Reverse-Engineering

of an X-ray Imaging Device


Continued from part 2.


Can any of the on-board components be subjugated and put under the control of a different microcontroller? To answer this question, we probably need to take a closer look at the Rad-eye reference design and at the circuit designed by Bioptics (this is the company that designed the sensor, now a subsidiary of Faxitron).

Faxitron DC-44 Sensor Analog Circuit Click on the thumbnail for a pdf version of the circuit diagram.

The dashed block that encloses the high-speed op-amp pair U31 b and c together with the chain of 1.5k resistors and the provision of a potentiometer (R43) is very close to Rad-Icon's reference design. R23 and R43 determine the gain. Potentiometer R43 is shorted on the board, if it was needed, a trace would have to be ripped up. I measured the gain resistor R23 in-circuit and came up with ~130 Ohms. If this is correct, they use an insane gain of 27 (Rad-Icon recommends 5). Of course, in- circuit measurements are always highly inaccurate due to the many paths the current can take.

The Bioptics board can accomodate a four- or eight-sensor configuration, and jumpers on the board (probably) allow to select either configuration. My board is configured for eight Rad-Eye modules. Clearly, Bioptics wanted to save some money by using one ADC for two modules. The ADC is probably the single most expensive component, so here they could probably shave $50 off their material costs. To service two modules alternatingly, analog switches (MAX4616) are provided that connect one of two modules to the amplifier. I did not follow the digital signals, but my guess is that the pixel clock also switches the analog switches in U12: one module during the high phase, the other module during the low phase.

In fact, the module's pixel clock turns out to be 1MHz (Rad-Eye claims that up to 2.5MHz is possible). I assume that each transition of the clock then triggers the ADC. The ADC itself is worth discussing. Bioptics chose a 14-bit device. Rad-Icon suggests a dynamic range of 1:10,000. This dynamic range would be covered with 13 bits, so the 14-th bit is mostly noise? Or perhaps they average several exposures. With n=4 exposures, the 14-th bit would be justified. It would be interesting, in fact, to examine the signal-to-noise ratio of an original module.

Another interesting feature is the module-dependent bias voltage (VAD in Rad-Eye's reference design). A similar analog switch toggles between two buffered digital potentiometers (U27 with U31a and U32 with U31d). Interesting that the buffers are also the expensive high-speed op-amps AD462. A fast step response is hardly required at this point.

Lastly, another analog switch detaches the upper voltage (OUTS) and connects the amplifier to an adjustabe reference voltage (R6, R7, and the analog potentiometer). This does not appear to be a precision reference. There may be two reasons for this switch. First (and actually less likely), it could serve to feed a known reference voltage into the ADC to adjust gain and offset. Second (more likely), this constant voltage services the Rad-Icon specific NDR mode (NDR = non-destructive read). According to the RadEye datasheet, NDR mode causes the differential output pair OUTS-OUTR to be at the same potential and no longer provide a differential signal. If this is the case, the resulting amplifier output voltage would be 3.2V-OUTR, fluctuating from 1.2V (dark) to 0.5V (saturation). By switching the ADC into unipolar mode, the reduced gain is taken into account, and a negative image is created, which can well serve for background subtraction and for auto-exposure. But this is pure conjecture.

On the Bioptics board, this circuit is found four times to service a total of eight modules. It becomes clearer at this point that the best approach is to build a new amplifier/ADC from scratch. In an early idea-collection phase, we decided to use the OPA2353 from the reference design, which has a better slew rate and better SNR than the AD462. For ADC we could choose the ADS805 with 12 bits depth. This is probably adequate for a home-made PCB. The clock could be 2 MHz, which allows for shorter exposure times than the Bioptics device allows. Stretching the integration time is always possible anyway.


Since we are in the process of examining the Faxitron device anyway, let's take a look at some other details. The x-ray tube is buried under three layers of metal: The outer shell, the driver cover, and the cover for the tube compartment. Warning labels are supposed to discourage disassembly, but this has just raised my curiosity.

Removing all covers reveals a TruFocus TFX-8050 microfocus tube. This is an interesting fact as TruFocus, Inc. in Watsonville, CA, still exists. The 8050 is out of production, but TruFocus sells the TFX-8100, which appears as if it could be used as a drop-in replacement tube. The tube has an 8 micron focal spot, can deliver up to 50kVp (the Faxitron is limited to 35kVp, though) and has a beam angle of approximately 61 degrees. The target is tungsten, and the output window is made of beryllium.

Inside the outer x-ray generator housing, we find a control board (not further examined) and the high-voltage source. The source, including three modules on the control board, are manufactured by Gamma High Voltage Research, Inc. in Ormond Beach, FL. The HV generator is labeled XR15-35P/M530A, where 15-35 probably indicates the voltage range, 15kV to 35kV. A view inside the generator housing is shown in Figure 6.

Figure 6: View of the x-ray tube, high-voltage generator and control units.

Of moderate interest are the control board and power supply. The control board is shown in Figure 7. It features an 80C55 microprocessor, connectors to the front panel elements, and an elaborate interlock system. Apparently, a RS-232 interface exists on the control board, but it is not connected. Instead, what looks like a RS-232 connection is routed to a header on the control boad that is labeled "camera". Since this connector is not used by the RadIcon camera, we can speculate that an external camera was once offered as an option, but discontinued in favor of the integrated RadIcon sensor.

Figure 7: Photo of the main control board.

The power supply provides +5V at 40A, +/-12V at 10A and 6A, respectively, and +24V at 3A. I did not investigate any deeper, but it appears as if the 24V supply is feeding the HV generator as this is the source that is shut off by the door interlock. There is a double interlock. Closing the door closes a pair of switches, which in turn closes a 12V circuit on the controller board. Conversely, opening the door feeds 12V to a pair of aiming lasers that indicate the center of the sensor module. In addition, the door needs to bridge a pair of pads to close a 24V circuit, without which the x-ray generator does not fire. These pads appear to be somewhat susceptible to dirt and wear, so if you have a nonfunctional unit, make sure to check the interlocks.

A closer look at the connections between the main control board and other parts of the device is given in Figure 8. The main board connects to a 24-pin connector, from which cables run to various parts of the device. A saparate 2-pin connector provides 5V for the camera and 12V for the fan. The interlock switches are indicated in Figure 8, and the interlocks also control the aiming lasers, turning them on when the door is open. The remaining signals are widely unknown at this time. The wires indicated in red (E-M) connect a 9-pin sub-D connector that is accessible at the back of the device to the header labeled "camera". The wires indicated in blue (N-Q, S, T, V, W) begin on the main board at the header labeled "AEC" (automatic exposure control?) and end in a large connector labeled J3.

Figure 8: Pinout of main board connectors. A 24-pin connector connects the Faxitron main board to the interlock system and two connectors. Separate power connectors provide 12V for the fan and 5V for the camera.

For completeness, the following table relates the pins of the 24-pin connector (red) to the sub-D connector (I not connected):

1 I   2 H   3 G
4 F   5 E   6 M
7 L   8 K   9 J


Enough analysis for now, let's get to the design task at hand. On to Part 4.