Readout Noise Non-Uniformities Hamamatsu Fusion BT

1

Hi everyone,

I just want to raise awareness about an issue I’ve been having for a long time with Hamamatsu’s ORCA-Fusion BT (C15440-20UP). It is something we already noticed in my previous institute with some colleagues back in 2020. And now in the facility where I work I have access to a number of models that I could test (so far 4 different models).

There seems to be issues with the readout noise calibration that they do on those cameras. As far as I understand, there’s a factory calibration that is done as part of the manufacturing process, and on top of that there’s a user calibration, that can be performed remotely. Last time I was in contact with Hamamatsu, we tried this remote calibration without success and I will update this thread if I hear about it again.

Below I want to share 1 image per camera taken with nearly identical settings and using the ExCap software that comes with the DCAM api. All images are taken with the camera cap on (no light, dark frames) and with the shortest possible exposure time (1.7632e-5s or 17.632us) in Fast scan mode. To keep the size manageable, I exported PNGs with the contrast scaled between 95 and 105 (the Fusion BT is biased to 100 graylevels). I have the full property list for each model but I wouldn’t know how to share it effectively (its stored as key-value pairs on our Omero sever). All models are air-cooled with the fan active, and they all read -8°C at the Sensor Temperature. Defect Correct Mode is ON and Hot Pixel Correct Level is STANDARD. The most recent model (firmware updated to 1.22, don’t know how to get the firmware version for other models) we have has two interesting properties: OFFSET RECALIB DATA and OFFSET RECALIB STATUS but they are set to OFF and NONE in our case (and as I mentioned when we tried their calibration it didn’t work).

model_1
model_13074×1859 2.55 MB

model_2
model_23074×1859 2.56 MB

model_3
model_33074×1859 2.56 MB

model_4
model_43074×1859 2.56 MB

Some models are doing better (model_2 and model_4) than others (model_1 and model_3, bright and dark patches), and unless you really adjust the histogram like I did in those images you’d be hard-pressed to notice this in your images. I just want to make people aware that this could be happening. My discussions with Hamamatsu haven’t lead to any conclusive fix, but I imagine one could do a background subtraction of an average stack of images as a form of homemade calibration.

Anyhow, I’m curious to hear other people’s experience. Have you ever tried to measure readout noise uniformity on your Fusion BT? Perhaps I’m hallucinating at this point…

Take care,

Omni

2

Hi @Omnistic , I just took noise pictures of our 2 Hamamatsu Fusion BT cameras, and they come out like this:

2XUPBKL
2XUPBKL1843×1843 1.48 MB

and

ZbUu8gE
ZbUu8gE1843×1843 1.48 MB

The noise averages at about 100. Display settings: min=90 max=110. The cameras were just switched on for 5 minutes before acquiring the images.

All I can see is some mildly interesting striping and a bit of charge pile-up (?) at the bottoms.

That being said, I felt compelled to take these pictures because I could have sworn I’ve seen a very similar pattern to your background noise before, in at least one of these cameras. Yet, it’s not there right now. A blotched area of noise that doesn’t share the same “feel” as the surroundings. Stripes interrupted.

1 Like
3

@JelleP

Thank you for doing this test on your end as well. Was it in Fast readout mode? I had never seen this charge pile-up at the bottom, interesting.

With my limited sample size, its not possible to give a precise number, but it does not affect all units for sure. Though to give more context, I was testing a newly bought camera (model 1 in the images I shared) for a group in our university (they wanted to know if they could pay the bill). That is when I noticed the pattern again. Then, I don’t know if that’s the reason, but Hamamatsu was willing to try this remote calibration. Since it didn’t work in the end, the group returned the camera and didn’t pay the bill.

PS: as far as I know the Hamamatsu cameras are all biased to a certain graylevel value to avoid reading negative pixel values. In the past, I believe the Flash 4.0 was 200, and now for the Fusion BT it seems to be 100 (maybe due to the comparably lower noise).

Let’s hear what others have to say about it.

Take care, and thanks again,

Omni

4

Yup this was as raw as I could get ‘em. Using Nikon software, I did fast mode, 16-bit imaging, but the exposure time was not as short as possible. This was limited to “1-frame” in the dual-camera mode I’m running. That’s going to be more like 10 ms.

1 Like