Predicting Total Harmonic Distortion (THD) in ADSL Transformers using Behavioural Modeling, J. Neil Ross & Andrew D. Brown S 1
Outline Introduction ADSL Where is the need for the transformer? What are the design Issues? Modeling and Simulation Basic Transformer modeling Non-linear modeling System modeling Hardware modifications Planar transformers Conclusions 2
ADSL: Introduction Asymmetric Digital Subscriber Line The technology is used as a high-speed modem link with asymmetric up- and downstream data rates The technology is important because it uses the existing standard POTS network infrastructure The mechanism only requires an ADSL modem at the local telephone exchange (CO) and a similar modem at the customer site (CP) Digital Broadcast Broadband Network Narrowband Network Network Management Central Office ATU-C ATU-C ATU-C ATU-C ATU-C PSIN Splitter Loop Phones ADSL Termination Unit Central (ATU-C) Customer Premises ATU-R TE TE TE TE ADSL Termination Unit - Remote (ATU-R) 3
ADSL Modulation Scheme ADSL is based on a broadband modulation scheme multiple carriers placed at 4.3125kHz intervals With 256 carriers a 1.1MHz bandwidth is required These sub-carriers may also be referred to as sub-channels. ADSL co-exists with POTS POTS Upstream Downstream frequency 4kHz 25kHz 160kHz 240kHz ~1.1MHz 4
ADSL Analog Interface The signals are transmitted using a form of QAM, either Carrier-less Amplitude/Phase Modulation (CAP) Quadrature Amplitude Modulation (QAM) The modulators and demodulators are interfaced to the line with: Channel splitting filters Impedance-matching wide-band transformers ISP ADSL Chipset Wideband Transformer High Pass Filter Twisted Pair <6 Mbps High Pass Filter Wideband Transformer ADSL Chipset PC PSTN Low Pass Filter Low Pass Filter Phone Central Office Modem Home Modem 5
ADSL transformer design issues The ADSL transformers have several basic requirements: 1. Wide Bandwidth 1. This implies low loss, low leakage and low capacitance 2. Low Insertion Loss 1. This implies low resistance and loss 3. Low Distortion 1. This implies good linearity 4. Compact Size 1. There is a design trade-off between distortion and size 6
Modeling Insertion Loss We can use a simple linear model, with accurate parasitics to predict the insertion loss over a wide frequency range This is well understood and works reasonably well Ciw 10 0 Cp 50p LLp 10uH Rp 0.5 50p TX1 Rs 0.5 LLs 10uH Cs 50p Insertion Loss (db) -10-20 -30-40 -50 1000 10000 100000 1000000 10000000 Frequency (Hz) 7
Predicting Distortion Distortion is a result of non-linearities in the transformer The source of the non-linearities is mainly from the ferrite core of the transformer Distortion is quantified for ADSL system designers using the Total Harmonic Distortion Measure (THD), which is closely related to the Signal to Noise Ratio (1/THD) THD in this context is usually calculated using: THD = 5 i = 2 V 1 V i Where: Vi is the harmonic (i) V1 is the fundamental 8
Ferrite Material Hysteresis Ferrite materials exhibit some form of hysteresis The trouble with signal transformers is that a convenient major loop cannot be assumed The loops may be: Assymetric Minor Modeling these effects is actually very difficult to accomplish accurately, as the standard models assume a major loop Major Loop Asymmetric Minor Loop B Symmetric Minor Loop Initial Magnetization H 9
Different Loop Types Small Minor Loop Linear No Losses Minimal Distortion 0.01 0.008 0.006 0.004 0.002 B (T) 0-0.002-0.004-0.006-0.008-0.01-1.5-1 -0.5 0 0.5 1 1.5 H (A/m) 10
Different Loop Types Medium Loop Small Hysteresis Low Losses Minimal Distortion 0.1 0.08 0.06 0.04 0.02 B (T) 0-0.02-0.04-0.06-0.08-0.1-15 -10-5 0 5 10 15 H (A/m) 11
Different Loop Types Major Loop #1 Significant Hysteresis Significant losses Large Distortion 0.5 0.4 0.3 0.2 0.1 B (T) 0-0.1-0.2-0.3-0.4-0.5-60 -40-20 0 20 40 60 H (A/m) 12
Different Loop Types Major Loop #2 Significant Hysteresis Slightly reduced Significant losses Relatively smaller Large Distortion Early Closure B (T) 0.40 0.30 0.20 0.10 0.00-0.10-0.20-0.30-0.40-60.00-40.00-20.00 0.00 20.00 40.00 60.00 H (At/m) 13
Mixed-Domain Model The approach used in this paper was to use a mixed-domain domain model of the transformer, with a modified Jiles-Atherton core model Single Lumped Model R1 U1 U6 U2 vs vp vout 50 classicja V1 MMF MMF R2 100 R4 winding winding 14
Handling Minor Loops There are three approaches to dealing with minor loops 1. Characterise the model over a wide range of operation 1. This is OK, but means accuracy is compromised for each specific case, in favour of a generally accurate model 2. Characterise for only minor loops 1. This works well, but is very specific 3. Modify the model to change behaviour for minor loops 1. In theory this sounds good, but the existing models do not implement this capability. Plus, the models need a-priori knowledge of turning points, which leads to innacuraccies. 15
Characterisation Software To handle multiple loop optimisation, software has been developed to take measured BH curve data and extract parameters that best fit all the curves imported from different signal levels 16
Characterised Waveforms Three Measured Loops Major Medium Minor Optimiser Settings 3 loop fit Equal weighting Genetic algorithm B(T) 0.5 0.4 0.3 0.2 0.1 0.0-0.1-0.2-0.3-0.4-150 -100-50 0 50 100 150 H(A/m) 17
Modifications to the Jiles-Atherton Model Modifications were also made to the original Jiles-Atherton magnetic model to improve the minor loop modeling, basically modifying the behaviour depending on the recent turning points Worked well statically Less effective dynamically Poor with arbitrary waveforms Relies on turning points B (T) 0.2 0.15 0.1 0.05 0-0.05-0.1-0.15-40 -20 0 20 40 H (A/m) Modified JA Model Measured Original JA Model 18
Predicting the THD of ADSL transformers In practice, there was no need to have overly sophisticated models to predict the performance of the transformer as an individual component, as the standard tests were undertaken at single frequencies at a time. Models were created for standard core types and materials, simulated and the resulting THD compared with measured results using the same components Toroid TN10/6/4-3E5 Low Profile E11R-3E6 Integrated Inductive Component IIC Custom planar Devices E18/E14 EP Cores (standard wire-wound) EP13 19
Toroid TN10/6/4-3E5 THD (db) 0-10 -20-30 -40-50 -60-70 0 2 4 6 8 10 12 Peak Voltage (V) Measured Simulated 20
ER11-3E6 0-10 THD (db) -20-30 -40-50 0 2 4 6 8 10 12 Peak Voltage (V) Measured THD Simulated THD 21
Packaging Options As well as wire wound, integrated inductive components (planar devices) were tested and compared with standard wire wound (EP13) cores: Ferrite Package Pins [a] side view Internal Tracks [b] plan view [c] 5:5T PCB Layout 22
EP13 and IIC Comparison The generic planar device is much worse than the EP13 core, and this is predicted by the simulation THD (db) 0-10 -20-30 -40-50 -60-70 0 2 4 6 8 10 12 Input Voltage (V) IIC EP13 23
Conclusions thus far Smaller ER and toroid cores are not good enough Basically too small saturate early Not enough inductance Not enough air gap relative to core size Integrated Inductive Components not good enough Too few turns No gap What about custom planar devices? More turns Custom gap Low profile = better density of ADSL modem in exchange 24
Custom Planar Transformers Using a customised planar device, with a large enough core (predicted by the simulation) to ensure low distortion, the tests were repeated using an E18 core 25
E18 Custom Planar Device The resulting THD figures were comparable with the standard EP13 wirewound device 0-10 THD (db) -20-30 -40-50 -60-70 0 2 4 6 8 10 12 Input Voltage (V) APC4199 E18 (no gap) E18 (Gapped) 26
Conclusions Behavioural modeling has been used in a range of ADSL transformers to predict: Insertion Loss Bandwidth Distortion Simulation has been used effectively to guide the design of the proposed transformers to meet the requirements of the ADSL design context New planar devices have been implemented that will greatly improve produceability and packing densities in exchanges. 27