Mobius

[BobDiaz]

Is there a way (for dev) to improve the graining with a firmware somehow?

Can you also explain why the frames at night are smeared? At day time, even when im driving 150KM/h the frames are perfectly sharp, but at night every movement smeard across the frame. In still cameras its because of longer exposure times at night. Why does it happen in the Mobius? Anything that can be done to improve it?

I could write a whole book about compression. (Just joking around)

Without going into a lot of technical details, the image is broken into little boxes, like an 8 x 8 pixel area. The exact value of each pixel is lost in the compression software, but the relative values between each pixel in the box is preserved as closely as possible, but it may be a bit off relative to the other pixels, but it's close. While the exact brightness is lost in compression, the fine details are preserved unless the compression is VERY excessive. Compression noise in video look like little boxes that change. MPEG-2 used an 8x8 area, but MPEG-4 allowed for different sized boxes. The "dancing boxes" first appear in solid colored areas with very little detail, not as a smearing of detail.

I've seen the smear in the movement in the night shots and to me it looks like a slow shutter speed. The camera must record under a wide range of lighting conditions and needs to adjust the shutter speed of the camera to adapt to the light. In bright light, many video cameras use a very short shutter, like 1/2000 of a second to capture the image. As the light drops off, the shutter speed must be longer to have a proper exposure. It is possible to increase the gain in the video amplifier to offset the loss of light, BUT as the gain increases, the background noise in the image chip is also amplified too. Some DSP (Digital Signal Processing) can be applied to the signal before compression to help reduce the noise, but because of the random nature of the noise, it's not easily performed. Most of the approaches I've seen generate a loss in sharpness.

I'm not sure how Panasonic pulled it off with their AG-AC90 video camera. This is NOT a car camera, but a camera for profession video production. Increases gain always results in increased noise, but on their AG-AC90, they have kept the noise way down, even at high gain levels.

The technology is always getting better and better, but for the near future, there's not a whole lot at the price points we are looking at. However, 10 years from now may be another story...

 

[jokiin]

However, 10 years from now may be another story...

that's about 5 lifetimes in electronics, can't even imagine how different it will be by then

 

[BobDiaz]

I measured the current draw when the Mobius is in the off state back in May and the 15μA you state is correct. I'll try and measure how long the supercaps hold their charge after an automatic shutdown, but that can take a while! Connecting an ammeter circuit to sound an alarm or stop a timer when the supercaps lose their charge would falsify the results, so the supercaps need to be tested by hand. If still holding charge after a set period of time, the supercaps need to be recharged and the test has to be repeated from scratch using a longer time period. I also like to confirm the results by repeating the test. I could also set up a simulation rig using a resistor to drain the supercaps faster, but since current leakage also comes into play I prefer to do it the correct way. I may have to temporarily abort the test if I need my test camera for other purposes so I have no idea how long the test will take. I think your estimate is very optimistic. The Mobius draws a lot of current to properly shutdown. But we'll see.

The discharge curve with the unit in standby must be like a long RC curve. As time progresses, the voltage on the super capacitors slowly drops until it's no longer high enough to support the real time clock. With a current of 15uA, it's like having a 233K ohm --> 246K ohm resistor across the capacitor. The RC time constant would be 582,500 USA notation (582.500 Europe notiaion) seconds --> 615,000 (615.000) seconds. However by the time the voltage reaches the 37% point, the real time clock stopped long ago. If I know the voltage of the capacitor at the point when the Mobius went to standby, and the voltage where the real time clock shuts off, I could come up with a reasonable estimate for how long the Mobius can go without external power before the clock stops.

The other way would be to go a day without power, if that works, charge it up and go 2 days without power...

I have a wild guess, I think it will last for 1 day, maybe 2, but after 3 days I'll bet the clock is off.

 

[Isoprop]

........
The other way would be to go a day without power, if that works, charge it up and go 2 days without power...
.....

That's the way I'm now doing it, but I started with hours - too pessimistic I guess. The big unknown is how much current the camera draws from the supercap when it writes the last clip and shuts down. When the camera loses the date I'll use the center point between the previous and current time period and start all over again. Just rather time consuming .
I let the supercaps charge for 10 minutes before starting the tests - just a random value I chose, much too long, I know, but it's a constant value and easy to remember.

 

[Mtz]

You can start with more than 2 days. My date/time was not reset after almost 2 days. The setting was 1080p, standard.

enjoy,
Mtz

 

[niko]

I could write a whole book about compression. (Just joking around)

Without going into a lot of technical details, the image is broken into little boxes, like an 8 x 8 pixel area. The exact value of each pixel is lost in the compression software, but the relative values between each pixel in the box is preserved as closely as possible, but it may be a bit off relative to the other pixels, but it's close. While the exact brightness is lost in compression, the fine details are preserved unless the compression is VERY excessive. Compression noise in video look like little boxes that change. MPEG-2 used an 8x8 area, but MPEG-4 allowed for different sized boxes. The "dancing boxes" first appear in solid colored areas with very little detail, not as a smearing of detail.

I've seen the smear in the movement in the night shots and to me it looks like a slow shutter speed. The camera must record under a wide range of lighting conditions and needs to adjust the shutter speed of the camera to adapt to the light. In bright light, many video cameras use a very short shutter, like 1/2000 of a second to capture the image. As the light drops off, the shutter speed must be longer to have a proper exposure. It is possible to increase the gain in the video amplifier to offset the loss of light, BUT as the gain increases, the background noise in the image chip is also amplified too. Some DSP (Digital Signal Processing) can be applied to the signal before compression to help reduce the noise, but because of the random nature of the noise, it's not easily performed. Most of the approaches I've seen generate a loss in sharpness.

I'm not sure how Panasonic pulled it off with their AG-AC90 video camera. This is NOT a car camera, but a camera for profession video production. Increases gain always results in increased noise, but on their AG-AC90, they have kept the noise way down, even at high gain levels.

The technology is always getting better and better, but for the near future, there's not a whole lot at the price points we are looking at. However, 10 years from now may be another story...

You right, mobius has to find a perfect shutter speed balance between day and nigh. From side by side night video of dod ls300w vs mobius I made we can see that ls300w is much better at night, I guess due to different shutter speed and maybe default EV value is shifter to (+) side. However day time mobius is better than dod ls300w. They both have similar major hardware components ( cpu, cmos, lens aperture, etc ) so for me it looks like something need to be adjusted by firmware in order to produce better night video recording.
I am not that tech technically advanced in knowldege like you, so those are my assumptions from practical use of different dash cams.
It always good to read your posts and educate myself.

 

[niko]

I will be very interested if we still see horror pictures of molten Mobius when powered by Supercaps. The Supercaps should generate no heat when running, unlike LiPos, and there is more space inside the casing for air to circulate.
Isn't the Australian summer coming up soon?

More on charging...
I don't think some parts of the following information have been previously posted, so here goes.
The USB 12V adapter can pose a problem to the Mobius' LiPo when the camera is used as a dashcam. If the adapter is under-dimensioned or flaky it won't provide enough voltage to properly charge the battery. The charger IC built into the Mobius will stop charging when the external voltage drops below 4.2V and the LiPo will start to discharge. This in turn requires less current from the external power so the external supply can now provide the needed min. 4.2V again. The process continues Ad infinitum. If the external USB 12V supply is unstable these fast charge/discharge cycles are extremely detrimental to the battery and can cause it to heat up and fail within a week. This may have caused those melted Mobius we've seen.
When used as a dashcam, the 12V converter should be capable of supplying at least 1A /1000mA with a stable voltage of +5V to +5.5V. Don't yell, I know I've posted many times in the RC groups that 5.5V is outside the USB specification, but for dashcam use it's acceptable and approved by the developer. He recommends the voltage never drops below 5V. Just don't go higher than 5.5V or you may be asking for trouble. 5.5V is safe though.
Another important factor for dashcam use is the charging cable. It should be a thick cable and of high quality. Poor quality and long cables reduce the voltage significantly. Don't use them. The car charger and the 3M USB cables sold by the Chinese sellers have been tested with the Mobius and are known to work properly.

Sorry this was a bit technical, but hope the information is useful to someone.

This non-stable "low" voltage ( 4.2v, 4.5v etc ) usually accures if using cheap usb cigaretter lighter adaptors, as my experinence shows. I am using 12v to 5v converter which was mentioned in few "hardwire" related threads. I havnt measured how constantly stable voltage provided there, but I can assume they are more reliable to use for having constand 5v due to being more dedicatedly built for this purpose and lenght of power cables from 12v loom to converter then to mobius ( or any other dash cam ) is about 20-30cm in total. This is way less then running 3 metre usb cable via A pillar to fuse box or cigarette lighter, so drop in voltage is minimized if using shorter thicker power cables / wires as you pointed.
Again, I might be wrong in my theory, you and BobDiaz are more into scientific knowledge. I make more self-teached ptactical conlusions and always happy if more wise guys correct me if I do understand something wrong.

 

[Isoprop]

This non-stable "low" voltage ( 4.2v, 4.5v etc ) usually accures if using cheap usb cigaretter lighter adaptors, as my experinence shows. I am using 12v to 5v converter which was mentioned in few "hardwire" related threads. I havnt measured how constantly stable voltage provided there, but I can assume they are more reliable to use for having constand 5v due to being more dedicatedly built for this purpose and lenght of power cables from 12v loom to converter then to mobius ( or any other dash cam ) is about 20-30cm in total. This is way less then running 3 metre usb cable via A pillar to fuse box or cigarette lighter, so drop in voltage is minimized if using shorter thicker power cables / wires as you pointed.
Again, I might be wrong in my theory, you and BobDiaz are more into scientific knowledge. I make more self-teached ptactical conlusions and always happy if more wise guys correct me if I do understand something wrong.

You are most probably right about using cheap USB chargers. The developer tested various low quality charges and found some that may well damage the battery and/or the camera. The 4.2V I mentioned is specific to the battery charge management IC used in the Mobius. A shorter cable will always be better than a longer cable because, like you say, it has less internal resistance, so a cable length of 20-30cm is ideal - as long as it's good quality. Some cheap cables aren't soldered properly with maybe only one or two strands of the wire connected to the USB contacts. I have thrown away numerous such cables. The problem is that it's very difficult to test them. I had a few that worked perfectly when they were new, but after a short time they started acting up. For the Mobius I make my own cables. Initially I had to because there was no dedicated cable available at the time. There wasn't even a casing, so we tested the prototypes using a dangling lens module and, if I remember correctly, the heatsink was strapped to the DSP with tape, or maybe there wasn't a heatsink - I don't remember.

 

[Isoprop]

You can start with more than 2 days. My date/time was not reset after almost 2 days. The setting was 1080p, standard.

enjoy,
Mtz

How did you measure the time? If you manually stop the recording and then unplug the cable the shutdown power required to update the SD card won't be included in the equation. This is thousands of times (comparing 15μA to ~350mA) more than required to keep the RTC running. All the same, I'll let my camera run for another day and see if it's lost the date and carry on from there. I doubt it will keep the time for 2 days, but it will be nice if it does. It all depends on how much the Supercap was drained by the shutdown procedure which roughly takes one second - also needs to be measured .

 

[BobDiaz]

That's the way I'm now doing it, but I started with hours - too pessimistic I guess. The big unknown is how much current the camera draws from the supercap when it writes the last clip and shuts down. When the camera loses the date I'll use the center point between the previous and current time period and start all over again. Just rather time consuming .
I let the supercaps charge for 10 minutes before starting the tests - just a random value I chose, much too long, I know, but it's a constant value and easy to remember.

There's a little bit of a trick to reduce the number of tests you have to perform. While you could go 1 day, if that works, go 2 days, if that works go 3 days, ... that takes a lot of tries if the each test works.

Start with 2 days of waiting. Does it work? If NO try 1 day. If yes, try 4 days.

If 4 days fails, try 3 days. If 4 days works, try 8 days. If 8 days fails, try 6 days (4+8)/2 = 6 ...

If you look at the numbers as follows: 0, 1, 2, 3, 4, 5, 6, 7, 8 we're always trying to find the middle point between where it works and where it doesn't work. Thus, if it worked at 2 days, it might work at 3 or 4. Thus, try 4 first and if that works, we know 3 will work.

Forgive me, but before a became a teacher at El Camino College and Compton College, I was an Engineer. Engineers can't help but wonder about all the stupid technical details, it's in our DNA.

 

[fatpanda]

You are most probably right about using cheap USB chargers. The developer tested various low quality charges and found some that may well damage the battery and/or the camera. The 4.2V I mentioned is specific to the battery charge management IC used in the Mobius. A shorter cable will always be better than a longer cable because, like you say, it has less internal resistance, so a cable length of 20-30cm is ideal - as long as it's good quality. Some cheap cables aren't soldered properly with maybe only one or two strands of the wire connected to the USB contacts. I have thrown away numerous such cables. The problem is that it's very difficult to test them. I had a few that worked perfectly when they were new, but after a short time they started acting up. For the Mobius I make my own cables. Initially I had to because there was no dedicated cable available at the time. There wasn't even a casing, so we tested the prototypes using a dangling lens module and, if I remember correctly, the heatsink was strapped to the DSP with tape, or maybe there wasn't a heatsink - I don't remember.

What are you connecting to , to only have a cable length of 20-30 cm?

 

[Isoprop]

There's a little bit of a trick to reduce the number of tests you have to perform. While you could go 1 day, if that works, go 2 days, if that works go 3 days, ... that takes a lot of tries if the each test works.

Start with 2 days of waiting. Does it work? If NO try 1 day. If yes, try 4 days.

If 4 days fails, try 3 days. If 4 days works, try 8 days. If 8 days fails, try 6 days (4+8)/2 = 6 ...

If you look at the numbers as follows: 0, 1, 2, 3, 4, 5, 6, 7, 8 we're always trying to find the middle point between where it works and where it doesn't work. Thus, if it worked at 2 days, it might work at 3 or 4. Thus, try 4 first and if that works, we know 3 will work.

Forgive me, but before a became a teacher at El Camino College and Compton College, I was an Engineer. Engineers can't help but wonder about all the stupid technical details, it's in our DNA.

Thanks for your tip. I'm very familiar with the algorithm and it's exactly how I started going about it, but then, after reading Mtz's post I decided I had started too small and would never finish! Now I've most probably started too high and will also never finish! So much for theory .
I learned the theory in BASIC some 40 years ago when we had to traverse sorted lists in BASIC. In those days we used a telex hooked up to a mainframe somewhere in Paris. How times, and programming languages, have changed!
I expect it will take several days until I have any results, and if i goof it will take even longer....

 

[Mtz]

How did you measure the time?

I just unplugged the cable. The dashcam users will not manually stop the recording but very rare. We want small dashcams just to forget them exist, but when we need them, the needed recording must be always there. A normal dashcam usage is to mount and forget. But in the first days users want to test, to watch their recordings, some of them are looking to their GPS trip in some PC software. But all of them will finish tired about watching these types of scenes, including some GPS track, which is just time wasted for the user. But always somewhere, hidden, in their mind they know the dashcam is always recording which add a little comfort to their driving period of live. This is when the camera is almost hidden. When not hidden they can't forget because is always in the front of their eyes, another reason do distract them from driving. The driver is driving more comfortable when he can't see its own camera.
I hope now you understood why Mobius can be used by many users. Compared to other good dashcam is nothing special, but the size.

Back to Mobius: the 720p night recording must to be much more improved. I still not understand why the 720p is zoomed compared to 1080p. For me is not a problem this "zoom", but maybe for other users will be when they will understand that 720p recording can be better for them because of more videos in the same Card space.

I am thinking to make a Mobius requests thread.

enjoy,
Mtz

 

[BobDiaz]

Back to Mobius: the 720p night recording must to be much more improved. I still not understand why the 720p is zoomed compared to 1080p. For me is not a problem this "zoom", but maybe for other users will be when they will understand that 720p recording can be better for them because of more videos in the same Card space.

The reason for the "zoom" is that they are only looking at 1280x720 out of 1920x1080 on the chip. Thus, it's a crop of what the whole image sensor sees.

The proper way to do it is to scan the full 1920x1080 and scale the image down to 1280x720. The scaling does require a bit of work, because a 3x3 pixel area has to be converted to a 2x2 pixel area. Whoever developed the software decided to take the easy way out and only scan part of the chip.

Now, maybe there's more to the story than what I know, maybe the processor wasn't fast enough to do all the steps required at a 30P rate; I don't know. I wish they scaled it down or at least hope they can scale it down in a future software version.

Bob Diaz

 

[Mtz]

I have no problem with the crop but with the quality at night for 720p. The feeling is like a zoomed imaged, or something like grain, but is not grain. During day recordings I don't have this feeling.
I will post a night recording tomorrow.

enjoy,
Mtz

 

[jokiin]

The reason for the "zoom" is that they are only looking at 1280x720 out of 1920x1080 on the chip. Thus, it's a crop of what the whole image sensor sees.

The proper way to do it is to scan the full 1920x1080 and scale the image down to 1280x720. The scaling does require a bit of work, because a 3x3 pixel area has to be converted to a 2x2 pixel area. Whoever developed the software decided to take the easy way out and only scan part of the chip.

Now, maybe there's more to the story than what I know, maybe the processor wasn't fast enough to do all the steps required at a 30P rate; I don't know. I wish they scaled it down or at least hope they can scale it down in a future software version.

Bob Diaz

actual imaging area on this sensor is 2304 x 1536, not sure if DOD are doing something like this to make their pretend 720p60 mode that they do with the same sensor but if it is then perhaps just cropping is a better option anyway if the processor is not up to it, the results they (DOD) deliver in 720p60 mode are woeful

 

[Isoprop]

Bob Diaz, Mtz and jokiin, don't worry about the 720p "Zoom" (Angle Of View). You'll be able to select between a Small and Large AOV for both 1080p and 720p in the upcoming FW version which is just around the corner. However, the DSP can't handle a large AOV with 720p@60fps, so if you use 720p@60fps you'll always have the small AOV.

There are many more functions planned for future releases, but these will take a long time to implement. It's becoming all but impossible to correctly test all the different combinations and it's extremely time consuming. As Tom (from RC groups) has posted many times, the next major release will include the WB and color settings similar to the #16, but that is still a long way away. There has also been a delay in FW development due to the Chinese National Day holiday week. The developer has TONS of ideas thanks to all the input from RC groups. He's so devoted to the Mobius project that I'm sure he'll implement everything that's possible, but these things take a tremendous amount of time and the developer only has limited resources.

 

[jokiin]

if the development is ongoing then I'm sure they'll do well with it, firmware makes a world of difference

 

[russ331]

Tom Frank posted this yesterday...

"The image quality, especially white balance and color saturation, are still being tweaked. A FW upgrade is also in the works that will give many user toggles to adjust image quality. It will get better..."

It's good to know I've bought a cam that will continue to improve over time. So many dashcam models have firmware issues, for which the manufacturer shows little or no interest in resolving.


Sent from my iPad using Tapatalk HD

 

[Isoprop]

Tom Frank posted this yesterday...

"The image quality, especially white balance and color saturation, are still being tweaked. A FW upgrade is also in the works that will give many user toggles to adjust image quality. It will get better..."

It's good to know I've bought a cam that will continue to improve over time. So many dashcam models have firmware issues, for which the manufacturer shows little or no interest in resolving.

Sent from my iPad using Tapatalk HD

Yes, the white balance and color control have been planned for 'Phase 2'. However, 'Phase 1' hasn't been quite completed yet .

The developer has an excellent reputation for his firmware updates so you've made a good choice in that respect. I don't know the situation with dashcams, but for 808/RC cameras I don't think any other manufacturer has come anywhere near his service and devotion. He's also a perfectionist.
If you want a universal 1080p camera as small as possible, then the Mobius can't be beat. If you need a camera just for dashcam use I can well imagine there could be better solutions. Since I don't use a dashcam that's for others to decide, but I have noticed that the dashcam market is well established and would expect a 'universal' camera will have a hard time to compete in this market segment.