I am a bit confused about spectral and temporal filtering of light.
I consider a single-photon source of a bandwidth of $1\,$nm at $1550$ nm.
That means, that the wavelength of the emitted photons should be gaussian distributed around $1550$nm. The bandwidth of $1$nm gives rise to a coherence time of about $8$ps.
That means, if I place a single-photon detector with a perfect timing resolution behind the triggered source, I expect to see an uncertainty of the photon arrival times of $8$ps.
If I now send the photons through a dispersive medium, for example $100$km of a dispersive fiber with $18$ps/(nm km), i expect the peak of arrival times to become widened to about $1800$ps.
Now, the spectral bandwidth of the photons has still not changed, their coherence time due to their energy-uncertainty is still $8$ps. Now i can divide the time-axis into bins of $8$ps and get much more spectral information about the photon. in fact, I can for example divide the $1800$ps - peak into 225 $8$ps bins. Would this type of "spectrometer" give me a spectral resolution of $~\sim 4.5$pm?
I am a bit confused about this, because the arrival of a photon at a very specific time lets me deduce its wavelength due to the deterministic chromatic dispersion. However, isn't this also the same as spectral filtering? If I had before filtered the spectrum down to $4.5$pm, the coherence time would be huge (in fact, about $1.8$ ns). Therefore my question: Where is the mistake in these ideas? Would this type of spectrometer work?
If I had a source of energy-anticorrelated photon pairs that I send through the dispersive medium, would I be able to find the pairs in the anti-correlated bins?