Camera settings and multicolor

[carap...@gmail.com]

Hello,
first of all thanks for making ThunderSTORM available to everyone. As a lonely PhD student in super resolution this has really made my life easier.

I have a question regarding camera settings. I use an Andor ixon+ DU897 EMCCD camera. I got all the specifications from the manufacturers in order to correctly analyze STORM data.
Unfortunately I always get an error message 'Camera base level is set higher than values in the image!'. If I lower the value of Base Level the frequency of the error is greatly reduced, but then the photon numbers are most likely wrong.
Does anyone else use this camera and could maybe share the values used? Is this a common problem?

Second, any plans on implementing multicolor techniques such as spectral demixing?

Thank you

Gora

[Christophe Leterrier]

Hi Gora,

Tese are the settings I use with the same Andor camera:

I think the Nikon N-STORM software specifies a base level of 100 A/D for this camera, but I found it to be a bit lower on mine. The 12.48 value is the A/D gain before EM amplification, corresponding to the 5.1x converter gain (I you use 1x it will be around 5 times lower). 

Hope this helps,

Christophe

[Martin Ovesný]

Just to clarify...that is not an error, it is just a warning. It means that some pixel in the image has lower value than offset. Some of the cameras don't have this value fixed and it is more like an average of different offset values which might differ by tenths of unit or even full units. This happens with one of our cameras too.

If you use lsq or wlsq fitting then you don't have to worry that it would affect results. It shouldn't affect the photon counts of molecules either, because the offset will be counted as an offset of individual molecules. Thus the background deviation should be also correct.

Best,
Martin

[Bram van den Broek]

Hi all,

We have the same Andor camera on our Leica GSD setup. The mean pixel value of dark images (no illumination) is about 101, so I assume the offset is really 100 in our case.
Yet the minimum value in any of those images is about 85. Readout noise is responsible for this, which is the reason why an offset is needed in the first place: to keep the noise distribution symmetric around the base level.

Indeed, ThunderSTORM then gives the warning 'Camera base level is set higher than values in the image' at virtually every frame. Lowering the base level reduces the amount of warnings, but it is bound to give you incorrect photon numbers.

What are your ideas about this, Martin? I guess it is good to automatically subtract a lower value if things are really off, but because of the readout noise you can never get it right: either you subtract a too small value (here 85), yielding too high photon counts, or you subtract the actual offset (here 100), yielding asymmetric noise because negative numbers don't exist.

By the way: many thanks for creating this wonderful plugin! It is by far the best and most user-friendly localization microscopy tool around.

Best,
Bram

[Martin Ovesný]

Bram,

I don't think I have any more ideas about this. Readout noise is usually tiny compared to measured photon counts. I don't think it is such big deal whether the value is subtracted or not. In your case you "lost" 15 units. I don't know what your A/D rate is but it doesn't seem much compared to photon count of a single molecule. Also, if I am right, the noise is Gaussian-like and correct way of handing it is LSQ estimation. Since the fitting algorithm already estimates an offset, it should not matter what value was subtracted as long as all photon counts in the image stay positive. Does this sound reasonable to you?

Best,

Martin

[carap...@gmail.com]

Thanks guys for the useful answers.
What I understood is that, given how the program estimates an offset, it is most important to have bright spots, as always. Even if I 'lose' 15 units, the program would slightly underestimate the photon numbers and hence overestimate the precision. But better to claim a lower resolution than achieved, in my opinion.

I'm still hoping for a response regarding spectral demixing!

Last thing, is there a way to do 3D cross correlation for drift correction?

Thanks again to everyone

[Bram van den Broek]

Hi Martin,

You're right of course: the photon count is generally much higher than these 15 gray values represent. It's only that when the warning message is displayed (at every frame), people may wonder what they do are doing wrong, whether they should change settings... while actually they have set the correct background.

Anyhow, it's good to know what ThunderSTORM exactly does with the data.

@Gora: Actually, by subtracting 85 in stead of 100 the photon count is a bit overestimated, leading to a slightly better precision than in reality.

Bram

[sandra...@mail.huji.ac.il]

Hi!

Great work with the ThunderSTORM! Thanks first for all the work you did.
This post really helped me.
We use the same camera (Andor ixon+ DU897) and we got the manufacturers information that the baseline AD count is 200. If insert this baseline count we get the same warning message:

'Camera base level is set higher than values in the image!'

Christophe said something interesting, the N-STORM Nikon software (NIS Advanced) puts the AD count to 100. We couldn't find this setting in the software. Does anyone know where to find it?

I know this is an old post, still, I would be glad if someone had a hint.
Best, Sandra

pixel size [nm]: 160
photoelectrons per A/D count: 16.2
base level [A/D counts]: 200
EM gain: 250