Using a chopper/lock-in-amplifier allows to measure only the signal caused by Luminescence. Otherwise you can also registrate "harmful" light which is not associated with Luminescence. The PL intensity can be significantly smaller than the intensity of the "harmful" light. Therefore, in this case, the photoluminescence may be only a weak background. On the other hand you want to be sure that the studied samples are quite good crystalline quality.

Thank you Prof. Gnatenko. This is helpful answer.

As Prof. Gnatenko has mentioned, while taking the PL spectrum, you have to be very sure to filter out the background light, the light from excitation source. This can be performed by adding filters at the output stages. One should always perform the PL measurements in a dark room so that the light from the environment should not affect the measurements. While measuring the PL for thin samples or for the samples with very less emission, you have to use amplifiers at the output stage or you can use APD or PMT as a detector. However, then you have to be very sure that you do not have any background noise otherwise it will also be amplified.

Thank you Kumar, We have added filters at the output. Only thing is that we do not have chopper and lock-in-amplifier.

Perhaps more general answer: lock-in (or synchronous) detection is nothing more than the method to considerably increase the signal/noise ratio. If you do not see the signal, but just the noise – try to reduce noise. Lock-in has very narrow bandpass, so it cancels contribution from the noise sources oscillating at other frequencies. You can use synchronous detection even without lock-in device physically present – the algorithm of a lock-in is simple. What you need: modulation of the excitation laser. If you use solid-state excitation laser (like CNI for example) then it might have modulation input. Put a sin(wt) there from any function generator. The signal from your detector (I suppose that you use a single channel detector with a monochromator, otherwise synchronous detection cannot be applied) multiply by a sin(wt+f), then integrate over at least 10 periods of your modulation function. Finally, the output DC signal will be proportional to the amplitude of the component of your signal oscillating at the given frequency. The components oscillating as sin(nwt) are removed.

Thank you Mr. Drachenko for a comprehensive answer.

I guess lock-in detection helps, and can also allow you to get the exciton lifetime in the sample from the phase shift if relevant modulation frequencies are used. I usually just use steady-state PL, in which case the tricks for low-signal PL are simple: block every other source of light in the room, use the maximum laser intensity that doesn't damage your sample (or cause Auger effects), arrange your optics to collect as much PL as possible, use filters to block the laser wavelength (you may need several identical interference filters) keep the spectrometer input slit wide open (you lose wavelength resolution but at least you get more signal), use the grating with the fewest lines/mm (more photons per pixel on the CCD) and increase the integration time drastically, there's no rule that says you can't acquire a spectrum for several minutes or even 1h.