What may come as a surprise is that class 'D' amplification is all around us, whether we realise it or not. Maybe the mobile phone, MP3 player, TV sound system, sub woofer, 5.1 surround sound system or computer sound card etc. almost all now use class 'D' technology, so it is not new and not unknown, just not realised.
The background
Back in the 1960's, the great Sir Clive Sinclair devised and sold an audio amp system comprising some very small transistors that would revolutionise audio reproduction, a brilliant idea that had it been practical at that time, may have made a serious impact,. However it didn't, as many of his units did not work well or blew up in a short space of time.
I think this, more than all else, put class 'D' out of mind as a serious contender. At that time it was more a case of no suitable transistors being available than any basic flaw in the principle or design. Since writing the original draft I have spoken to my ex employer Vic Cedar and to my surprise he knew Clive and with his connections at Texas Instruments helped Clive solve the problems, but I think by then many had lost interest in class 'D'. (Note the very first integrated circuit audio amp was announced shortly after the demise of the first class 'D' design, another original from the amazing brain of Sir Clive Sinclair!)
By about the 1990s many companies of good repute had joined the class 'D' revolution including Maxim, Texas Instruments, Zetex, Analog Devices, National semiconductor and Tripath etc., these are all major players in the chip industry. Texas Instruments were the first to create an integrated circuit, and Analog Devices are in the forefront of many of the best D/A converters around, quite apart from their numerically controlled oscillator chips that have changed the world of signal generation and radio. So I think these guys know what they are on about and are not developing a technology that has no future.
This is borne out of the fact that this system (class 'D') is now used in many bass subs, TV sound systems, in-car audio and even mobile phones and MP3 players, so I think to challenge the technology is a bit late in the day.
However in the cloistered world of esoteric hi-fi it appears that any technology that is less than 40 years old must be suspect. After all, transistor amps, never did sound as good as push pull valves, but they were never as good as class A single ended triodes, and they were never as good as bright emitters.
I recently downloaded a track from a cylinder recording made in the 1890s and I have to admit that it sounded truly awful compared to even a 78RPM recording from the 1950s, so I think that the art of Hi-fi is improving, and yes, I do believe early transistor amps did not sound that great.
Much of the reason for this was in their 'sonic signature'. By that I mean the spectrum of their distortion components. If one were to compare the highly regarded Leak valve amplifiers of the day, to their later introduced TL30 transistor amplifier, note that although the specs. may look similar on paper, the distortion spectrum shows a different picture, and one that is more in keeping with the hearing experience, so take this to mean that spectrum analysis is a good indicator of perceived performance, and probably the most accurate.
The spectrum of most valve amps is such that all distortion components reduce as the output level reduces, and at real listening levels, the distortion, largely second and third harmonic is below the noise floor, this also holds true for vinyl. However transistor class 'B' is another story where the distortion actually increases as the level is reduced and often what is left is the high and undesirable harmonics. Unfortunately a CD recording also has these characteristics, note I say recording, as this has nothing to do with the player, it is a characteristic of digital methods. To the point where adding a little background hiss masks the high order harmonics and can result in a better listening experience!
From my understanding class 'D' should also have the 'reducing distortion with level phenomenon', but this often gets masked by the measurement methods. From these examples I can understand why, for ultimate sonic nirvana, vinyl and valve are probably perceived as the only way.
So what is class 'D'?
Class 'D' is basically a continuation from A, B, and C class amplification.
Class 'A' means that the devices, (valve or transistor) are conducting throughout the complete audio cycle. I don't think anyone who has heard a class A amp would not agree with its merits, its warmth, fullness etc.
Class 'B' means that the devices are conducting for half the cycle, either on positive or negative halves of the AC cycle, and virtually all conventional transistor amps fall into this category, as did many valve units.
Class 'C' means the device is conducting for less than half a cycle.
Class 'D' means the output device is conducting for very much less than half a cycle, possibly microseconds.
It is as simple as that.
Class 'C' is therefore pulsing the positive or negative half cycles. This is of no use for audio but is used in radio transmission as a tank resonator restores these pulses to their original form. It can be 80% efficient in converting DC power into AC, whereas class 'B' cannot do better than 66% and class 'A' cannot better 25%, and that is assuming perfect theoretical devices. On the other hand, class 'D' is capable of some 95% conversion efficiency!
Class 'T' and class 'Z' are propriety names for individual company's slant on the theme and not a specific amplifier class.
Class 'D' means that a transistor, positive or negative, is always conducting at any one time, and that means that the speaker is always connected to something solid like a power rail of virtually zero impedance, and that means it is always heavily damped and that means that the speakers have no time to flap around resonating and doing their own thing, as they are under such tight control of the amplifier.
This, in my opinion, accounts for the tight and controlled response I have heard from these devices. It also does not have the transition from one conducting device to the other, at audio frequencies, and therefore no discontinuity in the centre, the one place where our hearing is the most sensitive.
Note that no natural sound, a string dying away, a oboe reed etc can ever have a discontinuity at the centre of its waveform. I believe that this accounts for the smooth, detailed and fat warm sound I have heard from this technology.
No doubt you are familiar with dimmer switchers. The normal household dimmer switch works by a switching principle. It switches power on after a pre-determined time from the zero crossing point of the mains 50Hz cycle, if this were to be 9 milliseconds after the zero point the light would be very dim, and if it were to be 1 millisecond after the zero point then it would be practically full on. I don't think you would have dimmers at home if they were massive heat producing rheostats as used by the theatre companies up until the 1960's.
So switching technology began to gain increasing acceptance in many fields, particularly in power supplies, to the point that almost no piece of consumer electronics does not have a switched mode power supply in it, be it a TV, DVD player, computer or even a low energy light bulb, and these units appear to be accepted by all, without question. Of course they are far more power efficient than linear supplies, but when it comes to top end Hi-fi, a switch mode supply or anything that might generate some high harmonics, is considered totally out of order, even though it may be amplifying the TV sound source, computer, DVD or CD player powered by a switch mode supply.
Class 'D' is nothing more than a switched mode supply that can perform both positive and negative transitions very quickly and in real time. As its transition frequency is high then its performance is very like the power supplies we all accept, and all harmonics it may generate are easily filtered out, as they are in any decent power supply design.
The principles
The basic principles of class D are very simple. The supply voltage is switched to the loudspeaker, either as a positive source or a negative one, very quickly, so quickly in fact that you would never know, in the case of a Tripath class 'T' amp (their name), about 500,000 to 700,000 times a second. That is a lot faster than a long wave broadcast frequency and therefore not difficult to filter out, and still preserve the audio bandwidth.
As an audio aficionado you might believe you can hear beyond this, however this is not so, what you may hear are down converted artefacts of these higher frequencies getting mixed in a non linear device. Nyquest theory proves that there is no information on a CD beyond 22 KHz and even on the best pressed vinyl, nothing beyond 30KHz.
The theory
Class 'D', often wrongly called 'digital', probably because it has a 'D' in the title, is simply a continuation of the 'A', 'B', 'C' theme, and is not really any more digital than the original FM radio, in fact very similar except for the base frequency and some subtle differences in modulation. Remember, in FM radio the IF frequency is amplitude limited to remove any AM components and that results in a 0 or 1 situation.
Hence it could be called digital, but the timing now carries the audio information. This is exactly what happens with class 'D', it is the timing that carries the sound information and not the amplitude.
The form it takes is Pulse Width Modulation (PWM) and this has been around for very many years, and was used extensively by phone networks to improve speech quality and reduce noise on long distance lines between exchanges before the advent of eight bit digital that is now used.
A small discourse into the use of the word 'digital': To me and others, digital means to convert a value, voltage etc. into a number as in the case of 8 bit telecoms or 16 bit CD, if it is not converted into a number then it is not really digital. PWM like FM radio is not therefore digital, and the name 'Class 'D' was coined well before digital systems even existed.
What is now making PWM come out of the woodwork and into the mainstream is largely to do with the speed of transistors - we all remember the days of 4.7MHz computers, now often 3GHz.!
In its simplest form, a PWM signal is created by comparing the incoming audio at low level, to a perfectly linear and fast triangular wave generated in the amplifier chip. Once the comparisons are made, the signal then becomes a pulse train of changing mark/space ratio, a time rather than a level thing. This is then amplified by power devices to the 10 or 100W level before being turned back into analogue audio by a simple low pass filter. Before the filtering it has all the characteristics of an AM broadcast spectrum and a replica of the original signal, the AM part being filtered out.
This is maybe a very simplified explanation but I hope it gives you an idea about what is going on with a PWM signal.
I believe some companies have taken the original CD 16bit data and using digital signal processing (DSP) have converted that directly into a PWM form, thus removing the need for the triangle wave generation and D/A conversion. This is still class 'D' but could be called digital, as we're back to dealing with numbers.
Note that there are not many stages in a class 'D' circuit where harmonic distortion can creep in. The first depends on our triangle linearity, and the second is in the need to add dead time, to the output switching transistors in order to avoid any chance of overlap and certain destruction.
With transistors getting faster and faster, the dead time cause of distortion is rapidly disappearing, and using the DSP route, ie removing the triangle wave comparison section, our linearity cause is removed.
However in an analogue (standard) amplifier, every device can contribute distortion, and then more feedback can make this distortion effectively disappear. Every capacitor, resistor, transistor etc. can affect the final sound, and whilst just adding more feedback does reduce the lower harmonics, (and improve the specs), all at the expense of creating higher harmonics and it is these which are the most undesirable.
A crucial point also, is in our perception of distortion, it is the actual harmonics that count not necessarily the total, 3% of second and fourth probably sounds actually better than none at all, hence the perceived beauty of single ended triode amps. But 0.1% of seventh and ninth would sound harsh and tiring. Such is the way we hear.
In a recent press release (Jan 2007) from Zetex, they stated that a new chip set was capable of up to 500W into 4R at 0.004% distortion, so we are now seeing figures that put many of the top conventional amplifiers to shame, valves being completely lost in the rush.
Remember the early chess programs, running on a Spectrum computer that would say “I didn't expect that” when all it had, sound wise, was a single port pin and no analogue circuitry. Now that was unbelievable that it could be done at all, although pretty awful sound quality, but another example of class 'D'.
My audio experiences
Some years ago (2002) I visited a hi-fi show in Kent and having listened to some very expensive and exotic valve gear the only sound I was impressed with came from a massive horn-loaded speaker with a Lowether drive unit.
At this point I enquired about the amp etc. and the guy showed me his home made Tripath class 'T' module powered by a small 12V battery in a beautiful wooden case and driven by a portable CD player, but the sound was awesome, and this was in mono!
My own personal experience of class 'D' was to find an amplifier chip that would not heat up and scorch the PCB it was mounted on, unlike the previous linear TDA2003 that we had used, and give us a bit more power, for the talk back system.
It had to run on 12V, we were already using 4R speakers, so a full bridge design was the only way to achieve this. This was not a Hi-fi problem but simply a logistics one. We wanted a change to the video assist talk back system and the following is the route that I followed.
After much searching on the web, a Texas Instruments class 'D' amp kit from Ramsey Electronics (USA) was found. Having built up a couple of these kits, for a test, I connected them to a pair of Genexxa speakers having ribbon tweeters, and fed them from a good Sony CD walkman player and…
Well I was amazed at the pin point 3D positioning and holographic soundstage I heard, plus the overall sound, and this was unbelievable from two £30 kits and two PP3 9V batteries. Even this 'bottom end' design had many of the subtle mid frequency nuances that I had only ever associated with top end hi-fi.
From this point on, I started to believe in the sonic merits of class 'D' technology. Since then I have purchased two amplifier boards from Autocostruire, one a low power Tripath 20W per channel kit, now built into a small case with 12V battery power, and the other a ready built Tripath 100W per channel module, and apart from the obvious power difference they sound exactly the same and that is sweet, exceptional, excellent, warm, cosy, natural, and better than anything previous, including my Radford or Leak amplifiers from the past.
The future
I strongly believe that over the next ten years, class 'D' will creep into all aspects of sound reproduction without anyone ever knowing it, as indeed has the use of integrated circuits (ICs) over the past twenty years. I know many believe the IC to be inferior but that is more imagined than real - after all an IC is only a collection of transistors in one bit of plastic, and not a different technology at all. However an IC being less than forty years old must be suspect with the esoteric Hi-fi community.
Please note that all I have said is an observation of what I've seen, heard, found and read, and all for my own progress towards the ultimate Hi-fi nirvana.
My business is more to do with squeezing the last drop of performance from a PIC micro controller or sending RGB video images down many metres of cheap twisted pair cable or creating an interesting lighting effect for a feature film or pop group, or just looking at an interesting problem and trying to solve it.
Please understand that I have no commercial interest in the Hi-fi trade, this is merely a hobby, as was writing this article.
Agree with John? Disagree? Or just want to know more? Leave your comments below.
Want to write a feature for AVReview? Send it to editor@avreview.co.uk
