How do I reduce noise of the output voltage?

[cks]

yellow line: output of photodiode which detects an LED
blue line: output voltage to the LED


newfile1: the file without measuring using average of the scope.

newfile6: the file with using average of 32 times of the scope.

How to filter the noise? It's noisy, I find it's ok to use average of the scope to get the best pattern. but I prefer it to use other method. I try to use bypass capacitor, but it doesn't work. quite disappointed.

 

[cks]

Later, I tried the RC circuit and a lot of noises have been depleted.

 

[Redbelly98]

A capacitor should work if you use an appropriate value of capacitance.

To find that value, we need to know how much resistance the photodiode is connected to?

I'll guestimate that those signal spike are about 1 us in duration (but you could verify that by expanding the time scale). So you'd want a capacitance where

RC > [spike duration]
or
C > [spike duration] / R

Moreover: the waveform has a 500 us half-period. The capacitor can't be so large as to noticeably distort the square wave. So:

RC < (1/20) * [500 us]
or
C < [500 us] / (20R)
C < [25 us] / R

So, for example, if those spike are about 1 us duration, choose a capacitor in between
[1 us]/R and [25 us]/R.

 

[cks]

Thank you Redbelly98, I understand what you mean and gain some other ideas of how to filter the noise.

Before I read your reply, I already tried out my new RC filter, and it works perfectly well.
I didn't design the RC filter by looking at the duration of the spike as mentioned by you and the period of the wavelength.

I designed the RC filter by looking at the time taken, T for the change from the positive voltage to the negative voltage (because my photodiode detects LED which is supplied by square wave, so the photodiode is supposed to detect square wave) (obtained from the photodiode without RC filter). To let you understand what I mean, I attach a bmp file for you to look at.

The time constant of RC filter should be rougly about T, which from the graph shown, is about micro seconds. (actually, I want to zoom in but too much noise, I can't see the change properly).

 

[cks]

After putting the RC filter,


there is a slant instead of an even straight line at the peak, which is due to the exponential decay of the RC filter, I didn't really realize this until I read your reply(the other question). It deserves a bit of attention when I design it next time,


But for my photodiode which is used to see the absorption profile of Rubidium, the range use is 50mV / div, so I guess this RC filter is ok.

Thank you.

 

[cks]

From my experience together with your answer,

I think there are 3 parts that need attention when we want to filter the noise,

Like what you said,
1) the half-period of the wave
2) the duration of the time for the spike.

what I observed,
3) the time response of the photodiode = change from +ve to -ve

 

[Redbelly98]

After putting the RC filter,

there is a slant instead of an even straight line at the peak, which is due to the exponential decay of the RC filter, I didn't really realize this until I read your reply(the other question). It deserves a bit of attention when I design it next time,

That slant makes me think there is a high-pass filter in there somewhere, filtering out DC and low frequencies. Apart from that, things are looking pretty good now.

Can you post a schematic of your photodiode circuit, including the RC filter?

 

[cks]

For my photodiode circuit,
http://sales.hamamatsu.com/assets/applications/SSD/si_pd_circuit_kspd1043e06.pdf
the 5th schematic design.


My RC circuit,
http://en.wikipedia.org/wiki/Image:Series-RC.svg

with capacitance = 100nF and resistance = 300Ohms

 

[cks]

I'm not sure why do you say there is a high-pass filter somewhere?

 

[berkeman]

I'm not sure why do you say there is a high-pass filter somewhere?

They're saying it because of the droop in your 2nd BMP waveform file. If it were only the result of a LPF, and the original signal did not have the droop, then that implies some capacitive coupling in series with the signal in the filter, and hence the HPF.

 

[Redbelly98]

For my photodiode circuit,
http://sales.hamamatsu.com/assets/applications/SSD/si_pd_circuit_kspd1043e06.pdf
the 5th schematic design.


My RC circuit,
http://en.wikipedia.org/wiki/Image:Series-RC.svg

with capacitance = 100nF and resistance = 300Ohms

Looks good, though I don't know where the droop comes from in this case. Try setting the 'scope to DC mode and see if that changes things appreciably. If not, then it would seem to be a real feature of the signal, not showing up when you had a 1 msec period but showing up in this 10 msec period signal.

The 30 usec time constant looks like a good value, filtering out the noise spikes yet not noticeably distorting the 1 to 10 msec period waveform. If you wanted, you could also try a 30 nF cap (for 10 usec time constant) but you could also proceed with what you already have.