Reverse Path Design for Docsis 3.0 and Beyond - Part 2: Optical Link Dynamic Range
This is the second in a series of articles about the set-up and operation of DOCSIS networks. The demands of DOCSIS 3.0 and 64-QAM (quadrature amplitude modulation) in the upstream make it necessary to revisit our network design approach. The network must be evaluated from an overall system perspective that considers the cable modem termination system (CMTS), HFC plant, and subscriber devices (modems and set-top boxes) as a part of an overall unified system design.
The optical link has the most restrictive dynamic range of all components in the overall HFC reverse path system. This is because of the limitations imposed by optical link clipping and optical link broadband noise (relative intensity noise, optical receiver noise, etc.). In this article we present data from measurements on two types of optical links in order to quantify the available dynamic range.
Most of the previous work on reverse path system analysis and system design has been based on noise power ratio (NPR). NPR is a useful tool for comparing the performance of competing types of links. However, attempts to use NPR to determine the nominal operating levels for actual RF carriers can become complex and confusing.
It is desirable to identify a technique to characterize a particular reverse path optical link that is more straightforward. Bit error rate (BER) performance of individual carriers is the ultimate figure of merit for the performance of reverse path optical links. Consequently, we characterized the performance of our optical links based on BER. For our purposes, BER will mean the uncorrected link bit error rate (no forward error correction, or FEC).
REVERSE OPTICAL LINK TYPES
For the purposes of this investigation, two types of links were considered. The first type utilized an analog distributed feedback (DFB) laser transmitter to convert the input electrical signals to an optical signal. The output of the transmitter was coupled into an optical link consisting of optical fiber and passive loss. The loss of this link was varied in order to measure the effect on performance. The output of the optical link was connected to an analog optical receiver.
The second link tested was a CISCO bdr® link. The bdr® transmitter contains an A/D (analog to digital) converter to change the input analog electrical signal into a digital signal. A DFB laser in the bdr® transmitter was intensity modulated by the digital signal. The output of the bdr® transmitter was coupled into an optical link consisting of optical fiber and passive loss. As in the case of the analog DFB link, the loss of this link was varied in order to measure the effect on performance. The output of the optical link was connected to a bdr® optical receiver. The bdr® receiver contains a photodiode to change the optical signals into digital electrical signals and a D/A (digital to analog) converter to change the digital electrical signals back into analog electrical signals.
It was our goal in this testing to present a loading scenario that was equal to or worse than the loading that will be experienced by the vast majority of systems in the foreseeable future. Consequently the entire 5-42 MHz band was loaded with carriers. Systems that are more lightly loaded (fewer carriers) than that described herein will exhibit greater dynamic ranges than our results.
The system loading consisted of a spectrum of four 64-QAM RF carriers at 5.12 Msym/s (6.4 MHz channel bandwidth) and three 16-QAM RF carriers at 2.56 Msym/s (3.2 MHz channel bandwidth). These carriers occupied the frequencies from approximately 5 MHz to 40.2 MHz. The amplitude of each carrier was individually adjustable. Initially the spectrum was adjusted for the same composite power for each carrier. The spectrum is shown in figure 1.
|Figure 1. Test Spectrum|
Note that on a spectrum analyzer with a resolution bandwidth that is set to a value less than the occupied bandwidth of the carriers (30 kHz RBW, for example) the 2.56 Msym/s carriers will appear to be 3 dB higher than the 5.12 Msym/s carriers. This is a consequence of the occupied bandwidth of the 2.56 Msym/s carriers being one half of that of the 5.12 Msym/s carriers—that is, the same energy is concentrated in one-half of the bandwidth.
For the purposes of this analysis, we also considered the worst case amplitude inequality scenario. In that scenario we considered a system in which the long-loop automatic level control (ALC) would correct for any amplitude error greater that ±4 dB as sensed at the CMTS upstream input. We also assumed that the frequency response of the splitter/combiner network between the optical receiver and the CMTS was less than ±1 dB in the band from 5-42 MHz. Consequently, carrier inequality would be less than ±5 dB. Based on experimentation, the worst case scenario occurred when the three 16-QAM carriers and three of the 64-QAM carriers are at a level of nominal +5 dB and one of the 64-QAM carriers is at a level of nominal -5 dB. A sample spectrum is shown in figure 2.
|Figure 2. Worst Case Carrier Amplitude Inequality|
Based on the experimental results the 64-QAM carrier at -5 dB with respect to nominal amplitude will be subject to the narrowest BER dynamic range. Note that this was an extremely severe worst case scenario.
LABORATORY TEST SETUP
A test setup was built in the laboratory to measure reverse link performance. A simplified block diagram of this test setup is shown in figure 3.
|Figure 3. BER Test Set-up|
A matrix generator was used to create three 16-QAM carriers at 2.56 Msym/s. These three carriers were centered at approximately 10 MHz. Three Webstar model DPX2203 cable modems were used to generate three 64-QAM carriers at 5.12 Msym/s. The cable modems were loaded with test software that permitted transmission of a continuous random bit stream for testing purposes. The three carriers were centered at approximately 17.8 MHz, 30.6 MHz and 37 MHz respectively. Finally a Rohde & Schwarz SFU broadcast test system was used to generate a 64-QAM carrier at 5.12 Msym/s centered at approximately 24.2 MHz. The generator was configured to produce a continuous test signal using DVB-C. All QAM carriers in this test carry uncorrelated bit streams.
The combined carriers were then applied to the optical link under test. At the output of the optical link under test the test signals were split between a Rhode & Schwarz EFA test receiver and a Sunrise telecom AT2500RQv spectrum analyzer. The Rhode & Schwarz EFA test Rx was used to demodulate the 64-QAM signal from the Rhode & Schwarz SFU Broadcast Test System and provided a raw BER measurement. The Sunrise telecom AT2500RQv was provided to permit monitoring of all carriers during the test. An HP 4378 power meter was also available for calibration of the carrier amplitudes.
The variable attenuators on the input and output of the optical link under test permitted simultaneous adjustment of all carriers. These attenuators were operated in concert so that the amplitude of the carriers at the Rhode & Schwarz EFA test receiver and the spectrum analyzer did not change. The carriers were raised and lowered to allow the operator to determine the BER dynamic range.
ANALOG DFB LINK PERFORMANCE
The initial tests were performed with an analog DFB link. A Cisco model 4013907.1610 analog optical transmitter with Pout=3.7dBm @ 1610nm and a Cisco model P2-HD-RXR-SA, 4012718 Prisma II analog optical receiver were used. In these tests 14 km of fiber was used for the optical path.
BER data was taken at several link losses (5 dB, 7 dB, 9 dB, 12 dB, 15 dB and 21 dB). The initial data was taken with all carriers at equal power levels, as per figure 1. A representative BER curve taken at 7 dB link loss is shown in figure 4.
|Figure 4. 7 dB Link Loss|
Note that when all carriers are at equal power levels the 1E-7 BER dynamic range is approximately 30 dB.
Similar data was taken for the scenario when six carriers are at 5 dB above nominal level and 1 carrier is at 5 dB below nominal level as per figure 2. A representative BER curve taken at 7 dB link loss is shown in figure 5.
|Figure 5. 7 dB link loss with Uneven Carrier Amplitudes|
Note that under the worst case condition of six carriers at +5 dB with respect to nominal and one carrier at -5 dB with respect to nominal a BER of better than 1E-7 is achieved over a 19 dB range. The carrier at -5 dB with respect to nominal exhibits the 19 dB dynamic range. All other carriers exhibit dynamic ranges significantly greater than 19 dB.
bdr® LINK PERFORMANCE
Testing was also performed on a bdr® link. A Cisco DWDM bdr® optical link with Pout=6.2 dBm @ 1550nm was used. Several link losses and receiver types were tested.
The initial data was taken with all carriers at equal power levels, as per figure 3. Representative test results at 34 dB link loss (including 100 km of fiber) are shown in figure 6.
|Figure 6. BER for bdr® Link|
Note that when all carriers are at equal power levels the 1E-7 BER dynamic range is approximately 29 dB. Similar data was taken for the scenario when six carriers are at 5 dB above nominal level and 1 carrier is at 5 dB below nominal level as per figure 4. That data is shown in figure 7.
Figure 7. BER for bdr® Link with Uneven Carrier Amplitudes
Note that under the worst case condition of six carriers at +5 dB with respect to nominal and one carrier at -5 dB with respect to nominal a BER of better than 1E-7 is achieved over an 18 dB range. The carrier at -5 dB with respect to nominal exhibits the 18 dB dynamic range. All other carriers exhibit dynamic ranges significantly greater than 18 dB.
BER DYNAMIC RANGE AS A FUNCTION OF LINK LOSS
The previous test results may be used to compare the performance of various link types. Consider the case of equal amplitude carriers. The 1E-7 BER dynamic range is a function of link loss for the analog DFB link. However, the bdr® link has the same dynamic range regardless of link loss so long as that loss does not exceed the maximum specified for a particular bdr® type. This is presented graphically in figure 8.
|Figure 8. 1E-7 BER Dynamic Range for Different Link Types|
The DFB link performance is equivalent or slightly better than bdr® link performance for link losses less than 9 dB. However, the DFB performance falls off at link losses greater than 7 dB. bdr® with a standard receiver has performance roughly equivalent to a DFB for link losses less than 9 dB. However, the bdr maintains this level of performance for link losses up to 20 dB. Similarly, a bdr® link with an APD (extended reach) receiver maintains its performance up to link losses of 27 dB. Finally, a DWDM bdr® link with an APD receiver maintains constant link performance at losses up to 34 dB. In summary, a bdr® link does not always provide greater dynamic range than a DFB link. A bdr® link does, however, maintain a high dynamic range over a much greater link loss.
Figure 9 shows similar data for the case of 6 carriers at +5 dB with respect to nominal and 1 carrier at -5 dB with respect to nominal.
|Figure 9. 1E-7 BER Dynamic Range for Different Link Types with Unequal Carrier Amplitudes|
When the carriers are at uneven amplitudes the available dynamic range is less than when the carriers are at equal amplitudes. However, the effects of link loss on dynamic range are the same.
These results indicate that the reverse optical links tested provide a substantial dynamic range even under full carrier loading and worst case carrier amplitude inequality. With full loading and worst case amplitude inequality the DFB link gave a 1E-7 BER Dynamic range of 19 dB. With full loading and worst case amplitude inequality the bdr link gave a 1E-7 BER Dynamic range of 18 dB.
For short links (less than 9 dB link loss) the analog DFB and bdr® (digital reverse) link give approximately the same dynamic range. The DFB dynamic range starts to decrease as the link loss is increased beyond 9 dB. However, the bdr® link maintains its dynamic range at much higher link losses than 9 dB.
In the final article of this series we will develop a link budget and set-up procedure based on the available dynamic range of the optical link. This dynamic range will be used to accommodate the other impairments of the overall system (ingress, RF plant thermal noise, optical crosstalk, etc.). We will present an overall system link budget that equitably allocates this dynamic range. We will also propose a system set-up and alignment procedure.