I had been tinkering around with amplifier circuits on bread boards for a number of years but ultimately my real aim of all this work was to build my own amplifier which hopefully would work out cheaper than buying an expensive piece of audio equipment and possibly sound just as good if not better.
Stereo 30W RMS integrated amplifier.
8 Ohm load
Inputs: CD, Tuner, Tape, Auxiliary, Mic.
Controls: Volume, balance, treble, bass, Mic Volume and Mic Balance.
I decided to do the design in 3 parts, keeping the power amplifier, power supply and pre-amp sections all separate. My approach to designing the power amplifier is as follows:
Starting with the output specification of 30W RMS into an 8 Ohm load I calculated maximum peak voltage swing and peak current. I designed a darlington pair class AB push-pull output stage which would deliver the required voltage swing and current. The design was to be a single ended design hence it required a dual split power supply. I chose the power supply rail voltages based upon the output voltage requirement with some voltage headroom for the driver amplifier stage to provide the required drive levels. I constructed the push-pull output stage on breadboard and then made measurements of the input voltage required to deliver the specified output voltage swing, the input impedance of the push-pull stage and also tested it at high-frequencies (100kHz) to assess its performance. I then designed a class A voltage gain stage to drive the push-pull stage. The class A voltage gain stage had some local feedback in the form of an emitter resistor to improve the linearity of the stage. I preferred for the voltage gain stage to have moderate gain with good linearity rather than having lots of gain with poor linearity and then relying on the overall feedback of the whole amplifier to linearise the gain. One of the reasons for doing this is it makes stabilising the whole design much easier. If the loop gain is too high and uncontrolled, achieving stability can be very difficult. With more moderate loop gain where the phase angle is better behaved, it is easier to obtain stability.
I had decided at the start of the design that the individual power amplifier stages would be AC coupled. There are pros and cons to using AC coupling in a power amplifier. On the down side there is a limit to the low frequency response of the amplifier and since there are normally 3 stages, there will be 3 RC poles. At low frequencies this can lead to instability and “motor-boating”, hence special attention must be paid to the value of the coupling capacitors. On the positive side one doesn't have to be concerned with each stage providing the DC bias for the next stage. When one has to take DC bias into consideration it also affects the voltage swing and gain capabilities of the preceding stage and the design becomes much more complicated. I added the class A driver stage to the bread boarded design and designed a differential amplifier to drive the voltage gain stage. Differential amplifiers are usually used on the inputs of power amplifiers since they provide a convenient way of supplying the negative feedback from the output stage. They also have better linearity than a single ended common emitter stage. I added the differential amplifier to the bread board and applied overall negative feedback.
The last step in the design process was to ensure the design was stable. I made use of a spreadsheet to plot the locus of the loop gain at low frequencies where the AC coupling would cause leading phase swings and at high frequencies where the device and circuit parasitics would cause lagging phase swings. In both cases I made use of dominant pole compensation to achieve stability. In a feedback circuit, if the locus of the loop gain calculated as the product of the total forward gain and the feedback ratio passes through or encompasses the point (-1, 0) in the complex plane, the circuit will be unstable and oscillate. Care must therefore be taken to prevent the locus of the loop gain from getting near this point. Each RC time constant in the circuit will be responsible for either a 90 degree lead or lag phase angle. In a 3 stage design there will be 3 RC time constants and so the lead/lag angle will approach 270 degrees, swinging through 180 degrees as it does so. The principal of dominant pole compensation is to make sure the loop gain has reduced below unity when the locus of the loop gain passes through 180 degrees. Preferably the locus should not cut the line x = -0.5 - the line of critical stability. This is achieved by making one of the poles operate (dominate) before the others do. In the case of the AC coupling this necessitated choosing one of the cut-off frequencies of the coupling circuits to be around 20 – 30Hz and making the other two cut-off frequencies much lower (say around 0.5Hz). For the high frequency stability I made use of the Miller capacitance effect on the voltage gain common emitter stage. By introducing a small base-collector capacitor, one can control the pole formed by this capacitor. Here again setting this pole sufficiently below the other high frequency poles made the design stable.
Once complete I ran some tests on the final design, such as full power tests to ensure it achieved the required 30W output. I also calculated the heat sinking requirements for the power devices.
With the power amplifier complete I designed the power supply and pre-amp. The power supply consisted of a bridge rectifier and smoothing capacitors and LM317/LM337 regulator circuits with TO-3 pass transistors to provide the required rail voltages for the power amplifier. The pre-amplifier design used fairly standard circuits for the tone controls, volume and balance controls, non-inverting buffers on the inputs etc. I designed the microphone input using a two stage discrete transistor design to achieve best noise performance.
I based the box design on a typical commercial Hi-Fi amplifier component shape. I designed 3 PC boards to go inside, keeping the power supply, power-amplifier and pre-amp on separate boards. I purchased a 180VA standard E-core type transformer for the design. I spent a fair amount of effort on the layout and appearance of the front panel, since I intended the design to look like a commercial product (like a “bought” one). I completed the construction of this project before commencing my third year at University.
On the whole the completed amplifier project was very successful and performed well against similar commercial designs in the same class and budget range. At the time I constructed this amplifier (being a university student) I did not even own a pair of speakers or even a CD player. A few years later I purchased my CD player and subsequently my first pair of loudspeakers. When I made these purchases I took my amplifier to the local audio shops and the owners allowed me to test it against a few of the models they had in stock as well as testing it with the CD player and speakers I wanted to purchase. My first pair of loudspeakers were B&W 305 floor standers. These I purchased from the local Hi-Fi store, making my acquaintance with the owner with whom I would work at a later stage at GGAcoustics.
Once I had a workable set-up of CD player, speakers and amplifier I started experimenting with the amplifier design. Having heard a range of audio equipment in the Hi-Fi shops I wanted to improve my amplifier. I tried larger capacitors on the main power supply, a larger transformer, thicker internal wiring and better quality capacitors with lower ESR. Each incremental modification made to the amplifier was accompanied by a change in the sound. With some of the modifications the effect on the sound was more noticeable than others. I assessed the effect of each modification, by making only one change at a time, then listening to the amplifier. Sometimes the effect was immediately noticeable and other times it took hours of listening over several days to me decide whether the change had been for the better or was even noticeable at all. On some occasions my initial impression following a modification was that there was an improvement in the sound, but after a few days I began to feel that something was lacking in the sound. Eventually I would revert things back to how they were prior to the modification and usually it was immediately apparent that the “goodness” had been restored. Each time I stumbled across an effect that appeared to have a positive influence on the sound I would attempt to apply a logical technical explanation as to why this would be. This way I developed my own set of design “rules” for high end audio designs.
Eventually I decided to rework all of the circuits and PC Boards. I wanted to do this because there were a number of technical issues with the original design that I was not happy with and I had some new ideas I wanted to implement which I knew would result in a significant improvement over the previous design. I also replaced the 180VA square E-core transformer with a 300VA toroidal transformer. The toroidal transformer occupied about the same volume as the E-core transformer and so fitted in to the space available quite nicely. Due to their geometry, toroidal transformers can deliver more power than E-core transformers of similar size and so are the preferred choice for audio amplifiers. Toroidal transformers generally also have lower flux leakage than E-core types. I reworked the power supply board making tracks thicker and replacing the reservoir capacitors with better quality and larger capacity components. I added a simple “soft start” circuit to the transformer to allow the power supply capacitors to charge up gently at switch on.
For the new power amplifier design I decided to use a DC coupled topology. I had also devised a unique circuit topology of my own where the voltage gain stage isn't just a simple common emitter stage, but is more of a push-pull design. This has numerous advantages in terms of the voltage swing capability of the stage as well as being inherently more linear. Also with the removal of the AC coupling capacitors I eliminated one possible source of distortion. All capacitors exhibit dielectric absorption – some more than others. This is quantified as the loss tangent of the capacitor and is more prominent at higher frequencies. Dielectric absorption is known to be non-linear- it is similar to the hysteresis effect in magnetism. It stands to reason that any component in an audio circuit that exhibits non-linear characteristics is undesirable. If these components can be reduced in number or removed altogether, so much the better. Where capacitors have to be used they should be selected for lowest ESR (equivalent series resistance) and loss tangent. The ESR is a lumped quantity which represents the series resistance of the leads and plates summed together with the losses in the dielectric. Therefore a low ESR capacitor will have a low loss dielectric. From my work in audio amplifiers, capacitors are the single biggest culprit responsible for the harsh, shrill sound of cheaper audio equipment. The manufacturers use standard grade components to cut costs. Whilst I was making modifications to my amplifier I also decided to see what I could change in my CD player. I replaced the standard grade capacitors in the audio output chain with low ESR ones and the improvement was obvious. I also replaced the filter op-amp with a high quality audio op-amp of the type used in professional audio equipment. Here again the improvement in sound quality was quite noticeable.
In calculating the component values for the new power amplifier design I also decided to take my analysis to a new level. I derived transfer functions for the voltage gain stage and input differential amplifier stage using the Ybers-Moll transistor model. These transfer functions contain exponential expressions and so cannot be solved mathematically. I wrote a program for my HP48G calculator to numerically solve these equations so I could estimate the gain for each stage. The program also gave me an estimate of the harmonic signal levels I could expect from each stage which I could incorporate into the feedback model of the entire amplifier to estimate expected distortion figures for it when it was being operated at full rated power.
For the pre-amplifier I kept most of the circuit the same, aside from changing the op-amps and capacitors I used. The main aim behind the rework of the pre-amplifier was to improve the board layout.
Internal view. Stereo power amp board at the back, Power supply bottom left, Preamp bottom right.
When it came to testing the new design I discovered I was getting some crosstalk between the left and right channels. I tracked this crosstalk down to the pre-amp layout. I had used a single ground for the entire circuit and return currents from each channel were inducing small voltages in the ground and thus getting into the other channel. I made some modifications to the board by cutting a few sections of the ground and providing wires from a few critical points directly back to the star ground point at the supply input to the board. This issue with the grounding was my first major encounter with grounding problems and it started me thinking about possible methods for suitable grounding. Subsequent experiments with ground layouts on other hobby projects and work related designs have lead me to the philosophy I currently use which is mentioned in the final part of the Airband receiver project.
The improvement in sound was certainly worth the effort of reworking the design. Even after I had completed the rework I made some further minor modifications in the form of a few component value changes and such like. However eventually I reached the point where I realised that I had taken the design in its existing form as far as I practically could and started thinking about embarking on a new design. This new design took the form of the 120W amplifier project. This 30W amplifier still performs well today.
First Design: Feb 1993 - Jan 1994
Further work: 1997