30 fps vs 60 fps

[Vortex Radar]

So I've been wondering how much of a difference 60 fps makes over 30 fps and decided to test it out using a pair of A329's.

Online I've sometimes seen people say that shooting at a faster frame rate automatically implies less shutter speed, but that's not necessarily true. You do get smoother videos, sure, but that doesn't always lead to improved detail. In fact sometimes things like HDR work better.

Anyways, in this video I go into some camera theory, a bunch of testing between different frame rates, noticing when each frame rate has an advantage (like how 60 fps only has an advantage in lower light before you switch over to HDR), comparing different implementations of HDR, etc.

I also did a comparison of file sizes with different dashcams. Interestingly some show larger file sizes when you double up the frame rate which is what I'd normally expect, some keep the file sizes constant, and one, the A119 Mini 2, actually lowers (!) the file size when you jump to 60 fps.

 

[kamkar]

I compared 2 similar systems over 10 years ago, one set to 30 the other 60 FPS, and it was extremely rare the 60 FPS camera got a capture the 30 dident or let me rephrase, rare the 60 FPS system got a plate capture that was less hard to read than the 30 FPS systems footage.
I was at the time convinced that 60 FPS was the holy grail in dashcams and had to be what i / we wanted, but my test said otherwise.
I am a little pleased my hopes for HDR seem to be a little more true.

One more thing people forget, if you have a bitrate limit and is at it, well spreading that out over 60 FPS instead of 30, well you will have a thinner coat of " paint " on your footage so to say.
i have always said though not really 100 % true, called bitrate for " the image quality multiplier ", BUT there do seem to be a point of diminishing return cranking up bitrate, this was also tested by me using the same two cameras.
In that case 28 mbit seemed to be just fine for the 1080p footage and 40 mbit dident really do much more but make the system run hotter.

Looking at bitrate you also have to be mindful of the encoder used, as i recall it was said that with H.265 you get the same image quality as with H.264 but at half the bitrate.

 

[Nigel]

@7:05. The 60fps has twice the motion blur of the 30fps, yes sometimes 60fps has less, but at other times it has more motion blur.

@10:00. The 'O' in DOL does not mean overlaid, it is referring to a temporal overlap of the exposures, as in they are exposed simultaneously, not at "different times" as you state. (The Sony published Clear HDR explanation has that wrong!)

@11:20. Clear HDR actually intentionally forces both exposures to be the same length, hence has far more motion blur than DOL HDR, it is not designed for freezing plates.

 

[Vortex Radar]

@7:05. The 60fps has twice the motion blur of the 30fps, yes sometimes 60fps has less, but at other times it has more motion blur.

@10:00. The 'O' in DOL does not mean overlaid, it is referring to a temporal overlap of the exposures, as in they are exposed simultaneously, not at "different times" as you state. (The Sony published Clear HDR explanation has that wrong!)

@11:20. Clear HDR actually intentionally forces both exposures to be the same length, hence has far more motion blur than DOL HDR, it is not designed for freezing plates.

Do you have any references about this? Sony is pretty explicit about how it behaves as I described. It’d be a pretty big mistake on their part if they got it backwards.

DOL HDR

When the digital-overlap (DOL) HDR feature is on, the image sensor captures two images in succession: one with a short exposure according to the bright region and the other with a long exposure adjusted to the dark region. HDR is realized by synthesizing these two images to complement one another.

However, this method involves a slight time lapse between the two shots and this can cause some artifacts, such as a blurred outline and chromatic aberration, if the target is in fast motion.

Clear HDR

When the Clear HDR feature is on, the image sensor captures two images simultaneously, one with a low gain level set to the bright region and the other with a high gain level adjusted to the dark region. The images are then synthesized.

This method has the advantage of delivering images of a moving target without chromatic aberration and other artifacts because the two images are captured at the same time.

The Clear HDR feature is suitable not only for security cameras but also for applications to capture moving targets, such as traffic monitoring systems and dashboard cameras.

Source: https://www.sony-semicon.com/en/technology/security/index.html

 

[Nigel]

Do you have any references about this? Sony is pretty explicit about how it behaves as I described. It’d be a pretty big mistake on their part if they got it backwards.

You could try this: https://medium.com/@svogisan/introduction-of-hdr-image-sensor-and-its-applications-83d93e50f4e6

It is still not all that easy to follow, the key thing is that the rolling shutter reads one line of pixels at a time, working its way down the sensor, or up in some cameras. It first reads the long exposure, then reads the short exposure for the same line, then moves on to the next line, so the short and long exposures are interleaved, as the lines of old TV images were. You do not get all the lines for the short exposure, followed by all the lines of the long exposure, with a whole exposure time between the two exposures as in older HDR cameras. With DOL, the short exposure is just one line (1/2160th of an exposure time) behind the long exposure, and the short exposure of each line is ahead of the long exposure for the next line of the rolling shutter image.

Hope that makes sense?

 

[LateralNW]

So I've been wondering how much of a difference 60 fps makes over 30 fps and decided to test it out using a pair of A329's.

Online I've sometimes seen people say that shooting at a faster frame rate automatically implies less shutter speed, but that's not necessarily true. You do get smoother videos, sure, but that doesn't always lead to improved detail. In fact sometimes things like HDR work better.

Anyways, in this video I go into some camera theory, a bunch of testing between different frame rates, noticing when each frame rate has an advantage (like how 60 fps only has an advantage in lower light before you switch over to HDR), comparing different implementations of HDR, etc.

I also did a comparison of file sizes with different dashcams. Interestingly some show larger file sizes when you double up the frame rate which is what I'd normally expect, some keep the file sizes constant, and one, the A119 Mini 2, actually lowers (!) the file size when you jump to 60 fps.
View attachment 81915

What are the bit rates between 30fps and 60fps? Ideally doubled?
If they are similar wouldn't that also affect detail?

It would be interesting to see what the ISO is set at or is it variable? Increases noise?
Change the lens to a lower f-stop? F1.2 but less depth of field which might make it even harder to capture number plates?

In the graph it shows the Viofo A329S 2Ch having exactly the same values but with 1 Ch it is different. Is it possible that 60fps is not working with 2x cameras?

Clear HDR I was hoping to be the answer to prevent blurred cars and licence plates but based on Viofo's findings it didn't work as Sony implied. eg 1 photo taken and the processing done on the 1x photo.

Thanks for looking into this topic.

 

[kamkar]

/me confused on clear HDR too, by Sonys words / material i was thinking " OMG yes ", but then others like Viofo say otherwise.
I am not sure if this is CUZ there need to be something " special " in the ISP to process clear HDR like Sony say / intend, and so that is why Viofo can not pull it off and get the desired results.
If it is any good surely it must be in Sonys interest to enable people to make use of it.

This is all off the deep end i feel, and we might risk just confusing people more, or at least the common dashcam shopper / user.
I read into things like this some times in the dead of night trying very hard to not be bored as that is a sign of a feeble mind, but i must also admit, these things are not something my brain retain very much off.

 

[Nigel]

I am not sure if this is CUZ there need to be something " special " in the ISP to process clear HDR like Sony say / intend,

Some of the Starvis 2 sensors will process the Clear HDR in the sensor, the processor can just treat it as an ordinary HDR image.

 

[Vortex Radar]

What are the bit rates between 30fps and 60fps? Ideally doubled?
If they are similar wouldn't that also affect detail?

Here's the raw data, including the percentage increase of the front cam:

It would be interesting to see what the ISO is set at or is it variable? Increases noise?

Yeah, I don't think dashcams offer us the option to see that. I'm definitely curious too.

Change the lens to a lower f-stop? F1.2 but less depth of field which might make it even harder to capture number plates?

Are there dashcams with variable aperture lenses? I thought they were typically fixed aperture and fixed focus.

In the graph it shows the Viofo A329S 2Ch having exactly the same values but with 1 Ch it is different. Is it possible that 60fps is not working with 2x cameras?

I'd reported to Viofo earlier that 60 fps videos on the A329S are only capturing at 30 fps even though the files are saved as 60 fps, and they're doubling up every frame. I haven't doublechecked with the latest firmware to see if this is fixed. With the original A329, it's actually capturing at 60 fps, so perhaps that's part of the difference?

Clear HDR I was hoping to be the answer to prevent blurred cars and licence plates but based on Viofo's findings it didn't work as Sony implied. eg 1 photo taken and the processing done on the 1x photo.

Definitely a bummer. IIRC, I've seen some other companies advertising Clear HDR, I think VanTrue for example, but I'm not sure if that's actually be verified. I'm guessing someone has though?

Thanks for looking into this topic.

It's fun nerding out over this stuff!

 

[Vortex Radar]

This is all off the deep end i feel, and we might risk just confusing people more, or at least the common dashcam shopper / user.
I read into things like this some times in the dead of night trying very hard to not be bored as that is a sign of a feeble mind, but i must also admit, these things are not something my brain retain very much off.

haha yeah, even though my video is done, I keep thinking of ways of how to explain things more simply and wish I had the chance to do it again, lol, because I have better explanations in my head now. Some of this stuff can be a bit complex, especially as you first start getting into it, but hopefully we can boil it down to "Feature ABC is useful in situations 1, 2, or 3, but not in 4, 5, or 6" without having to go off in the weeds, haha.

 

[Augustus]

Here's the raw data, including the percentage increase of the front cam:

View attachment 82002

Yeah, I don't think dashcams offer us the option to see that. I'm definitely curious too.

Are there dashcams with variable aperture lenses? I thought they were typically fixed aperture and fixed focus.

I'd reported to Viofo earlier that 60 fps videos on the A329S are only capturing at 30 fps even though the files are saved as 60 fps, and they're doubling up every frame. I haven't doublechecked with the latest firmware to see if this is fixed. With the original A329, it's actually capturing at 60 fps, so perhaps that's part of the difference?

Definitely a bummer. IIRC, I've seen some other companies advertising Clear HDR, I think VanTrue for example, but I'm not sure if that's actually be verified. I'm guessing someone has though?

It's fun nerding out over this stuff!

Nice video mate doing god's work by investigating what has divided the community for years haha. I find it interesting that file sizes don't increase for some models especially 70mai too, I wonder why 🤔

 

[Nigel]

Definitely a bummer. IIRC, I've seen some other companies advertising Clear HDR, I think VanTrue for example, but I'm not sure if that's actually be verified. I'm guessing someone has though?

Yes, it won't work though, Clear HDR is deliberately designed to maximise motion blur, great for avoiding LED lamp flicker and for avoiding HDR ghosting effects; it is not designed for reading license plates in the dark, which is what we are interested in.

 

[LateralNW]

Here's the raw data, including the percentage increase of the front cam:

Thanks for sharing your raw data.
I'm assuming the video length is the same for all examples, perhaps 1 Min?
It is interesting that the only dashcams that appears to work as expected is the Vantrue E360 and the single channel A329n

I wonder if there is some sort of interpolated or stronger compression being used on some of the units?

Yeah, I don't think dashcams offer us the option to see that. I'm definitely curious too.
I have no idea if there is an ISO setting. I was thinking more inline with SLR's really.

re there dashcams with variable aperture lenses? I thought they were typically fixed aperture and fixed focus.

Again I was only putting it out there. Just wondered if it might make a difference.

I'd reported to Viofo earlier that 60 fps videos on the A329S are only capturing at 30 fps even though the files are saved as 60 fps, and they're doubling up every frame. I haven't doublechecked with the latest firmware to see if this is fixed. With the original A329, it's actually capturing at 60 fps, so perhaps that's part of the difference?

The doubling up on every frame could that be interpolated? I'm guessing but I would think it would allow the file spec to show 60fps ?

It's fun nerding out over this stuff!

I agree. Unfortunately in my early days of dashcam testing, Dashcam manufactures didn't like their underbelly being exposed. I enjoy pulling them apart! Things seem to have changed recently.

 

[Vortex Radar]

Thanks for sharing your raw data.
I'm assuming the video length is the same for all examples, perhaps 1 Min?
It is interesting that the only dashcams that appears to work as expected is the Vantrue E360 and the single channel A329n

Yep, lol. I didn't wanna just look at the bitrates and extrapolate or anything, so I figured I'd just let each dashcam run for at least a minute and then just pull the numbers that way.

The doubling up on every frame could that be interpolated? I'm guessing but I would think it would allow the file spec to show 60fps ?

So if you go frame by frame in the "60" fps A329S/T clips, there's no interpolation in between. Just 2 identical frames in a row and then 2 more identical frames in a row.

I agree. Unfortunately in my early days of dashcam testing, Dashcam manufactures didn't like their underbelly being exposed. I enjoy pulling them apart! Things seem to have changed recently.

View attachment 82011

lol.. I used to do radar detector teardowns back in the day, but I stopped because I ultimately didn't understand enough to make it be sufficiently meaningful. There was some cool stuff I could glean about the layout and design, but I didn't know how to follow the path of the signal going from the antenna, through the various amplifiers and processors, and ultimately out to the speaker and display. Radar detector companies also used to complain that I could put things back together in a way that could compromise performance or some of the stealth characteristics (allowing signal leakage back out) and so just to be safe, I've opted to keep things factory and not have anyone wondering if I'm unintentionally and unfairly making something perform worse than it should.

 

[viofo]

Yes, it won't work though, Clear HDR is deliberately designed to maximise motion blur, great for avoiding LED lamp flicker and for avoiding HDR ghosting effects; it is not designed for reading license plates in the dark, which is what we are interested in.

That's right. This is what we got while we testing the CLEAR HDR.

 

[Vortex Radar]

That's right. This is what we got while we testing the CLEAR HDR.

Well that's a bummer...

 

[LateralNW]

So if you go frame by frame in the "60" fps A329S/T clips, there's no interpolation in between. Just 2 identical frames in a row and then 2 more identical frames in a row.

Do you find the video stutters?
Perhaps it looks normal but because it thinks it's 60fps it plays back that much faster disguises the duplication?

Just thinking out loud.

 

[Pack72]

Thank you for this video. I like to understand how things work, especially things I interact with on a daily basis, but I got into photography back in the days of film and mechanical shutters. I'm slowly moving from "dash cam as a black box" to "dash cam as an understandable device", and you have helped.

 

[Vortex Radar]

Thank you for this video. I like to understand how things work, especially things I interact with on a daily basis, but I got into photography back in the days of film and mechanical shutters. I'm slowly moving from "dash cam as a black box" to "dash cam as an understandable device", and you have helped.

Yeah if you have a background in still photography too, then understanding some of the camera and exposure basics will definitely come easier. 🙂 There's some additional factors when we start doing video instead of stills, but looking at just one frame at a time helps keep the discussion simpler.

 

[jsmith]

@Vortex Radar

Very good video, but there are a few points that were overlooked on 30 FPS vs 60 FPS.

1. In your comparison, you acknowledge that larger objects are often more discernible (like a passing car) at 60 FPS vs 30 FPS. In the case of a hit and run, where the plates weren't captured, a description of the vehicle is relevant to police and insurance companies. If you can state it looked like a 2007 Volvo, that's still better than nothing.

2. Not all accidents involve car. Wild animals (Deer, Moose, etc), pedestrians, Objects fall from a vehicle, etc. In these cases, more frames create a smoother video day and night, which leads to better documentation.

Yes, 60 FPS isn't a magic fix for license plates because 1/60 vs 1/30 is generally not a quick enough exposure for high action. However, don't fully discount the secondary benefits of a higher frame rate.