Camera board comparisons: Pi NoIR v1 vs Pi NoIR v2
The new 8 megapixel Raspberry Pi camera board has been out for just over a month, and we’ve been seeing some really impressive work being done with it. As many of you know, we also make a version of the camera board with no infrared (IR) filter: the Pi NoIR. Version 2 of the Pi NoIR has also been upgraded to use an 8 megapixel Sony sensor. People use the Pi NoIR to see in the dark (especially useful for wildlife camera traps and for security cams), to achieve some wacky camera effects, and to work on hyperspectral imaging.
What we haven’t seen so far is any comparison of the output from the Pi NoIR v1 and the Pi NoIR v2. So we were really pleased to find this video from Andr.oid Eric which demonstrates the cameras’ raw output side by side, alongside output from both cameras with a selection of IR and UV filters.
Let us know in the comments if you have your own comparison photos or video from v1 and v2 of either of the camera boards – we’d love to see more!
Alex Eames did his usually thorough comparison a while ago :-)
Is it just me or are the photo’s of the version 1 more sharp (more clear) ?
The first batch of camera boards (which has run out now) was calibrated by the sensor factory as they usually calibrate those sensors: as if they were mounted on the front of a camera phone with an expected focal depth of a metre or less. That means that they were less sharp at distance. That’s been addressed now, and they’re now calibrated at or near infinity, so focus with a camera board you buy today will be sharp.
Is there a replacement program for those of us who got the defective cameras? It was very frustrating to buy a “new” model and find out it is much blurrier than the old one. Waste of time and money! How does someone know if they are actually buying and getting a newer batch that has this problem resolved? Is there a model number change?
RS and Farnell have been accepting returns (and it was a very small batch, which has all sold through now, so you won’t be at risk of getting another). Of course, for a lot of uses a shorter focal length is actually desirable – and if you want to try to alter the focal length yourself (doing this will mean you can’t return it, of course – and you’ll need to use caution), you can break the seal and twist the lens unit on its thread with pliers to change its distance from the sensor.
It sounds like your statements are only correct then if using one of your primary distributors and not another intermediate reseller, which you have no idea what their inventory is?
Do RS and Farnell provide an exchange if the camera was purchased from some intermediate other reseller?
No, of course not. You’ll need to talk to whoever you bought it from.
Two weeks ago Eben said in the forum:
“we may also consider changing from glue to a viscous compound to allow users to manually refocus.”
Now you say you can’t return your camera if you alter your focal length.
I’m not really disappointed that you seemed to have made a choice. But it’s sad not to hear a single comment about it.
Tinkering around is what the RasPi and all the available modules are made for. But if you tinker with a camera module to use it for your own project you will definitely lose your guarantee!?
I ask RS HK, they don’t know about “accept return”.
I’m interested in using the new camera for astronomy purposes. The problem I’m finding is that the way the raspberry pie handles camera data makes it extremely difficult to focus the image.
If I use a local monitor (very awkward at a telescope) I can turn on preview mode for a few tens of seconds which lets me center the target of interest and focus it but I need to keep restarting the preview mode over and over again. If I use a streaming version that I have multiple seconds of latency which makes it damn near impossible to get a target for focus.
There should be some way to turn on a video stream with very low latency (i.e. under 50 ms) which will let me target and focus. Once that’s done then I can start streaming to capture images and I really don’t care what the latency at that point
Now maybe there’s a new version of software or something I miss saying but about a year ago when I was trying this, the control was not friendly for astronomy purposes. :-)
IIRC you can use a timeout of 0 (i.e. raspistill -t 0 ) to force an infinitely-long preview?
Just glancing over this it seems the camera is still fixed-focus, which is disappointing. I have 2 low-end Motorola phones (xt1505,xt1527) in which the cameras will actually focus, and over a respectable range from about 2 inches to infinity. I don’t know how it’s possible given the physical size constraints, and the driver has to know about it obviously. It seems like a worthwhile goal. With the OpenCamera app you just tap on what you want to focus on. The Kindle Fire 10 HD doesn’t focus.
I received both std and noir cameras the day after they were announced and both were short focused. Objects located at the optimal short distance (as in average mobile phone photography use) were perfectly in focus. Virtually all camera lenses require a change of focus from infinity to near-field.
Fortunately most cameras have lenses that allow the focal point to be adjusted for the required role. Use inside of a bird box requires a different focus to that used used for bottom of the garden observations.
There is only a tiny amount of glue locking the lens into place which can be scrapped off and then careful pressure applied to rotate the lens. As the lens holder is tapered I pressed the front of the assembly against a very SOFT rubber pencil eraser and used long nose pliers to hold the camera body and stop any torque being transmitted to the PCB mounting point. This makes it far more versatile and if you wish to use it with an adapter for telescopes, microscopes or telephoto lenses then you need to remove the lens for best results. The only danger is loosing the lenses so this procedure is carried out at your own risk. Remember to properly clean the lens after this procedure.
This is after all intended for experimentation and tinkering and it is not like you are going to destroy a £10K professional lens and camera system. All I can say is that the £10K cameras do not produce images that are £10K significantly better than the RasPi.
Got to be honest, if this represents what the v2 camera is doing then it’s not that impressive. The images look significantly darker, losing a lot of detail. I’m hoping it’s just a bad batch (Although why you’d go out of the way to show this off on the blog is a little confusing).
Perhaps it’ve been better off holding off until someone had done a more accurate comparison, especially if you’re still tweaking things with v2 (which, whilst neither here nor there, should have been made abundantly clear that you were going to be doing after release).
Surprised by the “darker” comment, are you referring to the YT video comparison at the top of this page, or something else? I see most shots exposed the same; a few v2 darker than v1 (justifiably, as it shows more detail that v1 washed out) and some lighter. What I found myself and also see here is less detail on v2 at the edges of the frame, but more detail at the center. If you choose to refocus the lens, you can affect that.
Hopefully the V2 NoIR allows the user to control exposure as the V2 shots need some adjustment for best visual quality. In fact to my eyes the V1 shots look way better currently. Be sure to post a new blog once these challenges have been resolved. Thanks in advance! ;-)
“Hopefully… control exposure”? Full exposure control is available with both v1 and v2 cameras, at least with raspistill / raspivid. What program are you using? If you’re referring to this video, can you mention at which specific mm:ss time you see V1 looking way better than V2? They all look a little different because of the wider angle of V2 (and there were many complaints when V1 was new that it should have a wider angle) but I’m not seeing anything consistently “way better” except for focus near edges. In fact I think V2 often has better exposure. Compare 0:36 with 0:41, to me clearly V1 is overexposed in this shot.
I’d disagree with that – V2 looks underexposed, to me, unless the tower blocks in the distance are what we’re supposed to be looking at. The texture on the side of the seat is not well defined.
Or maybe the gamma is wrong.
Of course a lot depends on the setup of the monitor.
Ok, that’s a fair comment. To me, usually when any part of the frame has detail blown out to full white I consider it overexposed even if I’m not “supposed” to look there. Too dark can often be improved afterwards; fully blown out cannot. Fortunately with both v1 and v2, you can select center-weighted metering if that’s your goal, or put in a manual exposure bias if there’s a consistent difference from your intention. Much more difficult is to correct focus if it’s good at center frame and bad near the edge, as many v2 cams tend to be. Another difference is v2 tuning seems to have less sharpening in the GPU, which is a judgement call but it makes v2 suffer compared with v1, on top of the lens differences.
Been following the pitfalls of the v2 NOIR and am surprised its not been withdrawn pending a real upgrade rather than a simple wider fov replacement for the obsolete v1 sensor. Focus across the lens remains a significant issue.
Looking forward to the upgraded v2 that is an actual upgrade!
I find the new Camera module worse than a V1.3
See my quick comp
also, using the module with the same orientation, the picture get’s rotated or flipped (I don’t remember)
“the picture get’s rotated or flipped”
That was a bug that got fixed in later versions of the rpi-firmware.
So you got the first batch with the focus problem?
Or are all v2 just worse then v1 and the story about the first batch is damage control?
This makes it far more versatile and if you wish to use it with an adapter for telescopes, microscopes or telephoto lenses then you need to remove the lens for best results. The only danger is loosing the lenses so this procedure is carried out at your own risk.
Not sure if anyone said this, but thank you and thanks to the Raspberry Pi team for providing a newer, versatile, inexpensive NoIR camera again… This was a handy video to share and, hopefully, more versions of comparison tests will show up as well. Have a lovely summer!
Comments are closed