Thursday 25 July 2013

The Modern Amplifier 

When you want to power up a concert, disco or even a garden party the amplifier is the essential backbone for your music. From my earliest days I have been fascinated with electronic circuitry and experimenting with components and aspired to create a device to power up two idling speakers. I have grown from an electronic hobbyist  to become an electronics student, enthusiast and an audiophile who enjoys learning about this widely established field of technology.

In my spare time I experiment with the circuits I build to try and achieve better fidelity at higher efficiencies. This blog explains the basic concepts of amplifiers to those of you who have a general interest in them and hopefully spark designs of your own. 

Class A

My first project was a Class A audio amplifier which I built in 2006. It was an extremely basic version as I began to step into the world of analogue electronics and I believe this will be the starting point of any, audio amplifier enthusiast. The principle behind the operation of Class A is relatively simple. A Class A audio amplifier will continually run a bias current even when there is no signal. This will ensure the driving transistor will always be in its correct quiescent state. To put it simply imagine a running tap. The tiny music signal will be used to reduce the flow or increase the flow rapidly just like a human controlling the valve. The variation in the flow of current, which in the analogy will be the flow of water, will then be AC coupled to the loudspeaker.

I started out with a simple general purpose transistor such as the C828 small signal transistor. I used a basic circuit as shown below:


Class A amplifier (1)

As you can see the entire circuit contains very few parts. The biasing of the transistor can be easily achieved because it is directly visible that the mid point is around 0v to allow for maximum swing. The quiescent current will be Ve/R (Where R is the load resistance). The bias current Ib is equal to the collector current Ic divided by the current gain, Beta. 
As simple as this circuit is, it has many pit falls. The first and most obvious is that the maximum output current is higher with smaller R, but gain with higher R, immediately requiring compromise. This can be surpassed by using a current source as the load resistance which forms the basic circuit for the Pass Labs, Zen amplifier.

What is the main benefits of Class A amplifiers? Apart from being simple to build, Class A amplifiers are the most linear of the 3 types of amplifiers. Linearity is what audio engineering is aimed at achieving as it means 0% distortion for perfect linearity. Class A amps have a constant quiescent bias current running. Intuitively, this implies that if output signal is only a fraction of the bias current, during a signal the transistor hardly moves off its quiescent point. Several parameters such as beta and the output resistance (which itself is a function of beta) will not alter significantly. this leads to high linearity which is why Class A is regarded by most as the standard to beat in terms of sound quality.

 Unfortunately the price to pay for this is colossal inefficiency. Since it wastes so much power in idle, they average an efficiency of around 20-30%. The low efficiency translates heat dissipation which therefore require massive heat sinks, and thus, are not feasible in extremely powerful set ups (upwards of 1KW).

 Very rare, high-end systems today use Class A operation in conjunction with valves (which injects  harmonics into the output) that appeal to audiophiles. Whether this sounds better than pure Class A amplifier or any other amplifier is a subjective matter.

My earliest design consisted of a pre- amplifier and an output stage (both of which were class A) for each of the two channels of the stereo input as you can see below:


The input was 4.8V from 4 rechargeable batteries and the total output power was about 2 watts, which admittedly isn't much. The drivers I used were two Panasonic 6 ohm house speakers.

Just to clear things up a bit, any references to Class AA or New A or any other form of class was down to marketing and really only fell into one of the three main amplifier topologies with slightly unconventional biasing.

Class AB

In order to main reasonable amounts of fidelity and make a significant improvement in efficiency it was required to move into the class AB type of amplifiers. Class B amplifiers turn on only when a signal is present. Their principle is quite simple: there are two transistors in common collector configurations with their emitters tied together. When the positive portion of the input is present the transistor at the top is on and the bottom is off, and vice versa for the negative half of the signal. They act as a sort of 'antagonistic pair'.

Class B amplifier (2)

However a Class B amplifier will suffer intolerable amounts of distortion because there is a portion of the input wave form for which the circuit is completely off. This can be fixed simply by adding two diodes in series across the two bases to keep the transistors barely on. This will mean that cross over distortion will be eliminated almost completely. Therefore you have a good compromise between efficiency and fidelity. 

Below I have a video of my simple Class AB amplifier in action.



                           

Some say Class AB amplifiers still struggle to produce the audio quality similar to that of Class A amplifiers which maybe due to the fact that the transistors in a Class AB experience a wider range of operating conditions (they have to start all the way from 'off' state), and therefore exhibit a lower linearity while others say a well designed Class AB will match a any Class A amplifier. I guess it is down to the customer as to what sounds more 'natural and rich' although in practice there are far more variables involved than just class. One thing remains for sure though, which is the fact that Class A amps are by far the easiest to construct by amateur enthusiasts as the circuit is much less complex.

The voltage gain given by a Class AB output stage is small, about one, which is not useful. Therefore you need a 'pre-amplifier' like stage at the input to give the necessary voltage gain. One of the most simple and versatile circuits I have come across was the one posted by Dino Segovis in his site Hackaweek.com. He explains the circuit diagram in detail and builds a working prototype. If you are an enthusiastic beginner I strongly suggest you start at his website or the one in the references (2). 


Simple Class AB Amplifier(3)


The above circuit is a very simple first design step. The circuirt simply uses resistors R5 and VR3 to attain a mid point bias for the output stage. The first stage is a simple Class A amplifier with its load being the bias network. Although this circuit covers the basics (including some 'elementary' feedback) it has many limitations and has to be improved sufficiently in order to become a competent Class AB design. One of the first and most obvious flaws is the joint input and voltage amplification (IPS and VAS) stage. This stage will simply be far too non-linear at high signal voltages when providing the necessary gain and the problem is magnified for high power, low beta Power BJTs not to mention its unsymmetrical current source and sink capability. One simple way to avoid this problem is to employ a compound/Szklai pair configuration as shown below.


Three common constituents (4)

All I did afterwards was replace the transistors in the basic Class AB (the BC 327) with the Sziklai pair. The gain of the Sziklai pair is: 

Where Beta 1 is the gain of the driver transistor and Beta 2 is that of the output transistor. As you can see such a gain would be useful for powering a stage which would require a decent amount of current gain. Further more, and emitter follower can be implemented in a situation where there is a very high output resistance such as cascaded differential pair with a current mirror.


I built a transformer that will act as a step-up as a step down. It has a ratio of 2:1 turns and my hope was to increase the effective resistance seen by the amplifier by (N1/N2)^2, which meant in my case by 4 times. 

Output Transformer

Although generally purpose built audio transformers have a relatively flat frequency response mine, being home made, certainly didn't. The core was not of very high quality so its magnetic saturation limit was low causing a maximum current limit below requirements. The sound was distorted and I felt a definite drop in power at low frequencies. 

I have mentioned a few basic topologies that enthusiasts use when building audio amplifiers but these are just the basics. To learn in depth about high performance design, such as feedback compensation, improved IPS, VAS and OPS stages and low noise design, I suggest you start reading Designing power Amplifiers by Bob Cordell (6).

In my Final Year Project I am currently exploring the various effects of incremental design improvements, example; with/without local and global feedback, on Total Harmonic Distortion (THD, see 'how to measure amplifier fidelity) of the amplifier. I aspire to create one of my own Class AB topologies. Additionally I am looking into high fidelity noise performance as well.

Class D

This is the most popular class of amplifier present. The Class D is a completely different species altogether. It uses Pulse Width Modulated (PWM) wave as the main input to the output stage rather than just an analogue signal. 
The PWM wave is square wave with varying duty cycle which switches between two terminal voltages at a frequency a much higher than what is audible. Effectively this PWM contains harmonics, which when filtered, gives just the output.  In order to create a PWM wave of a signal, the input wave is directly compared to a waveform of predetermined frequency. The most common type of this reference waveform is the triangular wave although variants of these are used as well. The PWM wave is then used to switch high and low side power MOSFETs from which the resulting output waveform is filtered.


Class D principle (5)


As you can see from figure, the MOSFET output transistors are in a half bridge topology and work as a switch being completely on or completely off. There is a major benefit to this method as there is almost no heat lost in the MOSFETs at all as they have very high conductance. It is in theory possible to have 100% efficiency but it is generally lower than that due to IR losses in the components and the filter. 

Full Bridge Topology

The Full bridge (H - Bridge) topology derives its idea from the H-Bridge motor drives. Effectively the load is driven by two identical Half bridge drivers working with a 'anti - phase' PWM.

H Bridge Class D topology (7)

However, note must be taken that now instead of two states to the PWM you could represent it with 3 states, fully positive, ground/neutral and fully negative. This means the the Signal now more closely represents a sinewave and thus presents less distortion (cancellation of 2nd order harmonic distortion). 
Furthermore, the half bridge requires two independent power supplies positive and negative, whereas the H bridge can operate on only one if necessary. 

I have a prototype of the half bridge Class D I built. It is still only 7 watts. Below is a video of it powering my Subwoofer which is running at the edge of distortion (although the video makes it sound heavily distorted, it is much clearer in reality and only exhibits small amounts of intermodulation).




This is only the first prototype. I am currently working on a full bridge design and later on implementing feedback. Additionally, I also conduct research into Transistor-Transistor Logic (TTL) topologies for gate driver outputs and Multi-Level Class D amplifiers.

How do you measure Amplifier fidelity?

The local community simply refers to amplifiers having clinical and warmth or dynamic sound. What does this mean in engineering terms? How good is 'good'?

Basically the end goal of any amplifier designer is to have a perfectly linear amplifier, but like anything in life, this is only approachable and not really achievable. A good, or more precisely a high fidelity amplifier can be defined as one that can reproduce sound as close to the original source as possible. But once again what is close? For that we need to analyse distortion.

The fidelity of any amplifier is measured by its Total Harmonic Distortion (THD) and its Intermodulation Distortion (ID). Designing an amplifier to minimise the presence of these factors is what all audio engineer aspire to achieve. Every amplifier produces some form of harmonics due to the fact that there is never a perfectly linear amplifier. A harmonic is constituent of the fundamental frequency and is always a multiple it. For example 10 Hz has harmonics at 20, 30, 40 Hertz and so on. The equation for Total Harmonic Distortion is given by the equation below:



At first glance the equation looks scary, but on closer inspection it isn't. The Vrms here is simply the amplitude of each harmonic and nothing more. When you square an amplitude, you remove the sign (positive or negative). The small numbers denote which harmonic it is, i.e. the first, second, third and so on. THD is a percentage and shows the percentage of harmonics to that of the fundamental frequency. Quite simply the lower the THD, the better the amplifier and sound quality. Human can here a THD of around 0.3% so anything below that is a very High fidelity amplifier.

Intermodulation
Simply put, harmonic content interferes (often destructively) with the fundamental signal. Effectively any harmonic content present in the original signal is distorted in amplitude due to the amplifiers non linearity which gives rise intermodulation.
In order to illustrate my point, take a look at the figure below.

Effect of Intermodulation (8)
The original signal is now devoid of some content. This effect is exaggerated as the fundamental signal becomes 'richer' in frequency and harmonic content.

Aesthetics

The aesthetics of a modern day amplifier is a bigger determining factor than most believe. High-end manufacturers quite often use expensive material to create a variety of glossy to brushed aluminium effects on the outer skin of an amplifier to attract potential customers.

I also enjoy trying to give my designs an 'Aesthetic feel' but unfortunately I don't have a vast budget and have only very limited resources and tools to achieve this. Therefore I used quite tough card board sheets to make the base of the amplifier then covered the rough edges with a single piece of shiny board. I also use a few small sheets of aluminium to strengthen the exoskeleton of the amplifier.




I have a BassFace 8.1 800W passive subwoofer which has a reasonably good low frequency response eve though I realise that there is a quick drop in SPL (Sound Pressure Level) after around 60 Hz.




Designing the case

I continued after completing the 20 Watt amplifier to search for more power. The answer came in the form of 3 Integrated Circuits called LM3886T. These were the most powerful ICs I have worked with and decided to start testing and building the circuit immediately. I found that these have quite a high bias (around 90mA which isn't too good) but under testing they gave clear adequate acoustic performance  This amplifier can handle +-35V in which case it will push 50 Watts into 8 ohms each and I have three of them! Unfortunately I only have +-12V but even this is enough to light up a room with well defined mid range and quite heavy bass. I wanted the amplifier in a small package so I added in a few heat sinks (they weren't very big) and since the space inside was all crammed with components, an active cooling fan to keep the air circulating. Finally I wrapped the circuitry styling a rather unconventional asymmetric shape.





As I have a passion for art I  like to envision designs of amplifiers that may or may not actually exist in reality or are yet to come. I feel that modern day amplifiers and especially AV receivers,  despite the many varieties available on the market still have more room for extravagant design. Most AV receivers are 'boxy' and a simple redesign can give much more appeal. Well that said, this is still my opinion.


References

(1) Rod Elliott . (1999). Class A. Available: http://sound.westhost.com/class-a.htm. Last accessed 29/10/2013.


(2) Paul McGowan. (-). Class B. Available: http://www.pstracks.com/pauls-posts/class/9087/. Last accessed 10/29/2013.


(3) Eric Coates. (2007). Class AB Power Amplifiers. Available: http://www.learnabout-electronics.org/Amplifiers/amplifiers55.php. Last accessed 25/11/2013.


(4) Emitter follower. Available: http://www.ece.umd.edu/~neil/306/ckts/ef.htm. Last accessed 29/10/2013.


(4) Long tailed pair. Available: http://en.wikipedia.org/wiki/Differential_amplifier. Last accessed 29/10/2013.


(4) Rod Elliott . (1999). Compound Pair Vs. Darlington Pair. Available: http://sound.westhost.com/articles/cmpd-vs-darl.htm. Last accessed 29/10/2013.


(5) Pulsed Amplifier. Available: http://en.wikipedia.org/wiki/File:Pwm_amp.svg. Last accessed 29/10/2013.


(6) Bob Cordell. (2010). Designing Audio Power Amplifiers. Available: http://www.mhprofessional.com/downloads/products/007164024X/cordell_ch01.pdf. Last accessed 29/10/2013.

(7) Elliot Rod. (1999). Class D Audio Amplifiers - Theory and Design.Available: http://sound.westhost.com/articles/pwm-f4b.gif. Last accessed 25/11/2013.

(8) Nelson Pass. (2011). Audio, Distortion and Feedback. Available: https://passlabs.com//download.php?download_file=https://passlabs.com/images/uploads/mypdf/distortion_and_feedback.pdf. Last accessed 5/12/2013.

Jeremy Cook. (-). 'Build your own class AB amplifier'. Available: http://hackaday.com/2011/08/14/build-your-own-class-ab-audio-amplifier/. Last accessed 10/29/2013.


Jacky. (-). Low cost amplifer.. Available: http://www.circuitstoday.com/low-cost-amplifer. Last accessed 29/10/2013.