I think we need more information (methods and materials type of info) to be able to answer your question:

1. I presume in smFRET you mean single molecule FRET? If so...

2. Are you performing smFRET on immobilized molecules or on freely diffusing molecules? Ifyou are measuring smFRET in freely diffusing mode, are you analyzing data in bursts or in time bins (or in binned bursts)?

3. What dyes are you using? What filters? Dichroics? Lasers wavelengths, type (pulsed, cw), excitation scheme (single laser, alternating two lasers), detectors (PMTs, MPDs, SPADs)? what laser powers are you inputting? what pinhole are you using (in the case of confocal microscopy in epifluorescence)?

4. Are you retrieving ratiometric FRET (ratio of photon counts from two detection channels), or donor-lifetime-dependent FRET (like in pulsed-interleaved excitation, PIE)?

5. What software are you using to analyze the data?

6. What are your control measurements in smFRET?

But even before all of these questions,what are your results of FRET in a bulk measurement, before going into smFRET measurements? 

It will also be very helpful if you can share the data in some way, so we can judge the results.

I am doing an smFRET assay on an immobilized DNA in a prism type TIR microscopy. Here is a link to the paper I am following  http://www.cell.com/action/showImagesData?pii=S0969-2126%2812%2900300-0

I am adding 150mM salt to stabilize the G4 structure. I am doing pretty much same experiment as this paper with Cy3 and Cy5 labelled at each end on the DNA. Fig 3 last panel shows you the peak of the histogram that I am talking about. I was geeting the peak initially but after a month I am suddenly getting a diffused signal though out the x coordinate. do you have any idea why that could be?

You said that: "Although I am using 150mM NaCl and getting a FRET peak at 0.8 but the peak is diffused through 0 to 1 and not a sharp peak". This may mean that some of your immobilized molecules yield the expected FRET trajectories, while others do not.

Since you are performing immobilized FRET and the FRET efficiency histogram you refer to is constructed from many molecules in your field of view, I would try to identify, which immobilized molecules in your field of view give a minimally fluctuating signal around E=0.8 and which not. This may be caused by imperfect immobilization or from problems in surface passivation. I really hope you do not keep the slides for months. If you do, please try to prepare a fresh sample of immobilized molecules.

Well we keep PEG coated slide for a week and add fresh neutra avidin just before adding oligo (biotin attached). But as you can see the structure of my oligo, it is partial doublestrand and so I am not sure if the problem is from imperfect annealing. But I tried annealing several times as I did before (when the signal was good). So was wondering if the dye can go bad or reagents or anything else. Any clue?

Before even thinking about a solution, I would like to understand better the problem. What does your molecule-by-molecule analysis yield(the one that I suggested above)? Is it that a certain fraction of molecules yield trajectories with small fluctuations around E=0.8 throughout the trajectories,while other molecules yield totally spurious trajectories?

Only some of the molecules show a stable trace at 0.8. Some others start at 0.8 and then fluctuates and most others start at random points and fluctuates. Does that make sense?

Yes, it makes sense that your sample got degraded with time, so that with time,less and less single molecules show up the expected trajectory signature.