be very happy and spending a $1000 for better crap and having something is better than having nothing.

Currently I have bee looking at listing approximate parts and trying to determine if these prices are suitable or if there are cheaper/better solutions

Projector - $100 (Aproxiamte)http://www.ebay.com/itm/Cine-film-proje ... 540023ec0a
Programmable USB LED Controller - $25 Fade candy http://www.adafruit.com/products/1689
Led Lighting - $120 10Xhttps://www.adafruit.com/products/518Bright White(Blue Led) 2X https://www.adafruit.com/products/322 (Overkill of =~1700 Lumens)
Power Retirements for lighting - $25https://www.adafruit.com/products/658
Camera Vision Camera- $315 http://www.ptgreystore.com/blackfly-12- ... 134%20$315
5m Super 8 film test RP-32 test panel - $67 http://www.wittner-kinotechnik.de/katal ... d_s8mm.php (expensive and not extremely necessary but is very useful)
Software $0-$100 (Guess) Software is either made from camera lighting and sensor SDKs or purchased from other sources
Camera Mount - $33 h[url]ttp://www.ebay.com/itm/4-way-Macro-Focusing-R ... 25906d2da0[/url]

[RCBasher]

Thomas has emailed me to ask about my software and invited me to reply directly or post here, will do both as I get these kind of questions on regular basis so perhaps easier to reach a wider audience from here.

I wanted to ask you about the software you developed. First I'd like to know how much it would cost for a license from you and if you think it's now possible to port it to other standards like GigE and/or if it's easily possible to accomplish the same functionality using SDK's provide by Machine Vision Manufacturers like Point Grey?

To answer this I must clarify that I have two software designs, both are for Firewire Machine Vision cameras and run under Windows.

The first is what I call a “freebie” in that I only pass on the cost of the ActiveDcam runtime licence. This has to be bought through me because it is tied to my ActiveDcam developer licence. I charge a rounded figure of £50 but as exchange rates jump up and down and there are often PayPal and currency transaction fees I sometimes win slightly, sometime lose slightly. What it will never do is help pay back for the developer licence, camera, time, etc. but then I’m not expecting it to. Dcam2Digits is my contribution to the DIY community for all the help and guidance I have received over the years. It is a stand-alone capture utility for Machine Vision Firewire cameras.

The second is my full Cine2Digits package which is an RGB lighting system and Capture system combined. Not appropriate to mention prices here (probably against forum rules or something. The ActiveDcam runtime price mentioned above is effectively public knowledge as listed on A&B Software website), but will say that I add a small mark-up over the component costs to buy my beer occasionally. The software requires the same ActiveDcam runtime and I do add a small fee for my application to help get back some of my development costs. Again, I mostly offer it as a service to the DIY community and any proceeds so far will certainly not put much food on the table!


The second part of the question is in regard to porting to other interface standards. I get asked this one a lot and when I say “no” some people get rather indignant about it, but let me explain why not:
There are many interface standards for Machine Vision cameras, common ones being Camera Link, Firewire, GigE Vision and now USB 3 Vision. Think about what I would need to do to develop for each platform – buy a developer licence for ActiveXXX, buy at least one camera for each platform, buy expensive interface cards in the case of Camera Link, etc. I’m not in business, I’m just an amateur and there isn’t enough call for what I do to ever payback the required investment. Then there is my time….best not go there, or mention MACs!

I decided to support Firewire because it was (probably still is) the most prevalent and suitable interface for our application. The promises of GigE do little for us and most do not realise that the only way to get close to Gbit speed is to have a dedicated GigE port with special driver and a single camera connected. Even then, we are talking 1000Mb/s versus 800Mb/s for Firewire 1394b, so not a big increase.

Now there is USB 3 Vision which offers a big jump in bandwidth, allowing for a serious increase in frame rates and/or sensor resolutions and bit depths. My “no” answer is soon to be a “yes for USB 3" answer as I have made a big investment in USB 3 and in the middle of porting my full Cine2Digits application. I will not be porting my DCam2Digits “freebie”, sorry.

Final part of the question is in regard to camera manufacturers SDKs. When I first started out I tried to use AVT’s SDK but at the time it was flaky to say the least and in truth I probably didn’t know enough about machine vision cameras to be able to work around the issues. So I went on the search for an alternative and decided on ActiveDcam. It was an unexpected extra expense (Granfer’s words on creeping expense above should be heeded) but a good solution. Big point is it is camera manufacture agnostic, who wants to redevelop their application for each SDK out there if one decides to switch brand? Second point is it has many features which are not easily available with most SDKs. For example, as far as I can tell, the PGR SDK does not have a deBayer function for 16bit depth, just 8 bit.

Regarding the “easily possible” bit of the question, all things are relative. As has been mentioned by other posters above, it depends on the skills and knowledge of the person trying to do it. As a rule, if someone needs to ask such a question then it is a fair bet that it is not easy for them, but may or may not be for someone else.

Hey, I’m getting towards a Carl length post! 8O While I'm here, my simple answer to the OP question is no, but I’d double the length of this post if tried to explain why. Most has been said already anyway. There is a “perhaps” of course, that being you get real lucky on ebay and pick up all you need for a song, you are able to design and build circuits, write software and understand the subject enough to know what to write and are mechanically adept!

Frank

[RCBasher]

My two-penny worth on the original post –

The post topic, qualified by the opening sentence, asks if Thomas can build a dual 8 HD digitiser for under $1000, I guess $999.99 is ok Bottom line (IMHO) is no, but double it to $2000 and then you will be in serious danger of achieving it!

Unfortunately the specification has been mixed up with a possible solution, leading to questions about HDR, motor replacements, etc. As with buying a house where the three most important things are location, location and location, it is my opinion that for this kind of film transfer the most important three things are the camera, camera and camera! Other design requirements will flow from this, for example:

1) Do you need HDR? With the right camera and illumination, most likely not. Don’t forget that HDR has its own problems and requirements, like pixel perfect frame alignment – no projector and camera assembly can be that rigid. If you doubt whether a single exposure can cut the mustard, take a look at this crop from a b&w 16mm frame, taken with the IMX174 colour sensor, RGB lighting:
https://www.dropbox.com/s/ov7gkz8sxkpgs ... g1crop.tif
Open it up in Photoshop or the like and apply some heavy gamma and watch the detail magically appear out of the blacks. If you are too lazy, here's one where I applied gamma of 4 (or 0.25 depending how you do your sums) during capture:
https://www.dropbox.com/s/4g10jo97t5ydn ... g4crop.tif


2) How fast can the camera and supporting system go? With the right camera and illumination, there would be no need to change the projector motor or even remove/modify the shutter blade.

In choosing a camera, sensors to consider are the ICX285, ICX674 and the new IMX174. Just starting to test the later and looks very good for the money. 16bit transfer (or packed 12bit) is essential and then an accurate gamma function in the capture software before optional conversion to 8bit if going straight to avi. If you can get 16bit transfer at or higher than your projector runs, then projector modifications are minimal.

To get sufficient lighting power and control the easiest solution is to flash the LEDs so the average power is miniscule. Some LEDS are allowed to be over-run for low duty cycle pulses. If you use RGB lighting then it is easier to dynamically control the length of the pulses than it is to control the current through the LEDs. This way we solve four problems in one: extra lighting power, lower supply power, white balance and exposure control. Such a solution rules out sensors without a global shutter though (Carl is not always 100% correct!).

My budget guide for a pretty good system:
Camera Point Grey GS3-U3-23S6C-C: $1295 + any taxes, shipping and USB 3 cable (but need to wait for a while for firmware to be sorted)
Lens: Schneider – Kreuznach Componon 2.8/50mm enlarger lens: expect to pay around $70 to $90 on ebay for a good one.
Extension tubes and adaptors: $30 to $35
XYZ camera mount: Pot luck here and depends on your own ability to improvise, but there’s some on eBay right now for $250. I built mine from 3 stages and it cost about half that.
Projector: $100 to £200 for a serviceable machine
Lighting and software: say $200 to $500 depending on what you can do yourself.
Skills and knowledge: priceless if appropriate 8)

Of course you will need a good PC in addition to this, but this was left out of the original target.

Frank

[carllooper]

To get sufficient lighting power and control the easiest solution is to flash the LEDs so the average power is miniscule. Some LEDS are allowed to be over-run for low duty cycle pulses. If you use RGB lighting then it is easier to dynamically control the length of the pulses than it is to control the current through the LEDs. This way we solve four problems in one: extra lighting power, lower supply power, white balance and exposure control. Such a solution rules out sensors without a global shutter though

Good point. If you are using an LED system for colour balance and exposure control (using the recommended PWM technique), then you can get variations in colour/exposure from one scanline to the next (on a rolling shutter camera). However this problem doesn't rule out using a rolling shutter camera (or rule out using LED lighting as it might be equally argued). It just requires a different sampling technique to that which might otherwise be used. And adjusting the PWM strategy used on the LEDs. On the camera side the general idea is to capture a number of frames and integrate the result (merge the frames together). A global shutter camera effectively does the same thing (a buffer accumulates photon detections over time, which averages the result). With a rolling shutter camera you just have to do this same thing yourself. So for any given sensor pixel, during a single exposure, a given LED light might be on or off. And that can be a problem. But only at that particular moment. Over time a PWMed LED will exhibit a constant ratio of ons to offs. Indeed PWM as a means of controlling brightness is based entirely on this very thing - that the brightness of light is not specifiable at any given instant in time, but becomes specifiable (and controllable) over time (approaches a steady value). By analogy, the toss of a coin at any given moment is not specifiable, but over time the ratio of heads to tails quickly approaches a steady or reliable value (50:50). Casinos make huge profits on the reliability of such things. In the same fashion you can program the camera software to sample the film (and lighting) in the same way: over time. The more frames you sample the more the integral result approaches the average value (the brightness). In this way one is effectively turning the rolling shutter camera into a global shutter camera. However some testing of various PWM strategies will still be necessary to avoid systematic interference effects. One doesn't want a situation where some sensor pixels are always registering heads (or always registering tails) or some systematic in-between result. Plugging some controlled random noise into the pulse widths, will go a very long way to remedying such. Some experimentation and theorisation should find a PWM strategy that is optimal for a particular camera.

But it does require more work on the design/software side. As previously mentioned (in relation to jello motion blur), the use of a rolling shutter is not so much a good idea, as a technical challenge. And that's not everyone's cup of tea, but it is my cup of tea.

C

ps. but it still remains true that I'm not always 100% correct. I'm simultaneously disturbed and flattered that anyone might even think that is true. I'm certainly the last one to think such a thing. But my ego certainly prefers that it be correct. It experiences acute pain whenever it discovers it is not as clever as it thought! The correction is always the answer.

[RCBasher]

Whoa there Carl, who said anything about PWM drive? Sorry to cause pain to your ego, but wrong assumption leading to wrong analysis and conclusion.
Exposures in my design are just a few tens of microseconds, usually in the range 10 to 100us and designed to be this short to allow use with a continuous transport system. The LEDs are driven from a DC constant current source, so they are full on for the length of the exposure flash.

The staggered start and end of exposure for each sequential line in a rolling shutter CMOS sensor means that the only way to expose all lines equally is to set the camera exposure to greater than the combined stagger/readout time (usually 1 / maximum sensor frame rate) plus the flash time of the LEDs. The sequence would be to then trigger the camera, wait until the last line begins collecting light (a frame period), fire the R, G & B flashes, terminate the camera exposure, wait for another frame period for sensor readout before starting again. Really not a practical solution.

There are some CMOS sensors which have an optional Global Reset mode. They still have a staggered end of exposure to each line but they start collecting light on all lines simultaneously. It could be possible to use one of these but there are not many around and their other performance characteristics are not good enough for our requirements. The CMOSIS CV2000/4000 sensors are a case in point, big claims by the sensor manufacture for dynamic range but it never materialised. In my tests, it was completely blown away by an ICX285 sensor and stands as a very good example why not to select a sensor based just on a higher pixel count as the image quality may be rubbish.

It now looks like the new IMX174 with true global shutter, adequate resolution and great dynamic range is finally a CMOS sensor to get excited about as it reduces the camera cost to about 65% of that for an ICX674 based camera. BTW, for anyone interested in these things, the ICX674 is a great sensor, slightly pipping the ICX285 for dynamic range.

Frank

[carllooper]

Whoa there Carl, who said anything about PWM drive? Sorry to cause pain to your ego, but wrong assumption leading to wrong analysis and conclusion.
Exposures in my design are just a few tens of microseconds, usually in the range 10 to 100us and designed to be this short to allow use with a continuous transport system. The LEDs are driven from a DC constant current source, so they are full on for the length of the exposure flash.

The staggered start and end of exposure for each sequential line in a rolling shutter CMOS sensor means that the only way to expose all lines equally is to set the camera exposure to greater than the combined stagger/readout time (usually 1 / maximum sensor frame rate) plus the flash time of the LEDs. The sequence would be to then trigger the camera, wait until the last line begins collecting light (a frame period), fire the R, G & B flashes, terminate the camera exposure, wait for another frame period for sensor readout before starting again. Really not a practical solution.

There are some CMOS sensors which have an optional Global Reset mode. They still have a staggered end of exposure to each line but they start collecting light on all lines simultaneously. It could be possible to use one of these but there are not many around and their other performance characteristics are not good enough for our requirements. The CMOSIS CV2000/4000 sensors are a case in point, big claims by the sensor manufacture for dynamic range but it never materialised. In my tests, it was completely blown away by an ICX285 sensor and stands as a very good example why not to select a sensor based just on a higher pixel count as the image quality may be rubbish.

It now looks like the new IMX174 with true global shutter, adequate resolution and great dynamic range is finally a CMOS sensor to get excited about as it reduces the camera cost to about 65% of that for an ICX674 based camera. BTW, for anyone interested in these things, the ICX674 is a great sensor, slightly pipping the ICX285 for dynamic range.

Frank

Hi Frank,

I assume from the way you describe your LED system that it must be using PWM to control brightness. If it's not using current, nor PWM, what else would it be using? And I assume from your reference to me being "not always being 100% correct" that you were making a side swipe at my proposal (ie. the use of a rolling shutter camera).

So in that context I'm simply pointing out that in my proposal (rather than in your proposal) mine does not require ruling out use of a rolling shutter camera. In other words the side swipe isn't really justified (although it might be for other reasons, such as the fact that I'm not always correct ). And certainly my proposal isn't by any means a "practical solution" however I've never claimed otherwise. Indeed, on the contrary, I've been at pains to say it is a "technical challenge".

But that is what I'm into. As are others I know. In other words it doesn't actually matter to me whether the result is practical or not. But for others, of course, it probably does matter, so that's why I spend a good deal of time clarifying that - just so that there is no misunderstanding, ie. that I'm not challenging how others (such as yourself) might theorise it, or otherwise do it. I'm just describing how I'm doing it. And if there is any fault in it I'm more than happy to find about it.

But if I'm mistaken in any of this, then I'll let my ego know it's wrong, so it can do some suffering.

Carl

[RCBasher]

I assume from the way you describe your LED system that it must be using PWM to control brightness. If it's not using current, nor PWM, what else would it be using?

There you go making assumptions again Carl! I just described how I drive the LEDs, through a constant current DC source, switched on just for the length of the flash. I also explained the flash periods of the red, green and blue LEDs are what determine the exposures on the sensor. I do not change the current through the LEDs to "change brightness". The LED brightness is fixed.

Think about how a flashgun controls its power - usually by terminating the flash when enough light has been produced. Now think this concept x3 for RGB and there you have it.

As I mentioned above, it is much easier to dynamically control the period of the flash than the brightness of the LEDs. I use precision hardware pulse generators in a microcontroller that have jitter in the small fractions of microseconds, so the pulse flashes are very accurate and repeatable.

I'll leave you to massage your ego....

Frank

[carllooper]

I assume from the way you describe your LED system that it must be using PWM to control brightness. If it's not using current, nor PWM, what else would it be using?

There you go making assumptions again Carl! I just described how I drive the LEDs, through a constant current DC source, switched on just for the length of the flash. I also explained the flash periods of the red, green and blue LEDs are what determine the exposures on the sensor. I do not change the current through the LEDs to "change brightness". The LED brightness is fixed.

Think about how a flashgun controls its power - usually by terminating the flash when enough light has been produced. Now think this concept x3 for RGB and there you have it.

As I mentioned above, it is much easier to dynamically control the period of the flash than the brightness of the LEDs. I use precision hardware pulse generators in a microcontroller that have jitter in the small fractions of microseconds, so the pulse flashes are very accurate and repeatable.

I'll leave you to massage your ego....

Frank

Yes - I make assumptions. Assumptions are not the same thing as mistakes. Only when the assumptions are incorrect are they mistakes. Since you pointed out my assumptions were incorrect I was seeking correction on such. I wasn't "making assumptions again". I was literally asking you how you were controlling it.

So thanks for the clarification. I think.

In my setups the film is held stationary so there's no requirement for short sharp bursts of light. Indeed it is precisely to avoid such a requirement that I've been using a stationary frame. However in my setup, using PWM LED with a rolling shutter camera, there would be the problem of which I spoke. And if what I've elaborated is unrelated to your problems with rolling shutter cameras, it's certainly not unrelated to my problems with such.

So not really quite sure what your complaint or criticism with my proposals might be ... I certainly don't have any with yours.

Carl

[RCBasher]

I wasn't "making assumptions again". I was literally asking you how you were controlling it.

So perhaps if you had written the assumption statement in the past context rather than the present (eg "I had assumed from....") then I would not have assumed incorrectly that you were still wrongly assuming!

Anyway, fun as this banter may be, my original posts were about thinking the requirements through thoroughly then picking the right design and components to meet those requirements.

Frank