Using LEDs To Determine A Video Camera’s True Framerate

Interpolation and digital cropping are two techniques which are commonly used by marketing folk to embellish the true specifications of a device. Using digital cropping a fictitious zoom level can be listed among the bullet points, and with frame interpolation the number of frames per second (FPS) recorded by the sensor is artificially padded. This latter point is something which [Yuri D’Elia] came across with a recently purchased smartphone that lists a 960 FPS recording rate at 720p. A closer look reveals that this is not quite the case.

The smartphone in question is the Motorola Edge 30 Fusion, which is claimed to support 240 and 960 FPS framerates at 720p, yet the 50 MP OmniVision OV50A sensor in the rear camera is reported as only supporting up to 480 FPS at 720p. To conclusively prove that the Motorola phone wasn’t somehow unlocking an unreported feature in this sensor, [Yuri] set up an experiment using three LEDs, each of which was configured to blink at either 120, 240 or 480 Hz in a side-by-side configuration.

As [Yuri] explains in the blog post, each of these blinking frequencies would result in a specific pattern in the captured video, allowing one to determine whether the actual captured framerate was equal to, less than or higher than the LED’s frequency. Perhaps most disappointingly about the results is that this smartphone didn’t even manage to hit the 480 FPS supported by the OV50A sensor, and instead pegged out at a pedestrian 240 FPS. Chalk another one up for the marketing department.

17 thoughts on “Using LEDs To Determine A Video Camera’s True Framerate

  1. I wonder if at any point that high fps mode did work, as it is quite possible it did but the software updates happened to bork it or it technically can do it, but only when the image processor is cooled far better than in the production model.

    Has to be said I don’t see why anybody would want more than the low hundreds on a phone camera anyway – you can’t really control the optics, exposures etc to actually get good super slow shots of the thing you are interested in. So going that into slow motion is just filling up your drive with low quality but very slow shots (if it worked). Making it another feature like all phone camera’s in general – useful more because it is simple and functional enough point and shoot that is convenient to carry than actually really good at capturing what you wanted how you wanted it.

        1. It looks like, if you assume an EV runs at 75MPH for its entire charge, EVs lose about 13% of their estimated range, with German luxury marques typically doing better than estimated and Tesla and Japanese makers being the worst offenders.

          A constant 75MPH is literally the worst case criteria relative to range estimates (except for going faster, of course), but maybe a semi-fair way to evaluate 300mi+ range estimates. American road trips are the most quoted reason for range anxiety, and you likely drive somewhere around that speed through the middle of the country. Not representative of a commute or a day/weekend trip on one of the coasts, but that’s also not what you need to buy range for in the first place.

          1. The whole issue is about what standard if any the auto-manufacturers use to report “estimated range”. It’s not necessarily the EPA standard. In the past, they mostly used the Japanese 10-15 Mode test which only goes up to 70 kph with an average speed of 22.7 kph. Tesla used to report range when driven up to 85 kph – now they use the EPA estimate.

            Also, no fans, no heater or AC, no winter tires, no headlights, no radio or anything… it’s kinda like MPG cheating by taping up door gaps, pumping up the tires and removing the charger belt from the engine. They give you the best numbers they can get away with.

          2. @Dude, I don’t see how interior fans, led lights, and/or a typical oem radio would make anywhere near the difference in range that you’re implying.

            Most people should know that an air conditioner takes power to run in any car, though over time it has become less significant than it used to be. And it’s good to point out to people in cold places that many EVs are not designed to stay outside in low temperatures except by using the equivalent of a block heater not only when parked but also to make up for heat losses while moving. Luckily, when it’s very cold it’s normal to drive slower for safety on icy roads, and if you also do it to make up for the worse fuel economy that’s fine too.

          3. > I don’t see how interior fans, led lights, and/or a typical oem radio would make anywhere near the difference in range that you’re implying.

            You fail to appreciate how little power the car can use to make the advertised range. For example, the Mitsubishi i-MiEV was claimed to get 160 km on just 16 kWh of batteries, so that’s 10 kWh / 100 km or about 10 kW at highway speeds. Also mind that the auxiliary power which runs the lights, fans, window heaters, power steering pumps, etc. comes through the 12 Volt system, which goes through a DC-DC converter with efficiency losses, so if you turn all that off you can potentially save a kilowatt or more.

            >”Auxiliary systems in modern cars represent substantial part in energy consumption. Sometimes even more than 15%.”

            It’s a sum of small things, where you cheat a little bit here, a little bit there, and the combined effect is that your car apparently goes twice as far for the money.

          4. @Dude, but that’s not those systems, it’s the sum of many systems combined. While headlights have grown brighter rather than consume as little power as actually required, I would expect the consumption to be noticeably less than 1 percent of the total at highway speed (aka less than 100 watts). The car’s computers are required, but the added power of running the radio screen and speakers versus turning it off for the test seems like it’d be even less; something between tablet and laptop levels of power. Doesn’t seem like it’d be necessary to do any of that to meet test figures either. The interior fans could be more, but still not enough to be the reason someone gets 10% worse economy – that’ll be squarely on things like tire pressure, aircon/heater use, etc.

            I could see the steering consuming more than that, but not while on the highway and usually not for very much of the time. The assist is usually lowered at higher speed, and you turn very gently. You’ll lose more to the actual act of steering than to the motor that makes it easier, when it’s electric instead of hydraulic – that’s probably part of why they even use it on my gas engine vehicle.

            An EV’s 12V DC-DC converter sounds like it would have less challenges to efficiency than for instance a server’s 120/240VAC to 12VDC power supply, for example. I remember seeing plenty of those which are rated for 94-96 percent, and their passive drain is pretty small compared to 10kW, so at most loads that shouldn’t be a big factor if you change the optimization from “server” to “car” unless you cheap out or have trouble doing it with automotive rated components with a long intended lifespan.

            To be clear I don’t disagree that various tests have been cheated. That’s tradition, they’ve been cheating MPG tests for generations. Just pointing out that there’s no need to turn all these minor things off thinking that you’re going to save a bunch of energy. It’s not like when the alternator fails on your car and you need to keep enough energy in the battery for the spark plugs to fire to get you home.

          5. >not enough to be the reason someone gets 10% worse economy

            Again, it’s a sum of many things. Lights 100W, blower fan 100-200 W, even the windshield wipers take about 50 Watts each. Rear window heater 100 Watts… you can count many things that easily total up to 500-1000 Watts if and when you need them, which can be ignored for the duration of the mileage test by turning off the DC-DC converter in software and running them on the 12 Volt battery. Also, these seemingly small loads hit you proportionally worse at lower speeds, because they tend to be on regardless of whether you’re moving.

            But the biggest problem of all is heat. A typical car’s heater core can put out 3-5 kW of hot air to get the windows defrosted and the interior heated up in the dead of winter. That’s a significant amount of power, and the reason why companies like Tesla have gone for heat pumps, which coincidentally don’t work in the freezing cold weather because the pump performs worse as the temperature difference increases.

          6. To appreciate how much heat you may need in the winter – unless you always start and stop in a heated garage – consider that some modern small diesel cars are already too efficient and don’t make enough waste heat to keep the interior properly warm.

          7. Well for the wiper motors, when you’re using them you tend to find more loss by the extra rolling resistance of driving thru puddles. And all the kinds of heater tend to go together with winter tires and slush and slow driving etc, so it’s a combination of things. I already talked about heat a few replies ago. But making the entire cabin sweltering hot like people seem to enjoy doing is silly when you don’t have ample waste heat. An insulated cabin or a more efficient way of heating just the occupants (heated steering wheel, heated/vented seats, etc) can help with the other. Heat pumps for houses can run down to a lot colder than they used to while maintaining a worthwhile COP, so the manufacturers just need to do the same if they find it worthwhile to keep a good COP in extreme cold.

            But the point I’m trying to make is that we’re talking about different things here. You’re saying that turning off as much as possible and running the 12V stuff off of the lead battery during the test can make for a better fake score. I’m saying that driving an electric car doesn’t mean giving up basic things like the ability to see at night. The EV driver can’t and shouldn’t get the same range as the fake score by turning off the radio and the lights. That’s just like expecting the gas car driver to remove belts from the engine in order to get the rated MPG. What might make a difference to the EV driver is the bigger stuff, of which heat is a great example.

      1. A device can be run outside of its reported spec – it is quite possible that if you drive it at the higher clock it will function just fine, or possibly when dealing with sensor array you can sort of fake it – perhaps for instance this particular array can be read in R G and B or alternate lines separately so you can actually stagger the start of each frame in each colour/line and get something that is kind of higher frame rate…

        But ultimately In exactly the same way many computers ship with CPU being clocked well beyond the stated manufacturer specification, the stated spec of a part doesn’t actually mean anything. Beyond it ain’t broken and we wont warranty it unless it performs worse than this.

  2. The classic “LED chaser” light effect can also be used for this. Have a long string of LEDs, of which one at a time illuminates for e.g. 1 ms each. Then you can see exposure time and frame interval by counting how many LEDs are lit in each frame and how much the starting point moves between frames.

    For rolling shutter cameras you need to position the LED string horizontally to avoid the time difference between top and bottom of rolling scan.

    1. That’s an interesting point. CMOS sensors don’t expose the whole frame at once, they have a rolling shutter with a blanking line and a reading line sweeping across the sensor, and nothing stops you from rolling two or more “shutters” across the frame to get higher frame rate.

      The sensor might then be however many FPS but the actual contents of each frame overlap in time.

  3. Likely, but not conclusive!

    Phone camera set the image brightness by mostly adjusting exposure time. Typically only fixed optical aperture, so no other means present.

    So – for a given scene the phone might decide “I’ll need 5 msec for the image to look okay” and this would limit the possible frame rate. In this example, the video looks like typical bench light.

    I suggest trying to at least 5x the light (as is necessary for any high speed video) or go outside into direct sunlight.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.