N4 Pro's Rear Camera - The Details

[DigitalCorpus]

It is mechanical. It is based off of the expected M12 lenses. As is reasonable, Vantrue (or the ODM) secured it properly considering the "extreme" thermal cycling, the harsh vibration environment of cars, and I will not fault them here. Unfortunately, they way they applied the adhesive is as a thread lock. And it works ***really*** well as such since it feels similar to a hardened version of hot glue.

TL;DR
The focus problems are a because of a poor QA/QC processes in place. If the lens is screwed in too much, then you could shim the mount with *thin* paper and fix the problem. Too little, you have to replace the mount and the lens.

I have nylon jaw pliers. Though they can lack grip, I consider this a non-diy-able focus adjustment as I was applying a lot of torque. Here is the most dramatic frame grab:


And while I was there, PCB shots:


Edit: The SOIC-8/SOP-8 is a 4MiB flash memory IC where the firmware and config for image processing are stored. A couple of transistors, standard assortment of caps, resistors, and a few inductors and transistors. Then the SoC.

The lens housing is machined metal. The mount is glass fiber reinforced polymer, probably nylon. Instead of a could of dabs of adhesive to affix the lens in a static location, the vibration heavy and thermal cycling of an automotive environment means they wanted to fix the lens in place. This means they used the adhesive as a thread lock. It can be cut with a razor and some effort. Based on that amount of force experienced, removing this lens [from the mount] will mean something gets broken.



I'm going to check the dimensions of the mount and its alignment pins to see if an eBay part will be able to be dropped in. If so, then the focus problems can be resolved assuming it is a front focus problem. If it is a back focus problem, then you could remove the mount + lens and put some very thin paper between it and the PCB.

Now, given how secure the lens is mounted, your last ditch resort, which I *highly* advise to not do, is to use some files and remove material between the lens mount and the PCB. If you don't know what a micrometer is and calipers are the finest measuring tool you have, this isn't for you. We're in the lapping realm of material removal.

 

[Agie]

It is mechanical. It is based off of the expected M12 lenses. As is reasonable, Vantrue (or the ODM) secured it properly considering the "extreme" thermal cycling, the harsh vibration environment of cars, and I will not fault them here. Unfortunately, they way they applied the adhesive is as a thread lock. And it works ***really*** well as such since it feels similar to a hardened version of hot glue.

TL;DR
The focus problems are a because of a poor QA/QC processes in place. If the lens is screwed in too much, then you could shim the mount with *thin* paper and fix the problem. Too little, you have to replace the mount and the lens.

I have nylon jaw pliers. Though they can lack grip, I consider this a non-diy-able focus adjustment as I was applying a lot of torque. Here is the most dramatic frame grab:
View attachment 68794

And while I was there, PCB shots:
View attachment 68795View attachment 68796

Edit: The SOIC-8/SOP-8 is a 4MiB flash memory IC where the firmware and config for image processing are stored. A couple of transistors, standard assortment of caps, resistors, and a few inductors and transistors. Then the SoC.

The lens housing is machined metal. The mount is glass fiber reinforced polymer, probably nylon. Instead of a could of dabs of adhesive to affix the lens in a static location, the vibration heavy and thermal cycling of an automotive environment means they wanted to fix the lens in place. This means they used the adhesive as a thread lock. It can be cut with a razor and some effort. Based on that amount of force experienced, removing this lens [from the mount] will mean something gets broken.
View attachment 68797


I'm going to check the dimensions of the mount and its alignment pins to see if an eBay part will be able to be dropped in. If so, then the focus problems can be resolved assuming it is a front focus problem. If it is a back focus problem, then you could remove the mount + lens and put some very thin paper between it and the PCB.

Now, given how secure the lens is mounted, your last ditch resort, which I *highly* advise to not do, is to use some files and remove material between the lens mount and the PCB. If you don't know what a micrometer is and calipers are the finest measuring tool you have, this isn't for you. We're in the lapping realm of material removal.

Thank you for getting to the bottom of this rather sticky issue. All that remains now is whether they've taken this QA issue to heart and rectified it with production units going forward...

 

[DigitalCorpus]

I'd like to have a proper solution, from the manufacturer or DIY, before I consider this closed. Due to the IQ stuff I posted about, I cannot verify if the focus on this lens is set properly, but first sniff says that it is set to infinity.

That said, the manual says the rear lens is f/1.8 w/ a 165° FOV; don't know if that is horizontal, diagonal, or from the image circle. I have a hard time believing that with the selection of lenses I've looked at, albeit, not a conclusive list. Unless they are using a lens for a 3 MP application, which then we might have something.

Edit: Not going to inflate my post count, so I'm just edits atm. The OEM rear camera objective has a 14.25 mm hole in the case for a 14 mm max diameter objective. The current objective's body length is 13 mm. The sensor is slightly smaller than 7.1mm diagonally, thus making it likely a 1/2.7", 1/2.8", or 1/3" sensor. The lens is matching focal lengths from 1.9-2.8 mm with f/2.0 apertures. The mount, btw, is 8.25 mm tall with a hole spacing of 18 mm.

Edit: I know it doesn't "fit", but this should stick out of the case and theoretically would be brighter than the stock lens. My curiosity is piqued:
Interesting possibility: https://commonlands.com/products/wide-angle-3mm-m12-lens-cil327
This one is close to stock: https://commonlands.com/products/small-2mm-m12-lens
This one too: https://commonlands.com/products/wide-angle-3mm-m12-lens-cil329
And this one: https://commonlands.com/products/m12-fisheye-lens-cil222


Edit: Yup, 1/2.8", Vantrue N4 Pro rear camera sensor is the IMX307: https://dashcamtalk.com/forum/threads/49238/post-601873

 

[DigitalCorpus]

Okay, so I'm sure there are others products out there, but I haven't found them after a few hours of searching; even on aliexpress et al. By the way, these are known as "M12" or "S-mount" lenses.

The best alternative S-mount/M12 mount would be a cast or machine piece of metal (aluminum or zinc are commonplace), but again, I cannot find an appropriate one. The common hole spacing is 20 mm, sometimes 22 mm, but very few are 18 mm, and none are in any metal alloy.

There are 3 various plastic types you'll find:

• Plastic - PA (polyamide, nylon, nylon 66/6,6/6-6 et al), PC (polycabonate), PP (polypropylene), ABS, and other polymers by themselves, but these can be blends/mixes, but not in this context
  • Theses are cheap and fine for basic things, but not our applications because they become too soft when heated or are too brittle for the vibrations in the mix
• Polymer + GF30 - Pick a polymer from above, usually PC or PP, with an additional fill of fiberglass measured, by volume, added into the blend
  • The glass fill adds substantial rigidity and tensile strength, especially in higher temperature environments
• LCP - Liquid Crystal Polymer
  • Not much I can say here except that there is a computer fan manufacturer (Noctua) that uses this material heavily for fan blades due to long term dimensional stability, especially when heat and consistent forces are involved

I was looking for a non-US site since this is an international audience, and there are GF30 mounts, but for 20 mm & 22 mm screw hole distances.
Correction, there is this one on Alibaba, but is is a buy in bulk situation: https://www.alibaba.com/product-detail/LCP-Plastic-M12-Board-Lens-Holder_62219606882.html

I initially thought that the mount was a GF polymer mount, but it looks surprisingly close to the CH5117A. The aforementioned is 0.4 mm taller, but the similarities are remarkable. Possibly able to be purchased here.
A US distributor sells something similar with the model number CLA018-LCP, though this one has no alignment pins, and in single unit quantities.


I'm still going through lenses from many manufacturers. I'm actually looking for >f/1.8 and possibly lower distortion, because that is also a typical sign of higher optical quality. A lot of lenses have IR cut filters at 650 nm, but some at 850 nm too. For those unaware, visible light is considered 400-700 nm. An argument for 850 nm filters is extra illumination, but the CCM table might not be able to be correctly altered. I'm shifting my attention here this weekend such that I can narrow down a couple lenses to try in the rear camera as the sharpening algorithm and NR/3DNR algorithms seem to be applied similarly between the 4K & 1080p cameras.

 

[kamkar]

PC fan s are a little like dashcams, slowly getting better and better.
One of the new things are to make fit more tight, and so you cant have fan blades that distort after some use

In PC fans i am looking forward to the alpha cool fans launching soon, they seem to accidental having stumbled on something.

 

[DigitalCorpus]

@Agie why do you favor the IMX323 over the IMX307?

 

[DigitalCorpus]

Btw, tried more for removing the lens from the stock mount. Unsurprisingly, I unscrewed the retaining ring for the front element. So, focus problems will only be resolved with a warranty/RMA claim or replacing the lens + mount.

Now, to narrow down some selections, would folks favor FOV or f-stop? I have a couple of lenses that purportedly faster than f/1.8, but they almost undoubtedly a narrower FOV over stock. Also, should I take Vantrue’s 160° measure as horizontal or diagonal, as that matters quite heavily? Ill break out trig maths with my laser distance measurement tool, but that’s a last resort. I don’t want to purchase more than 3-4 lens options for the sake of waste.

 

[Agie]

@Agie why do you favor the IMX323 over the IMX307?

Haha lol wut I don't remember saying that. You can delete the post by clicking on the Edit tab and it should say Delete underneath mate

 

[DigitalCorpus]

Haha lol wut I don't remember saying that. You can delete the post by clicking on the Edit tab and it should say Delete underneath mate

I know, and it was the IMX335 when I looked it up again. However, that sensor has almost 2 stops less sensitivity. I’m curious how clear it’s 720p pixel binning is, but that would be similar [sensitivity] to 1080p on the IMX307.

 

[Agie]

I know, and it was the IMX335 when I looked it up again. However, that sensor has almost 2 stops less sensitivity. I’m curious how clear it’s 720p pixel binning is, but that would be similar [sensitivity] to 1080p on the IMX307.

Maybes. Granted the IMX335 is a decent sensor but it's old news now of course.

What bugs me is all these rear view mirror and bike cameras that are now using the IMX335 and either quoting 4k (interpolated fake claims of course), and/or charging exorbitant prices for what is essentially old tech.

 

[wibbly]

So is there general agreement that the focus error is static? Some have reported that the degree of focus shift is heat dependent, so worst case could even move between a front and rear focus error as the temperature changes, I guess. This would point to a more fundamental design change being required. Maybe a VERY narrow depth of focus in this set up is the underlying issue?

 

[Agie]

So is there general agreement that the focus error is static? Some have reported that the degree of focus shift is heat dependent, so worst case could even move between a front and rear focus error as the temperature changes, I guess. This would point to a more fundamental design change being required. Maybe a VERY narrow depth of focus in this set up is the underlying issue?

Pretty much.

Wasn't heat dependent for me. If you watch my camera comparison videos with the A229 Pro it's obviously focus issues and that was in like 18 to 20C heat.

 

[wibbly]

Btw, tried more for removing the lens from the stock mount. Unsurprisingly, I unscrewed the retaining ring for the front element. So, focus problems will only be resolved with a warranty/RMA claim or replacing the lens + mount.

Now, to narrow down some selections, would folks favor FOV or f-stop? I have a couple of lenses that purportedly faster than f/1.8, but they almost undoubtedly a narrower FOV over stock. Also, should I take Vantrue’s 160° measure as horizontal or diagonal, as that matters quite heavily? Ill break out trig maths with my laser distance measurement tool, but that’s a last resort. I don’t want to purchase more than 3-4 lens options for the sake of waste.

I can't see why actually, but changing the FOV or aperture will possibly change the amount of light reaching the sensor compared to stock. I assume exposure control is independent for the rear camera, so this shouldn't matter.

I would favor a solution that errs towards a greater depth or focus so the adjustment should be less critical. And I think I would favor FOV over aperture. At night in lowest light conditions, when a car is following you with headlights on, it's difficult to make out too much anyway.

 

[wibbly]

It is mechanical. It is based off of the expected M12 lenses. As is reasonable, Vantrue (or the ODM) secured it properly considering the "extreme" thermal cycling, the harsh vibration environment of cars, and I will not fault them here. Unfortunately, they way they applied the adhesive is as a thread lock. And it works ***really*** well as such since it feels similar to a hardened version of hot glue.

TL;DR
The focus problems are a because of a poor QA/QC processes in place. If the lens is screwed in too much, then you could shim the mount with *thin* paper and fix the problem. Too little, you have to replace the mount and the lens.

I have nylon jaw pliers. Though they can lack grip, I consider this a non-diy-able focus adjustment as I was applying a lot of torque. Here is the most dramatic frame grab:
View attachment 68794

And while I was there, PCB shots:
View attachment 68795View attachment 68796

Edit: The SOIC-8/SOP-8 is a 4MiB flash memory IC where the firmware and config for image processing are stored. A couple of transistors, standard assortment of caps, resistors, and a few inductors and transistors. Then the SoC.

The lens housing is machined metal. The mount is glass fiber reinforced polymer, probably nylon. Instead of a could of dabs of adhesive to affix the lens in a static location, the vibration heavy and thermal cycling of an automotive environment means they wanted to fix the lens in place. This means they used the adhesive as a thread lock. It can be cut with a razor and some effort. Based on that amount of force experienced, removing this lens [from the mount] will mean something gets broken.
View attachment 68797


I'm going to check the dimensions of the mount and its alignment pins to see if an eBay part will be able to be dropped in. If so, then the focus problems can be resolved assuming it is a front focus problem. If it is a back focus problem, then you could remove the mount + lens and put some very thin paper between it and the PCB.

Now, given how secure the lens is mounted, your last ditch resort, which I *highly* advise to not do, is to use some files and remove material between the lens mount and the PCB. If you don't know what a micrometer is and calipers are the finest measuring tool you have, this isn't for you. We're in the lapping realm of material removal.

How easy was it to get the assembly out of the case - have they used glue to keep the case assembled?

 

[Julian23]

So is there general agreement that the focus error is static? Some have reported that the degree of focus shift is heat dependent, so worst case could even move between a front and rear focus error as the temperature changes, I guess. This would point to a more fundamental design change being required. Maybe a VERY narrow depth of focus in this set up is the underlying issue?

It isn't related for me either.
15°C or 35°C while letting the camera bake in the car didn't make any difference.
The front camera is working flawlessly.

 

[Agie]

It isn't related for me either.
15°C or 35°C while letting the camera bake in the car didn't make any difference.
The front camera is working flawlessly.

100% this.

I think that's why Jeff has been silent cos they know they've f ucked it, and same with S1 Pro which hasn't been released yet but didn't get too many brownie points with testers here on DCT

 

[DigitalCorpus]

So is there general agreement that the focus error is static? Some have reported that the degree of focus shift is heat dependent, so worst case could even move between a front and rear focus error as the temperature changes, I guess. This would point to a more fundamental design change being required. Maybe a VERY narrow depth of focus in this set up is the underlying issue?

Pretty much.

Wasn't heat dependent for me. If you watch my camera comparison videos with the A229 Pro it's obviously focus issues and that was in like 18 to 20C heat.

It isn't related for me either.
15°C or 35°C while letting the camera bake in the car didn't make any difference.
The front camera is working flawlessly.

100% this.

I think that's why Jeff has been silent cos they know they've f ucked it, and same with S1 Pro which hasn't been released yet but didn't get too many brownie points with testers here on DCT

The error is static, yes. But I do not think it is as simple as what you guys are interpreting it as. I say this because I don't have the CTE (coefficient of thermal expansion) for the LCP and thus I have run the numbers. I can tell you that practically every diagram of an M12 lens lists a tolerance of +/- 0.3 mm for where the focal plane is. Additionally the range mentioned in the example, which is reasonable for outside air temp, had a mid point of 25 °C and the extremes stated are only a 10 °C excursion. There is a temperature rise of the rear camera, though I haven't measured it, it is still present. And we know car interiors can easily see 60 °C, which is 35 °C from that aforementioned mean, thus the expansion from this higher temp will be ~3.5x that of your example.

This does not preclude Vantrue from fault though. They did have a QA problem and mine that was purchased just under a month ago doesn't appear to have this issue along with having very secure adhesive applied in a manner akin to thread lock. This speaks to having fixed the problem, but not a resolution for the affected units.

I can't see why actually, but changing the FOV or aperture will possibly change the amount of light reaching the sensor compared to stock. I assume exposure control is independent for the rear camera, so this shouldn't matter.

I would favor a solution that errs towards a greater depth or focus so the adjustment should be less critical. And I think I would favor FOV over aperture. At night in lowest light conditions, when a car is following you with headlights on, it's difficult to make out too much anyway.

FOV will affect sharpness and contrast in the lens. However, the only real metric that give some semblance of meaning for the quality of these lenses over the years is the megapixel rating; 5 MP is kinda what I'm aiming for, but a good 3 MP one might be just fine too. It is manufacturer dependent.

Exposure control is all on chip or DSP. There are no mechanical means of adjustment for these types of cameras, generally. That said, literally have of my driving for the past decade and a half is well after sunset and in some dimly lit freeways. Plenty of secondary light pollution, but that is a different topic. Going from f/1.8 to 1/1.4 is ~2/3rd of a [photographic] stop thanks to rounding, but is considered 1-stop (twice or half as much light depending on point of reference). It may not see like much, but every little bit helps when there isn't much light around. As for affecting depth of field, it isn't an issue with these sensor sizes.

An aperture of f/1.8 on a 1/2.8" sensor is equivalent to... ~f/6.9 and the standard f-stop used in regular photography is f/7.1 in that area if on a 35 mm sensor. Opening that up to f/1.4 is equiv. to ~f/5.4, to which the standard measure is f/5.6 in this range. As for depth of field between those two, it is negligible. Using a hyperfocal distance calculator, 0.2212"x0.1252" sensor size & 2.2 mm lens, says that the distance for these sensors changes from 90 cm @ f/1.8 to ~113 cm @ f/1.4. If you're going to the trouble of ordering a new mount and lens, you're generally the type to fidget with lens to have appropriate focus, imho.

Edit: A shallow depth of field for this application would be a 35 mm equivalent aperture of f/4. We're dealing with almost twice the depth of field while ignoring hyperfocal guidelines which make it more than twice the DoF.

How easy was it to get the assembly out of the case - have they used glue to keep the case assembled?

It would take 5-10 minutes with the right tool and a single screwdriver. 2 minutes to change the lens/mount. Another 5-10 for a fist timer to fidgit witht the focus and the 360-480p live view from the app. 3-4 min to snap and screw everything back together.

 

[DigitalCorpus]

matweb says ~10-19.8 um/m * °C. There was ~4 mm of aluminum threads in the mount. Assume the contact point is mid-way, thus 2 mm. The IC-cut filter was ~4 mm from the PCB, so 6 mm in total.
6 mm * 20 um/m = 0.00120 mm
0.00120 mm * 35 °C = 0.0042 mm

Focal plane tolerance is 0.3 mm. 0.0042 mm is ~1.4% at worst.

Aluminum's CTE is ~24 um/m, about 20% higher than LCP, with a lens length of 13 mm, 2 of which is removed for the mount to be fair, thus 11 mm.
Napkin math of 20% more over 2x the length makes that 0.0042 mm become 0.01008 um for the lens + the 0.0042 mm for the mount, or ~4.76%. Assuming a rigid body.

Thread pitch is 0.5 mm, thus 360° of rotation moves the lens in or out 0.5 mm. In order to account for the tolerances of the lens, you'd need to rotate the lens to somewhere within +/-216° from the mean. With a 4.76% max deviation from temperature alone, that is ~17°, which is just shy of 1.8 mm between 2 marks you make on the lens because of temperature. If you include the tolerances of the lens, that 1.8 mm adjustment is inside of the +/- ~22.6 mm distance between 2 tick marks on a lens you'd have to adjust the focus for.

Yeah, Vantrue had bad QA/QC checks for the rear camera focus. Temperature may have caused the adhesive they used to do fun things, but the focus being off isn't from the mount or lense housing material. It could have been a lens QA problem from that manufacturer, whomever it was, and Vantrue didn't catch it.

I'm sleep deprived, my math might be off so don't count this as gospel. Feel free to double check.

 

[wibbly]

As for depth of field....

I was refering to depth of focus, so the the tolerance of placement of the image plane in relation to the lens. I.e. the tolerance of the sensor's displacement within the camera. Not the dept of field (the range of focus in front of the camera lens)

 

[DigitalCorpus]

Even though we don't have that information directly, the math still stands for the low tolerance of these types of lenses and that the CTE is at worse, less than 10% than the mechanical spec.