Georgia Institute of TechnologySchool of Electrical and Computer Engineering

ECE 4006CSenior Design

Dr. Brooke/Dr. Jokerst

D-Link/OCPOptoelectronics - Group 6

Design Report

By:Benjamin King

Alisha McClintonNakeya Johnson

Date:February 12, 2002

Introduction and Overview of Design Project

Our design project consists of two major projects. The first project is to setup a testbed

for the D-Link DGE-500SX PCI Gigabit Ethernet card pictured below in Figure 1. The purpose

of the testbed is to accomplish the same results as last semesters Intel/Agilent OE groups. The

second project is to design and build our own optoelectronic circuit with an 850 nm Vertical

Cavity Surface Emitting Laser (VCSEL) and PiN photodetector. This circuit will interface to the

Maxim Tx and Rx boards via SMA coaxial cables.

Figure 1. D-Link DGE-500SX PCI Gigabit Ethernet card

In implementing the D-Link test bed, we will disconnect the optoelectronics module,

located on the D-Link card, by de-soldering techniques and mounting it onto an external PCB.

Next, the external optoelectronics module will be reconnected to the card using an SMA cable.

The cable will be approximately 1 ft to 11/2 ft long and will be cut in half and partially stripped

back to expose the wires within the cable. These visible wires will be connected directly to the

OE module within the PCB and should be no longer than 1cm.

The second portion of the design project is to compose our own optoelectronics module

by designing a laser and a photodetector. Several companies were considered for the selection of

laser and photodetector components. The components chosen have to agree with the

specifications of the 3287 Maxim testing kit used by the receiver and transmitter group. Some of

these specifications are as follows:

Adjustable DC Bias Current

Adjustable Modulation Current

30 mA Laser Modulation Current

20 psec of Jitter (left to right)

1.5-1.6 Threshold Voltage

3-3.5 Forward Supply Voltage (from Maxim Chip)

Monitor Diode (run from Chip)

Several components from the companies of Lasermate and Honeywell met the Maxim

specifications. For example, the HFE4380-521 and HFE4384–522 series of VCSELs form

Honeywell have relatively low threshold voltages, slope efficiencies, 2.5V forward supply

voltages, 200 psec rise time, and 1.25 Gbps. VCSELs with these such specifications will be

compatible with the Maxim board.

Testbed

The testbed for the D-Link card will be the same as last semester is shown in Figures 4

and 5. The optoelectronics module, Agilent HFBR 53D5, on the D-Link card is the same as the

one used last semester on the Intel/Agilent card. The OE module is the little metallic box on the

left side of the D-Link card in Figure 1. It contains a VCSEL emitter and PiN photodetector

along with some electronics to drive the emitter and amplify the output signal. A schematic of

an Ethernet card that uses the Agilent HFBR 53D5 optoelectronic transceiver is shown in Figure

6.

Figure 4/5. The front and back view of last semester testbed (Intel/Agilent)

The first step in making our testbed circuit will be to de-solder the OE module from the

D-Link card. We will consult with Edgar on how to properly de-solder the module and then

solder it on the PCB shown in Figure 7. Once the OE module is properly mounted, we will then

build our circuit on the PCB consisting of surface mount (SMT) resistors and capacitors. We

will use SMA connectors/cable to interface our testbed circuit with the D-Link Ethernet card.

Figure 6. Schematic of Ethernet card that uses the Agilent HFBR 53D5 transceiver

Figure 7. PCB used for the testbed circuit

Some possible problems may be encountered. One key problem is the transmission line

effects of the components and connectors. To avoid this we will use surface mount components

and high-frequency SMA connectors with coaxial cable. We are using two separate 5V power

supplies, one for the receiver and one for the transmitter. The reasoning is that we will not need

to have a power supply filter on the board to eliminate any noise, which might cause

interference.

The Emitter Module

In designing the emitter portion of the optoelectronics module, a vertical cavity surface

emitting laser (or VCSEL) is the most practicable device. We will be operating in the 1 gigabit

range, so our lasers need to be 1 Gbps or better. We investigated various Honeywell, Lasermate,

New Focus, and Mytel VCSELs. Each laser examined by these manufacturers was 1250 Mbps

or 1.25 Gbps.

The product line carried by New Focus was extremely expensive when compared to the

others, and we did not want to incur those types of costs. There was not much background

information on the Mytel line of lasers. Even though they were relatively cheap, we were not

sure of how dependable they would be in shipping and overall laser quality. The final decision

was then made between various lasers made by Lasermate and Honeywell. This decision was

not as easy to make.

The final decision came down to the HFE4380-521 and HFE4384-522 both manufactured

by Honeywell, and the TSC-M85A4X manufactured by Lasermate. There were many

characteristics that were key in our decision. The most important characteristic is the threshold

component. The threshold current is the amount of power needed to turn on the VCSEL. A

decreased threshold current is desirable, because it also leads to an increased mean time to failure

(MTTF). The MTTF is the average amount of time the device can be operated above threshold

without dying out, or in essence, failing. The threshold currents of each laser can be found in the

parameter charts illustrated below in Figures 9, 10, and 11. Similar to the threshold is the slope

efficiency. The slope efficiency is a measure of how many amps will be needed to give you a

certain amount of light out of the laser. Higher slope efficiency is desirable. We want a

decreased amount of current to produce an increased amount of light. The typical slope

efficiency of the HFE4384-522 was .15mW/mA. This was the highest of all three candidates

and as a result, the most expensive. Due to monetary constraints, we were left with the

Honeywell HFE4380-521 and the TSC-M85A4X by Lasermate. Due to the type of board we will

be employing (Maxim 3287), the VCSEL that we chose will not need to include a back monitor

diode. The back monitor diode is a photodetector that monitors the VCSEL functionality and

output power. If the power decreases, it shows that input power needs to be increased.

Some other characteristics of each laser that were taken into consideration include the

peak optical power, forward voltage, rise and fall times, and deterministic jitter. The peak

optical power of the Lasermate TSC-M85A4X and the Honeywell HFE4380-521 was .31mW

and .35mW respectively. These values did not differ greatly enough to base our decision off of

output power alone. We then looked at forward voltage. The forward voltage of the Lasermate

TSC-M85A4X and the Honeywell HFE4380-521 was 1.9V and 1.8V respectively. Again, these

values are extremely comparable. Next, we looked at rise and fall times. The inverse of the rise

and fall time result in frequency. The rise and fall times of both devices range from 130ps to

230ps, all of which result in frequencies well into the gigabit range. The deterministic jitter of

the devices is the determining factor of hoe wide the eye on the eye diagram will be. We need to

minimize this attribute in order to achieve a clear (“open”) eye diagram. In weighing all the

characteristics, the decision was made to purchase the Lasermate TSC-M85A4X.

Figure 9. VCSEL parameters of the HFE4384-522, manufactured by Honeywell.

Figure 10. VCSEL parameters of the HFE4380-521, manufactured by Honeywell.

Figure 11. Laser characteristics of the TSC-M85A4X, manufactured by Lasermate.

The Photodetector Module

In designing the photodetector component, we will be researching the P-i-N photodiode.

Some of the important attributes include the responsivity [units: A/W], which is the amount of

current generated for each watt of light power. In addition to responsivity the capacitance of the

P-i-N junction must be low to accommodate for a high frequency, which corresponds to the data,

transfer rate of over 1 Gbps. The dark current [units: A] of the detector must be very low (~

1nA) so the device does not produce any current when there is no light on the detector. Finally,

the photodetector should be connectoriezed with SC connectors so that the fiber optic cable can

be easily attached.

The OSI FiberComm FCI-H250G-GaAs-100 is a 2.5Gbps GaAs Photodetector with a

transimpedance amplifier (TIA). The characteristics of this photodetector are

Figure 12. Elecro-Optical characteristics of OSI FiberComm FCI-H250G-GaAs-100

shown in Figure 12. The responsivity of the OSI photodetector meets our specifications, but the

price is expensive compared with the Lasermate device.

The Lasermate RST-M85A-306 photdetector meets all the design specifications for

Gigabit Ethernet. Figure 13 summarizes the electro-optical properties of the photodetcotor.

The responsivity is typically 0.4 A/W, which means the device puts out 0.4 Amps of current per

1 watt of light detected. The device comes connectorized with SC fiber optic connectors and

pins to mount on a PCB. The price is around $15 per detector, well below that of the OSI

photodetector ($55-65).

PARAMETERS  SYMBOL  MIN TYP MAX  UNIT   TEST CONDITIONS Responsivity (1) R 0.35 0.4 - A/W VR=5V, = 850 nm

Dark Current ID - 1 2 nA VR=5V

Breakdown Voltage VBD 50 85 - V IR=10A

Capacitance (2) C - 1.2 1.5 pF VR=5V, f=1 MHzFigure 13. Elecro-Optical characteristics of Lasermate RST-M85A306

Link Budget

The VCSEL and photodetector will be connected to transmitter (TX) and reciever (RX)

circuits, respectively. The specifications of these boards were analyzed to determine the

appropriate values for the design parameters of the VCSEL and photodetector. The purpose of

the link budget is to make sure that the TX circuit will be powerful enough to drive the laser, and

that the photodetector output current will be powerful enough to drive the TIA on the RX. As

mentioned earlier, two different Honeywell and one Lasermate VCSEL was compared. Each of

these options was SC connectorized. The slope efficiency varied for each of these emitter

devices. For the photodetector, a Lasermate SC connectorized model was used. The design

specifications that will be discussed in calculating the link budget were included in Figures 9, 10,

11, and 12.

To determine whether or not the TX could drive the laser, the DC bias or drive current of

the TX was obtained from the MAX3287 specifications. The Maxim board has an adjustable DC

bias current, between 0 and 300 mA. Next, the drive current of the VCSEL was obtained. The

threshold (drive) current, Ith, of the VCSEL must be DC biased, from 1.2 to 1.5 times its value. In

other words, if the maximum Ith of the VCSEL is 6mA (as it is for both Honeywell VCSELs),

then it needs to be DC biased to 7.2 mA. This is well within the range expected for this project,

4-8 mA. This value of 7.2 mA is also well below the DC current coming into the laser from the

TX (max 300mA), meaning that the TX will drive the VCSEL. Only 7.2 mA are needed to drive

the VCSEL at fast switching speeds (lasing speeds, rather than LED). These high witching

speeds are required for gigabit applications.

The modulation current of the TX is adjustable between 2 and 30mA. This means that the

range of current that will be modulated lies between 7.2 mA and 30 mA. The light output of the

laser is calculated using Equation 1.

Ith * slope efficiency = Light [Watts] (Eq. 1)

Using a low slope efficiency of 0.04 mA/mW (HFE4380-521), 0.288 mW of power are

generated from the VCSEL. Using a laser with a high slope efficiency of .15 mW/mA

(HFE4384-522), 1.08 mW of power are generated. The maximum power or range of emitted

light from the laser is maximized at 1.2 mW for the low slope efficiency VCSEL, and at 4.5 mW

for the high slope efficiency VCSEL.

This range of output power is not all incident on the photodetector. Power is lost at the

SC connectors and in the fiber. The exact losses have not been estimated, but it is thought that

the maximum loss will be around 3dB. For purposes of computing the link budget, a 3dB loss

will be assumed. This factor is converted to attenuation using Equation 2. For 3 dB of loss, the

attenuation is 0.5, or half the power. Taking half the range of output power will provide the

actual amount of light incident on the photodetector.

Attenuation = 10-(dB loss/10).

The responsivity of the photodetector is also very important. It is expressed in A/W. The

new range of output power is multiplied by the responsivity of the photo-diode produces the

current that will go into the transimpedance amplifier. Based on the specifications of the

transimpedance amplifier on the MAX3266 , this current should be greater than 80 micro Amps.

This is the minimum value needed to drive the RX. Below are the calculations taken to generate

the link budget and the results, both for the high slope efficiency and low slope efficiency

VCSEL.

From this table, it is evident that to drive the RX at 80 micro Amps, the high slope

efficiency VCSEL will be much more reliable, having a range of currents beginning at 189 micro

Amps as shown in Table 2. If the losses in the fiber and the connectors turn out to be less than

1/2 the output power, then the low slope efficiency VCSEL might still be a good option (and a

cheaper one). Another alternative could be to find a photodetector with higher responsivity.

Nonetheless, the HFE4384-522 VCSEL seems like a great option. It can be driven by the TX,

and it can drive the RX without a problem, even after considering an attenuation of 1/2. The

prices have not been obtained, so a final decision cannot be made until this information is

Parameters & Equations HFE4380-521 HFE-4384-522Max Ith (Threshold current) 6 mA 6 mADC bias of laser = 1.2 *( Ith) 7.2 mA 7.2 mASlope efficiency 0.04 mW/mA 0.15mW/mA Power @ DC bias = Slope efficiency * DC bias 0.288 mW 1.08 mWModulation current (TX) 30 mA 30 mAPower @ Mod. current = Slope efficiency * Mod. current 1.2 mW 4.5 mW

Range of Emission from VCSEL (without losses) 0.288 - 1.2 mW 1.08 - 4.5 mWdB Losses (or Attenuation = 10^-(dB loss/10)) 3dB (or ½) 3dB (or ½)Range of Emission from VCSEL (with losses) 0.144 - 0.6 mW 0.54 - 2.25 mWTyp. Responsivity of Lasermate's PD 0.35 A/W 0.35 A/WMin. current from PD = min(Range of Emission)*Resp. 50.4 uA 189 uAMin. current from PD = max(Range of Emission)*Resp. 210 uA 787 uA

known. In addition to cost, an exact estimate of the losses incurred through fiber is yet to be

determined.

Conclusion

A Fiber optic cable will be used to connect the VCSEL laser to the P-i-N photodetector.

An ideal fiber optic cable would have no power loss but in reality all fiber optical cables has

some power loss per distance. The power loss is expressed in dB/km, which means the decibel

loss per kilometer of cable. The Infinicor 600 fiber optic cable by Corning is

telecommunication grade fiber that has 50 um diameter. The fiber optic cable attenuation is 3.5

db/km for the 850 nm VCSEL that will be used. Finally the fiber optic cable will employ SC

connectors to allow for easy connection to the devices.

The link budget for this optical circuit is calculated by measuring the current required to

drive the laser that is then converted to light. The light sent down the fiber optic cable is then

detected by the photodetector and converted back to an electrical current. The operating current

of the Lasermate TSC-M85A4X VCSEL is 12mA. That is converted to light and sent through

the fiber optic cable with a low loss of 3.5 db/km. Finally the current at the photodetector is

equal to the input power multiplied by the responsivity.