Pros and Cons: A119S v2 VS A119 v2

[Dont Knowler]

You'll just have to be patient. The problem is not being ignored.

Dear OCD Tronic,
I appreciate your reaction, and your calls for patience. Actually, I am patient, and I am ready to wait for Viofo engineers to resolve the hot pixels problem. My sincere opinion is that the Viofo A119S V.2 is not a bad cam, and with hot pixels fixed (and a metal lens holder, perhaps) it will become a good cam. Otherwise I would have returned my both cams. As I said, I am still waiting for the "good" firmware revision to come - I have not given up hope.
What I hate is when someone (not you) tries to make me believe that white dots all around the screen are tolerable, even after Viofo engineers admitted that this IS a problem.

 

[OCD Tronic]

The second generation auto Calibration technique is new. The first attempt didn't work. Novatek SDK moves slowly. The delay is not on purpose I can assure you of that.

 

[jokiin]

Well when you find a DVR without pixels let this forum know.

all CMOS sensors have them, it's impossible to make a sensor that doesn't using currently available technologies, the reason they are not seen in most products is the pixels get hidden from view using a software process, right now if the automated process is used with the Sony sensor when this feature is enabled it has some negative impacts on picture quality, the algorithm used for this needs further improvement which is something Novatek/Sony are working on, it has been improved already but still needs further work

 

[Nigel]

That exactly is what is NOT being published. What IS being published is how altitude, and geographic location, can affect the relative frequency of white pixels as a result of cosmic radiation.

Using cosmic rays as an explanation for white pixels may or may not be valid (I have no idea what the actual cause is), the fact is they are a reality and happen even in high end dSLR cameras (as anyone who does long exposure night photography knows). Likewise every LCD/LED monitor or TV I've ever purchased has contained some disclaimer to the effect that 'hot' pixels are normal and not covered under warranty. The fact that the Sony sensor which, as I understand, is optimized for low light imaging has them more than others comes as no surprise to me given it's optimization.

The 'practical value' of the relative incidence at 3,000 meters is to show that people living in, or traveling between, Breckenridge, CO (2,926m), Taos Valley, NM (2,841 m), Mammoth Lakes, CA (2,414 m), or Leadville, CO (3,109 m) could have a higher incidence of white pixels than in Miami, FL (1.8 m) or San Diego, CA (19 m). I'm sure a camera at the top of Mt. Everest would fare even worse.

I don't think the issue we have with warm pixels has anything to do with cosmic rays, cosmic rays either cause a brief flash or permanently destroy pixels leaving them completely white or completely black. The amount of cosmic ray damage starts off at zero when the sensors leave the factory and increases over time.

The warm pixels issue is caused at the factory during manufacture and has to do with leakage where some pixels are brighter or maybe darker than they should be, normally by a small amount only be visible in dark conditions, it does not change significantly over time.

The most important thing for a dashcam is the amount of detail it can capture in an image, if the firmware for hiding the warm pixels also hides or damages some of the detail then it is better to just live with the warm pixels. If calibrating a sensor is only hiding the warm pixels then it is not fixing anything. Viofo is correct to concentrate on giving us detail first and dealing with warm pixels in dark conditions as second priority.

For an action camera it is a different issue, then you want the video to look good, the accuracy of the detail is not so important.

 

[Dont Knowler]

The most important thing for a dashcam is the amount of detail it can capture in an image

This is very true. However, this aspect is much more limited by recording bandwidth, rather than by sensor defects or even the max. sensor resolution.
My experiments showed, that with h.264 or similar compression you can expect the 1080p sensor data to be fully (= without noticeable omission of details) "fitted" into the data stream written to a data carrier only if the data stream is ca. 50 (!) MBit/s or more. All the dash cam-typical data rates, which are somewhere in the range between 10 and 20 MBit/s, imply, that a vast amount of data (= details) is being constantly omitted from the video in order to make the video fit into the available data recording rate. No hot pixels can decrease true video quality as dramatically, as this issue does. Just have a look at the trees both sides of the road, recorded by, say, the A119S while you are driving fast - you will not be able to see the individual branches or leaves, even where the 1080p resolution would have been more than sufficient to depict those. The bigger the difference between the current video frame and the previous frame - the less the overall "true detail intensity" of the current frame being saved. Which is too bad, since dash cams are by design intended to hold road accidents details. For road accident moments it is typical, that a vast part of the dash cam field of view changes dramatically within a fraction of a second. And this is exactly the situation where the dash cam recording algorithms tend to omit the most details, in order to fit the video into the relatively narrow recording bandwidth.
This is not a Viofo defect - this problem is pretty much common to all dash cams I know.
However this is (IMHO) the biggest drawback modern dash cams have. I wish there were a dash cam capable of recording the "true" 1080p with 30 or 60 fps to a fast SSD, rather than to an SD card (with its limited recording rates). When it is about big traffic accident money - I want all the details captured, and I am not necessarily willing to let a relatively "stupid" dash cam decide on what details to drop for the sake of its own long-term recording stability.

 

[Dashmellow]

This is very true. However, this aspect is much more limited by recording bandwidth, rather than by sensor defects or even the max. sensor resolution.
My experiments showed, that with h.264 or similar compression you can expect the 1080p sensor data to be fully (= without noticeable omission of details) "fitted" into the data stream written to a data carrier only if the data stream is ca. 50 (!) MBit/s or more. All the dash cam-typical data rates, which are somewhere in the range between 10 and 20 MBit/s, imply, that a vast amount of data (= details) is being constantly omitted from the video in order to make the video fit into the available data recording rate. No hot pixels can decrease true video quality as dramatically, as this issue does. Just have a look at the trees both sides of the road, recorded by, say, the A119S while you are driving fast - you will not be able to see the individual branches or leaves, even where the 1080p resolution would have been more than sufficient to depict those. The bigger the difference between the current video frame and the previous frame - the less the overall "true detail intensity" of the current frame being saved. Which is too bad, since dash cams are by design intended to hold road accidents details. For road accident moments it is typical, that a vast part of the dash cam field of view changes dramatically within a fraction of a second. And this is exactly the situation where the dash cam recording algorithms tend to omit the most details, in order to fit the video into the relatively narrow recording bandwidth.
This is not a Viofo defect - this problem is pretty much common to all dash cams I know.
However this is (IMHO) the biggest drawback modern dash cams have. I wish there were a dash cam capable of recording the "true" 1080p with 30 or 60 fps to a fast SSD, rather than to an SD card (with its limited recording rates). When it is about big traffic accident money - I want all the details captured, and I am not necessarily willing to let a relatively "stupid" dash cam decide on what details to drop for the sake of its own long-term recording stability.

I think you make some good points.

As prices for SSD devices begin to come down and as they get smaller in physical size it would be cool to see them incorporated into future dash cams. It might even be possible to have tiny swap-able external hard drives for dash cameras.

You can now buy a very small 240 GB SSD drive for only $99 USD. (430 MB/s seq. - Read and 400 MB/s seq. - Write speeds.)

That's about the same cost as a 200 GB microSD card but MUCH faster.

 

[Rayny2e]

This is very true. However, this aspect is much more limited by recording bandwidth, rather than by sensor defects or even the max. sensor resolution.
My experiments showed, that with h.264 or similar compression you can expect the 1080p sensor data to be fully (= without noticeable omission of details) "fitted" into the data stream written to a data carrier only if the data stream is ca. 50 (!) MBit/s or more. All the dash cam-typical data rates, which are somewhere in the range between 10 and 20 MBit/s, imply, that a vast amount of data (= details) is being constantly omitted from the video in order to make the video fit into the available data recording rate. No hot pixels can decrease true video quality as dramatically, as this issue does. Just have a look at the trees both sides of the road, recorded by, say, the A119S while you are driving fast - you will not be able to see the individual branches or leaves, even where the 1080p resolution would have been more than sufficient to depict those. The bigger the difference between the current video frame and the previous frame - the less the overall "true detail intensity" of the current frame being saved. Which is too bad, since dash cams are by design intended to hold road accidents details. For road accident moments it is typical, that a vast part of the dash cam field of view changes dramatically within a fraction of a second. And this is exactly the situation where the dash cam recording algorithms tend to omit the most details, in order to fit the video into the relatively narrow recording bandwidth.
This is not a Viofo defect - this problem is pretty much common to all dash cams I know.
However this is (IMHO) the biggest drawback modern dash cams have. I wish there were a dash cam capable of recording the "true" 1080p with 30 or 60 fps to a fast SSD, rather than to an SD card (with its limited recording rates). When it is about big traffic accident money - I want all the details captured, and I am not necessarily willing to let a relatively "stupid" dash cam decide on what details to drop for the sake of its own long-term recording stability.

Going to make some popcorn, this is very interesting stuff.

 

[Nigel]

I think you make some good points.

As prices for SSD devices begin to come down and as they get smaller in physical size it would be cool to see them incorporated into future dash cams. It might even be possible to have tiny swap-able external hard drives for dash cameras.

You can now buy a very small 240 GB SSD drive for only $99 USD. (430 MB/s seq. - Read and 400 MB/s seq. - Write speeds.)

That's about the same cost as a 200 GB microSD card but MUCH faster.

View attachment 31376

Alternatively you can buy a 256GB micro SD that will fit in current dashcams that will do 90Mb/s, plenty fast enough for 2x 1080 resolution. Not sure what price a non-fake one is these days?

 

[Dont Knowler]

Already now there are SSDs with both storage capacity and recording speed large/high enough to store many hours of video, which are tiny enough to be integrated or insertable into dashcams of usual sizes. E.g. SSDs with the M.2-interface: https://en.wikipedia.org/wiki/M.2
And those SSDs are not that expensive - not any more.
One more thing: dash cam manufacturers do not even have to integrate those SSDs into their products themselves - a push-in slot, just like on many computer motherboards, would be enough.
Neither is it a technical problem to manufacture lenses, which would have a true in-focus area larger (and not smaller - greetings to you, you typical modern dashcam designers) than the actual sensor size. Such a lens would presumably only cost a couple of bucks more than the current Viofo lens, which only gives good focus in the center of the sensor, while at the picture edges you get fuzziness...
I am not a marketing man, however I can imagine that thousands, if not millions, of perspective customers all over the world would be happy to spend a couple of dozens bucks more on a product, which would be a quality cam, and not an endless-quality-vs-cost-compromise-piece-of-scrap.

 

[Nigel]

Just have a look at the trees both sides of the road, recorded by, say, the A119S while you are driving fast - you will not be able to see the individual branches or leaves, even where the 1080p resolution would have been more than sufficient to depict those.

Motion blur - happens on all cameras, even ones that record raw video!

 

[Dashmellow]

Alternatively you can buy a 256GB micro SD that will fit in current dashcams that will do 90Mb/s, plenty fast enough for 2x 1080 resolution. Not sure what price a non-fake one is these days?

Hopefully future dash cams will feature true 4K resolution and much higher bit rates than we live with today. At that point small SSD drives will have their day.

 

[Dont Knowler]

Alternatively you can buy a 256GB micro SD that will fit in current dashcams that will do 90Mb/s,

You are right, when you say, that switching to SSD is not a must: good SD or MicroSD cards could do the same job. But tell me, which dash cam can make real use of those 90 Mb/s? Most dash cams I know do not even come close to that, with their 10 to 20 Mbit/s data rates. Those data rates are ridiculous, and are by far not sufficient to accommodate even the 1080p (to say nothing about 4K etc.) data flow without dramatic details omission.
P.S. Please do not mention the Russian DataKam dash cams with their declared 50 Mb/s - those cams have other problems, so I would never want to buy them.

 

[Dont Knowler]

Motion blur - happens on all cameras, even ones that record raw video!

There is huge difference between the normal and pretty unavoidable motion blur and the compression artifacts, resulting from attempts of the codec to fit a huge amount of data into a narrow data rate. Try downloading a 50 to 100 Mb/s raw video from the internet, have a look at it, and feel this difference.

 

[Nigel]

You are right, when you say, that switching to SSD is not a must: good SD or MicroSD cards could do the same job. But tell me, which dash cam can make real use of those 90 Mb/s? Most dash cams I know do not even come close to that, with their 10 to 20 Mbit/s data rates. Those data rates are ridiculous, and are by far not sufficient to accommodate even the 1080p (to say nothing about 4K etc.) data flow without dramatic details omission.
P.S. Please do not mention the Russian DataKam dash cams with their declared 50 Mb/s - those cams have other problems, so I would never want to buy them.

To fully use the 90 Mb/s you will need one of the next generation 4K 60fps chipsets in your dashcam, and then wait for the following generation that include AV1 codecs so that you can get the quality you are asking for into that 90 Mb/s.

Can't see the dashcam manufacturers rushing to fit 4K chipsets though, currently we are going backwards in resolution back to 1080 because the low resolution sensors are more sensitive.

 

[Hombré]

@jokiin ............. As I follow this thread I wonder if a speck of dust has infiltrated one or more of the layers of glass and would love to see a manufacturing photo

I know the painstaking tasks that Intel and others use to prevent any initial foreign contaminants............ Just thinking out loud

 

[jokiin]

@jokiin ............. As I follow this thread I wonder if a speck of dust has infiltrated one or more of the layers of glass and would love to see a manufacturing photo

I know the painstaking tasks that Intel and others use to prevent any initial foreign contaminants............ Just thinking out loud

these types of components are manufactured in dust free environments, quite a lot of effort goes into keeping them that way

 

[The Deacon]

@jokiin ............. As I follow this thread I wonder if a speck of dust has infiltrated one or more of the layers of glass and would love to see a manufacturing photo

I know the painstaking tasks that Intel and others use to prevent any initial foreign contaminants............ Just thinking out loud

If not dust, then possibly one or more minute air bubbles trapped between the glass and plastic layers of the windshield. Things so tiny they'd never be an issue for a driver looking through the windshield, but will be noticed in videos taken with them an inch or less in front of a camera's lens.

 

[Dont Knowler]

one or more minute air bubbles trapped between the glass and plastic layers of the windshield

As I was shooting my test footage to prove hot pixels, I had both cameras lying on my table, with no windshield in front of them.
One more thing was that some of those hot pixels were shimmering - that is something a bubble cannot do.
Another thing was that those pixels were moving synchronously with the camera field of view, as I pointed it different ways - that was a sign that these defects were located inside my both cams, and not outside.

 

[Dont Knowler]

Can't see the dashcam manufacturers rushing to fit 4K chipsets though, currently we are going backwards in resolution back to 1080 because the low resolution sensors are more sensitive.

IMHO "fair", or "true", 1080p footage, with the FOV sharply focused from one edge to another, at a decent data rate of, say, 50 to 100 MBit/s (to get all the details captured rather than omitted due to insufficient data rate) is more than enough to make any dash cam user happy. To record 4K in "true" 4K quality, without dropping too much details, one would need a data rate which would in my view be unrealistically high (for a small windshield device).

 

[jokiin]

50 to 100 MBit/s

never going to be practical in a dashcam using currently available technology, bitrate isn't everything either, just increasing the rate doesn't make up for poor encoding, M2 does up to 45mbit and there are cameras out there doing a better job with less than half the bitrate