EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: sgt_bear on December 07, 2017, 07:39:11 am
-
Hello
I'm having a input from a silicon photomultiplier. My goal is to bring the pulse output to a "known" form and time constant.
See attached a image of the signals (persistant screen). How can i transform the pulses to a fixed time, and a gaussian shape?
My ouptut should be a pulse that is always X- seconds long, and only varies in height. The ouptut pulse height should be in relation to input pulse height.
Thank you for your help
-
Your attachment is 0kB file
-
Here is the correct attachment
-
How can i transform the pulses to a fixed time, and a gaussian shape?
Why gaussian? You could just amplify it and use a low pass filter.
Or combine a peak detector circuit, some treshold detector and a timer and you'll get a nice square wave signal.
-
A gaussian shape is nice for calculations.
The problem here is, that an ADC has to read the height of the pulse. I agree that a Square-Wave Signal is a nice idea, but it has the disadvantage that you need a very fast ADC with high resolution in bits the measure 5000 different voltages here. Devices >50MSPS @ 16bits are very expensive.
In case of a Gaussian shape you can use a lower resolution ADC, e.g. 8bit, and then calculate the height based on geometry. For example, if you have a sine wave with fixed frequency, you can calculate its amplitude by only having 4 samples. 2 in each rise/and fall.
-
The problem here is, that an ADC has to read the height of the pulse. I agree that a Square-Wave Signal is a nice idea, but it has the disadvantage that you need a very fast ADC with high resolution in bits the measure 5000 different voltages here. Devices >50MSPS @ 16bits are very expensive.
Uh, no. If you use a sample and hold and a peak detector circuit you can "hold" (stretch in time) the pulse for as long a time as you want.
-
The standard way seems to be a CR filter followed by multiple RC filters.
-
If you want the area of the pulse, why not integrate it with an op amp? Then you can read the voltage with a slow ADC and reset the integrator. That eliminates any restriction on the pulse shape. Really fast op amps are cheap relative to ADCs of similar speed.
I'd suggest spending some time with "The Art of Electronics" by Horowitz and Hill. The 3rd edition came out recently and is quite marvelous. It was written to provide guidance to people doing things such as your project.
-
The problem here is, that an ADC has to read the height of the pulse. I agree that a Square-Wave Signal is a nice idea, but it has the disadvantage that you need a very fast ADC with high resolution in bits the measure 5000 different voltages here. Devices >50MSPS @ 16bits are very expensive.
Uh, no. If you use a sample and hold and a peak detector circuit you can "hold" (stretch in time) the pulse for as long a time as you want.
Theoretically yes, practically no. In the time you are holding the pulse, you miss new pulses comming in..
All commercial applications uses a gaussian shape. If we are looking at "Typical applications" image, there is a pulse like mine (only its positve) and then gets shaped to gaussian.
https://www.fastcomtec.com/index.php?id=172 (https://www.fastcomtec.com/index.php?id=172)
They provided a schematic, but sadly they removed the values :(
-
How can i transform the pulses to a fixed time, and a gaussian shape?
Why not just buy a commercial product? It sounds like there is a suitable one here: https://www.fastcomtec.com/index.php?id=172 (https://www.fastcomtec.com/index.php?id=172)
The second paragraph in https://entertaininghacks.wordpress.com/library-2/good-questions-pique-our-interest-and-dont-waste-our-time-2/ (https://entertaininghacks.wordpress.com/library-2/good-questions-pique-our-interest-and-dont-waste-our-time-2/) might help guide everybody in this conversation.
-
In a hobbiyst environment it is hard to keep up with the prices of commercial products. Commercial sellers sell products that have gone through several tests and specific norms to comply which makes them expensive. Also the company selling them has to give a certain guarantee on the products etc.
The prices of these products are mostly way above hobbist budgets.... thats the reason why a commercial product is not always an option
-
They provided a schematic, but sadly they removed the values :(
You should be able to determine the component values using Octave or MATLAB. The circuit is a simple low pass filter. I'm sure the actual device is more complex.
However, what you want is:
B = G/A
where:
A is the Fourier transform of the output pulse
B is the Fourier transform of the impulse response of the desired pulse shaper
G is the Fourier transform of the desired output pulse
That will show you what frequencies you need to amplify and in turn the various component values for the Sallen-Key filters.
The Fourier transform of a narrow Gaussian in time is a broad Gaussian in frequency. The big limitation is how broadband an op amp are you prepared to pay for. The UPC2710 suggested by H&H is a ~$2 from Digikey.
Dump a bunch of pulse outputs to CSV format and post them. I won't make any promises, but it's an interesting question. I'm an instrumentation junkie, so I may well take a poke or two at it. Meanwhile familiarize yourself with Octave. It's free and very capable. And watch out for divide by zero in the equation above.
You should also look at the "Art of Electronics" by H & H. It includes a couple of photomultiplier circuits (cf pp 842-843) as well as an excellent section on filter design. The first circuit suggests using a 1 GHz BW amp such as an NEC UPC2710.
The filter you want is a low pass filter with a skirt that is approximately Gaussian. The transformation is apply gain and then filter. You need an integrator to provide the low frequency content of a Gaussian. A rough guess is that you'll wind up with a cascade of several filters with decreasing bandwidth.
It may well be that your current amplifier circuit is not broadband enough. I should also note that a Gaussian is a zero phase wavelet. That is not physically possible unless there is a delay. The real world is minimum phase. Systems do not produce output before T0.