UMP 5.2 vs 6.0 low light/FPN comparison tests

Hey! In a number of reviews I've heard people commenting that the new 6.0 firmware has improved the low light and fixed pattern noise on the UMP. I was just wondering whether anybody had actually tested these claims by shooting the same conditions with firmware 5.2 and 6.0 and comparing the results? As a UM4.6 user currently stuck on firmware 4.8, I'd be very interested in seeing a comparison (before I continue lamenting the lack of updates) if any UMP owners can be bothered? Thanks in advance!

Another claim of eliminated FPN from respected cinematographer Ed David. Also, his comparison footage of CinemaDNG vs BRAW:

Another claim of eliminated FPN from respected cinematographer Ed David. Also, his comparison footage of CinemaDNG vs BRAW:

Thanks Stephen!

It's definitely enlightening to see the difference between cDNG and BRAW on the new UMP firmware. However, I'm more interested in seeing the difference between cDNG on 5.2 vs. cDNG on 6.0. I think that will give a better idea as to whether there are any improvements to low light/FPN or whether these improvements are only occurring in the BRAW format.

In the tests that I have done it seems that it’s not the braw codec that has fixed the fpn but the new color science 4 that came with the update.

The cinema Dng files still have way more information in them then any of the braw codecs. This is even on the braw website. Where cinema Dng files are 548mb/s and the highest quality braw is 274mb/s

Which is why I believe the new color science 4 is what has fixed the fpn in low light, and not the braw compressed codec.

The cinema Dng files still have way more information in them then any of the braw codecs. This is even on the braw website. Where cinema Dng files are 548mb/s and the highest quality braw is 274mb/s

I wouldn't necessarily look at that chart and say that something has x amount more information just because it's also x amount bigger of a file. In order for bitrates in CDNG to hit 548 MB/s (that be has to be uppercase or you're saying Megabits not bytes) you'd need to be filming something that's completely uncompressible. In most scenarios, 4.6K footage at 30 fps would hover around 391 MB/s meaning that there's a fair amount of repeat data in there (probably because the most significant bits will be the same for large amounts of values). This is relevant because I believe even lossy codecs will attempts to first do some lossless compression before degrading quality. So they still have more information but it's not necessarily twice as much.

Anyway. There's really nothing about color science that would fix fixed pattern noise. If that were true then people would have been able to get rid of the FPN in Resolve before firmware 6.0 came out by just switching footage to color science 4.0.

What's more likely the case is that the sensor profiling info included in BRAW files allows the decoder to compensate for FPN. If it turns out that FPN is gone even in CDNG then it's more likely a change to what the DSP is doing. I haven't seen any test footage that has FPN but I have seen tests that involve regular noise and BRAW is significantly cleaner and that's probably a mix of the split debayering and the sensor profiling.

I shot some none cientific test at night time and was able to see some FPN if the blacks were lifted. If anyone likes i could upload the footage for viewing.