Hello,

Could a sample-hold process work?

Bertus

 

Hello,

Could a sample-hold process work?

Bertus

I was thinking the same when I asked what the important part of the signal was, he foiled that idea when he said the rise time and decay time was important - I think.

 

If you have data stored in software, the reference and signal are all collected by PD-A, why do you want to add additional noise and uncertainty by introducing a second Photodetector that has to be calibrated vs first detector? Also, the path and cell that contain the sample are not the same. YOur single-path setup is perfect. Now, you just need the software to subtract a reference signal mm editable before it.

If your sample is causing a time delay (fluorescence or phosphorescence), then you can still use a single beam but you need to collect data with a precision time signal as a second stream of data - using that second stream of data to trigger your laser pulses.

Thank you for the comment, GopherT.
In principle, using only single PD should work with a lock-in amplifier. Practically, in usual pump-probe experiment (where signal to noise ratio S/N ~ 10-3), people use a second PD as a reference BG. Then after subtracting this BG from PD-A, a higher S/N can be obtained easily by 10 times. Noise can be reduced very much. This is well-known concept there. But unfortunately, in my experiment, I expect that signal/BG is 10-4 or can be less than that. I think that without subtracting BG properly, the measurement will not be easy. Also I am not sure if BG can be eliminated by the software in my situation as signal/BG is too small. Is it possible ?

About the time delay which can come from the path difference between PD-A and PD-B or sample, it will not be critical. 1 nanosecond corresponds to 30 cm length by the speed of light. And the timing between odd and even is not necessary to be perfect. Maybe ~ 100 nanosecond temporal difference is acceptable.

As you suggested, I am trying to look at the software to see what can be done properly. Because I am not that good at those kind of circuit problem and software as well.
Thank you for the comment, GopherT.

 

That's a great description of your setup.

I think device differences and tiny delays you are trying to measure could be a challenge.

Two possible approaches.

First, have one beam and one sensor. Measure the phase difference between the signal driving the laser and the sensor signal.

Do the measurement twice, one without sample and one without sample. Delta the two.

You will need to control laser brightness. If you don't want to do that, use two beams plus two sensors and Delta the signal from the two sensors. And then swap the light path.

Alternatively, you can use the signal driving the laser to trigger a counter, and use the signal from the sensor to stop the counter.

A two laser two sensor version of this approach is also possible.

You probably need a high speed counter for this.

Thank you for the comment, really. That looks nice to use the counter.
If the error tolerance of temporal shift is about 0.1 us to 1us, can a counter work well ?
And can you give some references for the circuit of the counter ? Because I am a newbie in handling circuits, really.

 

How are you collecting and storing your data from the detector(s)? An off-line manipulation of your data is the normal process. Identify which is a background signal and which is a "sample" signal. Subtract and move on.

Thank you for the kind reply, GopherT.

I think that an off-line data manipulation after collecting them is more or less same to the case of one-beam and one-photodiode scheme which we discussed. It is because the essential of the pump-probe technique is on the simultaneous subtraction of background response between PD A and B. Then this simultaneous subtraction can eliminates noises 'coherently' from various sources like a laser, mechanics, electronics, which are contained nearly equally in two photodiodes. Once data of each PD are collected, the noise source becomes usually incoherent and noises from each PD are not eliminated effectively (it is reduced very much, but noise level is higher by 10 times compared with the simultaneous subtraction scheme). I am afraid so -_-a.
Thank you very much, GopherT for the comment.
Any help/comment will be appreciated.

 

I don't know if this will work for you, but this is the technique that was used in the system that I designed a chip for about twenty years ago (and some of the details are fuzzy after that amount of time).

The image (I forget the size, perhaps 256x256, though it might have been as small as 64x64) stared straight into the probe beam through a region of combusting gases. The pump beam was pulsed at the frame rate, but internally there were two frames per external frame. The first frame integrated the photocurrent onto a sampling capacitor and the second frame integrated the photocurrent off the sampling capacitor. What was left on the sampling capacitor at the end of an external frame was just the difference between the two. We could also run multiple frames without resetting the capacitor in order to average out the noise.

Because each pixel used a single photodiode for both measurements, most of the nonuniformity automatically canceled out. We got rid of most of the rest by flat-fielding the imager using a two-point correction per pixel.