Blurry images from 322, but not all the time

[jollino]

Hello all,
in October 2021 I bought a Nextbase 322GW dashcam and an original Nextbase polarizing filter from Amazon.it (since I'm in Italy).
I've been quite happy with both, however the camera randomly records blurry footage. This happens without any pattern whatsoever, and will persist across all the 3-minute clips it records until it powers down. The next time I start the car it may happen again or not. Essentially, every time it powers on it's like flipping a coin.
The problem is that there's no way to know how it works until I move the SD to the computer and look at the files, which is quite frustrating since I often use the footage for short videos I work on and I have no way of knowing if a whole trip's footage is usable at all until I'm back at the computer . Further, the low quality would make it quite hard to discern license plates in case of an accident.
I'm not even sure if it's a focus issue, or if for some reason the dashcam just "forgets" to apply a sharpening filter. The fact that it may fix itself — or break itself — with a simple reboot would suggest a software issue. Then again, I have no idea if the camera is fixed focus or not.

Attached are a couple of examples from after I updated to the latest firmware (21.5) — I had to do it manually because for some reason the app on the phone insisted that 19.x was the most recent one — taken from the same spot. I saved them as high-quality JPEGs to avoid compression artefacts. While the light conditions are slightly different, it’s obvious that one is considerably less sharp than the other. Notice the bricks on the short side of the wall closest to the camera, or the wheels of the minicar, or the gate of the building on the left.

If anyone is curious, here are a few unedited clips; I just removed the audio, but did not re-encode the video (ffmpeg -c copy -an).
https://www.jollino.it/_tmp/nextbase/nextbase_1.mp4 (1st screenshot)
https://www.jollino.it/_tmp/nextbase/nextbase_2.mp4 (2nd screenshot)
In this one from today (and the whole “session" thereof) the blurriness seems even worse:

https://www.jollino.it/_tmp/nextbase/nextbase_3.mp4

But on the next session, i.e. on my way back, it was pretty much sharp again:

https://www.jollino.it/_tmp/nextbase/nextbase_4.mp4

Since I've seen occasional reports of focusing issues, I was wondering if anyone had encountered a similar problem, where the camera is fine most of the time but then randomly goes and stays out of focus. The third clip actually seems to show an extreme case of short-sightedness, since the wall on the right at the start seems pretty sharp as long as it stays nearby, but then everything else remains out of focus while I drive away and nothing gets that close.

I'm actually quite happy with this dashcam (pending how support handles this :-D I've already opened a ticket with them over the weekend, so they haven't had a chance to reply yet) and I was actually planning to get a rear camera to go with it.
Should they tell me that it can't be fixed and I need to return it to Amazon for a replacement, I may actually go with a 522 so I can keep 60fps from the front camera.

In the meantime, any suggestions would be most welcome.

Thank you in advance!

 

[Nextbase Support]

Hello,

Sorry to hear that you're having a focus issue with your 322GW. Having looked at your footage, there seems to be a lot of vibration from the vehicle impacting on the quality. The blurred parts of the image also seem to be those in shadow.
Are you using the suction mount or the adhesive mount please, and does the blurring ever occur to objects out of shadow please (e.g. in direct sunlight)?
Have you tried adjusting the polarising filter please?

Kind regards,
Millie
Nextbase Technical Support

 

[jollino]

Hello Millie,
Thank you very much for your quick reply.

I am currently using the suction mount, and at this time I'm unable to install the adhesive mount (I still have the one from the old dashcam, and if I put on the new one, I won't be able to remove the old one only — I'll get around to doing that soon.) Still, while it is true that the suction mount is more prone to vibration, I'm not sure I understand how that would lead to the problem happening randomly.
Moreover, I always let the car stay in ACC/ON mode (i.e. dashboard on, engine off) for a few seconds before starting the engine, as it's a diesel car and it needs a little time to settle. That way, the dashcam comes on and starts recording well before any vibration has a chance to start. By that point, it's already either in focus or not, and will remain that way until the dashcam shuts down together with the car.
As for vibration during travel, that's mostly due to the abhorrent state of roads in my area. Still, my experience as a photographer makes me think that such vibration would certainly lead to motion blur, not lack of focus. Besides, 60 fps in daylight are plenty to freeze the action with a high shutter speed, especially considering the nominal f/1.6 aperture; in other words, when moving frame by frame one can certainly see that everything's shaking compared to the frame before, but the frame itself does not show motion blur, just out-of-focus blur (or non-sharpened). Furthermore, once it's short-sighted, it happens even when the car is completely still, i.e. before the engine is started at all.

You can see this in these two videos I recorded just earlier today:

https://www.jollino.it/_tmp/nextbase/nextbase_5.mp4

https://www.jollino.it/_tmp/nextbase/nextbase_6.mp4

The first clip is not terribly OOF compared to others I've had, while the second is basically flawless. In the first one (_5), I kill the engine at 1:13. From that point on till the end, there is no vibration.
In the second one (_6) it takes 2 seconds for the image to come to life, which I supposed is to be expected, and I start the engine at 0:36. Up to that point, therefore, there is no vibration.
Still, if you compare those two parts of the videos, you can notice how the metal structure has much more detailed in the second clip, and even more detail is lost in the first clip for far away objects: the shingles on the orange roof of the little house on the left third, the windows on the tall yellow building in the distance, the brown beams right under the rooftop of the small white house, the vertical bars of the railing on the wide brownish building.
None of those objects are in shadow.

I understand that this may seem like I'm pixel-peeping, but I only got to notice these details because in some cases the video footage looks incredibly washed out, much more so than this, and it's simply unusable and unsatisfactory. It's especially frustrating given how, other times, the quality is pretty much perfect.

Again, compare the first few seconds of clips _1 and _3 I linked in my original post, before any vibration is added (0:16 in _1, and 0:32 in _3), and notice how in _3 the wall itself look much less sharp (not just the bricks for which the light has changed, but also the posters), or the parked cars (the two white ones and the yellow one hadn't moved), or the trees (not just the tall ones, but also the smaller ones in the distance and the branchy bare ones just behind the streetlamp), or even the buildings in the distance. It's a huge difference that remains for the whole recording, not just for the first file. It's pretty much the same difference between how I see without my glasses versus with them.

Changing the angle of polarization seems to have no effect on focus, it just makes the reflections worse (as is typical of all CPL filters, after all.)

My question at this point is: does the 322GW even have a lens that's able to change focus at all, and if so, does the software trigger a focus search and lock upon starting up? (And presumably never again for the duration of the session.) Or is the lens fixed-focus and set to a hyperfocal distance?
In the former case, my hunch is that it just doesn't always focuses properly for some reason, but telling why is impossible for me. I can provide more videos, or log files if the dashcam is able to generate any, and I'm more than happy to help address the issue.
In the latter case, I'd reckon that the issue is either that for some reason some element of the lens has come loose, or some bug in the firmware fails to properly sharpen the camera input in software (and again I'd be happy to provide any material to figure it out). I'd lean more towards the latter, in that case, because if an element has come loose then it would generally not change focus until hitting a pothole or something, yet in my case all I have to do is to turn the dashcam off and on again and boom, the focus has changed.
I'm also keen on trying any beta firmware you may have. I've been beta-testing software for a long time, both open-source and commercial, and I have no problems doing so.
As a developer, however, I understand that this glitch is not easy to reproduce, but the fact remains that the dashcam's IQ is randomly sub-par, and at this time I can't say I'm that satisfied with the product.

I do hope that this can be sorted out somehow, as I was (and still am) considering getting a rear camera add-on, or even getting a 522 if this 322 is unfixable and ought to be returned.

I'm looking forward to your suggestions, and thank you again!

Daniele

 

[kamkar]

Heat related focus shift, that are changing back and forth with temperature is most often indicative of a loose lens element, so in that case you can not refocus the camera but have to replace the whole lens.
But a rattling camera can also mess things up of course, though i would expect a jello appearance more than a fuzzy one then.
With the vehicle stationary you can rule out any issues related to a unfortunate mounting of the camera i would think.

Never experienced this issue tough some times i have had cameras mounted in lets say creative ways.

 

[jollino]

That makes a lot of sense, I hadn't thought about heat. Now that you mentioned, it seems like it's marginally more likely to happen when I first take the car, i.e. on my way out. I don't have any hard statistics however, I'm only now starting to address the issue as I've been dealing with some family stuff in the last few months and this was the last thing on my mind.

It definitely does happen with the vehicle completely still, as the videos above show. I've tried advancing them frame by frame (anyone can confirm that by downloading them) and they all seem to be fine, with no micro-blur.
With my previous dashcam (an Azdome) I wound up designing and 3d-printing a custom mount for the exact angle I needed it for my car, and used 3M VHB tape to secure it. That definitely improved the shaking I was getting before, but then again the Azdome's suction mount was much looser compared to the one that came with this one (that's also why I'm sticking with it for now, pun intended).

Speaking of which, what's the best way to remove VHB tape from a windshield? I was thinking of using a hairdryer from the outside to slowly warm the area up, and then possibly a scraper with a blade to detach it from the inside. No idea if that's appropriate, however.

Thank you very much for your input, and should this not pan out with Nextbase (I really hope it does though, I honestly really like this little thing!), I'll be sure to ask for your opinion on which dashcam to go with as you seem to have a lot of knowledge in the field.

 

[kamkar]

i just roll it off with a thumb, that is the parts i am not able to cut off with my pocket knife, some times i also use some glue cleaner on my window, but only as i test so many cameras and with leftover residue that can have a effect on new cameras mounted in the same place.

If you are mounting to the clean snmooth glass sur´face and not a dotted pattern on the top of the windscreen, the static sticker some provide can be used, then there are no cleaning of windscreen as the sticker then go onto that static film.
I have just used that ( a little 30 X 50 mm piece ) on the Viofo T130 i have tested, it was the first time i tried it and it worked just fine

 

[jollino]

@kamkar, thanks for the suggestions! I had to be a little more aggressive and use a Stanley scraping blade to ease the tape off the glass and then clean it up, but it's all smooth now. I don't think I've ever saw those static stickers you mentioned; I'll do some research on them. Ideally I'll just attach this (or another dashcam) and be done for a while.

@NextBase Support (Millie), I tried gently shaking the dashcam to hear if anything sounded loose, but it was absolutely quiet. I re-enabled the dual video recording, if anything to see if both appear to be OOF when the issue shows up. I will be keeping an eye on it in the next few days and report back. I'm not anticipating much driving for now, and I only have a handful of brief recordings from my tests re-positioning the suction mount (and all appear to be fine). I'll be happy to try anything you may suggest in the meantime. Thanks in advance!

 

[kamkar]

the static sticker are transparent plastic, the one viofo have is about 10 x 10 cm in size, i just cut out a little piece as the footprint / mount of the T130 system is small.

 

[Nextbase Support]

@NextBase Support (Millie), I tried gently shaking the dashcam to hear if anything sounded loose, but it was absolutely quiet. I re-enabled the dual video recording, if anything to see if both appear to be OOF when the issue shows up. I will be keeping an eye on it in the next few days and report back. I'm not anticipating much driving for now, and I only have a handful of brief recordings from my tests re-positioning the suction mount (and all appear to be fine). I'll be happy to try anything you may suggest in the meantime. Thanks in advance!

Hi,

What computer software are you using to play the footage please? We've played all the clips here and are struggling to see which blurriness you're referring to. Have you tried playing the footage using MyNextbase Player or VLC please?

Kind regards,
Millie

 

[kamkar]

There are indeed vast differences in how different player software's handle the same footage, the default windows media player are one of the bad ones.

 

[jollino]

@NextBase Support I've used MyNextBase Player, VLC and IINA, and even Chrome, Safari and Firefox when opening the URLs above, on my iMac 27" 5K, my MacBook Air M1 and my 9th-gen iPad. I even copied the clips to a USB drive and played them on my Sony TV, using both VLC for Android TV and the default Sony video player app; ditto for using AirPlay to stream the footage from a Mac to the TV. The videos play great everywhere, but the OOF clips are OOF and the sharp ones are sharp.
After all, it's just a basic H.264 stream in a MP4 container with a relatively low resolution, so any decoder from the last ten years handles it just fine. Besides, it's not a matter of decoding artefacts — it's the image itself that lacks optical sharpness.

I'm honestly very surprised by the claim that nobody can see the difference in overall sharpness between e.g. the first and third clips, or the end of the fifth and the beginning of the sixth, especially in the areas I mentioned above. For the record, I just went for a drive and had the same end-result as the last two clips I posted yesterday (first one is OOF, the second is tack sharp), so enabling the dual file recording makes no difference with the problem.

Still, if you are unable to acknowledge the problem (which is fully within your rights of course, I'm not "demanding" anything), I'm just not sure what else I can do.
Everyone to whom I showed the 1st and 3rd clips — I initially thought I was being overcritical — immediately noticed the difference, without me even pointing out where to look. A colleague literally said "I just had to look at the trees".
But, well, fair enough.

I really like the dashcam and how it works, but I'm just not satisfied with the unreliable image quality... which is arguably the main point of a dashcam, especially for safety reasons. I tried to provide all the information I could think of, as I always try to do when asking for support to troubleshoot an issue, and asked questions about how focusing works in the device (without receiving any answers) to try and figure out how that may play into it. Whatever the problem is (maybe it's a vampire and doesn't like being turned on when facing sunlight?) and as much as I love tinkering with things, this is closed source hardware and software, and I can't do anything else about it.

At this point I can't see any other option than packing the camera and the filter up and returning them to Amazon.it for a full refund under warranty. It may just be a defective specimen rather than a widespread problem, and being the nerd I am, I've had my fair share of lemons (a wifi security camera from a well-known manufacturer happened to come with damaged RAM, a very expensive Sigma lens for SLRs had its autofocus mechanism die on me for no reason, and so on). Maybe this 322 is really just a bad copy. The point is that I'm not sure I feel like gambling on a new 322, or upgrade to a 522 so that when I get the rear camera I don't have a drop in fps. If it happens again, I'll have to deal with the returns again and I'd be back to square one. Maybe other 322 owners can chime in to see how widespread the issue is and help me decide?

I'm open to any advice you might have, if any. I'm also available if you want to talk directly, e.g. on the phone.

Thanks,
Daniele

 

[jollino]

Hi again,
for the record, last night I completely reset the dashcam to see it makes any difference. I'll keep it for a few more days and check how it behaves. Should it still exhibit the issue — which is indeed there — I can't do anything else than returning it to Amazon and replacing it with something else.

However, just for the sake of it and since I'm waiting for a client who's late, I've looked at the files again. I'm genuinely not sure how one may "struggle to see the blurriness" after seeing the sharp one; I understand that without seeing the sharp one the others may appear decent enough, but when looking at them side-to-side the difference should be obvious.

To kill some time I've actually gone pixel-peeping on the earlier clips and noticed that this is indeed a physical focusing issue rather than just a non-sharpened image.
I'm sure I'm saying nothing new to engineers and photographers here, but the small sensor of the dashcam would make it hard to get any bokeh (this is not a critique: these things ought to be small). It's just a law of physics: a small sensor means an almost pinhole aperture, which means an extremely large depth of field. It's the same reason why smartphones have to fake bokeh via software: most of them have an aperture so small that a "real" bokeh effect can't be obtained.

The spots of reflected light here are actually pretty wide, instead of being tiny as one would expect from using a small sensor and a lens at hyperfocal distance (see: circle of confusion). Yet we do get a hint of bokeh, with a wide and "fluffy" reflection, suggesting that the lens is indeed out of focus. In the sharp image, even though the reflections are bigger due to the slightly different light, they have very defined edges and follow the shape of the car. In the blurry ones, the reflections are round and wider, with non-sharp edges, as is typical of OOF bokeh. Furthermore, the blurry ones are actually shortsighted: near objects, such as the nearest part of the brick wall or the speaker reflection in the windshield (far right of the frame), are perfectly in focus. You can actually see the individual holes of the speaker grill.

Now, whether this is due to heat creep as @kamkar suggested, that's something I'm unable to test properly but it would be quite worrisome if February's "heat" were able to do that. What would happen in July if that were indeed the reason?
One of the reasons I immediately liked Nextbase's dashcams is that they can be removed relatively easily compared to others, but one still needs to re-aim them after reattaching. If the heat causes this focusing issue, then I'd have to keep it out of the car anyway, not just off the mount — and do so every single time I stop the car, whether I'm back home or elsewhere.

As I said, I'll keep testing it to see if the reset did any good but at this point I highly doubt it. It really looks like a hardware issue, and whether it's just a bad copy of the lens or a broader design issue, that's not something I can evaluate in any way. I like to think that I was just unlucky (your generally good reviews suggest that) and I've said this many times and I mean it, I really like the dashcam and its features, so I'm honestly disappointed that I'm disappointed by this IQ issue!

If @NextBase Support has any updates or wants me to run other tests, I'll be happy to.
In the meantime, what is the best way to properly and safely remove the CPL filter? I bought it separately, so I'd have to also return it separately to Amazon/iZilla.

Thanks,
Daniele

 

[Nextbase Support]

Hello,

The main difference in the above images is the sun position. In the initial images, the sun is directly towards the rear of the car, allowing full focus and a clear image. In the secondary image, the sun is illuminating the side of the car out of picture, and causing shadowing on the rear light and near-side of the vehicle. As the area of the car is now in full shadow, the detailing is not as sharp. if you do a direct comparison of an object in the same illumination (e.g. the wall on the right which is in constant sunshine for every image), the detailing remains sharp. if there was a physical problem with the dashcam, we'd expect the whole image to become blurry, which is not exhibited here. This is of the quality we expect from a 1080p dashcam. If you are looking for a more detailed image, we would recommend one of our dashcams with 1440p or 4K.

Regarding the polarising filter, it's adhered to the dashcam using an adhesive ring. I would recommend twisting and pulling the filter away from the body of the dashcam. If you were to upgrade to a 522GW, this comes with a built-in polarising filter.

Kind regards,
Millie
Nextbase Technical Support

 

[jollino]

@NextBase Support I mean no disrespect but that is simply not true: the light source does not affect the focus distance. If you don't trust me (I've been doing photography for just two decades after all, both outdoors and in studio...) ask any other photographer of your choice: if light affected focusing, then long timelapses like this one wouldn't be possible. The wide circle of confusion is an immediate giveaway that the lens is near-sighted, which is why the wall is always in focus: it's just close, and happens to be in focus; looking ahead, things get progressively out of focus. That's why infinity focus is such a big deal that lenses often have a scale precisely for that. It's frankly impossible not to notice the difference in focus between the first and second or third clips, as it persists for the whole duration of the recording.
Whether the lens goes back to being sharp(er) after I stop the car because it cools down or because it somehow refocuses on startup (which I can't confirm without tearing the thing apart, and I won't do that) is something only Nextbase knows.

A wide aperture (f/1.6 as per your specifications) and a misplaced focus point will still cause a lens to be near- or far-sighted, despite a small sensor: the f-number is a ratio, so f/1.6 is pretty wide regardless of actual size. I've actually dismembered an old Logitech C310 webcam to deliberately make it near-sighted when I repurposed it as a 3d printer monitor, as I needed a closer focusing distance than the original person-in-front-of-a-screen distance that it was originally built for. That webcam has a sensor that's probably even smaller than a dashcam's, and it was still feasible to re-focus it.

At this point, I reckon that apparently Nextbase's labs operate on different laws of physics than the rest of the universe, and I'm just going to give up trying to explain the problem, taking more pictures and screenshots and tests to provide proof. At least I can return it to Amazon for a full refund, but I can't help but feel more disappointed by the lack of acknowledgment of the issue than by the issue itself. That's why, despite my original enthusiasm, I'm not going to upgrade to a 522GW or buy/recommend any Nextbase products.

Manufacturing issues may and do happen — remember the Hubble telescope in 1990? it also had a focusing problem… — but denying them despite a customer providing proof is just odd: what's the point? Had the response just been "yeah that doesn't look good and it's weird that it happens randomly" I'd have returned it to Amazon and re-ordered it with full confidence that it was just a lemon, like I did when Tp-Link owned it that their security camera was having RAM problems and acknowledged them in the diagnostic logs I had sent them.
In this case the suction mount, the car's vibrations, the video media player and even the position of stars are being blamed for a lens problem that's there for anyone to literally see... so nope, sorry, unless Amazon or someone else for some reason offers me a discount to give it another go, any remaining confidence is simply gone. Granted, I could return it again if the replacement also had problems, but is it worth the gamble? Not really, I already have too many things to deal with and too little time.

I'm really disappointed because, as I said many times now, I really liked this little thing and I thought it would be a good investment; I mean, I was even going to spend another €80 on it for the rear camera, or sell this one and get its bigger brother! Why would I make this up or be a pest to such an award-winning company? and I hope it's clear I mean no disrespect with anything I've said; I just want the thing I paid for to live up to its own standards, but so be it. As we say here, not all donuts come out with a hole. I may change my mind, but right now that's my position about the whole ordeal.

Thank you anyway and best of luck on your future products.
Daniele