T54PRO-ZE versus T5442-S3 in low light

[alekk]

TL;DR is I would recommend people use Firmware 0731 ... but be CAREFUL with it because you can easily push the camera too far and end up with a hot noisy mess!


So fortunately it was moderately cloudy again tonight and the moon hasn't quite come out yet ... so I think the light level is about the same. So we can compare the image taken last night with the T54PRO-ZE running firmware 0928 versus tonight using FW 0731 ... with the ONLY difference is the 0928 has Gain=100 and the 0731 has Gain=62. So yea, this seems to confirm that the Gain was "re-scaled" and the 0928 100 is about the same as the 0731 62.

BTW, I wrote earlier "I can kinda see/understand why from a marketing/sales perspective that you "redid" the scale since yea, once you go past 62 (on the 0731 firmware), the noise really does increase ... and I can envision people doing this, seeing the noisy image, and saying "this camera sucks" ... so maybe have something on the gain bar that says "using high values will brighten an image, but may introduce undesirable noise." ... so depending on your tolerance for noise, this wasn't a totally unreasonable limitation.

But don't get too excited about "turning it up to 11", as I'll again warn you that pushing the gain above 62 (with FW0731) is ... pushing it ... especially if you are at the light gathering limit of the camera.







So that wraps up FW0731 versus FW0928 ... but how about 5442-S3 versus T54PRO-ZE w/0731?



My front yard is pretty dark ... so as mentioned before, I have to use some fairly ridiculous parameters on the 5442-S3 to light it up. In this comparison, the 5442-S3 is at 160msec (I'm going to use that value - rather than 1/6s shutter - since that is what the menu provides) with gain=100 and 2D-NR & 3D-NR set to 50. The T54PRO-ZE has the SAME parameters ... BUT ... wait-for-it ... wait-for-it ... the shutter value is 20msec - i.e. that is TWO STOPS faster ... so the aperture/processing is having to make up for that ... although we don't know if the "gain (ISO) scale" of the 5442-S3 and T54PRO-ZE are the same.

Pretty impressive to turn it up to 11 (or maybe 12!) - which do you like better - the 5442-S3 or the T54PRO-ZE?

LOL the "right" answer is neither ... as both are a noisy/smeared mess ... but if I had to choose, I like the T54PRO-ZE ... even though the shutter speed is 4X faster than the 5442-S3!








Finally, I'll show you what happens when you don't crank it to 11 ... but just go a little past 10. Here's two images from the T54PRO-ZE (the clouds were clearing to the East, so I was worried about the Moon popping out and messing up my exposure ... so I didn't run out there to pose for the pictures) that used identical setting of 40msec (1/25 shutter speed), 2D NR and 3D NR of 50 ... with the difference being the gain was changed from 62 to 65.






I.e. that little extra push does degrade the image much, but results in just over a half stop difference in exposure (as measured by the Photoshop Color Sampler - sorry I said Eye Dropper in a previous post) which is significant! BTW, I haven't decided on what I'm going to end up running ... since FOR SURE I would like to have a shutter speed faster than 1/25s, but I have to work with the light I have and the camera's capabilities.



So that's why I recommend Firmware 0731 (based these quick tests) because it gives you some extra room to play with to improve your images ... albeit be careful, because you can also destroy them!

P.S. Sorry I didn't upload any video to look at motion ... but obviously even in the 1/50s (20 msec) picture, you are still going to get motion blur ... on top of a lotta noise & compression blur.
Before someone says it, I'm NOT saying you should run your cameras this slow ... and if you NEED to identify people, by all means "light it up" ... but I'm just trying to compare cameras that I happen to have.

PPS. If you click on an image to see large and then "arrow" across, that seems like a good way to compare - note the file names to help you out.

 

[EMPIRETECANDY]

Thanks @alekk i think the camera's external LED light is not light up when you make the compare right? The smart illumination can be triggered from IR to LED to get a better full color pic when the place is dark, this is the adavantage of the dual illuminations.

 

[bigredfish]

OR...

You could do what we used to do back in the old days, just after the discovery of electricity and Home Depot, and add 3000-5000 lumens to the scene for about $30


View attachment Home_ch4_20240320221535_20240320221612.mp4

 

[alekk]

@EMPIRETECANDY : The on-board light is NOT turned on ... as I'm interested in assessing the camera's sensor/lens/processing capabilities.

@bigredfish : LOL bringing "3000-5000 lumens to the scene" ... and yes, that's why I wrote "if you NEED to identify people, by all means "light it up."

Here's what the house looks like with that turned on. As mentioned, wife would not be a fan of it (don't think I am either) and a bit annoying to neighbors walking by.

But on the other other hand, if you want to "show" that you have a camera (and discourage the dirtbags), that's a "good" thing.

My very quick impression is the onboard illumination works pretty well ... albeit this situation (mounted high with nothing up close would be over-exposed) is about as easy as you get.
And yes, before someone says it, you need the camera to be lower to identify faces, etc.

 

[Darrin Morrow]

OR...

You could do what we used to do back in the old days, just after the discovery of electricity and Home Depot, and add 3000-5000 lumens to the scene for about $30


View attachment 229482
So first light is the Motion flood light you referenced and then the light on the camera? guessing the cam light dialed way down. Sorry just now saw 5442 at the bottom of the video, using a trigger to change to color? Didn't think they could do this

 

[bigredfish]

Setting a 5442 to “Auto" under the Day/Night tab and adjusting the sensitivity on that tab, allows the camera to switch from B&W to Color and back based on its internal light sensor. Always been that
way.

Normally there's not enough light to switch, but when the motion flood kicks in the camera senses the light and switches to color

The "second" light you see is a cheap Mr Beams directly over the door that often gets triggered by the motion floodlight

 

[alekk]

Some more pictures from tonight - very similar lighting conditions as previous couple of nights as it was an hour+ after sunset and the moon is not out yet ... albeit tonight is clear versus the previous cloudy weather, but I don't think making much difference. But should have some decent moon shadows later.

@EMPIRETECANDY asked if I could try using a 25 msec (1/40s) shutter with gain=62 (remember I've got Firmware 0731), and 2D & 3D NR at 50. While I'd love to run with the faster shutter speed, I was pretty sure that was going to be too dark for my liking ... but easy enough to test.

So I put on my "color checker" outfit! ;-)

All of the images (except last one) are taken by the T54PRO-ZE (0731 Firmware) with 2D & 3D NR at 50. The only thing I changed was shutter (again, in msec, not 1/s) and gain. In order of appearance (and brightness), the settings were 25-62 (per Andy's request), 40-62, 40-65, 40-70, and 50-65 - again, those are shutter-gain.

As suspected, 25-62 was too dark for me. As previously mentioned, there is about a half-stop difference (for this scene at these setting) moving from a gain of 62 to 65 - shutter speed is 40msec - which is significant. So lets turn it up to 11 ... or actually just 40-70 ... which really lightens it up, but now the noise is really getting ugly.

The dilemma is that I've hit the wall on the exposure triangle of aperture, shutter, and gain ... albeit processing (noise reduction) is another tweak, but with tradeoffs.

I've already got the aperture opened up all the way. And even at a shutter speed of 40msec, I'm having to use a lot of gain to brighten the image ... albeit I can knock some of that down by cranking up the NR ... but you can't crank that too much because otherwise it's a smeared mess.

What I can do is slow down the shutter ... which obviously increases motion blur ... and 40 msec is already very slow ... but lets try 50msec. And then I'll reduce the gain back to 65 so it's not as noisy - albeit it's still a mess - check out the road ... but I might be able to "tune" that with more NR - good for static, not so good for motion.

And that's what the last T54PRO-ZE picture shows. Bonus is a couple were walking their dogs which you "can make out" in the top-left ... but yes, no surprise that it's quite blurry.

Finally, for comparison, the very last picture is from the 5442-S3 that is also using 2D & 3D NR of 50 ... but gain=100 and exposure of (YIKES!) 160msec.

So I think it is pretty impressive that the T54PRO-ZE provides a "better" picture than the 5442-S3 ... and is doing it at a 3X faster shutter speed.

All of this applies to THIS fairly dark scene - yours may not be so challenging. And yes, if I REALLY needed a more reasonable shutter speed (say, 1/100sec), my only option (with the current Dahua camera's I'm testing) would be to light it up. But this progress bodes well since another 3X improvement would get me to 1/60s shutter ... which is starting to be reasonable.

 

[EMPIRETECANDY]

Looks bit good now without the LED on the IPC-T54POR-ZE, i think when the build in led light on, the pic will be much better.Aha you only want the full color with the current external light.^^

 

[alekk]

Looks bit good now without the LED on the IPC-T54POR-ZE, i think when the build in led light on, the pic will be better.

Yea, the improvement (WITHOUT the LED) over the 5442-Z3 is impressive (I'm thinking of ordering another!), especially the color handling ... since that much more easily shows noise.

Here's a few pictures. I put the NO LED light first ... so you can easily compare to that. The 2nd image is the same settings (shutter=50, gain=65 and 2D & 3D NR at 50) and I only changed the shutter speed on the last image to 10msec - that's 1/100s.

I'm standing about 55 feet away (slant range) and this is a pretty big area for that LED to light up ... and we can see the inverse-square law in effect as the light fades the further back (but it "makes" it across the street), but I'm guessing the image processing is trying to boost the exposure ... but obviously it can only do so much.

I'm not sure what happened to the focus on the last image - hit the AF button several times, but it hit basically the same value ... and even when I tried an area focus where I was standing.
Another aspect I noticed in the Live Feed is (as everyone has seen), if there are any bugs/particles flying around, you'll see a blurry object pass through the frame - surprising number of these!

 

[alekk]

BTW, do NOT judge the on-board light based on just these pictures.

First, it's basically impossible to evenly light an entire scene using a single point source of light (if I was inside taking pictures, I'd do bounce flash off a ceiling ... LOL that's kinda hard to do outside!) and even more so because the light has to be on the camera.

Plus I don't have any experience using the on-board camera light, so I'm sure I could improve upon this.

I would look to others much more experienced than me to assess that ... and since I spun off this thread to specifically discuss the low-light performance of the T54PRO-ZE versus the 5442-S3, check out the main thread for on-board light, IR/BW, and other discussions.

 

[alekk]

Here's a quick timelapse showing the Moon Shadows from last night (along with some daytime footage) from the T54PRO-ZE and 5442-S3.

While the T54PRO-ZE is much better than the 5442-S3 (and the shutter speed is 3X faster!) in (this!) low-light situation, there's not a whole lotta difference between them during the day (with good light!) ... and I actually like the saturated colors of the 5442-S3 a bit more - both are set to 60% saturation - I think the White Balance (both are set to Auto) is being handled differently.

 

[bigredfish]

Why is it the YT player wont allow anything past 1080P

Can you not upload the raw files in 4MP and full resolution?

I know I can

 

[alekk]

Good question and I should have explained that it's 1920x1080 because my program parses the ffmpeg stream and saves that size as image files.
So to create the time-lapse (do you really want to watch 12+ hours in real-time?!?), I grab an image/minute at 60fps and use ffmpeg to generate the movie.

 

[Ookie]

Thanks for taking the time to do these tests for us.

Do you happen to have the full settings you are using on the T5442-S3? I have a lot of white light here at night. Trying to get a color at night setup that doesn't wash out during the day.

 

[wittaj]

Thanks for taking the time to do these tests for us.

Do you happen to have the full settings you are using on the T5442-S3? I have a lot of white light here at night. Trying to get a color at night setup that doesn't wash out during the day.

That is what a day and night profile setup is for.

 

[Ookie]

That is what a day and night profile setup is for.

Thanks. That's how i had it set up in the past. I'll go back to it. Any idea if there is a way to set anti-flicker to 60 Hz? It seems to be locked at outdoor with no drop down. It appears that the devs never heard of floodlights.

 

[bigredfish]

Anti flicker limits your exposure choices if I recall to 1/30, 1/60, and 1/120?

 

[wittaj]

^+1 this.

You want to avoid using the antiflicker because all it does is cap faster shutter speeds.

If you are getting a flicker due to LED lighting, you either accept it or find an odd shutter speed that is off cycle the flicker.

But if you are pushing a camera into color without enough light for a said shutter speed, then you either need to go to IR or add more light or accept the flicker.

 

[bigredfish]

I will say in the 5442H-ZHE bullets I have as well as the 4K/X bullets, I get rolling shutter at anything faster than 1/120 or 0-8.33ms

Shouldn’t happen but it does

 

[steve1225]

I will say in the 5442H-ZHE bullets I have as well as the 4K/X bullets, I get rolling shutter at anything faster than 1/120 or 0-8.33ms

Shouldn’t happen but it does

Because you don't see classic "analog" flicker from incandescent lamps and 50/60Hz (in reality 100/120Hz cuts over zero voltage) AC power supply frequency.

Today problem with flicker is different - it's flicker from LEDs and LEDs DC power sources. Depending of LED model and PWM frequency You have flicker from 500Hz (very low-end models, very visible to cameras) to 20KHz (profesional LED systems done for video recording).

Those LEDs flicker are usually visible on anything faster than 1/100 (1/120)..
In theory AFSA should help here - camera should find lighting visible frequency and shift frames in time to match lights.

But Dahua never correctly implemented this feature. It's available on modern SOCs.
Modern mobiles / sport cameras have those much better implemented.