UVC Lamps, Part 3: Photodiode Measurements
Internal Reflection – Source of High Reading Photodiodes
- How Is Responsivity Measured?
- Reflections Inside the Photodiode Package
- A Demonstration of Internal Multiple Reflections
Introduction
Earlier parts of this series on UVC germicidal lamps focussed first on modelling that related the irradiance at locations near the lamp to the lamp properties, and second on power measurement results for a couple of particular lamp assemblies. In this part I report on the results of photodiode sensor measurements made on these same lamps and lamp assemblies. The first two parts of this series are predecessors to this part. The models are necessary so that the sensor measurements can be compared to them, and the power measurements are necessary so that the models can be accurately parameterized. Since the sensor measurements to produce irradiance values rely upon manufacturer’s responsivity specifications, the models together with power measurements are being used as a “poor man’s” way of providing calibration sources to test the manufacturer’s specifications by comparing sensor reading to model prediction.
What has enabled this testing is the availability of affordable UVC photodiodes from Genicom Co, Ltd, a Korean company. These are Schottky-type photodiodes using aluminium gallium nitride based material and are available from the distributor Digikey at affordable prices.
Photodiodes
Three sensors were used in the measurements. Two of them contain the same die but have different packages, which produce different angles of view. All of them include fused quartz windows. The GUVC-S10GD is a surface mount device with a 150 degrees viewing angle. The GUVC-T10GD is mounted in a TO-46 can and has a 60 degree viewing angle. This sensor die has a UVC response and is blind to any other wavelengths (see below). The third sensor is GUVC-T21GH and has exactly the same UVC response as the other two but the die is assembled in a large TO package together with the op amp and feedback components. It has three terminals: 5 VDC supply, Ground, and Output. This particular device is therefore very easy to deploy.
The two sensors have photocurrent responsivity calibration constants somewhat different from each other (but nearly identical to Figure 1) and the third sensor has a responsivity in the form of a voltage output. From the data sheets:
GUVC-S10GD 39 nanoamperes at 1 mW/cm2 at 254 nm irradiance
GUVC-T10GD 34 nanoamperes at 1 mW/cm2 at 254 nm irradiance
GUVC-T21GH 0.355 volts at 1 mW/cm2 at 254 nm irradiance
Shown in Figure 2 is the surface mount device and one can see why the package provides a wide FOV (field of view). In Figure 3 mounted in a housing is shown the TO-46 device with its much narrower FOV.
The third sensor is shown in Figure 4 and one can see portions of the built-in transimpedance amplifier as well as the photodiode die mounted to a printed circuit board.
Transimpedance Amplifier
Except for the GUVC-T21GH, the other two photodiodes need to be integrated with a transimpedance amplifier. This circuit can be very simple as shown in Figure 5.
The circuit is configured to operate the sensor in the photovoltaic mode. Op amp negative feedback keeps the voltage at the IN- input zero as illumination generates increasing dark current. This current flows through the feedback resistor network as the output voltage swings negative. Often to achieve a large gain resistance such as shown in the schematic one needs to put in series smaller values. The op amp needs to be a single supply rail-to-rail version. The feedback capacitor value can be varied, but circuit analysis shows that any value greater than, say, 10 pF will make the circuit stable and there will be no gain peaking. One of these assemblies is shown in Figure 6 mounted in an aluminum housing.
The GUVC-T21GH was made functional in the assembly shown in Figure 7.
Sensor Assemblies
In Figure 8 is shown all of the sensor assemblies that have been constructed. Not all of them will be discussed here. Only those indicated as sensor1, sensor2/4, sensor 3, and sensor5 were used to create the results reported here.
Except for sensor3 which has its own amplifier built in, the others are contained in aluminum housings that are grounded to the 5 volt power supply in order to eliminate 60 Hz pick up. The 5 volt power supply is a simple cellphone USB wall charger.
Method
The basic testing was to put the photosensor on the centerline that bisected the lamp and record the measurement as a function of distance from the lamp. In this manner one can compare the result to one of the models. Normally the minimum distance from the lamp was constrained by the FOV of the photosensor, that is, the full length of the lamp being measured should lie within the FOV. Measurements were made when the lamp stabilized, normally after 5 minutes.
The two lamp sizes were a 36 watt rated bi-tube that was always positioned vertically, and a 6 watt rated tube that was normally positioned horizontally. It was found that it was important that the horizontal lamp be elevated away from the tabletop on which it sat because UVC reflecting off the tabletop caused the readings to be higher than they should have been. The 36 watt rated lamp did not have this problem because the photosensor placed on its centerline was always sufficiently elevated away from the tabletop.
36 Watt Rated Lamp Results
These results are presented first because this testing was the simplest since there was only a single electronic ballast that drove two different lamps, one ozone-producing and the other not. This lamp is shown in Figure 9 plugged into the base that contains an electronic ballast operating at 34 KHz and driving the lamp at 30 watts. Since the results in Part 2 of this series show that the two lamp types were virtually putting out the same power, the data was gathered on the non ozone-producing lamp.
All three photodiode types were used and the results are summarized in Figure 10. Note that sensor2 and sensor4 are the same photodiode (GUVC-S10GD) but mounted in different assemblies.
In the graphs the solid lines represent three different models calculated for a lamp consuming 30 watts, not 36 watts, since separate measurements of tube power and wall plug power demonstrated this value (see Part 2 of this series on UVC Lamps that also shows this lamp’s wall loading suggests a 40% conversion efficiency).
The sensors GUVC-S10GD (sensor 4 and sensor 2) and GUVC-T10GD (sensor 1)agree with each other while the sensor GUVC-T21GH (sensor 3) measures somewhat less than these other two (graph on the left).
In the graph on the right the results for sensors GUVC-S10GD and GUVC-T10GD are divided by 1.4 and the results for GUVC-T21GH are divided by 1.22. These factors force the measured results to fall on top of the models for distances 20 centimeters and greater. The data below 20 centimeters progressively falls a bit under the models and this can be shown is the result of the field of view of the sensors no longer seeing the full length of the lamp.
Since the testing in Figure 10 was done at various times, the following data, shown in Figure 10a, was measured at the same time as a re-check.
One sees that the factor for sensor4 changed to 1.6 from 1.4, for sensor1 and sensor3 the factors remained the same, 1.4 and 1.22, respectively, and the factor for the single data point for the assembly sensor5 is also 1.4. Interestingly sensor5 has the same photodiode as in sensor4. These factors will be discussed in a section that follows.
So, in a way, this lamp-ballast combination serves as a reference since its power is certain at 30 watts and its efficiency for UVC output is likely 40 %. One can then re-calibrate the two sensors using the factors shown. The adjustment factors of 1.4 or 1.6 and 1.22 initially were somewhat mysterious in regards to their cause.
6 Watt Rated Lamp Results
The presentation and analysis of these results are a bit more complicated. Two different ballasts were used with two different lamps. (See Part 2 of this series for images of these ballast assemblies.) Again, one lamp was ozone-producing and the other not. The magnetic ballast drove both of these lamps with around 6 watts of power into the arc while the electronic ballast operating at 24 KHz drove the lamps around 4.5 watts for the ozone-producing one and 5 watts for the non ozone-producing lamp. To keep things a bit more simple only sensor3 and sensor2/4 was used. In the following what is called “old lamp” is the non-ozone-producing lamp and “new lamp” is the ozone-producing lamp. Also, to give some reference to these results the Keitz model is plotted for a 6 watt lamp with an arc length of 6 inches operating at 33% conversion efficiency. Therefore, the solid curve is reference only, and is not intended to represent that actual output of the lamp/ballast assembly when the data was taken.
In Figure 11 the results for sensor3, the one with the built-in amplifier, are shown for each lamp.
Note in both graphs data for the magnetic ballast measured higher irradiance than for the electronic ballast. There are two possibilities to explain this. The one perhaps more likely is the magnetic ballast assembly with lamp is housed in a plastic housing with a metallic foil behind the lamp. Reflection from this foil is likely responsible for the higher irradiance. The other possibility is the magnetic ballast drives more power into the lamp than does the electronic ballast. However, if one discounts the high electrode losses at 60 Hz the radiance power out may be the same or even lower.
The other point to take note of is that both sets of measurements indicate irradiance lower that of 6 watts at 33% efficiency. It has already been pointed out that for the electronic ballast only 4.5 watts is delivered to the arc, and for the magnetic ballast, a large portion of the 5 watts of measured arc power is going into electrode losses whereas for electronic ballasts electrode losses are much less.
In Figure 12 are the results for the sensor2/4 photodiode presented in similar fashion with “old lamp” on the left and “new lamp” on the right.
While the results for the “new lamp” show the same relationship where the irradiance for the magnetic ballast is higher than the irradiance for the electronic ballast, this is not true for the “old lamp”. The irradiance results appear to be the same for both ballast types. This testing would need to be repeated in order to explain the result.
Where the data is labelled “adjusted” the values have been multiplied by a measured factor representing the elimination of the table reflection. Since the 6 watt assemblies are operated horizontally, an intial mistake was to not elevate lamp and sensor away from the table top, which contributed to the irradiance due to reflection. An adjustment factor was experimentally determined and applied to the affected data in order to salvage its usefulness.
The last item to address is to see if the two sensor types behave as they did with the 36 watt rated lamp where the results for sensors GUVC-S10GD and GUVC-T10GD were 1.4 times higher than expected and the results for GUVC-T21GH were 1.22 times higher than expected. Therefore, additional testing was done where sensor3 and sensor5 (GUVC-S10GD) at approximately the same time measured the output of the ozone producing lamp and the non-ozone producing lamp in the electronic ballast assembly as a function of distance.
Seen in Figure 13 is this lamp assembly along with sensor5 aligned for a measurement. Note that the arrangement minimizes any reflection surfaces seen by the sensor. Sensor3, seen in the figure to the left, was re-mounted so it could be positioned in a similar manner to sensor5. Also the figure shows the closest distance of 9 inches for the series of measurements. This presents the lamp subtending a field of view (FOV) of about 36 degrees. For sensor3 the FOV is 60 degrees and for sensor5 is 150 degrees. In order to get a good comparison between the two sensor types the lamp was well warmed up, distances carefully measured, and at each distance one sensor measured, then sensor switched to the other and measured without any delay. Sensor positioning was precise. All results are summarized in Figure 14.
For the ozone lamp the arc power, as mentioned earlier, was found to be 4.5 watts, and wall loading suggested that its efficiency is 40 %. So the Keitz model with these numbers is also presented in the figure. Interesting it appears that sensor3 matches the model quite well. Sensor5 reads higher by about 15 %. If one divides sensor5 data by 1.15 Figure 15 results.
So what has been done so far is to assume the arc power and conversion efficiency correctly acts as a calibrator, just like what was done earlier with the 36 watt rated lamp results. Now if one proceeds with this to the results for the non-ozone producing lamp the data for sensor3 cannot be fit to the 5 watt arc power earlier measured unless one assumes a conversion efficiency of 28 %. See Figure 16.
So how could the efficiency be so low? One may speculated that this lamp’s envelope certainly absorbs the 185 nm mercury line and may well be less transparent to the 254 nm radiation than other non-ozone producing lamp envelopes are. This could explain the low output.
Photodiode Comparisons
In all of the above testing the attempt to use arc power measurements of the three different lamps has not been too successful in trying to use irradiance models to calibrate the different sensors. Table 1 summarizes the factors by which the sensor result would need to be divided to match the given lamp that was used as a reference.
| photodiode/model | 30 watts, 40% non-ozone | 4.5 watts, 40% ozone | 5 watts, 28% non-ozone |
| GUVC-S10GD (sensor5) | 1.4 | 1.15 | 1.15 |
| GUVC-S10GD (sensor4) | 1.4, 1.6 | n/a | n/a |
| GUVC-T10GD (sensor1) | 1.4 | n/a | n/a |
| GUVC-T21GD (sensor3) | 1.22 | 1.0 | 1.0 |
From the table one can see that the factors are somewhat different for the bi-tube lamp and the short lamps. Furthermore, sensor4 did not quite repeat after being tested again some time later than the first test. What remains consistent, however, is that sensor3 with the GUVC-T21GD, the one with the built-in amplifier, always reads 15% lower that the other two photodiodes. This testing seems the best that can done short of using a standard, certified UVC source.
Internal Reflection – Source of High Reading Photodiodes
How Is Responsivity Measured?
The method above to use measured arc power together with models to compare photodiodes was good enough to show that these diodes from Genicom consistently give results too high, that is, consistently higher than what the model predicts. It calls into question the photodiode specifications provided by Genicom. It seems not too plausible to accuse the responsivity in the specification to be so wrong, but it does beg the question how is the responsivity measured by the manufacturer. How is the one milliwatt/cm2 of 254 nm radiance delivered to the sensor when the manufacturer characterizes the device? For example, does a fiber optic deliver a narrow calibrated beam onto the die only, or is the beam wide and floods the whole window on the package? If it is the former, then internal reflections inside the package are not accounted for. Actual use such as in this study will always provided results higher than they should be because the illumination falls upon the whole window into the device.
Reflections Inside the Photodiode Package
The inside cavity of the device where the die sits including the fused quartz window must be considered. Rays entering the window but not falling directly onto the sensor will be both absorbed and reflected by the walls of the cavity, including the quartz window, which has a reflectance of about 10%. This level of reflectance has been computed using Snell’s Law and Fresnel’s equations along with the index of refraction of fused quartz at 254 nm wavelength. Consequently, the cavity creates a multiplier effect. This is not unlike an integrating sphere where the receiving sensor will have a much higher incident radiance than the radiance entering the sphere. Nor is it unlike what a collimator does to the rays emanating from an aperture.
A Demonstration of Internal Multiple Reflections
This effect has been experimentally observed with a fabricated collimator that was made in order to measure the exitance from the lamp at various positions along the tube. It is shown in Figure 17, a 1/2 inch diameter aluminum tube with a 0.147 inch diameter aperture at one end, 8.7 centimeters long, with sensor3 mounted at the other end.
A measurement was made with the aperture against the 6 watt rated non-ozone producing lamp at its middle. The result was 0.211 milliwatt/cm2 . If the aperture did not have the aluminum tube and the radiance is a diffuse source, then the expected radiance at the sensor would be given by the application of the following equations in Figure 18.
The cosines equal one since the axes of the sensor and aperture are coincident. The distance r is given and the area of the aperture dAs is also given. It was shown in another part to this series that the exitance of the lamp is about 28 milliwatt/cm2 which in turn implies an exiting radiance Lo of 9 milliwatt/cm2 . d2Φ/dAd is the incident radiance measured by the sensor and the result calculated to be 0.013 milliwatt/cm2 . Therefore the measured value is many times the expected value if the tube was not there. The shape of the tube helps reflect many of the rays toward the sensor, however, multiple reflections within the cavity and the eventual absorption of the ray energy in the walls of the cavity create a power balance, the result being the sensor receives a lot of this energy passing out of the aperture.
The sensor package too has to be viewed as a cavity with walls that both absorb and reflect incident rays. So, for this reason one can expect the Genicom sensors to give high readings such as described in Table 1.
Conclusions
The affordable UVC photodiodes from Genicom, Inc. work well. The use of UVC lamps that have been characterized by measuring arc power together with well-developed calculational models for lamp irradiance provides only an approximate way to calibrate these photodiodes, but is good enough to demonstrate that they read higher than predicted by the manufacturer’s specified responsivity. It has been shown the likely cause for this results from multiple reflections within the photodiode package of rays entering the window but not falling onto the die.
