Optoelectronic Detector and Related Circuit
The present invention relates to an optoelectronic detector and related circuit arrangement .
Optoelectronic detector circuits are widely used in modern high-speed Internet packet switching systems. Increasing demands on Internet package switches result in the need for high density, high speed, photo receiver arrays . One potential barrier to achieving higher packing densities is the choice of detector device, with many designs relying on the use of externally connected PIN diodes or APD's.
Conventional photo receivers compatible with monolithic integration with CMOS digital logic offer data rates of only a few Mbit/s. This is because a p-n junction photodiode is used since the standard CMOS process does not offer an intrinsic layer necessary for a p-i-n diode. The thi depletion region of the p-n junction diode results in only a fraction of the incident photons being captured in the depletion region, and it is here where there is a large electric field • to assist the separation of the photo generated electron-hole pairs and their sweeping out of the depletion region. Photo generated minority carriers created outside the depletion region either recombine elsewhere and therefore do not contribute to the photocurrent or must diffuse to the depletion region to contribute to the photocurrent . This is a slow process and results in poor quantum efficiency (i.e., the number of electrons contributing to the photocurrent per incident photon) and, more importantly, a slow component of photocurrent which
compromises speed. That is, in response to an incident light impulse, the photocurrent exhibits a fast component that rises and decays rapidly and which is superimposed on a slow component that decays slowly and forms a "tail" .
The use of transistors as detectors is unsatisfactory at high frequencies, predominantly because of such unwanted "tails" in detected pulses caused by the slow-moving diffusion of carriers generated in the base of the transistor.
Currently, there are two approaches to overcoming this problem:
First, fast" photodiodes employing compound semiconductor technology are hybridised with silicon CMOS. This typically requires flip-chip solder bump technology. The disadvantage of this approach is that the parasitic capacitance of the large solder bumps compromises speed; the solder bonds and the different thermal expansion coefficients of the chips reduce reliability and yield and hence increase cost. Also, the area taken up by the bonding pads prevents the achievement of small photo detector array pitches (circa 10 urn) which are thought to be required for high-density optoelectronic interconnects for integrated circuits applications .
Secondly, spatially modulated light detectors have been provided. The requirement is to remove the slow current component by detecting the difference current between interdigitated electrodes on monolithically integrated
(CMOS) photodiodes. The disadvantages are a reduced quantum efficiency due to the interception of the light by the metal electrodes on the photo detector surface; the large photo detector area required increased receiver complexity and bit rates limited to in the region of 300 Mbit/s.
The present invention seeks to provide for an optoelectronic detector, and related circuit arrangement, exhibiting an advantage over known such detectors and circuits and particularly which allows for a relatively high quantum efficiency and related speed of operation.
According to one aspect of the present invention, there is provided a p-i -n photodiode comprising a reversed bias base- collector junction of a silicon based integrated HBT device.
Speed and quantum efficiency of such a device represents an improvement over currently available devices .
A Si/Ge HBT BiCMOS offers a process compatible photodiode that can include an intrinsic region. The emitter of the HBT structure is in fact omitted. The base acts as the p- layer, the collector acts as the intrinsic layer, and the sub-collector acts as the n-layer. The whole of the intrinsic region can be depleted by a reverse bias that its less than the structure's breakdown voltage. The width of this depletion region is well matched to the absorption length of light in Si-Ge at IR wavelengths compatible with VCSEL sources (0.8-1.0 urn) . This can lead to a high quantum efficiency and little diffusion component to the photocurrent. Moreover, the reverse biased parasitic diode between the sub-collector layer and the substrate sweeps
away any carriers generated by photons captured in that region further reducing the potential for a diffusion component to the current .
One aspect of the present invention relies on the use of the reversed biased base-collector junction of an integrated SiGe HBT as the detector. The base collector and subcollector layers of the HBT in question form the p-i-n layers of the photodiode detector. Advantageously, the device can be formed as part of a current AMS 0.8 m SiGe HBT foundry process. Instructions are given to the foundry engineers not to grow the emitter layer of the device.
The basic photodiode structure has a response extending to circa 40 Gbit/s. Currently the speed limitations are caused by the receiver circuitry although improvements can be expected through better circuit design and the use of reduced feature size processes.
According to another aspect of the present invention there is provided a monolithic semiconductor structure comprising a p-i -n photo diode formed of a reversed bias base-collector junction of an integrated HBT device and a transimpedence amplifier arranged to be driven by an output from the HBT device, the transimpedence amplifier including a negative feedback buffer.
An advantageous aspect is that the circuits required can be very simple, take up little chip area and have low-power dissipation and so enable the formation of high density arrays .
As mentioned, the output current from the photodiode device can be used to drive the base of another HBT connected to form of a transimpedance amplifier.
It should be appreciated that SiGe can prove to be a better choice than Si for use as a p-i-n photo diode operating at 850 nm because its absorption at this wavelength is almost an order of magnitude greater.
As will therefore be appreciated in order to minimise the tail effect, it is important that most, if not all, of the carriers are generated within the junction's depletion region. Increasing the depletion region width in this way also improves performance by ensuring that the carriers drift at, or close to, saturation velocity.
The invention is described further hereinafter, by way of example only, with reference to the accompanying drawings in which:
Fig. 1 is a schematic perspective view of a SiGe HBT photo diode device employed within an embodiment of the present invention;
Fig. 2 illustrates a SiGe HBT for use in the present invention;
Fig. 3 illustrates a bandgap diagram for a SiGe HBT structure as employed in accordance with an embodiment of the present invention;
Fig. 4 is a first example of an equivalent small signal
diagram for a photo receiver embodying the present invention;
Fig. 5 is a circuit diagram of a monolithic device according to an embodiment of the present invention and comprising a photo resistor and a transimpedence amplifier;
Fig. 6 is a trace illustrating the dynamic resistance of the load illustrated in the circuit arrangement of Fig. 5;
Fig. 7 is a trace of the dynamic resistance of the .feedback device employed within the circuit arrangement of Fig. 5;
Fig. 8 is a trace illustrating the photo receivers dynamic load as plotted against frequency;
Fig. 9 is a trace illustrating the transimpedence game of certain arrangements such as that illustrated in Fig. 5;
Fig. 10 illustrates a trace of the peak output voltage plotted against frequency of the circuit arrangement • of Fig. 5 ; and
Figs. 11 and 12 illustrate traces of the transients in relation of the output circuit arrangement of Fig. 5 at point 5GHz and 1GHz respectively;
Fig. 13 is a circuit arrangement of a monolithic device including a photo receiver and transimpedence amplifier according to an alternative embodiment of the present invention;
Fig. 14 is a similar circuit diagram according to a yet further embodiment of the present invention;
Fig. 15 illustrates the transient response arising in relation to the circuit of Fig. 14;
Fig. 16 is a trace representing the AC analysis arising in relation to the circuit of Fig. 14;
Fig. 17 is a second example of an equivalent small signal diagram for a photo receiver embodying the present invention; and
Fig. 18 represents the bandwidth of a detector employing the photodetector of Fig. 17.
In accordance with a particularly advantageous aspect, the invention seeks to provide for a photo receiver design employing a small surface area, low power consumption, low heat dissipation and should be suitable for mass production and for operation at >1 GBps .
Turning first to Fig. 1, there is illustrated a schematic perspective view of a SiGe HBT photodiode which can be employed in accordance with the present invention. As noted, no emitter region need to be provided for the device and such emitter region is commonly omitted at the time of device formation. In Fig. 2 there is illustrated the general structure of a SiGe HBT for use in an embodiment of the present invention.
Fig. 3 illustrates a band gap diagram for the emitter, base and collector of the SiGe device and which, as is illustrated, includes a Ge doped region that extends only a relatively short distance into the collector region. While the advantages of the present invention can be achieved by means of such a silicon absorption layer device, particularly if wavelengths" in the order of 650nm are considered and as is described in further detail later, it should be appreciated that an increased concentration of Ge in the collector region can lead to use at a wavelengths in the order of 850nm.
For ease of understanding of the calculations that follow, and that relate to a SiGe absorption layer for use with wavelengths in the order of 850nm, Fig. 4 illustrates an equivalent small signal model 10 disposed between a cathode connection 12 and anode connection 14 and which comprises parallel connected elements current source Iph • 16, capacitors Cj 18, an ideal-value diode 20 and resistor Rsh 22. These parallel connected elements being connected in the series with a resistor Rs24.
In a preferred embodiment, the diode's dimensions have been calculated to accept a typical IR LASER beam (Beam Area - 10 m2, Pλ=1.25rnW). The depletion region width required in order to minimise the "tail" effect can be calculated as follows,
W ≥ a(iS0nm)
SiGe has higher absorption at 850nm than Si , Table 1 . SiGe has higher absorption at 850 nxn than Si, Table 1.
This property helps reduce the necessary depletion region width needed,
The minimum electric field required to achieve saturation velocity in SiGe is IxlO7 V/m. At 1.25 μm, the total reverse bias voltage (including the built-in potential of 0.97V) required is given by,
V„ = rlO7 x 1.25x10* = 11.5V
Therefor, the reverse bias voltage required > H.5V and is limited by the junction's breakdown voltage (14N).
Using Poisson's equation, for a given total reverse bias voltage of 12.5 N, the depletion width can be calculated as follows,
This biasing- arrangement results in a depletion thickness almost equal to the collector depth, this greatly minimizes the unwanted 'tails' in the output pulses.
The carrier transit time in the depletion region is given by,
The junction capacitance and series resistance are determined as follows,
, - . ..6 1.6 x 10^ , _
Rs = - = 6.25 x 10 β x rτ- = lΩ
• I x lO-"
Given the large depletion width, a quantum efficiency of 90% can be expected (assuming appropriate anti-reflection coatings) and the diode's responsivity obtained,
In this application, the light source is provided by a VCSEL diode with an output power of 1.25 mW, positioned close to the detector (no power loss from transmitter to receiver is assumed).
Thus, the photo-generated current for an incident power of 1.25 mW is as shown,
Detector Junction Parameters
Current from the reversed biased junction detector feeds a transimpedance amplifier formed by a HBT common-emitter amplifier and a MOSFET negative feed back buffer as illustrated in Fig. 5.
Fig. 5 illustrates a small signal equivalent circuit 10 such as that illustrated with reference to Fig. 4, and, as mentioned above, the current produced is arranged to feed the transimpedence amplifier formed by the HBT common emitter amplifier Ql which' is itself connected to a MOSFET load Ml, and which arrangement also includes a MOSFET negative feedback buffer M2.
- 28.35 μA, 7C » 210.8 μA, CQ=3.278V
The circuit's total power consumption = 210.8 μA x 5V s 1 mW
The emitter's series emitter is approximated by,
0.027 rc= ■ " - =I28Ω * 211x10-*
The DC current gain is obtained from,
β. I..2™8.35 χ!l£0~6 m 7Mx
The dynamic resistances of both the MOSFET load and feedback device can be obtained by simulation as shown in Figs . 6 and 7. With- reference to Fig. 5, Fig. 6, shows the dynamic load resistance of device M-. and Fig. 7 the dynamic load
resistance of device M2.
Fig. 7 illustrates a trace of the dynamic feedback resistance of device M2 of Fig. 5.
The amplifier's voltage gain is given by.
j _^ 4000
I2S
The diode ' s load resistance can also be obtained by simulation as shown in Fig. 7 and calculated as below,
The diode's dynamic load Vs Frequency is illustrated in Fig. 8.
The detector junction's bandwidth is dependent upon its junction capacitance and the load resistance used to convert the detected current into voltage .
I
BW = I
= 426χI0"Hr
2-R. , 2x .142x450 0.83x10''
The HBT's model parameters are given in the table below.
Table 3, AMS HBT NPNl 11 Model Parameters.
The AMS foundry's own BSIMv3 MOSFET models are used, dimension and resistance values are given in the table below.
AMS NMOS Dimensions.
The circuit was simulated using MiCROSIM's PSPICE A/D package's transient analysis. A pulse generator was used to simulate the input photocurrent.
The simulated peak output for a given peak input current of
775μA, over the entire frequency of operation, was obtained and graphs of transimpedance gain Vs Frequency and peak output voltage Vs frequency were plotted as illustrated in
Figs. 9 and 10 respectively.
The measured bandwidth is much less than that calculated for the photoddector.
The overall bandwidth is reduced by parasitic capacitances in the Mx load, M2 feedback resistance and the HBT.
Although intended for operation to 2 Gbps, the amplifier's capability to switch a light CMOS load to above a + 1.5V logic threshold at greater than 2 Gbps is shown in Fig. 11 and with the transient analysis of the output at 0.5 GHz and at 1 GHz being shown in Figs. 11 and 12.
Should greater sensitivity be required, one or more CMOS buffer amplifier stages (M3x and M4x) , of preset gain, can be added to the design as shown in Fig. 13.
In Fig. 13, there is again illustrated an equivalent small signal model 11 for the photo receiver device again driving a transimpedence amplifier Q2 connected to a negative feedback MOSFET buffer M2 and a MOSFET load Ml but, as mentioned, an additional two CMOS buffer amplifier stages M3A/B and M4A/B also being provided.
Fig. 14 illustrates a further variant of the photo receiver circuit which advantageously employs inverter elements Ml and M2 which serve to reduce the charge/discharge time as seen by the photodiode .
Figs. 15 and 16 illustrate the transient response, and AC
analysis respectively of the alternate circuit arrangement of Fig. 14.
It is also viable to operate the detector at 650nm and with the absorption layer comprising Si since the absorption of Si at 650nm is almost a factor of ten greater than that at 850nm, and this therefore improves the photodiode ' s responsivity. Although not presently available, red VCSEL sources for 650-70nm and are likely to represent appropriate sources.
Thus, although the foundry process employed at manufacture could be varied so as to increase the Ge concentration in the collector region by adding a custom diffusion step to the standard process, and so as to greatly improve the detector's absorption for operation between 650 and 850nm as discussed previously, operation at wavelengths in the order of 650nm is also available as explained in the analysis below.
Some of the formulae used for the calculation of parameters in the simulation have been taken from a PSPICE model in order to ensure the smallest margins of error. The results of this analysis are as follows and arise from the same calculations and simulations noted above for operation with 850nm source.
At an input power of lmW, the following results arise.
16
Aτea = 22°μμ m η = 0.6 IS = 0.027»10"
00.= 1.071 fF Cj = 1.85MF Responsivity = 0.314 A/W Rs = 2.4- 103
IΦ = 313.91 l«>μA
VR= 14 Xd = 1.362«μπ_ ι\ - 6.808«pS BW -= 85.60K-Hz (for 1 KOhm resistive load)
However, since it is unlikely that a loss-less link will be 0 implemented, it is necessary to include a minimum-6dBm (75%) power loss at the receiver for the following point-to-point link simulations. The re-calculated values for the new input power of 0.25mW are listed below.
5 Area = 22«μμ m η =0.6 IS = 0.027-10" 16
CJO:= 1.071 fF Cj = 1.859°-F Responsivity = 0.314 A/W
RS S 2.4-103 IΦ = 78.478yA
VR= 14 Xd = 1.362°μm τt = 6.808°pS BW = 85.601«GHz (for 1 KOhm resistive load)
Using model parameters from the foundry process PSPICE model files, it is possible to produce a new PSPICE photodiode model 110 for use in PSPICE simulations and this is illustrated in Fig. 17. The arrangement of the elements of the model is generally similar to that in Fig. 4 but here: resistor 124 has a value of 0.0024 Ohms; resistor 122 has a value in excess of 100 MOhms; diode 120 remains ideal; capacitor 118 has a value of 1.071 f and the current source 116 produces a current in the region of 70μA.
Using this model, PSPICE simulation can be used to verify the detector's bandwidth (~80GHz or 160 GBps) obtained by analysis and a trace of which is illustrated in Fig. 18.
5 As noted, the resultant photo-detected current is 78μA and this value can be readily used to drive a low input impedance/high gain monolithic transimpedance amplifier, fabricated using MOSFET ' s and HBT's from the same BiCMOS process . 10
Thus, as will be appreciated, while SiGe's higher absorption coefficient at IR frequencies enables the SiGe diode to outperform a similarly sized biased Si device, Si nevertheless offers a viable option for wavelengths in the 15 order of 650nm. At wavelengths in the order of 850nm, without the reduction in the necessary depletion width attributable to SiGe's absorption of IR, the bias voltage required to enable the device to operate correctly would have been larger than the device's reverse breakdown voltage 20 in this case (e.g. a Si device in this process would not work) . Even if Si devices are fabricated, in other processes, that can take the extra bias required to achieve a similar level of performance, the replacement of any given Si diode with a similar SiGe device (for a set bias) will 25 always be an improvement .
Advantageously the overall cost of fabricating SiGe devices is far less than that of GaAs devices as it approximates to the cost of Si fabrication.
_. 0
At the wavelengths mentioned, the direct integration of this
photo detector diode in any design will inevitable result in a reduction complexity cost and of the required wafer space when compared with designs implemented using a flip-chip bonded (group III-V material) diode and an increase in bandwidth when compared with designs based on Si diodes/phototransistors .