Design of the hottest burst mode transceiver chips

Design of burst mode transceiver chipset

1. Preface

due to the increase of new broadband applications and demands, it has developed rapidly, and bandwidth has gradually become the key to restrict the broadband performance of DSL users. PON (passive optical network) can solve this problem. It has the characteristics of high bandwidth and transparent network protocol. It is a point to multipoint optical fiber transmission and access technology. It has become the preferred solution to deal with the last mile of central office connecting to end users

from the structure of the passive optical fiber network system in Figure 1, it can be seen that there are only passive components such as optical fibers and optical splitters between the OLT and the ONU. It saves optical fiber resources, eliminates the need for active devices and equipment, and can effectively reduce network costs. At the same time, because of its simple structure, it greatly saves operation and maintenance costs

at present, passive optical network (PON) has become the main network structure of fiber to the home FTTH (fiber to the home). The downward direction of the system is from OLT to ONU. During downlink communication, TDM mode is used to gather all data packets into a data stream, which is transmitted by broadcast through optical fiber. At the receiving end, the user selects the appropriate address to obtain the data packet he needs. Uplink communication (from ONU to OLT) is quite difficult because multiple users use the same optical fiber for transmission. Only one user is allowed to transmit data packets to OLT at any time. TDMA protocol can be used to guarantee the above conditions. That is, the ONU is required to be in the off state when there is no signal transmission, and to be turned on quickly when transmitting signals, which requires the ONU to support special burst mode transmitters and receivers. Therefore, uplink access is the key of system design, and the transceiver supporting burst mode has become the focus and difficulty of the whole system

at the local end (OLT), the laser driver LDD is in continuous working mode, while the transimpedance amplifier TIA and limiting amplifier La need to support burst working mode. At the user end (ONU), the transimpedance amplifier TIA and limiting amplifier La support continuous working mode, and the laser driver LDD needs to support burst working mode

from Figure 1 and Figure 2, we can know the general configuration of passive optical fiber network. Some chips need to adopt burst mode, such as bm-tia, bm-la and bm-ldd. This article will talk about the principles and characteristics of these three chips. The focus is on burst mode laser driver bm-ldd

the following mainly describes the main differences between burst mode transimpedance amplifier (bm-tia), burst mode limiting amplifier (bm-la), burst mode laser driver (bm-ldd) and continuous mode laser driver

II. Burst mode TIA and burst mode la

1 The main difference between burst mode TIA and general TIA is that it must deal with the input signal without DC component (DC coupling) and the output amplitude change between bursts is required to be greater than 30dB. The problem of output offset voltage can be solved together with the subsequent bm-la. It can also be solved by the offset detection circuit shown in Figure 3 below

offset control: the mechanism by which the offset control circuit can remove the burst output offset one by one is to adjust the offset before each burst arrives. This circuit is also called "adaptive threshold control circuit"

how does this circuit eliminate the output offset voltage? Now let's ignore the current source IOS first. Before the burst message arrives, the peak detector is reset to the two output terminals of the amplifier, and both outputs common mode voltage. At this time, the differential output voltage is 0V. Then, the first "1" of the burst arrives, VOP rises and von falls. The peak value of VOP is stored in the peak detector and fed back to the input of the amplifier. When the next "0" bit arrives, the voltage value of the peak detector will appear at the von output. Because there is no voltage drop (no input current) on RF ', there is no voltage across the input terminal of the amplifier, and there is no voltage drop across RF (no photocurrent). Therefore, the peak values of the two output signals are equal, that is, the output offset is eliminated. The differential transimpedance of this burst mode TIA is about 2 · RF

chatter control: for bmtia, in addition to offset control, there is another trouble. That is, the interval between bursts is too long, and the burst signal is not received for a long time, that is, the peak detector reset time of TIA remains too long, then the noise of the amplifier will overlap with the output signal to produce a sequence of random codes, which is called chirp

one way to solve this problem is to artificially use the current source IOS to introduce a small offset voltage, as shown in the dotted line in Figure 3. This bias voltage must be greater than the peak noise voltage to control the chirp, but it cannot be too large, otherwise the sensitivity will be reduced

2. What is the difference between burst mode La and continuous mode La? Fig. 4 (a) shows a single side continuous mode La, whose input signal passes through the AC coupling network. The data signal is clipped by the average DC of the AC coupling network. Because the continuous mode signal is DC balanced, its average value corresponds to the vertical center of the eye diagram. If we use AC coupling to process burst mode signals, as shown in Figure 4 (b), the clipping level is still above the average level of the signal, but there will be many bit errors, pulse width distortion and some undesirable consequences in the vertical center of the current eye diagram

in burst mode, "Dr. Huimin Cao, chief scientist of DSM functional materials business department, the TIA and La (also including CDR) of the receiver must be DC coupled, and the various offset voltages generated by these devices must also be controllable. For bmtia mentioned above, if the offset caused by La is not large, there is no need to add an additional offset control circuit. On the contrary, if TIA has no offset control or single ended output, LA must have the function of completing offset control

Fig. 5 shows a single ended output LA with clipping control. The symbol VSL in the figure represents the slice level (which can be understood as the average DC of the single ended output signal). A pulse peak and valley detector is used to determine the maximum and minimum values of the input signal respectively. The average value is obtained from the matching resistors R and R 'and is used as the clipping level VSL. This circuit always clippes the input data signal at the center of the vertical eye diagram, which is independent of the average value of the input data signal. In order to accurately obtain the pulse amplitude of each burst data signal, the peak/Valley detector must be reset during each burst interval. It must be borne in mind that in burst mode systems, it is common for the signal amplitude to change greatly from one burst to the next

the loud/soft ratio shown in Figure 1 is large

III. burst mode laser driver

laser driver has two working modes: continuous mode laser driver, abbreviated as LDD; Burst mode laser driver, abbreviated as bm-ldd. Abbreviations are used below. The so-called burst is to send a batch of data information suddenly and instantaneously, and stop sending after sending, and the stopping time can be long or short, with randomness. The main difference between bm-ldd and LDD is:

(1) bm-ldd requires a high extinction ratio between burst and stop burst

(2) the automatic power control (APC) of bm-ldd has a hold function, which can ensure the correct transmission of burst data signals

1. Extinction ratio between bursts

burst mode transmitters are usually used for multi-user networks. Therefore, it is required that the unwanted light output (interference light) during the burst interval is very low. For example, in the PON system, there are 32 users, and the optical data signal output of one company will be "polluted" by the interference of the other 31 companies in the common media. These background lights will reduce the extinction ratio of light output. In general, for bmldd, the extinction ratio between bursts should be greater than 30dB. The extinction ratio during burst is ~10db, which is similar to that of continuous mode LDD transmitter

in order to achieve such a high extinction ratio, the bias current of the laser should be as low as possible (that is, the operating point should be just a little lower than the threshold current of the laser) or zero between the secondary bursts. In case of burst transmission, the bias current can be quickly increased to reduce the conduction delay and jitter. In addition, in order to reach the extinction ratio of 30dB before the next burst emission of any other user, bmldd often connects a transistor in parallel with its laser diode to quickly remove the carrier (darken) at the end of the burst emission

2. Automatic power control

in order to meet the burst mode, the burst mode laser driver bm-ldd adopts an automatic power control circuit (APC) with hold function. Giving the automatic power control circuit can maintain the DC bias value of the automatic power control circuit of the last burst sequence, so that when the next burst sequence occurs, a stable APC loop can be established quickly. The following three methods can be applied to burst mode APC

（1）. As shown in Figure 6, the current from the power monitoring photodiode is converted into voltage through broadband TIA, and then the peak detector detects the peak corresponding to the emission power. The laser output power can be adjusted until the peak output voltage is equal to the reference voltage Vref

however, the method in Figure 6 has some disadvantages: 1) when there is no burst emission for a long time, before the next burst arrives, simulate the output of the peak detector. Does that mean that the plastic material is out of date? It will not deviate from the actual peak value, which will cause error in output power. 2) When running at high bit rate, a fast photodiode is needed. 3) When running at high bit rate, the power consumption of TIA and peak detector is quite prominent

（2）. The burst mode APC is realized by a circuit of integration and discharge and a digital memory

this APC can overcome the above shortcomings, as shown in Figure 7

the power level is stored in a digital up/down counter, and the count stored in the counter will not change with time. When the power supply is turned on, it can maintain a precise power level. Therefore, in May 2012, DSM announced that it would launch 100million euros of knowledge and innovation investment in the Netherlands at the beginning of the emergency, and the error could be avoided. During a burst, the "burst clock" signal can be used to increase or decrease its power level. The increase or decrease of the count can be eliminated by the high and low temperature experimental chamber. These factors are determined by the rising/falling signal "u/d". The generation method is as follows: before the burst starts, capacitor C (node x) is discharged by reset switch Sr and simply discharged. During burst emission, the photodiode current IPD (T) charges (integrates) the capacitor C, and the desired peak current Iref is modulated by the transmission data through the switch SD to generate Iref (T), that is, the discharge current that discharges the same capacitor C. When the burst emission ends, the voltage generated by the capacitor charging is as follows:

where IPD is the peak current of the photodiode, the number of ones during the burst period, and B is the bit rate. The comparator compares this voltage with ov (reset voltage). It can be seen from the above formula that the result of this comparison is only affected by ipd-iref and has nothing to do with C, B and N1. If the difference is positive, count