Close

Designing and simulating the final version

A project log for Gamma PIN - Semiconductor Radiation Detector

An attempt to design and deliver universal and accessible radiaton analyser to be used in nuclear spectroscopy

Marcin WachowiakMarcin Wachowiak 05/03/2019 at 19:420 Comments

After a long time spend testing and thinking about the right solution to all the problems, I have made multiple alterations to the current prototype. Some of them require a different design, so the new board including these signal processing units should be produced soon.

Noise figures

Almost all of the simulations I performed in LTspice were noiseless. The output signals were neat and easy to process. To consider a more real case I added a white noise signal to the output of the preamplifier so that it resembles 60keV pulses. After picking the right values I ended up with pretty dense noise of 8 mVpp and exponentially falling pulses of amplitude 2.5mV below noise floor. Such subtle changes rendered the old simulation completely useless. The old smooth signal was replaced with spiky and indistinguishable fluctuations. In the old version of the circuitry you could easily see that the peak value was repeatable, now it will be a lot harder.

Exemplary real case pulse (notice bandwidth limitation to 20MHz, reconstructed pulse has proper noise value)
Reconstructed pulse in LTspice

Updated approach to signal processing

Because of all that noise present in the signal I had to redesign the whole signal path. It will no longer have any subtractors or rectifiers which would have to be set individually. Instead there will be some new blocks, which will help to extract and measure the peak from the ever-present noise. Here is a detailed schematic:

Schematic

 Below the analog signal processing path is the semi-digital segment responsible for pulse detection:

Comparators response signals

This double comparator block job is to detect pile up and send trigger signal to the MCU. If there appear at least two rising slope pulses while pulse duration is still triggered the MCU will not measure such pulse, as its height certainly was affected by occurring pileup. The exact values of components in the third prototype have to be yet properly calculated and checked. Voltage sources present in the simulation are for ease of altering the parameters and will be replaced by voltage dividers stabilized with a capacitor in the practial design

Detector resolution with noise

Observing the output signals that will be later sampled by an ADC it is easy to notice that even the same energy signals end up represented by different voltage levels. It is due to significant noise amplitude that affects the actual peak value. Even with the low-pass filters the fluctuations in the signal remain and error is still present. This causes the spectrum of certain energy to be spread and not visible as a single bar in the graph. The main source of the noise are PIN diodes and the preamplifier. By changing the detector to a photomultiplier tube with scintillator I might be able to improve the resolution, but that Is yet to be tested. Currently the circuitry of the detector will be tuned to minimize the noise impact on the pulse readout, having in mind that semiconductor detectors have their limits.

One way to minimize the distortion in the pulse is to increase the feedback resistance in order to extend the pulse duration and eliminate the high domain frequencies using a low-pass filter. It proves itself effective as far as simulation is concerned. On the one hand it reduces the amplitude spread and noise distortion, but on the other increases dead time of the detector. The detector is also extremely sensitive to electromagnetic interference and RF signal coming from the radio stations and BTS. Radiation pulses are in low frequency domain and in order to measure them without greater distortions a low pass filter is absolutely crucial.

FFT of preamplifier output signal

Preamplifier with LMP7721 test

To check if a lower GBW and lower noise operational amplifier would improve the TIA stage I performed a few tests with LM7721. The presets were similar to the 1 diode tests with bootstrap and LTC6268. The noise frequency was lower due to the bandwidth limitation as LMP7721 GBW is only 17MHz. The overall amplitude of the pulse was about 2-2.5mV below noise floor which was quite the same in comparison to LTC6268. The parameters have not changed in any favorable way. The TIA design will remain with LTC6268.

LMP7721 single radiation pulse

LMP7721 radiation pulse collection

Triple amplifier tests

To verify how simple amplification affects noisy signal I mounted amplifiers in the prototype and set overall gain to about 250V/V. From the screens taken you might see that the dense noise is not amplified because of the limited GBW of used amplifier (LT6232). Furthermore, there is an undershoot visible as amplifiers are connected through a capacitor. On the whole, even common amplifiers provide some kind of filtering that removes the high frequency noise. Sometimes the simulation in LTspice may provide too ideal data, and signals resembled without including the imperfections of the components, design, amplifiers or parasitic effects. I hope that some imperfections may actually improve the signal to noise ratio, but that will be verified in the third detector design.

Pulse after 3 stage amplification

Pulse after 3 stage amplification

Noise amplitude comparison top: 3x amplification bottom: TIA output

Discussions