The Azimuthal BCAM Side Head (A2049) is dual laser driver circuit with two lasers mounted on the printed circuit board. The lasers can be the LDP65001E, DL3147, or any elecrically and mechanically equivalent device. The two lasers we name emit ruby-red light in a rectangular cone. The base of the cone is a tiny light-emitting surface roughly 10 μm by 50 μm. A BCAM is an optical surveying instrument (Brandeis CCD Angle Monitor) consisting of digital cameras and laser diodes acting as optical point sources.
The following versions of the A2049 exist.
|F||CSC Rear Laser Pair|
|G||CSC Front Laser Head|
|L||Black Azimuthal BCAM Side Head (mirror image of A2049R)|
|R||Blue Azimuthal BCAM Side Head (mirror image of A2049L)|
|M||Black Azimuthal Side Head with ZIF Connector (mirror image of A2049N)|
|N||Blue Azimuthal Side Head with ZIF Connector (mirror image of A2049M)|
All varieties of the A2049 will work with the Azimuthal BCAM Head (A2048), the Azimuthal Side Head (A2051S), and the CSC Source Head (A2050G).
The Azimuthal Side Heads are used in the Azimuthal Source Plates, also known as Saloon Door Sources, in the ATLAS end-cap alignment system. Each source plate provides two pairs of red lasers and is controlled by an Azimuthal Source Head (A2051S).
The A2049F and A2049G are designed to work with the CSC Source Head (A2050G). The ZIF connectors on the boards are oriented so that the A2049F will work only when connected to A2050G-J2, and the A2049G will work only when connected to A2050G-J1.
For A2049F and A2049G board dimensions and hole locations, see this drawing. To see how the A2049F and A2049G connect to the CSC Laser Head (A2050), look at this drawing. The A2049F connects to J3 on the A2050G, and provides source elements 1 (LD1) and 2 (LD2). The A2049G connects to J2 on the A2050G, and provides source elements 3 (LD1) and 4 (LD2).
The A2049 provides two laser drivers, as shown in the schematic. Each driver has two terminals, but one of these terminals, V+ is common to both drivers. Laser A will turn on when LPA is six or more volts below V+ and laser B will turn on when LPB is six or more voltes below V+. If we apply more than twenty volts across the driver terminals, transistor array U1 will overheat and suffer permanent damage within a few minutes. If we apply more than forty volts, transistor array U1 will suffer breakdown and permanent damage. Otherwise, the driver is immune to the signal applied to its inputs, and it is impossible to damage the laser.
The two drivers share a common positive terminal, and this terminal is in turn connected to the metal cans of the laser packages. The lasers are housed in the industry-standard 5.6-mm laser can, as shown here. When can drive the A2049 with a fixed power supply, or with a power supply controlled by a switch, or with any open-collector logic switch, provided that we don't apply more than twenty volts to the driver terminals. But we designed the A2049 for a specific purpose: to work with the Azimuthal BCAM Head (A2048). The A2048 is a LWDAQ Device. It provides ±15V power for functions such as laser light sources. The most obvious connection between the two circuits would be to place V+ at 0V and drive LPA and LPB to −15V when we want to turn on the lasers. But that's not what we do.
Suppose we connected V+ to 0V. Now the two laser cans are connected to 0V. That sounds like a good thing. We put the lasers into two mounting holes in an aluminum BCAM chassis. The packages are connected to the chassis. But now we have violated the LWDAQ grounding rules: the chassis must have no direct current connection to the local 0V. If there is to be a direct current connection to 0V, this connection must be carried by the structure upon which the BCAM is mounted. Therefore, the laser packages must be isolated from the chassis in all LWDAQ devices. In BCAMs, we andoize the aluminum chassis, and rely upon the non-conducting surface to isolate the lasers from the chassis.
If this anodized isolation should break down, or be damaged during construction, or be imperfect when it is first applied, we will have electrical contact between the laser packages and the chassis. This contact will not be obvious if we connect the laser cans to 0V. A short between 0V and the chassis may generate noise and ground loops when the BCAM is installed in an experiment, but it will not stop the BCAM from flashing its lasers during testing and calibration. The BCAM chassis will almost always be grounded through the structure upon which it is mounted, even in the laboratory, and this ground has a habit of being connected to the 0V power supply of the LWDAQ, especially when the LWDAQ is running out of a VME crate. Once a BCAM is installed in a system, there will be no way of detecting whether or not its lasers packages are connected to the chassis, and no way to track down the source of system-wide ground loop noise.
The A2048 connects V+ to 15V through a 100-Ω resistor (actually, it's two 0.5-W 47-Ω resistors, see here). In a large experiment, we can rely upon the BCAM chassis to be grounded through its support structure. If the packages are shorted to the chassis, we will see 150 mA flowing through the 100-Ω resistor whenever we wake up the A2048 device inside the BCAM. The lasers will not flash. By means of the LWDAQ's power supply monitors alone, we can identify such short circuits in the system. The Analyzer Tool makes such identification automatic.
With V+ is connected to +15V, the A2048 turns on lasers by connecting LPA and LPB to 0V with a simple mosfet switch driven by a logic level. In general, if we connect V+ to a common positive voltage, we can turn on the lasers with an open-drain or open-collector logic output.
The laser packages are metal and they are used for the anode of the laser diode itself and for the cathode of its photodiode. Thus the metal can must be connected to the laser driver circuit. There must be a direct-current path from the laser can to one of the power supplies in the circuit that provides power to the laser drivers. If we use the laser driver with any LWDAQ device, the LWDAQ grounding rules prohibit any direct-current connection between the device chassis and the device power supplies. Thus we must ensure that the laser cans are isolated from the chassis.
In devices like the BCAM, we provide isolation by means of non-conducting anodization of the chassis surface. The nominal diameter of the laser can is 3.55 mm. We use 3.7-mm diameter holes with 45° chamfers on both ends. These holes are snug enough to locate the laser package within ±80μm of their nominal position, and spacious enough to avoid any scratching of the anodized surface by stray fragments of aluminum oxide when we press the laser into the hole. We check every BCAM for a short circuit between the laser can and the chassis after assembly. No fresh BCAM chassis with a 3.7-mm hole has yet failed the test. But some recycled azimuthal BCAM chassis, from which we have knocked out lasers glued into their holes, have failed the test because the anodizing was damaged by the violent removal of the lasers. For more information about recycling BCAMs, see the BCAM Repair Manual.
Given a step input on the LDA of −10V with respect to V+, the A2049 will drive laser A to full power within 5 μs. When LDA returns to V+ again, the laser turns off within a fraction of a microsecond. The 5-μs delay in the turn-on is designed to cooperate with the 4-μs command transmission time of the LWDAQ. When the A2049 is connected to a LWDAQ device like the Azimuthal BCAM Head (A2048), we turn on the lasers with a LWDAQ command word, and turn them off with another command word. The minimum time between the completion of the ON command and the completion of the OFF command is 4 μs. After 4 μs, the laser will not yet be at full power. Thus we are able to obtain more control over low intensity pulses of light than we would with a faster laser driver.
The power output of the lasers is inversely proportional to the values resistors R5 and R11. With R5 and R11 equal to 10 kΩ, the nominal power output from the LDP65001E or the DL3147 is 1 mW. Across two thousand DL3147 lasers, we found the output power varied from 0.5 mW to 2 mW with R5 and R11 equal to 10 kΩ. Our first one thousand LDP65001E lasers produced the same range of power. But our second thousand LDP65001E lasers produced between 2 mW and 8 mW. To reduce their power below 4 mW, we changed R5 and R11 to 33 kΩ, which we see in the schematic.
Laser diodes provide light of a single wavelength from a single point. The half-power width of the DL3147 output is a few nanometers. The dimensions of its emitting surface are roughly 20 μm by 50 μm. The single wavelength and point source make it possible to gather all the light from a laser, or almost all of it, and focus it into the core of a single-mode optical fiber, or to focus it into a tiny point on a DVD disk to identify individual digital bits. It is not possible to focus the broad spectrum of an LED with its simultenously large emitting area, into a point small enough for use with CDs, let alone DVDs.
But laser diodes are more complicated to drive than LEDs. We can provide power to an LED simply with a voltage and a resistor, and be pretty sure that we are getting close to the expected power output. We can flash an LED by supplying it with a pulse of power. The same is not true for laser diodes, or not if we want our laser diodes to survive through many millions of such pulses. Light amplification through stimulated emission of radiation (lasing) within a solid state laser diode is a sharp and varying function of the current through the laser. The following graph shows the typical power output versus forward current for the DL3147 laser.
At 25°C we see the typical DL3147 emits <0.05 mW at 20 mA, but by 40 mA it is already emitting 3.5 mW. The power output starts to increase steeply with current at around 32 mA at 25°C. We call this current the threshold current of the laser. The threshold current increases with temperature. Suppose we designed a circuit that drove 40 mA through the laser. The laser would produce 3.5 mW at 25°C. But it would warm up after a few minutes, and by the time it reached 40°C, it would be producing only 0.6 mW. In order to maintain a power output of 3.5 mW, we must increase the current to 50 mA.
If we apply 50 mA to this same laser at 25°C, the power output will rise far above 5 mW, which is the maximum rated power output for the DL3147. When we exceed the rated power output, there is a danger that the mirrors at the end of the lasing chamber will over-heat and come off, thus ruining the laser. This overheating of the mirrors need not cause a significant increase in the temperature of the entire device, so we cannot rely upon a temperature-induced rise in threshold current to protect the mirrors.
Furthermore, the curve we show above is for a typical DL3147. Individual DL3147 diodes can have threshold currents ten miliamps higher or lower at the same temperature. If we want to drive our lasers at 3.5 mW at 25°C, each laser will have to have its own drive current.
All lasers suffer from uncertainty in threshold current. Without some way to measure the optical power output from the laser, there is no way we can be sure we are driving the laser at its rated output power. Thus all lasers include an integrated photodiode that provides us with a current proportional to the laser output power. This photodiode tends to be mounted at the rear end of the lasing chamber. We build a circuit that adjusts the laser current until the photodiode current indicates full power. That is to say: we implement a feedback control loop that drives the laser power output to a pre-determined value.
When it comes to designing a feedback control loop, we have to consider its behavior when we turn on power. Most feeback circuits will over-drive a laser on power-up, especially those with more than one power supply voltage. One way to over-come these power-up problems is to provide additional circuits that raise the power supplies slowly. But these circuits can in turn be thwarted by certain actions, and the resulting laser driver is bulky, inefficient, and slow. When using such laser drivers, we have found ourselves unable to generate light pulses of less than 10 ms, let alone the 10 μs we want for our BCAMs.
The A2049 provides fast and robust control of laser current with a wide range of power supply voltages and pulse durations. Referring to the schematic, we see that U1-1, Z1, and R1 form a voltage regulator. The voltage on the lower end of R2 is 5.5 V below V+. We can apply anything from 6V to 20V across the terminals of the laser driver (V+ and LPA in this case) and obtain the same voltage applied to the feedback loop that makes up the right side of the circuit in the schematic. This circuit has its own local power supply of 5.5V between V+ and what we will call V−, connected to U1-6.
The photodiode current of the laser passes through R5 and U2. The transistor serves only as a diode. Its voltage drop is added to the voltage across R5. The diode ensures that the voltage we apply to U3-5 is at least 0.6V above V−, so that transistor U3-1 has enough voltage between its base and collector to operate outside saturation. The diode makes the feedback loop more linear and therefore more stable.
The voltage on the top end of R5 increases with photodiode current. The transistors in U3 are arranged as a differential amplifier. The base of one transistor, U3-2, is connected to Z2 and R2. The voltage across R2 is 1.6 V. The current through R3 is roughly 3.2 mA. When the voltage at the top of R5 is less than 1.6 V, the differential amplifier will steer this 3.2 mA through R4. The voltage across R4 provides the base current for transistor U2-1, which sinks current out of the laser. As current increases through the laser, the photodiode current increases also, and so the voltage on the top end of R5 increases. The feedback loop reaches equilibrium when the voltage at the top of R5 is 1.6 V, which means the current through R5 is 30 μA.
Now we encounter another problem of variation between laser diodes. The photodiode current corresponding to full power output varies by a factor of 5 from one LDP65001E or DL3147 to another. One solution to this problem is for the manufacturer to measure the photodiode currents, put each laser in an envelope, and mark the photodiode current on the side. During assembly of the laser driver, we could pick R5 to suit each particular photodiode. But our solution for the manufacture of BCAMs and other laser diode light sources for the ATLAS experiment is to pick a value of R1 that guarantees the power output will be between 1 mW and 5 mW. We started with R5 at 10 kΩ. Within batches of 100 lasers, we found that output power varied by only a factor of two, and all lasers were below 5 mW. Eventually, however, we obtained a batch of several thousand LDP65001E lasers with far less sensitive photodiodes, and we increased R5 to 33 kΩ, which is the value we see in the schematic.
Capacitor C1 is the dominant single low-pass element in the feedback loop, and provides the 5μs turn-on delay in combination with the 3.2 mA maximum charging current available from U3-1. We could decrease C1 and so make the driver respond more quickly, but the delay is deliberate because it works well with the 4-μs command transmission time of the LWDAQ (see above). With C1 equal to 1 nF, we can generate 100 ns pulses of light. With C1 equal to 100 pF, the circuit becomes unstable.
If we want to flash a laser even faster than 100 ns, we can work harder on the feedback loop and stabilize it for 10-ns pulses. Faster pulses may be possible, but we have not achieved them. And yet lasers are used to communicate at many gigabits per second. High-speed modulation of lasers is different. We begin with a feedback loop like our own, with which we control the average power output of the laser. We might pick 2.5 mW as the average power output. To modulate the laser power we then apply a ±5 mA current square wave to the laser current. If we use the graph above as our guide, we see that modulating the current from 35 mA to 45 mA at 25°C will cause the power output to go from 1 mW to 5 mW. We don't need a feedback loop to modulate the current. With a fast enough square wave generator, we can modulate a laser's power output at several gigahertz. But the modulation does not reduce the power output to zero. Nor does it produce fast pulses at random times.
Note: All our schematics and Gerber files are distributed under the GNU General Public License.