Maps from TODs related to detectors observing the sky in different frequency bands

[eudigiorgi]

Hello,

I have been using litebird_sim to study the impact of electrical crosstalk on the power spectra, as this effect introduces "mixing" between the TODs of different detectors.

Starting with a multiplexing frequency schedule, I assigned a frequency to each detector and constructed a crosstalk matrix for the LF12 design. From this, I selected N neighboring detectors in the multiplexing frequency domain and computed their TODs and maps.

I aim to analyze the crosstalk effect in two cases:

• between detectors observing the sky within the same frequency channel (e.g., 40 GHz).
• between detectors observing different frequency bands (e.g., 40 GHz and 60 GHz), where significant electrical crosstalk may occur.

This leads to the question: is it appropriate to simulate a single map from multiple detectors operating in different frequency bands? Does litebird_sim have a specific procedure for handling this?
I tried this, and the computation works, but I’m not sure if it actually makes sense from a coding perspective.

I hope this is the right place to ask; if not, please let me know, and I'll close the issue.

Thank you!
Eugenia

[ziotom78]

Hi Eugenia, your analysis is very interesting! I think I understand your needs. I’ll try to explain my point of view to you.

Map-making should always be run on detectors that see the same signal coming from the sky. Map-making algorithms exploit the redundancy in the observation (either because different detectors observe the same spot in the sky or because the same detectors observe the same spot at different times) to characterize the parameters of the noise and disentangle it from the actual signal.

This means that you cannot ask a map-maker to combine the data collected by detectors observing the sky at different frequencies, because they are not seeing the same sky. (One might argue that even two detectors at the same frequency do not observe the same sky, because their bandpass responses cannot be exactly the same… And in fact, somebody advocates producing just one map per detector.)

In your case, the right approach would be to generate timelines for all the detectors, considering the correlations between different bands at every frequency as you plan to do, and then call the map-makers N times to generate N frequency maps, where N is the number of frequency bands you simulate. (Each call to the map-maker should thus only use the TODs for detectors in the same frequency band.) The result will be a set of N maps that should be fed to a component-separation code: the output will be a set of foreground maps (CMB, dust, etc.), and it is at this stage (or later, if one goes on and extracts power spectra) that one can finally assess the impact of inter-frequency correlations on the scientific products.

[eudigiorgi]

Hello Maurizio, thank you so much for your answer!
To summarize and to be sure I understood: I can generate the TODs including the correlation due to crosstalk and than generate unique maps for detectors at the same frequency or, to avoid any trouble, single maps for single detectors.

The result will be a set of N maps that should be fed to a component-separation code: the output will be a set of foreground maps (CMB, dust, etc.), and it is at this stage (or later, if one goes on and extracts power spectra) that one can finally assess the impact of inter-frequency correlations on the scientific products.

At this stage I am not performing any component separation, since I am only feeding the maps with single components; for example at 40 GHz I am generating maps with only synchrotron radiation, without CMB or dust. And from this I am computing power spectra. Does it seem correct from a simulator point of view?

[ziotom78]

To summarize and to be sure I understood: I can generate the TODs including the correlation due to crosstalk and than generate unique maps for detectors at the same frequency or, to avoid any trouble, single maps for single detectors.

Yes, that’s right.

At this stage I am not performing any component separation, since I am only feeding the maps with single components; for example at 40 GHz I am generating maps with only synchrotron radiation, without CMB or dust.

Ah, ok. Let me repeat this to verify I understand your procedure.

1. You generate N maps of synchrotron radiation at N different frequencies.
2. You scan these maps and add correlated cross-talk systematics, producing TODs at different frequencies.
3. You use a map-maker to create back N synchrotron maps. These maps will differ from the input maps created in step 1 because they are affected by cross-talk systematics.

How do you plan to quantify the impact of correlations between the i-th and the j-th synchrotron map produced as output?

[eudigiorgi]

Yes, the procedure is pretty much the one you reported.

1. Choose the detectors from the IMO
2. Generate the input maps, create the observation and inject crosstalk on the TODs
3. Generate the maps with a binned map-making, at this stage one for each detector (from the TODs with and without crosstalk).

We are estimating the crosstalk level by studying the residuals between the maps with and without crosstalk. From this residual maps we calculate the TT, EE, BB power spectra.