Quick viewing(Text Mode)

Class D Audio Amplifier Application Note with Ferroxcube Gapped Toroid Output Filter

Class D Audio Amplifier Application Note with Ferroxcube Gapped Toroid Output Filter

Class D audio Application Note with Ferroxcube gapped toroid output filter

Class D audio amplifier with Ferroxcube gapped toroid output filter

The concept of a Class D amplifier has input signal shape but with larger bridge configuration. Each topology been around for a long time, however amplitude. And audio amplifier is spe- has pros and cons. In brief, a half bridge only fairly recently have they become cially design for reproducing audio fre- is potentially simpler,while a full bridge commonly used in consumer applica- quencies. is better in audio performance.The full tions. Due to improvements in the bridge topology requires two half- speed, power capacity and efficiency of Amplifier circuits are classified as A, B, bridge , and thus more com- modern semiconductor devices, appli- AB and C for analog designs, and class ponents. cations using Class D amplifiers have D and E for switching devices. For the become affordable for the common analog classes, each type defines which A Class D amplifier works in the same person.The mainly benefit of this kind proportion of the input signal cycle is way as a PWM power supply, except of amplifier is the efficiency, the theo- used to on the amplifying device: that the reference signal is the audio retical maximum efficiency of a class D Class A 100% wave instead of the accurate design is 100%, and over 90% is Class AB Between 50% and 100% reference. achieve in practice. Class B 50% Class C less than 50% Let’s start with an assumption that the Other benefits of these amplifiers are input signal is a standard audio line level the reduction in consumption, their The letter D used to designate the class signal.This audio line level signal is sinu- smaller size and the lower weight. D amplifier, is simply the next letter soidal with a ranging from 20 Their advantages are obvious in low after the C, and does not stand for dig- Hz to 20Khz. This signal is compared power battery operated personal ital. Class D and E are sometimes mis- with a high frequency triangle or saw- audio players and laptop . takenly defined as digital because the tooth waveform to create the PWM However they are also progressively output waveform superficially resem- signal.The input signal is converted to a displacing more traditional linear bles a pulse train of digital symbols. sequence of pulses whose averaged designs in mainstream applications value is directly proportional to the such as home entertainment systems, amplitude of the signal at that time.The automotive sound systems, and pro- Class D amplifiers frequency of the pulses is typically ten fessional installations where high qual- or more times the highest frequency of ity audio is important. A class D amplifier is basically a interest in the input signal. switching amplifier or pulse width modulator (PWM from now on) This PWM signal is then used to drive Audio Amplifiers amplifier. In this kind of amplifier all the power stage, creating the amplified power devices are operated in on/off , and finally a low pass filter An electronic amplifier is a device for mode, reducing the power losses in is applied to the signal to filter out the increasing the current, voltage or the output devices significantly. PWM carrier frequency and retrieve power of a signal. It does this, by tak- the sinusoidal audio signal. The result- ing power from a power supply and Class D amplifiers can be categorised ing filtered signal is then an amplified controlling the output to match the into two topologies; half bridge and full replica of the input. Figure 1: Block diagram of a Class D audio amplifier

CLASS D Audio Power cy by –20dB per decade.The switching Amplifier with GAPPED frequency of the amplifier can influ- FERRITES ence the choice of the filter order, the higher the fs, the lower the order One of the most important parts in a required to achieve a given attenua- class D, audio power amplifier is the tion within a specified passband. This output filter.The overall efficiency, reli- would seem to dictate the use of the ability and audio performance depends highest switching frequency possible. on it. A reference circuit has been The trade-off is that increasing fs modified in order to substitute an increases the switching losses and the inductor by a gapped .This com- EMI, thus decreasing the efficiency of ponent is designed to behave with the the amplifier. Figure 2: Example of the transfer function of a low pass filter over normalized frequency (f/fc). same performance as the first one. Several order filters are shown.

Output Filter Design

The overall efficiency, reliability and audio performance depend on the quality and linearity of the output fil- ter.The main goal of the output filter is attenuation of the high frequency switching component of the class D amplifier while preserving the signals in the audio band.

The order of the filter determines how many poles exist at the same fre- quency, with each order increasing the attenuation above the cut-off frequen-

0 0.2 0.3 0.4 0.5 0.6 0.7 0.8 1 2.0 3.0 4.0 5.0 6.0 7.0 8.0 10 A second order LC filter is the com- sated for by means of proper feedback mon approach, as it is lossless and has network design. Some manufacturers of the output filter a –40dB/decade slope, allowing for a simply leave it that way, so the reasonable rejection of the carrier if response is strongly dependent on the The linearity of the filter versus fre- the parameters of the filter and the load. Surely a non-desirable situation. quency and output power strongly switching frequency itself are properly influences the performance of the designed. More information about designing the amplifier and quality of the audio sig- output filter as a function of the cut off nal.Therefore, care has to be taken in frequency and load impedance is pre- order to select the optimal magnetic sented in annex 2. core at a reasonable cost.

The output inductor withstands the Iron powder cores are the cheapest whole output current without satura- solution, but the switching frequency tion, as well as keeps the for (several hundred kilohertz) is too high the off cycle, as in any non-isolated to make them suitable for this applica- switching converter (Class-D half tion. Figure 3: Second order low pass filter structu- bridge design is in fact analogous to a re. , its reference voltage Powder alloy cores are widely used, being the audio signal). mainly because they are very popular as output chokes in DC/DC convert- The first thing to do is to design the The ideal inductor (in terms of linear- ers and SMPS.The drawback is that the transfer function for the filter. Usually, ity) is an air-core one, but the size and curve is not flat, but with a a Butterworth or similar frequency number of turns required for typical constant slope drop.To overcome this response is chosen, with a cut-off fre- Class-D operation makes it impracti- problem the designer has to over quency slightly above the audio band cal, so a core has to be used in order specify the product in the sense that (30-60KHz). Have in mind that one of to reduce turns count and also pro- only a small region of the saturation the design parameters is the termina- vide a confined that curve is used. Losses at high frequen- tion load, that is, the speaker imped- reduces radiated EMI. It can be use cies are also a thread, implying the use ance. Usually, a typical 4 or 8 ohm either powder cores or ferrite cores. of low permeability materials wound is assumed, but that would Ferrite cores, must have a “gap” where with some 30 to 50 turns. produce variations in the measured energy is stored. size should also frequency response in presence of dif- be carefully chosen to keep DC and ferent speakers.That must be compen- AC losses low (requiring thick wire). µ Effective Permeability Effective

Figure 4: Saturation curve: ferrites have lower saturation, but flat response, while powder cores show high saturation with a continuous slope. Figure 5: International IRAUDAMP1 demo board. Features 500W + 500W peak stereo power, efficiency 93% at 1 kHz and 100 watts.Voltage supply ±50 (allowed ±25 to ±60). Compact size: 100 x 140 x 38 mm. Ferrite cores are the third option. They have the lowest losses at high , and though the saturation is lower than in powder cores, the curve is completely flat. This curve allows the designer to optimize the core volume and gap size to the required energy to store.The result is very low distortion with a low turns count (10 to 25) and, finally, at the most affordable cost.

A practical Class D audio amplifier has been evaluated to measure the influ- ence of the magnetic core on the dis- tortion of the audio signal.The amplifi- In the original design, the output filter only difference between them is the er is driven by an International of each channel is made with an AVX magnetic material of the inductor and Rectifier IR2011S device, delivering up (BF074E0474J with 0.47uF) the winding count. to 500 + 500 watts peak. and a Micrometals inductor core (NPT0104, core T106-2). It has an The measuring equipment used to test value of 18 µH and it is the amplifier performance is Audio The reference design wound with 37 turns of AWG18 mag- Precision System Two. This equipment net wire.This filter cut off frequency is allows easy and fast measurements The reference design is the 50 kHz (see annex 2 for details). under several conditions. It is widely IRAUDAMP1 from International Rec- used in the audio industry. Each chan- tifier. Complete details of the amplifier In one of the channels, the original nel has been loaded with a 5 Ohms can be found at International Rectifier inductor has been replaced by a FER- resistive load. website, in this application we will ROXCUBE TN26/11-3C20-A113. It focus on the output low pass filter has also an inductance of 18 uH, but it This first graph shows the total har- design. is only wound with 13 turns of the monic distortion and noise power same AWG18 wire. percentage versus the output power The switching frequency is 400 kHz; of the amplifier. The percentage this frequency will be modulated by Distortion and noise versus frequency increases with the output power due the low frequency audio signal through and power are the key parameters to to longer swing over the saturation the PWM (Pulse Width Modulator). define the quality of an audio amplifier. curves (see figure 4). Here we can see The MOSFET’s (International Rectifier The following curves compare the that the performance of the ferrite is IRFB23N15D) are driven by the performance of the amplifier on each similar and slightly better under high IR2011S high speed . channel, taking into account that the power conditions. Figure 6: Distortion over output power percentage ver- sus output power comparing powder alloy and inductors. Amplifier input 10 kHz, load: 5 Ohms, voltage supply ±50 Volts.

This second graph shows the same cy sweep over the audio spectrum at a slightly better performance of the fer- parameter (total harmonic distortion fixed output power (in this case 100 rite core over the powder core. and noise percentage) over a frequen- watts). Again, the same result shows a

Figure 7: Distortion over output power percentage ver- sus frequency comparing pow- der alloy and ferrite core inductors. Output power 100 watts, load 5 Ohms, voltage 5Ω Loading, ±Vcc = 50V, 100W supply ±50 Volts Ferrite core design with Ferroxcube gapped toroids

Ferroxcube offers a wide range of standard gapped toroids fitting in this application. The procedure shown below is intended to help the design engineer to choose the optimal (and therefore the most cost effective) product size and gap. It is important to note that custom cores and gap sizes are availble upon request.

The procedure starts with 2 basic parameters: the inductance of the out- put filter (see annex to calculate the value out of the cut off frequency and load impedance) and the maximum current that will flow through the optimize the core to in the sat- where L is the inductance in µH and inductor (derived from the maximum uration limit.The following graphs help AL is in nH/T2, and N is the number of power of the amplifier). to choose the right core able to stand turns. the power. For each core size it is These parameters will be used to cal- plotted the amount of energy capable The wire gauge is fixed by the maxi- culate the energy stored in the induc- to be stored and the maximum AL fac- mum DC resistance allowed for the tor: tor of the core. In annex 1 plots for inductor as well as by the current den- every standard size can be found. sity that the wire can handle. DC 2 E = L x I , where L is the induc- Note that it is possible to make cus- resistance can be estimated with the tance in µH, I is the maximum cur- tom sizes and custom AL factors upon following formula: rent in , and E is energy in µJ request. After the selection of the core size R = (1+OD-ID+2xH)xRLxN, The next parameter to be taken into and AL factor, the next step is to cal- where R is the DC resistance of the account is the size of the core in com- inductor in mOhm, OD, ID and H are culate the winding. The number of bination with the amount of energy to the outer, inner diameter and height turns is derived from the next rela- be stored in the inductor. In most of of the core in mm, RL is the resist- tion: ance per meter of the selected wire the cases the size of the core is limit- gauge and N is the number of turns. x ed by the board layout and mechanical N = L 1000 constraints, so the designer has to AL Annex 1: I2L curves for The following graphs show the energy Product dimensions are described as Ferroxcube gapped storage capability of standard Ferroxcube TN (toroid coated with Nylon) 13 toroids gapped toroids. There are more sizes and (outer diameter) /7.5 (inner diameter) AL values available and custom products /5 (height). can be made upon request. Annex 2: Now calculating the L and C values: Filter design basics

The most common and recommend- able output filter is a simple second order .They count a minimum number of components and show optimal flatness response while keeping a sharp cut off frequency.

The transfer function for a second order Butterworth filter has to follow These equations show that the induc- this equation: tor and capacitor values depend on Figure 9: Simple filter with bipolar supply. the loudspeaker impedance.Therefore it is recommendable to have a switch to choose between 4 or 8 Ohms.This means that the inductor should have 2 windings in series, so the switch will The cut off frequency for this filter is select the first winding for the low located at w0=1 rad/sec, so we should inductance (and low impedance loud- scale it by wc=2pfc, being fc the speaker) or both windings for the high desired cut off frequency: inductance filter.

Depending on the polarity of the amplifier voltage supply (bipolar ±Vcc or simple +Vcc) the output filter should be implemented accordingly, as shown in the following scheme: The basic construction for such a filter Figure 10: Double inductor filter for simple is: supply. Figure 10: Double inductor filter for simple supply.

These equations and figures derive a table with and inductance values for several cut off frequencies and loudspeaker impedance values.

fc (kHz) R (Ohm) L (µH) L Half (µH) C (µF) Figure 8: Second order low pass filter. 20 4 45 23 1.41 25 4 36 18 1.13 Calculating the transfer function of the 30 4 30 15 0.94 circuit in the Laplace domain: 50 4 18 9 0.56 20 8 90 45 0.70 25 8 72 36 0.56 30 8 60 30 0.47 50 8 36 18 0.28 Product range and Core type Dimensions (mm) Effective parameters Other specifications Outer Inner Height Core Effective Effective Effective Mass Isolation diameter diameter H (mm) factor volume length area (g) voltage D (mm) d (mm) (mm-1) (mm3) (mm) (mm2) (V) The cores are coated with Polyamide 11 (PA11), flame retardant in accor- TN13/7.5/5 13 6.6 5.4 2.46 368 30.1 12.2 1.8 1500 dance with UL94V-2, UL file number TN17/11/6.4 17.5 9.9 6.85 2.24 787 42.0 42 3.7 1500 E45228 (M).The inner and outer diam- eters apply to the coated toroid. TN20/10/6.4 20.6 9.2 6.85 1.43 1330 43.6 43.6 6.9 2000 Contacts are applied on the edge of TN23/14/7.5 24 13 8.1 1.69 1845 55.8 55.8 9.0 2000 the toroid for isolation voltage test, which is also the critical point for the TN26/15/11 26.8 13.5 11.6 0.982 3700 60.1 60.1 19 2000 winding operation.

Core type AL (nH/T2)µeff TN13/5-3C20-A40 40 ± 15% 90 TN13/5-3C20-A56 56 ± 15% 125 TN13/5-3C20-A67 67 ± 15% 147 TN13/5-3C20-A72 72 ± 15% 160 TN13/5-3C20-A79 79 ± 15% 173

Core type AL (nH/T2)µeff TN20/6.4-3C20-A68 68 ± 15% 125 TN20/6.4-3C20-A81 81 ± 15% 147 TN20/6.4-3C20-A87 87 ± 15% 160 TN20/6.4-3C20-A96 96 ± 15% 173 TN20/6.4-3C20-A109 109 ± 15% 200

Core type AL (nH/T2)µeff TN17/6.4-3C20-A52 52 ± 15% 90 TN17/6.4-3C20-A72 72 ± 15% 125 TN17/6.4-3C20-A88 88 ± 15% 147 TN17/6.4-3C20-A92 92 ± 15% 160 TN17/6.4-3C20-A104 104 ± 15% 173

Core type AL (nH/T2)µeff TN23/7.5-3C20-A65 65 ± 15% 90 TN23/7.5-3C20-A90 90 ± 15% 125 TN23/7.5-3C20-A106 106 ± 15% 147 TN23/7.5-3C20-A115 115 ± 15% 160 TN23/7.5-3C20-A124 124 ± 15% 173

Core type AL (nH/T2)µeff TN26/11-3C20-A113 113 ± 15% 90 TN26/11-3C20-A157 157 ± 15% 125 TN26/11-3C20-A185 185 ± 15% 147 TN26/11-3C20-A201 201 ± 15% 160 TN26/11-3C20-A217 217 ± 15% 173

Top View