Figure: Sample Voltmeter trace of the Eberline output voltage. Note the plateau region from t=1.3 to t=1.9 seconds.
The Eberline Pic-6A and Pic-6B are battery powered, handheld radiation detectors consisting of an ion chamber and an analog meter with logarithmic output. Each reads output from 1 millirad to 1000 millirad or 1 rad to 1000 rads; a rotary switch powers the device and toggles the operating range. The two meters only differ cosmetically and can be used interchangeably for our purposes, and references to "the Eberline" in this document refer to either.
The Eberline has an internal analog voltage that is proportional to the logarithm of the radiation dose. This output is predictable across six orders of magnitude and is easily read and transmitted via existing LWDAQ Voltmeter equipment. The Voltmeter's daq_source_commands field allows us to remotely fire our pulsed dental radiation source. We store the resulting voltage and plug it into a formula outputting absorbed radiation dose, in rads.
Every ion-chamber based radiation detector that we have observed has three potentially limiting factors: range of detection, response time, and spectrum dependence. The Eberline Pic-6A provides advantages over newer sensors when testing x-ray emissions in the quantities we require.
The Eberline has a specified saturation limit of 1000 rads per hour, or 270 millirads per second. Other hand-held radiation sensors we have observed saturate at much lower levels. For example, the Inovision 451p used by Brandeis for safety inspection has a manufacturer specified maximum reading of 5 rads per hour, or 1.4 millirads per second. In our radiation testing lab, we use a Heliodent dental x-ray to irradiate our circuits in 2-second bursts that exceed 150 millirads per second at a range of 50 centimeters. Our initial attempts to use the Inovision 451p to verify that output were unsuccessful, as the device does not operate properly at those levels.
The 451p and the Eberline both have limitations of response time. The 451p has a stated response time of 2.7 seconds from background radiation to 4 rads per hour. So, in a Heliodent short pulse exceeding 500 rads per hour, the Inovision will not give an accurate output regardless of saturation levels.
The Eberline's documentation reports its response time on the rads per hour setting to be "essentially instantaneous", which is inconsistent with our observations with the dental source. By reading the internal voltages with the LWDAQ Voltmeter, we can directly observe the response time, and verify that the circuit reached equilibrium during the x-ray emission.
The Eberline uses a charge pump to step up the voltage from 9V to approximately 3000V. That voltage is connected to ground via 1000-megaohm (R14) and 1-megaohm (R15) resistors in series. The voltage across the 1-megaohm resistor, approximately equal to the high voltage divided by 1000, is used to drive the gate of an n-channel JFET (Q4).
We have modified an A2057 to both power the Eberline and measure the voltage at the gate of the JFET via 3.5-meter twisted pair wires. We have observed that this voltage rests at slightly over 3.7 volts and drops approximately 97 mV for each factor of two increase in the absorbed dose.
The Voltmeter instrument records a series of discrete voltages. For the purposes of our experiments, the Toolmaker script sets the Voltmeter to trigger when the Eberline's output voltage (R16) drops below a threshold voltage (default = 3.6V). It then records a certain number of samples (default = 1200 samples) over a length of time defined by user input (default = 2 seconds). The toolmaker script then isolates a subset of the samples and averages them. This voltage is used to calculate the radiation dose.
To account for the time delay of the circuit, the Toolmaker script disregards all readings above the threshold voltage and creates a Tcl list containing the remaining samples. It truncates the last 50 samples from the list and averages the 50 samples previous to that. This allows the script to isolate the voltage plateau that occurs when the ion chamber circuit has adjusted to the new radiation level, as seen below.
For either radiation range, it is this voltage that is proportional to the logarithm of the radiation dose. We observe that when the distance between source and sensor is doubled, the voltage drops by 198 +/- 6 millivolts. From this, the radiation can be converted from voltage to millirads with the formula:
where A is determined by the resting voltage of the sensor, and the preferred units of absorbed dose.
Every ionization chamber will have a different response to x-rays of different energies, due to the design of the chamber itself and surrounding housing. Any radiation sensor will have a manufacturer reported responsivity graph, generally from 1 keV to 100 keV. Response curves for the three radiation sensors we use in-house are shown below:
We lacked a verifiable version of the Heliodent X-Ray source's energy spectrum. We used a simple method of repeated attenuation to approximate the output of the device over discrete energy levels. Radiation output from the radcal has an approximately constant energy response for photons from 1 keV to 100 keV (see above). Radcal samples were taken at constant exposure and distance using machined squares of aluminum of area density of 1.8 g/cm3 to reliably attenuate the output by a constant factor. Using the NIST's extensive mass attenuation data, we set up the radiation output (R0...Rn) at each level of attenuation as a linear system with the attenuation factor of aluminum for each photon energy level (a - g).
The resulting energy levels (E1...E7) represent a first approximation of our x-ray spectrum. The aluminum plates are relatively thick, and provide little information about the low-energy x-rays below 15 keV. More discussion of the energy spectrum of the Kevex DC source can be found here.
Subsequent observations corrorborated this spectrum. A modified A2075 was placed backwards during radiation testing, and survived approximately twice as long as a comparable circuit without the attenuation provided by the FR4 laminate of the circuit board. Using NIST values for a serviceable analog (Pyrex glass), our method of discretizing the spectrum predicted a 44% transmission (R1 = 0.44 * R0) of X-ray radiation through one printed circuit board. A transmission of 40% was observed.
We have observed the standard deviation of the voltage measurement of the Eberline sensor to be 4 millivolts. This is equivalent to saying the resolution of our measurement of dose with the Eberline sensor is +/-3%.
We have observed a periodic variation in the Eberline measurement that shifts the average voltage by 7 millivolts. This is equivalent to a multiplicative factor of 1.05. While this increases the standard deviation, we cannot separate the two errors, so we account for them independently. So, we claim the precision of our Eberline measurement is +/-8% on the ranges we have observed.