Ionizing radiation creates ion-electron pairs as it passes through neutrally charged matter. If this matter is air, you can pass a current through the ionized air and measure the relative intensity of the radiation. Typical smoke detectors use ionization chambers containing Americium-241 sources. The air in the chamber is ionized by the aplha particles produced by the Am-241 which allows for a small current to flow. When smoke enters the chamber, that current is changed which triggers the alarm. Ionization chambers are a common tool used in radiation measurement and detection, and like most scientific instruments are quite expensive off the shelf. As part of my job, I occasionally test electronic components in proton beams to assess their suitability for use in low earth orbit (the Van Allen belts are full of trapped electrons and protons), so I decided to see if I could build an ionization chamber to record exactly when the proton beam is turned on and off along with its relative intensity.
The currents produced by ionization chambers are quite small (pA to nA range), so the real challenge isn't building the ionization chamber itself since something as simple as a coffee can makes a fine ionization chamber; the real challenge is designing a circuit to measure that current which has good linearity, low offset, and reasonable dynamic range.
My goal for this project was to only use parts available from an electronics distributer like Digi-key, and that included the ionization chamber itself. I did not want to fiddle with buidling the ioniation chambers because I wanted them to have repetable performance. I considered a number options inclusing EMI shield cans (and I actually initially started to design a photodiode based circuit), but after coming across this blurb in the Littelfuse Gas Discharge Tube product guide, I realized that gas discharge tubes would likely work in this application: "For voltages below the breakdown voltage, the gas provides a good insulator. Very low leakage currents (10-12A) occasionally encountered result from ionization by cosmic rays, high energy photons, etc"
TDK B88069X5641T902 1.2kV GDT next to a Bourns 2093-300-SM-RPLF 3kV GDT
Gas discharge tubes are a type of surge arrestor. They have very low paracitic capacitance and very low leakage (under normal conditions) because they are fundamentally just air gaps. Large transient voltages cause these tubes to arc over into a low impefance state where they can hangle extremely large transient currents. The gas inside the tubes is typically neon, argon, or a neon argon mixture and it is typically at less than atmospheric pressure. The tubes themselves are hermetically sealed with glass or ceramic bodies sealed to the metal ends with good quality gas to metal seals.
TDK B88069X5641T902 1.2kV GDT and Bourns 2093-300-SM-RPLF 3kV GDT Broken Open
The physical characteristics of the tubes along with gas pressure and electric field determine the current that will be seen for a given amount of ionizing radiation. With higher gas pressure, more ions-electron pairs will be generated in a given volume by a high energy charged particle. The greater the volume of gas, the more electron-ion pairs will be formed, but the electric field must be sufficiently strong to collect the ions and electrons before they recombine through random collisions. Thus, a tube with greater gas volume due t a greater air gap may produce a lower current than a tube with a smaller air gap if the applied voltage is the same. Ideally, an ionization chamber is operated in the saturation region; that is to say that the applied electric field is sufficient to collect effectively all of the ion-electron pairs. Applying more voltage will then not increase the current through the tube until the applied voltage is so great that gas multiplication results in a single electron-ion pairs cascading into many more.
Electric Field vs. Pulse Amplitude - Knoll Radiation Detection and Measurement
The above plot is for proportional counters which provide a pulse response to single particle events, but the saturation region plateau is the same ofr ionization chambers. I did by first tests with a TDK B88069X5641T902 1.2kV GDT and had pretty good results. I then tried a Bourns 2093-300-SM-RPLF 3kV GDT, but it produced significantly less current at the same flux with the meger 6.5V bias voltage (measurement circuit explained in later sections) I applied to it. I found that it had resistive leakage that was significantly higher than the ERCOS tube as well, so I did not attempt to characterize it at higher bias voltages. The resistive leakage of the TDK tube, on the other hand, was below the my measurement limit (100fA) even with a 600V bias. Greater sensitivity may be achieved with different tubes, but I have found that B88069X5641T902 is well suited for my application.
Applied Voltage vs. Measured Current at Different Proton Flux Levels
The above plot shows my actual current measurements with this tube under a three different flux levels. The tube maintained reasonable linearity even at a 6.5V bias so while it is certainly more sensitive at higher bias voltages it is entirely workable to use this tube to measure proton flux with only a 6.5V bias.
Multifunction devices capable of measuring very small currents, very high resistances, and charge are called elecrometers. At a typical particle accelerator, you will find at least one electrometer set up to measure the charge accumulated by an ionization chamber. I am exclusively interested in instentanuous current, so it is just the picoammeter functionality of an electrometer that I need to replicate. There are three general types of current sensors with the necessary sensitivity to measure picoamp and nanoamp range currents:
Ordinary shunt current sensors will work, but there are major downsides. If you place a 1Mohm resistor in series with a 1pA current source, you'll only get 1uV. Up that to 1Gohm, and you'll get 1mV. That is certainly a voltage that a sensitive amplifier can deal with (ignoring input impedance and bias issues discussed later), but you run into problems when you also need to sense larger currents. A 1nA current produces 1V across the shunt which means that the voltage across the DUT is 1V lower which can impact the measurement. This voltage is known as the burden voltage.
The solution to this problem is to place the resistor in the feedback path of an OP-AMP. For a given resistance, current to voltage gain is the same as with a shunt current sensor, just inverted because the OP-AMP must lower its output voltage in order to keep its inverting input at a constant voltage despite current being injected into that node by the DUT. Because the OP-AMP keeps its input voltage constant, the voltage across the DUT can be held constant. Feedback ammeter is a term used in the Keithly Low Level Measurements Handbook, but a more common term, at least in my experience, is Transimpedance Amplifier (TIA). This type of circuit is often used as a first stage amplifier in photodiode circuits.
Coulombmeters measure electric charge, not current, but they can act as discrete time current sensors and can provide very good resolution. These can also be implemented as shunt and feedback circuits: shunt coulombmeters monitor the voltage produced by the DUT charging a capacitor, and feedback Coulombmeters place that capacitor in the feedback network of an OP-AMP. The problem with Coulombmeters is that they must be reset after their amplifiers rail out when their output reaches a supply rail. Any reset circuit connected to the input of the amplifier must have extremely low leakage in the off state. For reasonable performance, the reset circuit must use an electromechanical relay. "Low leakage" MOSFETs still have leakage currents >10nA with only a volt or two drain to source, so reed relays are typically used in sensitive electrometer applications for resetting integration capacitors and for reconfiguring circuitry and changing first stage gain values.
There are three main contributers to DC error in transimpedance amplifiers: feedback resistor tollerance, OP-AMP input bias current, and OP-AMP input offset voltage. I found that LMP2231 had the best balance of these parameters for this application despite its relatively low cost ($1.70).
1Gohm is the largest value resistor that can be purchased for less than $1 with a 1% tollerance:
A 10Gohm resistor would provide even more sensitivity, but it would add cost.
Input bias current is the parasitic current that flows into or out of the input of an OP-AMP. Zero current flows into the inputs of an ideal OP-AMP, and OP-AMPs can generally be treated as ideal in lower impedance circuits, but wherever there is a voltage potential in real circuits some leakage current will flow. In a CMOS OP-AMP, most of this current is from the ESD protection diode structures. Input bis current is common mode voltage dependent because the effective input bias current is the sum of the leakage from the top side and bottom side ESD diodes. With identical ESD diodes, you would expect the common mode voltage that corresponds to zero leakage to be at half the supply voltage since the reverse voltage across each ESD diode would be the same. Real parts have variation, so while you can reduce input bias current by adjusting your common mode voltage, you cannot eliminate it.
Input Bias Current
Since input bias current effects both inputs to an OP-AMP, much of its effect can be cancelled out by connecting both inputs to nodes of similar impedance. What cannot be compensated for with this method is input offset current -- the difference between the input bias currents of the two inputs. If both inputs are sinking 1pA with a 1Gohm source impedance, the voltage at both input nodes will be 1mV lower cancelling out the effect. If one input is sinking 1pA and the other is sinking 1.1pA, then there will be a 100uV difference in voltage at the two pins. Assuming a TIA circuit, the output voltage would then have a 100uV offset.
LMP2231 has typical input bias current of 0.02pA and a typical input offset current of 0.005pA. Those paramters only hold at 25C die temperature and with a common mode input voltage of half the supply. The maximum input bias current (inclusive of all input common mode voltage and operating temperatures) is 2500x higher: 50pA. From the performance characteristics plots, you can see how common mode voltage really impacts input bias current:
LMP2231 Input Bias Current vs. Common Mode Voltage Over Temperature
With a common mode voltage of 0V (as you would have with a typical shunt ammeter in a system without negative supply rails), the input bias current at 25C can be more than 0.5pA which is 25x the value of the 'typical' input bias current. Thus the feedback ammeter topology has significant benefits beyond just eliminating burden voltage since it inherently maintains a constant non-zero commn mode voltage. For this particular amplifier, the best common mode voltage for operation over a wide temperature range is 2 Volts. Input offset current for this device is so low that it was likely impractical to characterize it, but assuming that input offset can be modeled as a ratio of input bias current added to a constant offset, it should be minimized at the minimum input bias current point as well.
I did not bother matching the input impedances for by initial circuit, and at room temperature for how I use the sensor it is not really required. The output voltage with only background radiation the output voltage is 0.0mV which is as low as my Fluke multimeters go. If this sensor were to be used over a wide range of temperatures, it would be wise to match input impedances, though. The theoretical output offset resulting input offset current in a matched impedance circuit is equal to the input offset current multiplied by the feedback resistance. So for a 0.005pA offset and a 1 Gohm feedback resistor, the typical offset should be 5uV which is of course far below the 100uV sensitivity of my multimeter.
Input offset voltage is the difference in voltage that can be measured between the two inputs of an OP-AMP. This offset is the result of mismatch in the input transistors of the OP-AMP and the circuits that bias them. Input offset is a function of temperature, but the nominal value can varry significantly on a part to part basis. The temperature coefficient also varries siginificantly between parts: LMP2231A is screened to have a maximum temperature dependent drift in input offset voltage of +-0.4uV/C while LMP2231B may have a value as high as +-2.5uV/C. Since I am using this part at room temperature, I opted for the cheaper LMP2231B.
LMP2231 Input Offset Voltage Distribution
In a proper electrometer, input offset voltage is compensated for in a calibration step, but temperature dependent drift can still cause issues if the electrometer warms up during operation without recalibration.
LMP2231 Input Offset Voltage vs. Common Mode Voltage Over Temperature
As with input bias current, there is a sweet spot in common mode voltage for input offset voltage. In this case, it is around 3V. This voltage is a bit higher than the sweet spot for bias current, so the ideal common mode voltage depends on the range of temperatues over which the detector needs to operate. I elected to use a 2.5V common mode voltage because it is exactly half rail and is an average of the ideal common mode voltage for input bias current and ideal common mode voltage for input offset voltage.
I need to test with the gain of 1000 amplifier before saying anything here.