Professors: R. Nelson J. Larue Mentors: Allen Kine George Horansky Faruk Leric Jr. Brian Gorda Eugine Rubin March 19, 2007 1 Low-Noise Photodiode Simulation Writeup With their low-input currents, FET input operational amplifiers are quite commonly used in monitoring photodetectors. The most common of these configurations are usually applied to photodiodes. The photodetector bridges the gap between a physical phenomenon, such as light in our case, and the electronics used to detect and interpret these phenomena. Selection of the appropriate amplifier architecture depends on ones choice of noise, bandwidth, offset, and linearity. A traditional topology that is quite commonly used in photodetection is as follows: 1 Here the photodiode is in series with the input of an operational amplifier where ideally zero current flows. R IN and C IN are used to reduce DC and AC errors caused by input bias currents. C IN shunts signal so that the noise inverting input approximates AC ground. However, at the same time R IN also increases noise at the output by a factor of 1 Bonnie C. Baker, Comparison of Noise Performance Between FET Transimpedance Amplifier and A Switched Integrator, Burr Brown, 1993 2 !4KTRB times the noise gain of the circuit. The op-amp has feedback set by R 2 . This resistance in a current-to-voltage converter such as the one presented above, largely determines noise and bandwidth as well as gain. The spectral density of the resistor is given by !4KTR, excluding non-ideal resistor behavior, and appears directly at the output of a current-to-voltage converter. As we can see from the mathematical relationship, increasing R not only increases output noise, but it also increases the output signal by a direct proportionality. Consequently, signal-to-noise ratio tends to increase by a square- root relationship. Noise from the operational amplifier also influences the output, with an additional effect introduced by the high feedback resistance R 2 as well as the capacitance C 2. Amplifier noise sources are usually modeled by the input noise current, i n , and the input noise voltage e n shown as follows: 2 The current noise i n flows through the feedback resistor R experiencing the same gain as the signal current I p . The DC noise voltage gain is given by 1+R/R D . This gain is 2 Photodiode Monitoring With Op Amps, Burr Brown, January, 1995 3 kept small by the high diode resistance R D . However, at higher frequencies the above the current bandwidth capacitance large values C D of the photodiode compared to the feedback capacitor starts having a significant effect on the noise gain. This in turn influences the gain of e n in a proportional manner. Since the feedback resistance is commonly made large, the effect can take place at relatively lower frequencies for small values of C in parallel with the parasitic capacitance. The noise gain gradually rises and eventually terminates at the intersection of given bandwidth curve. The feedback capacitance C, shunts the feedback resistance R and results in a pole leveling response at 1+C D /C. For large area diodes, the capacitance C D can be in the picofarad range. This could have a detrimental effect since the noise gain could be amplified by a factor of a hundred as well. As such, gain peaking phenomena are a common side effect of high- feedback op amp resistors and need to be carefully studied to maintain good behavioral characteristics of the circuit. The effects of not properly selecting the feedback capacitance can result in overshoot, response peaking, poor settling and quite possibly oscillation. All of these effects are highly undesirable and need to be accounted for in the architecture and design of the proper circuit topology. Besides the DC and transient performance of the circuit, which is a function of the overshoot, settling time and oscillation, we must also account for the noise characteristics of the circuit. In understanding the current-to-voltage noise performance, it is important to recognize that the signal current and noise voltage encounter different frequency responses. The current-to-voltage is flat with frequency until the feedback impedance is rolled off by the stray capacitance. The majority of the op amps bandwidth quite commonly serves to amplify the noise error and not the signal. This is also typically the 4 dominant source of noise for high feedback resistances. A typical noise versus feedback resistance plot will usually behave in the following manner: 3 Upon inspection of the plot, we note that as the feedback resistance of a current-to- voltage converter increases, the dominant noise source changes from the op amp to the resistor and back to the op amp under gain peaking conditions. These effects are shown for several op amps. Nameley, the OPA 404, the OPA 128 which we will utilize in our design, and the OPA 111/OPA2111. The OPA 128 is a highly desirable op amp in our case because it exhibits low noise characteristics in the middle range that is due to the amplifiers low bandwidth that is 1 MHz (typically). It also excels in its very low DC error characteristics as its input currents are a mere 0.075pA. Thus it is also a very low input current device that we can take advantage of in our low noise amplifier design. To understand the detrimental effects of noise even further, it is important to give a brief overview of the governing principles behind low noise amplifier systems. The first stage of amplifier design is the most prone of all to noise disturbance because noise and signal are related at the input in an application. Noise when referred to the input is 3 Photodiode Monitoring With Op Amps, Burr Brown, January 15, 1995 5 independent of amplifier gain or input impedance. The Noise Figure is defined as the ratio of the signal to noise at the output to that of the input. This can be expressed as follows: NF= (s/n) I /(s/n) O , where (s/n) I is the signal to noise ratio at the input, and (s/n) O is the signal to noise ratio at the output of the device under test conditions. Hence, the noise figures can be referred to the input. We know that noise can be modeled as a series of Gaussian distributions and that uncorrelated noise combines in quadrature. Mathematically, this can be expressed as follows: e 2 tot = e 2 1 + e 2 2 + e 2 3 +..+ i 2 n R 2 n where e is a voltage input. This simple relation demonstrates that the total noise accumulates as the sum of the squares of individual noise contributions. Consequently, the success of the design will largely be influenced by the ability to curb and control noise figures. To illustrate the significance of the phenomenon, we will proceed with a relatively simple model as follows: Here G 1 , G 2 , and G 3 serve as hypothetical amplifiers of several stages in our circuit topology. It is noteworthy to point out that it is impractical to put all the gain in one location because of power, heating, material, and size related issues. To calculate the signal and noise at the output the following is done: 4 EE Lectures, 2002 6 S out = G 1 G 2 G 3 S in
This can be extended to an N th order system for theoretical purposes. We then have the following at the output: (e out /S out ) 2 = (e 2 1 + e 2 2 /G 1 + e 2 3 /(G 2 1 G 2 2 ))/S 2 in Therefore we can deduce from this relation that it is desirable to maximize gain at the input stage. The first stage would boost the signal enough for the transmission down a cable and should be large enough such that environmental noise is not significant enough to affect the functionality of the circuit. To control noise, it is imperative to recognize that gain peaking effects are the primary noise limitation with a commonly configured high feedback resistance topology. To limit gain peaking, additional capacitance is commonly added to bypass the detrimental effects posed by the feedback resistor. Simply adding feedback capacitance to the photodiode amplifier however, reduces the circuits high-frequency noise gain. This does not actually reduce noise bandwidth, but reduced gain peaking decreases the high-frequency effect of the op amps noise voltage. The amount of capacitance will vary depending on the impedance posed by the resistor. Further, although adding feedback capacitance is an effective means of reducing and controlling noise gain, it also decreases signal bandwidth by the same factor. That bandwidth is already low with high feedback resistance. Consequently, the end result could be a response of a kilohertz or less. Circumventing the bandwidth problem is accomplished by utilizing a composite amplifier. It consists of two op amps with the added one for phase compensation control. A common topology may look as follows: 7 5 With this composite structure, internal feedback controls the frequency response of the gain added by the second op amp A 2 . At DC, we know that capacitors block the feedback and this is the purpose of C 1 in the schematic. The overall open-loop gain will be given by the product of the two amplifiers. The gain will be rolled off by the open- loop pole of A 1 and by the response established by for A 2 by C 1 and R 3. This configuration is also a two pole roll off; therefore it is necessary to reduce it to establish frequency stability and avoid oscillations. It is interesting to note that minimal noise, or offset, is added by the second op amp A 2 since this amplifier is preceded by the high gain of A 1 . This topology is especially useful in situations where low level signals have greater sensitivity to noise such as in our case. However, the feedback approach adds simulation complexity to optimizing the design and little practical advantage over current to voltage stage followed by a modified two pole filter stage. Another design consideration that has to be accounted for is the bandwidth of the circuit. This is especially important for current-to-voltage converters since the total output noise increases in proportion to the square root of system bandwidth because a 5 Photodiode Monitoring With Op Amps, Burr Brown, January, 1995 8 broad noise spectrum is encompassed. Typically there is also the added conflict of an optimum signal-to-noise ratio and the bandwidth. This optimum occurs at a very high gain. However, for current-to-voltage converters the situation is dire because they are bandwidth limited. For instance, 0.5pF of stray capacitance around a 100M! feedback resistor pulls the signal bandwidth from a megahertz level to a kilohertz level. To minimize this, know as stray shunting, low capacitance resistors and assembly precautions are used. On a practical level, mounting feedback resistors on standoffs reduces capacitive coupling with printed circuit boards. This is one possible source that usually adds more capacitance to the overall circuitry. Bandwidth beyond parasitics such as capacitance requires lower feedback resistance and accompanying lower converter gain. Although it is very important to improve the bandwidth through voltage gain, output noise increases which is undesirable. As such, a right compromise must be struck for a given design. The traditional transimpedance amplifier topology is preferred because it gives us a real time representation of light excitation at the output of the amplifier. Alternative topologies on the other hand, such as the switched integrator, give a time-averaged representation of the input information from the photodetector. Despite the limits to the solution in real time, such as bandwidth limits and settling time problems set by the amplifier, this approach is optimal for low and medium bandwidth applications where information about the amplitude and shape of the input signal is important such as in our case. The switched integrator requires careful analysis of the frequency oscillation and the effect of its characteristics as the switched integrator. Additionally, the switched integrator topology is inherently limited in terms of bandwidth because of restrictions imposed by the slew rate, settling times, and switching times of the integrator itself. The 9 only advantage it has over the traditional topology according to the Burr-Brown article is its relatively better noise performance. Much of the discussion so far has dealt with the architectural design and topology most advantageous for the low noise amplifier to be used in photodetection. Since topology issues have been resolved, we are now ready to discuss concrete simulation outcomes. The simulation software that we utilized for our purposes was Linear Technologys (LTI) TINA Spice Simulation software. It is a user friendly software available free for download from Texas Instruments web-page. The decision to use the software is partly influenced by its ability to simulate noise figures that are not as easily obtained in PSPICE, and the inclusion of OPA 277 and OPA 128 in simulations model library. The requirements for the low noise amplifier are as follows: root mean square circuit noise equivalent to one picoampere of photocurrent for the Hamamatsu S1336-44 BQ photodiode. The overall circuit response for a 100 Hz photocurrent square wave varies dynamically from 1 pA to 25 mA. The output voltage for the specified photocurrent must be roughly 10 Volts. For a 100 M" feedback resistance the required RMS noise would be 1 pA * 10 8 "/!2 or 70.7 V. The design will use rechargeable Nickel metal hydride batteries that have a 9V case configuration . The primary circuit operation will be between 16.6 and 14 volts. For simulation purposes the circuit analysis will use plus and minus 15 volt supplies. The amplifier schematic for the low noise amplifier is as follows: 10 The schematic consists of a model photodiode which is composed of a diode shunt resistance, a current source for the dark current, a current source for the photo current, amplifier and a junction capacitance. The first op amp is an OPA 128KM. It is a monolithic amplifier that is commonly used in photodetection schemes. A noise free cascode and low-noise processing give it superb low-level signal processing. The choice of this particular amplifier is also based on its ultra-low bias current of roughly 75fA max, its low offset voltage, which is 500V max, its low noise currents, and low bandwidth. The amplifier is also optimized for low power consumption purposes. This will allow our circuitry to last longer for longer time durations and will require less charging frequencies as light spectra are being studied. Its side terminals are tied to 15 V voltage sources that are then grounded. At the positive terminal of the OPA 128KM we have also modeled the parasitic capacitance of the resistor. The feedback capacitance is the parasitic capacitance and the capacitance required to suppress circuit oscillations. The output stage of the first op amp 11 is followed by an RC filter as well as another amplifier. The second amplifier chosen for photo detection purposes was the OPA277P. The choice of this op-amp was influenced by the following operational characteristics: it has an ultra-low offset voltage of roughly 10!V, high open loop gain of 134dB, high common-mode rejection of 140 dB, and ultra-low bias currents. Asides from this, it also has a wide voltage supply range spanning from 2 V up to 18 V. Its positive input polarity has been grounded while the negative input is connected to the output of the first stage via a resistor. A modified 800 Hertz unity gain RC filter has been utilized for the purposes of noise control. The filter was initially designed to be used with a higher bandwidth operational amplifier which would have required additional filtering due to capacitive gain peaking of the first stage. The filter is followed by the dual version of the OPA277 , that is the OPA2277, which is used to create two unity gain differential signals which are 180 degrees out of phase. The signals then go to a LT1167AC instrumentation amplifier with gains of either two or one hundred. The conversion of the differential signal to a single phase signal by the instrumentation amplifier results in a gain of two so that with the instrumentation amplifier overall gain is 4 or two hundred. The output of the instrumentation amplifier goes to two amplifiers which are used as RC filters and for gain. The filters were designed so that their combined bandwidth was approximately 600 Hertz. The first filter has gains of 1, 3, 5, 7, and 10. The second will has gains of 1, 10, 20, and 40. Following the two filters there is a voltage follower that is basically used as load buffer. The three amplifiers following the instrumentation amplifier are the OPA277P. In light of the background information and our structural topology, it is only befitting to continue the discussion of low noise amplifiers with simulation outcomes. The simulations have been conducted in the following manner: capacitances C 1 , C 2 , and 12 C 3 have been varied in a manner to yield the most desirable transients, output and total noise curves. The frequencies at which responses were noted are 100 and 400 Hz respectively. Unless otherwise noted the photocurrent is one picoamperes, and a dark current of 3 picoamperes. For comparative purposes, plots will be included in our discussion to elucidate the subject matter. In the first attempt, we used the following physical set of data: Total Noise (uV) Output Noise (uV/!Hz) Cap. 1 Cap. 2 Frequency (Hz) Cap. 3 13.96 1.41 4p 4p 100 10n 27.45 1.29 4p 4p 400 10n The corresponding curves of the transient response, output noise and total noise looked as follows: 13 Transient Response Output Noise 14 Total Noise 15 From the table and plots we observe the following: the transient response is well behaved. It is regular, periodic and experiences no oscillations, indicating that the circuit is stable for the capacitance values listed. In a similar fashion, the output noise and total noise experience flat responses with no peaks or dips. From the table we observe that the output and total noise are 1.41 uV/!Hz and 13.96 uV for 100 Hz respectively. For 400 Hz, these figures are 1.29 uV/!Hz and 27.45 uV. As such, we observe that at higher frequencies, the output noise decreases while the total noise increases by a factor of roughly 2. If we perform the same analytical procedure for the following capacitive parameters and frequencies we find that: Total Noise (uV) Output Noise (uV/!Hz) Cap. 1 Cap. 2 Frequency (Hz) Cap. 3 13.99 1.41 1p 1p 100 10n 28.44 1.43 1p 1p 400 10n Transient Response Output Noise 16 Total Noise Upon completion of the simulation with lowered capacitance, we find that the transient response is well behaved once again. In a similar fashion, the output and total noise 17 experience no dips or peaks. The respective responses indicate that gain peaking has not become an issues as of yet. However at the same relative frequencies of 100 Hz and 400 Hz we find the following: at 100 Hz, the output and total noise are 1.41 uV/!Hz and 13.99 uV respectively. Essentially the output noise has remained unchanged while the total noise has increased by 0.03 uV. At 400 Hz, the output and total noise are 1.43 uV/!Hz and 28.44 uV. Therefore lowering capacitance has increased the output noise by .02 uV/!Hz while the total noise has increased by 0.99 uV. These changes indicate that lowering the capacitance makes the circuit more prone to disturbance due to noise. As such, oscillations, which are highly undesirable, are more likely to occur. This observation will be illustrated by the following results: Total Noise (uV) Output Noise (uV/!Hz) Cap. 1 Cap. 2 Frequency (Hz) Cap. 3 13.99 1.42 10f 10f 100 10n 28.5 1.44 10f 10f 400 10n Transient Response Output Noise 18 Total Noise 19 After lowering the capacitance even further, we ultimately find that the circuit starts oscillating. This is particularly obvious in the output noise plot, which shows an abrupt peak around a frequency of 4kHz. Due to the high capacitance of the photodiode, the amplifier becomes subject to instability and oscillation with lower feedback capactitance. The perceived significance of this observation shows that controlling the capacitance values is an important factor in determining whether the circuit topology will be well behaved. Therefore we constrict ourselves to capacitance values in the picofarad range. This observation will be illustrated by the following results: Total Noise (uV) Output Noise (uV/!Hz) Cap. 1 Cap. 2 Frequency (Hz) Cap. 3 13.99 1.42 10f 10f 100 10n 28.5 1.44 10f 10f 400 10n 20 Transient Response:
Output Noise: Spot noise for 10fF capacitance 21 Total Noise: 10fF capacitance Circuit simulations for conditioning where oscillations begin after a step input are difficult to simulate accurately as the software programs tend not to converge to a 22 solution. The oscillation condition is better shown using phase plots. However, noise plots also indicate the potential for oscillation. This is particularly obvious in the output noise plot, which shows an abrupt peak around a frequency of 4 kHz. The perceived significance of this observation shows that controlling the capacitance values is an important factor in determining whether the circuit topology will be well behaved and yet meet the bandwidth and transient response requirements. While simulations shown above used had a photo current input of 1 pA, the following simulations use a photo current input of 1nA in order to match the bench test results which will be discussed later. The output of the OPA128KM is 0.1 volts which is increased to 0.4 V by the differential gain of two and the instrumentation amplifier gain of two. The circuit simulation circuit topology looks as follows: 23 24 The amplifier bandwidth shows that our low noise amplifier will meet the bandwidth specification. With a lower, but stable capacitance values the bandwidth increases. Very low picofarad capacitors have higher percentage tolerance limits as well as less tolerance for the effects of parasitic capacitances. For an effective one picofarad feedback capacitance, the bandwidth is as follows: 25
For bench testing we do not have the ability to verify the circuit response photo current response with any accuracy to compare to the simulation results. However, we can measure the circuit response to a voltage input by attaching a 50 meg Ohm resistor to the non-inverting input which modifies the circuit input to a unity gain. The topology of this circuit looks as follows, however the second schematic sheet is not shown as it is unchanged from the photo current response shown above: 26 The above figure contains the input waveform and the circuit response. However, the input waveform has been multiplied by four to match the amplitude of the circuit response. 27 Asides from the aforementioned simulation responses, we also need to explore the bandwidth response of the circuit. The simulations include the photodiode box and the instrument amplifier in the amplifier box. For the OPA128KM, the time response has a 1pA current which yields 100uV. The two out of phase amplifiers add a gain of two. Further, the instrument amplifier has a gain of 100 so that the output of the instrumentation amplifier is 20 millivolts. Amplifier Bandwidth Response The amplifier bandwidth shows that our low noise amplifier will be a low bandwidth circuit. This is one of the specifications that our circuit has to meet. Hence, this requirement will be satisfied as shown. The bandwidth response will be restricted 28 between roughly 200 and 400 Hz. It is also important to explore the amplifier time circuit. The topology of this circuit is looks as follows: In the amplifier box it is important to point out that the detector box and amplifier do not have a common ground. Additionally, the instrument amplifier LT1167 gains are set by the resistor between pins rg 1 and rg 2 . R 10 and R 11 are selected by a switch and are not actually in parallel. The actual time response itself of the circuit is given as such: 29 From the simulation we deduce that the circuit is well behaved and experiences no gain peaking which could potentially make the circuit unstable or oscillate. Given the architecture discussion, it is also important to give a brief overview on some of the practical issues on the concrete design level. The considerations to be accounted for include external noise sources for instance. With its high resistance, a current-to-voltage converter is fairly sensitive to noise coupling from electrostatic, magnetic and radio frequency sources. As such, they will require great care. Care can be exercised by means of shielding, proper grounding, and the physical location of the actual components. Electrostatic coupling can emanate from the power line for instance in practical design. Therefore voltage differences between objects are impressed on various capacitors; consequently any voltage difference is coupled as noise current. To avoid this commonly encountered phenomenon, shielding is commonly resorted to. Shielding allows one to intercept the coupled current and shunt it to ground. However, it is important to avoid parasitic capacitances that are created by the shields themselves. Therefore, the shields must be returned to the signal common to avoid the coupling effect. Therefore, this concludes our design approach analysis. PCB Board Layout, Assembly, and Test 30 PCB board layout design is an important aspect in the correct functionality of a low noise amplifier. There are several different areas that need to be addressed when laying out the components which are necessary for our circuit. Both layout design and assembly are vital parts in successfully creating a functioning PCB board. The layout portion of the PCB board is initially constructed with the computer software program, ExpressPCB. Through this program the components we chose in the design and simulation stage can be physically placed onto a PCB board. This is accomplished by placing the necessary pins in the board which correspond to the correct sizing of the component that is being placed on the board. Yellow silkscreen outlines the body of the component, in-between the pins, which acts as a visual aid when laying out and assembling the board. The name of the component can also be written in the silkscreen to further aid in organization of the board and component placement. The components we used for mounting on the PCB board are resistors, capacitors, operational amplifiers, instrumentation amplifiers potentiometers as variable resistors, and a photodiode.
31 This picture represents what our final board layout looks like as a PCB file. Each component is traced with the yellow silkscreen and labeled, and each label corresponds to the component part used in the design/simulation. Furthermore, we chose to use a PCB board with two independent layers. A two layered board allowed us to use the top layer for component placement, and the bottom layer was used as the common ground, and for connections that would have required crossing lines on the top layer. Since we used only analog components we were able to use a single common ground to connect all the components that required grounding. These are pictures of our actual board after fabrication. The picture on the left is the top layer of our board before the components were soldered on, and the pictured to the right is of the bottom, or ground layer. Another important consideration of layout design is spacing. Initially we designed two separate boards which would be connect through a cable, however through the design/learning process we found that combining our two boards into a single board would reduce noise, and more importantly cost. 32
This picture represents an early design where the photodiode socket , the photodiode amplifier, the unity gain filter stage, the dual amplifier used to create a differential signal are separate from the other portion of the circuit. Reducing the two boards into one did reduce cost and also reduced the area, thus creating a more difficult task in laying out the circuit. Placement of the components is important when considering noise. Running connection wires next to other wires causes coupling and parasitic capacitance, therefore the placements of each component needs to be located in an area which is within ideal proximity to the other components connected to it. In addition to spacing, the use of our more complex parts in ExpressPCB required custom pin configurations to be created. The photodiode, the variable resistors and the standoff required custom creation and placement of the pins of these components required calculation down to the thousandth of an inch. While this task was not overwhelmingly difficult it did require precision as many of our comment pins are closely arrange and the correct placement of the pins is vital for a successful PCB board design. The photodiode also posed a unique challenge in the layout process. Our first stage of amplification is 33 extremely sensitive to noise created by leakage currents and the current noise generated by the photodiode. The board specifications could not be obtained from the manufacturer which created a design problem that needed to be addressed. Pico ampere leakage currents between the OPA128KM input pins where the photodiode is connected, or Pico ampere current leakages in the areas where the photo diode, high impedance resistors and OPA128KM input pins join. Leakages could occur on the board surface between the PCB material and the conformal coating, through the PCB board. Guard rings could mitigate the surface problem however board cleanliness and solder flux residues would still be a concern. Another consideration is the parasitic capacitance of the high impedance resistors, and the parasitic capacitance of the resistors to ground which could limit the design bandwidth. To solve both the leakage, parasitic capacitance and cleanliness problems standoffs would be used as support points for the high impedance resistors and also supporting the connection of the photo diode and the inputs to the OPA128KM amplifier. One of the selection criteria for the OP128 amplifier was its availability in a TO-92 case which would allow its input pins to be bent and not be inserted in the amplifier socket, and connect directly to the photo diode, which was ideal for our design. The standoffs are a non-conducting material which allows the photo diode, high impedance resistors, and amplifier inputs to be independent from the board, thus solving both the leakage and capacitance problems. Also as the PCB board and parts are new their connection could be achieved with minimal flux residue. 34 Once the layout design was completed and the board was fabricated, assembly posed the next challenge. The first difficulty we encountered in assembly after part mounting but before the insertion of the photodiode or amplifiers was a layout error which resulted in insufficient clearance between a routing hole and the ground plane. One of our pins was not completely etched from ground; this caused a direct connection between ground and the power supply, thus causing our entire board to be shorted out. For time constraint reasons we decided to sacrifice one of our extra boards and disconnect the power supply in sections of the board, therefore we were able to deduce in which general area of the board the short was occurring. Once we discovered where the faulty connection was, we carefully drilled the short out and soldered a jumper lead to complete the circuit connection. Additionally the plus and minus supplies were interchanged for one amplifier and to solve this connecting lines were severed with a small tool and jumper wires were installed. Another assembly conflict we encountered was flux. In some area there was a build up of flux and residue. The flux residue could create undesired conditions on our board which caused our board to function incorrectly. The flux was not a problem in the highly critical areas of the PCB. To remove the flux from our PCB board we cleaned the 35 This is a picture of how the op-amp should be placed on standoffs and guarded, to avoid the leakage and grounding problems surface with an alcohol solution which removed the flux sufficiently. The close proximity of the components on the board also caused some assembly disagreement. Close attention needed to be paid to ensure each pin was correctly connected to the proper lead of its component. The clearance between the hole area to solder a lead and the surrounding ground plane on the back of the PCB was minimal. The layout used the manufacturers standard clearance which we in many cases should have enlarged. In addition to that, the leads of each component needed to be isolated from the other leads surrounding each component, and with some components confined to a small area contact between the leads was difficult to avoid. Only one small surface mount integrated circuit part type was used in the design and it was soldered using a binocular microscope. To make sure each lead had clearance from another lead or the ground plane board was often closely inspected. Shielding the PCB board from outside noise, such as electrostatic and magnetic pickup was another important assembly consideration. To ensure our circuit maintains a low noise signal exposure to outside interference was necessary in producing a low noise signal. To account for outside pickup a fully enclosed box was purchased to encase the board, and to mount switches and other connections. The top, bottom, and two sides of the box were 0.050 inch steel and two sides for mounting switches and other components was .060 inch aluminum. The box would be the primary shielding from any electrostatic or magnetic pickup, with any required additional shielding being experimentally determined. The box did offer sufficient shield, however the photodiode requires exposure to light, and thus to accommodate the diode a hole was drilled through the box which allowed light exposure to the surface of the photodiode. 36 The hole necessary for the diode allowed a strong signal to be created from the photodiode, however the hole also exposed the circuit to noise (pickup). To allow the photodiode light, while shielding the rest of the board from noise a non-reflective steel tube was placed through the hole in the box. The photodiode was placed in one end of the tube while the other end was open to accept light for the diode. This offered a partial solution to our noise concern. However, it is anticipated that there will have to be a conducting mesh screen over the end of the tube. The design of the box and PCB did not have any common ground points. The battery power is contained within the box. A problem arose with the addition of the tube, and when the photo diode was inserted into the tube. The photo diode had little response to light. The cathode of the photo diode is connected to the TO-92 case, and the cathode is also connected to the inverting pin of the OPA128KM. The placement of the photodiode in the tube created a connection between the diode and the box, which caused our circuit to function improperly. After some experimentation it was determined that the isolation between the box and the PCB ground was approximately 30 meg ohms. Since the photo diode shunt impedance is typically 600 meg ohms the circuit was essentially converted to a voltage amplifier and not a trans- impedance amplifier. To avoid the connection between the diode and the box, an insulting 37 Picture of the tube in which the photodiode fits into for shielding. The tube is connected to the box dielectric was placed between the photodiode and the tube to avoid the electrical connection. The testing portion of our project was an important area which allowed us to make the correct modifications to our circuit. Since we encountered several problems during assembly testing our circuit was necessary in determining exactly where a short, oscillation, improper connection occurred. We found how close our measured gain values were in comparison to the theoretical values.
The error percentage of the measured value in comparison to the theoretical value is very small, which is important for our circuit to function properly. In addition to measuring the gain and voltages across the components of our circuit we also measured the output noise as a Gaussian distribution. 38 The output noise of 52.8 micro volts is well within our ceiling of 70 micro volts of output noise. Furthermore, we also took the FFT for the fast Fourier transform of the noise, which gave us our frequency response of the noise.
The test confirm that our circuit is overall well behaved and within our low noise specification. Testing was a vital portion of our project which allowed us to both adjust our circuit in order to obtain our desired results, and confirm the noise and gain values were within the appropriate range. In conclusion our project was a success. We completed our project on time with all parts functioning correctly. In addition to that we met all of our initial design 39 Picture of the noise at the output with a gain of 1600. specifications which were set by our professors. The output noise signal, as a function of the input, measured to be 50 micro volts which is under our maximum of 70 micro volts. The numerous gain settings on our box all produce and output gain that is very near the expected value. Through this project we learned several valuable lessons, which pertain to not only the technical side of engineering, but also the teamwork and communication aspect of engineering. Our group learned to work together and communicate to solve problems and meet deadlines. This project has taught us real life engineering skills which we will be able to demonstrate as we make the transition from student to employee. The future advice we would like to extend to next years low noise photodiode amplifier group is, first, design is the most crucial step. Make sure your design is correct as even small mistakes in the design take much more time to solve and correct once the board is fabricated. In addition to that, time management is another important topic for this project, and creating a time schedule in the first quarter would be a good idea. Lastly, make sure everyone communicates and is assigned specific areas. Our group successfully completed our project because we divided the workload, and made sure to communicate and work together as a team. 40 Appendix ! Software (Board Design): " Express PCB ! Software (Simulations): " TINA " PSPICE ! Parts purchased " www.digikey.com " www.newark.com 41