Why do high speed cams cost so much?

[Ben321]

Considering the simplest camera is a monochrome one, and also because it doesn't output 3 color channels (only 1 byte per pixel), it seems monochrome would be the best, so lets see what the bit rate would be. Starting at 1920x1080 at 240fps, that gives us a byte rate of 497,664,000 bytes per second. If the resolution is reduced to 640x480 that gives a frame rate of 1620fps. Further reducing the resolution to 320x240 that gives us a frame rate of 6480fps. And even further reducing the resolution to 160x120 gives us a frame rate of 25920fps. That is very impressive. But what is the bit rate? The byte rate is 497,664,000 bytes per second, and a byte is 8 bits. So the bit rate is 3,981,312,000 bits per second, which is 3.98GBits/s. That's a truly high bit rate, but it's nothing that a cheap FPGA and a few gigs of RAM can't handle. All you need is a simple array of RAM chips (usually these will be on a small memory board like those in a PC). An 8GByte RAM board is pretty cheap. It costs less than $100, and would give you about 17 seconds of recording time at the above mentioned byte rate. FPGAs have a very wide range of prices. Some cost thousands of dollars, while I've seen others for under $10.

It should EASILY be possible to manufacture a simple high speed camera, with the above mentioned specs, for between $200 and $300, and then make a decent profit by selling it at a price between $500 and $1000. Yet no company has done so. They all go overboard with specs (producing devices that cost well over $10,000, and some even over $100,000), using application specific (and quite expensive) in-house manufactured components (rather than cheaper off-the-shelf components) to get insane specs like full RGB color video at 640x480 at 4000fps. Yes, there is probably some scientific demand from major universities, the military, and other similar research facilities for such equipment, but it completely ignores another sector of the economy, the average consumer who happens to have a hobby/interest in science. In my opinion, major manufacturers of high speed cameras need to branch out, and in addition to their continued operations that support the professional scientific community, they need to create a separate product line that targets the average consumer. Such a cheaper product line for the average consumer, could be accomplished by making a high speed camera with specs similar to those that that I have mentioned at the top of this thread.

Thermal imaging camera manufacturers like FLIR have already branched out, and made cheaper consumer-level options available. Some companies like Seek have even started up with the sole purpose of creating cheaper thermal imagers that the average consumer can afford. Yet there are no such companies popping up making cheaper high speed cameras, nor are any of the major high speed camera manufacturers creating separate consumer product lines. There's a demand out there for such equipment, but no company seems to be trying to fulfill that demand. In my opinion, they could make a lot of money by targeting consumers with cheaper versions of the products they make.

[thm_w]

http://www.krontech.ca/store.html#!/Chronos-1-4-high-speed-camera/p/92268927/category=21981408

https://www.kickstarter.com/projects/1623255426/fps1000-the-low-cost-high-frame-rate-camera/description

[Ben321]

http://www.krontech.ca/store.html#!/Chronos-1-4-high-speed-camera/p/92268927/category=21981408

https://www.kickstarter.com/projects/1623255426/fps1000-the-low-cost-high-frame-rate-camera/description

Neither of these target consumer level buyers though. They clearly target scientists on a tight budget (which is still more than the average consumer's budget). Both of these sell for a couple thousand dollars. To me, a consumer-level camera is one that costs between $500 and $1000. As I explained in my above post, I believe it possible to create a high speed camera cheaply enough that it can actually make a profit when sold for between $500 and $1000.

[Bassman59]

http://www.krontech.ca/store.html#!/Chronos-1-4-high-speed-camera/p/92268927/category=21981408

https://www.kickstarter.com/projects/1623255426/fps1000-the-low-cost-high-frame-rate-camera/description

Neither of these target consumer level buyers though. They clearly target scientists on a tight budget (which is still more than the average consumer's budget). Both of these sell for a couple thousand dollars. To me, a consumer-level camera is one that costs between $500 and $1000. As I explained in my above post, I believe it possible to create a high speed camera cheaply enough that it can actually make a profit when sold for between $500 and $1000.

Sensors that can read out that fast are not cheap. Your budget is blown by the sensor alone.

[ataradov]

I believe it possible to create a high speed camera cheaply enough that it can actually make a profit when sold for between $500 and $1000.

Well, you've got yourself an idea for a startup. Start by finding a sensor in that price range.

[joeqsmith]

I have a Sony RX10 that I bought just for this reason.   It's a decent all around camera but I am not a camera buff.   I wrote the guy about the Chronos prototypes when I first saw it to see if he would sell me one.  In retrospect, I don't think it would have been good enough for what I want to use it for.  I think I need something much faster. 

Video showing a frequency counter being clocked at 1KHz compare the autorange speed of the 121GW prototype and the Gossen Ultra.   
https://youtu.be/13nv-NsQXDs?t=117

[thm_w]

Neither of these target consumer level buyers though. They clearly target scientists on a tight budget (which is still more than the average consumer's budget). Both of these sell for a couple thousand dollars. To me, a consumer-level camera is one that costs between $500 and $1000. As I explained in my above post, I believe it possible to create a high speed camera cheaply enough that it can actually make a profit when sold for between $500 and $1000.

Phones are capable of 720p at 240fps (S8 $500), or gopro 120fps. No average consumer will care that much about higher FPS as you need crazy lighting setups and special triggering to catch shorter events (if buffer memory is not sufficient).

960fps for $800: https://www.amazon.com/Sony-Cyber-shot-DSC-RX100-Digital-Camera/dp/B00ZDWGM34
or https://www.amazon.com/gp/product/B01DLLJ8CU/

[Ben321]

Neither of these target consumer level buyers though. They clearly target scientists on a tight budget (which is still more than the average consumer's budget). Both of these sell for a couple thousand dollars. To me, a consumer-level camera is one that costs between $500 and $1000. As I explained in my above post, I believe it possible to create a high speed camera cheaply enough that it can actually make a profit when sold for between $500 and $1000.

Phones are capable of 720p at 240fps (S8 $500), or gopro 120fps. No average consumer will care that much about higher FPS as you need crazy lighting setups and special triggering to catch shorter events (if buffer memory is not sufficient).

960fps for $800: https://www.amazon.com/Sony-Cyber-shot-DSC-RX100-Digital-Camera/dp/B00ZDWGM34
or https://www.amazon.com/gp/product/B01DLLJ8CU/

Lighting is fairly simple. If you have a proper LED light source that has the DC output of its power supply filtered properly, you will get a steady light output. If you get a cheap 10 watt LED lightbulb that's designed to replace a 60 watt standard bulb, it may have a lot of flicker, but one can easily modify the power supply by cutting open the base of the bulb. Then desolder the filter capacitor on the powersupply circuit board (make sure to discharge it first), and solder on another capacitor with a much higher capacitance. That should reduce or even eliminate the flicker. To reassemble the plastic base of the bulb, just hold it together with duct-tape.

No phone actually does 720p at 240fps (not sure if GoPro does). They all cut corners. They actually use a lower resolution, like 360p or even 240p at a frame rate of 240fps, and then post-process it. They take the tiny video that is either 360p or 240p, and upscale it to 720p. You can tell this when you pause the video and look at it, because the finest detail that you can see is actually like 3 pixels wide, not 1 pixel wide. So the video's true resolution (the size of the image for which the smallest detectable detail is represented by a single pixel) is nowhere even CLOSE to 720p.

[1anX]

Nikon 1 series offer high frame rates. I have the Nikon 1 V3 and it can shoot 720 and 1200 fps in smaller formats.
You can buy these cameras for under $1000 and second hand even cheaper again.

[AG6QR]

Lighting is fairly simple. If you have a proper LED light source that has the DC output of its power supply filtered properly, you will get a steady light output. If you get a cheap 10 watt LED lightbulb that's designed to replace a 60 watt standard bulb, it may have a lot of flicker, but one can easily modify the power supply by cutting open the base of the bulb. Then desolder the filter capacitor on the powersupply circuit board (make sure to discharge it first), and solder on another capacitor with a much higher capacitance. That should reduce or even eliminate the flicker. To reassemble the plastic base of the bulb, just hold it together with duct-tape.

The problem isn't just flicker, it's intensity.

At high frame rates, you need a high "shutter speed".  For 1000fps, each frame gets to use the sensor for a maximum of 1/1000 of a second.  That time period has to include not just the time to capture light, but also the time to set up the sensor and get the output from it.

As the shutter speed gets faster, you need a greater intensity of light in order to give the sensor enough light to see properly.  You can somewhat help the situation by making bigger pixels, but that means either lower resolution or a bigger sensor overall.  A bigger sensor increases the cost of the sensor, and also may increase the cost of the lens.

[Mechatrommer]

Lighting is fairly simple. If you have a proper LED light source that has the DC output of its power supply filtered properly, you will get a steady light output. If you get a cheap 10 watt LED lightbulb that's designed to replace a 60 watt standard bulb, it may have a lot of flicker, but one can easily modify the power supply by cutting open the base of the bulb. Then desolder the filter capacitor on the powersupply circuit board (make sure to discharge it first), and solder on another capacitor with a much higher capacitance. That should reduce or even eliminate the flicker. To reassemble the plastic base of the bulb, just hold it together with duct-tape.

The problem isn't just flicker, it's intensity.

use higher intensity light, i can make one cheap that can overblown any sensor.

[bugi]

Lighting is fairly simple. If you have a proper LED light source that has the DC output of its power supply filtered properly, you will get a steady light output. If you get a cheap 10 watt LED lightbulb that's designed to replace a 60 watt standard bulb, it may have a lot of flicker, but one can easily modify the power supply by cutting open the base of the bulb. Then desolder the filter capacitor on the powersupply circuit board (make sure to discharge it first), and solder on another capacitor with a much higher capacitance. That should reduce or even eliminate the flicker. To reassemble the plastic base of the bulb, just hold it together with duct-tape.

I am certainly not experienced on those circuits, but the little I have seen, the circuits are pretty simple... In the negative interpretation of simplicity, meaning that they are not designed and/or made to be very flexible and/or robust against changes. And many a power supply circuit I have read about (in general, not about LED light bulbs) specifically have also limits on how large a cap they can use on the output and still be stable. Also, depending on the circuit, one needs to consider inrush current.  Many a modern controller chip starts to have convenient features to handle both issues (large caps and inrush current limiting), but without investigating the bulb circuit first, it is a gamble.

An over-simplified "replace a cap and use duct-tape" approach could easily lead to blown fuse, or worse. (At least when executed by a non-professional who might think everything written on this forum is a fact from an expert.)

I would probably go simpler way, by using LED strips arranged on a panel (or buying a nice ready made one) and build the supply myself with known components (and plenty of reference/example circuits in the web to choose/adapt from). Might also implement a combined "flash" effect support to it, for short period extra brightness boosts. Of course, depends on how well the case can cool, how high light output is needed, etc.

[EEVblog]

https://theamphour.com/the-amp-hour-325-an-interview-with-david-kronstein-tesla500/

[Fungus]

Lighting is fairly simple. If you have a proper LED light source that has the DC output of its power supply filtered properly, you will get a steady light output.

I am certainly not experienced on those circuits, but the little I have seen, the circuits are pretty simple... In the negative interpretation of simplicity, meaning that they are not designed and/or made to be very flexible and/or robust against changes. And many a power supply circuit I have read about (in general, not about LED light bulbs) specifically have also limits on how large a cap they can use on the output and still be stable.

Yep, and this is doubleplustrue for LED drivers.

LEDs are current-driven, not voltage driven. Capacitors across the output can easily fool the driver circuit, use an inductor in series instead (DIY: wrap some wire around a ferrite ring).

I would probably go simpler way, by using LED strips arranged on a panel (or buying a nice ready made one)

You can get stupidly bright $12 LED panels on eBay these days:

eg. https://www.ebay.com/sch/i.html?_nkw=70w+led+panel

Big Clive has looked at one:


I might get some of those for lighting my bench. The large surface area should give really soft shadows - much better for photography than the six LEDs I have at the moment.

[Berni]

There are lots:
1) High speed image sensors are expensive
2) They need large lenses to cover the sensor and collect enough light.
3) Lots of light is needed to film at this speed so you need 100s of W worth of LEDs to light even just a tiny scene.
4) Only scientific applications need frame rates this high. For cool sports slow motion shots frame rates in the low to mid 100s are plenty fast enough
5) No high volume market to justify making production runs in the 1000s that is needed to make it price competitive to consumer cameras.
6) Development cost has to be recovered in these small markets

Nobody is stopping you from making your own high speed camera for that price point, but expect some challenges along the way.

[bugi]

From the article linked above:
"David also desigend a 5 kW LED rig but won’t be selling it. He may open source the waterblock design or sell just that piece."

Emphasis mine. But... 5 kW o.O  That would be some serious light output. Better not look at it directly.

[coppice]

• Find a high speed camera application that millions will pay a few hundred dollars for
• Use social media to stir up interest in that application
• Wait

With a demonstrably sizeable market it shouldn't take too long for something interesting to appear.

[Berni]

Tesl500 has a video on YouTube about that monster LED light setup. While 5kW might sound dangerous it's actually spread acros multiple water cooled panels and each panel has multiple COB LED modules. Yet all this power is only enough to bring a part of his garage floor to the brightness of noon sunlight.

There is a reason the guys at the SlowMoChannel always use there fancy Phantom camera outside in there back yard on a nice sunny non overcast day. Direct sunlight is the easiest way of getting enough light for a high speed camera.

[james_s]

I was also going to suggest watching those Tesla500 videos on high speed cameras. The amount of light required for really high frame rates is huge, a very long way from hacking consumer LED light bulbs. In traditional applications of high speed cameras like automotive crash testing and such it is common to use hundreds of kilowatts of tungsten halogen floods.

The camera in my phone can do 240fps and it's a relatively low end phone. I'm sure we will see incremental increases there however I doubt most consumers care about more than a few hundred fps. This is the classic situation where very limited market means relatively high prices. Typical consumer electronics items are dirt cheap because they build literally millions of an item. If you're building a tiny volume like a few hundred or even several thousand units the cost rises dramatically.

[filssavi]

A lot of people here rightly pointed out that the sensor and optics are very expensive, but that is only the tip of the iceberg in terms of material costs, after the signal comes out of the sensor you will have a big ass (TM) FPGA to show all that data into a big ass frame buffer, which for example in the phantom 4K 1000fps camera they have 64/128 GB of what I presume is DDR4 that alone blows the budget, the fpga is also multi thousand dollars, then you have to have some fast SSD (lets say at least 2/300 gigs PCI express) as sd cards to avoid spending half an hour saving that to sd card also not cheap

Then you have to do all the firmware/has to have everything work
Also don’t forget that you will have to develop software in order to at least transcode the raw video to a editable format

And I could go on and on with this list

[thm_w]

Lighting is fairly simple. If you have a proper LED light source that has the DC output of its power supply filtered properly, you will get a steady light output. If you get a cheap 10 watt LED lightbulb that's designed to replace a 60 watt standard bulb, it may have a lot of flicker, but one can easily modify the power supply by cutting open the base of the bulb. Then desolder the filter capacitor on the powersupply circuit board (make sure to discharge it first), and solder on another capacitor with a much higher capacitance. That should reduce or even eliminate the flicker. To reassemble the plastic base of the bulb, just hold it together with duct-tape.

Its not simple in a consumer device unless its built in.

No phone actually does 720p at 240fps (not sure if GoPro does). They all cut corners. They actually use a lower resolution, like 360p or even 240p at a frame rate of 240fps, and then post-process it. They take the tiny video that is either 360p or 240p, and upscale it to 720p. You can tell this when you pause the video and look at it, because the finest detail that you can see is actually like 3 pixels wide, not 1 pixel wide. So the video's true resolution (the size of the image for which the smallest detectable detail is represented by a single pixel) is nowhere even CLOSE to 720p.

Gopro claims 1080p at 240fps: https://gopro.com/help/articles/block/available-video-resolutions-for-hero6-black
Samsung S8 claims 720p at 240fps: https://www.samsung.com/global/galaxy/galaxy-s8/specs/
S9 is claimed 1080p 240fps, or 720p 960fps: https://www.gsmarena.com/samsung_galaxy_s9+-8967.php but is limited to a capture time of 0.2s

If it were some second tier or no name brand I wouldn't trust the spec, but in these cases I would believe it.