Are Class D amplifiers digital?

I am getting close to publishing my memoir! It's called 99% True and it is chock full of adventures, debauchery, struggles, heartwarming stories, triumphs and failures, great belly laughs, and a peek inside the high-end audio industry you've never known before.

[mr.bennett]

The analogy that did it for me was Flappy Bird. In the game, when you tap the screen, the bird jumps up, and if you don't tap the screen, the bird falls, tracing out a wave. A PWM signal chops up the continuous waveform into "frames, " and pulses a certain number of times each frame, just like tapping the screen on your phone to move the Flappy Bird up and down, and traces out a continuous waveform. There is some "stepping" in the wave, but that's in the same way there is stepping on your computer monitor when it draws a curve across square pixels. Thanks to Integrals, the farther out you zoom (in this case, chopping it into small enough frames, ) the smoother it gets. Eventually, you can't actually detect the aliasing, it just looks like a smooth curve.

[tribbinvw]

It's pulse-width modulation on an audio signal to drive the transistors that drive the speakers, because transistors are most efficient when fully open or fully closed. If the transistor outputs 5 volt and (at a specific point in time in the waveform) needs to put out 1 volt, then the transistor will be rapidly turned on 20% of the time. This happens at frequencies far beyond human hearing. The output is smoothed out by an electric circuit to suppress the RF.

The output (before filtering) is 'binary' in that it is outputting 0 volt and 5 volt, but is not considered digital, because the (what would wrongly be considered) 'digital signal' is like a stream of 0101010101, but with alternating intervals to get the desired mean (within hearing-range intervals) output-voltage.

Most times there are two transistors; one for above 0 volt input and one for below 0 volt input. For example; one transistor outputs 0 volt and +5 volt and the other 0 volt and -5 volt, similar to the push-pull mechanism of class AB amplifiers.

[MindBodySoulOk]

I just discovered your video series a month or so ago. I could seriously consider coming by and hearing what real sound sounds like. I would assume that any new found success you are having is due to these excellent videos.

[ThinkingBetter]

A class-D amplifier performance definitely depends on the output stage and how the filtering and feedback operates. The speaker load together with the output stage capacitors and inductors contribute in how the signal is shaped relative to the PWM timing. Class D amplifiers need some sophistication in the output stage to reproduce the analog input signal with difficult speaker loads or it performs like if you do hand drawing with closed eyes. Nowadays class D amplifiers are made using chipsets from ICEpower, NXP, TI or others and most of the magic is done within that chipset. Using a great enclosure and better quality components than the original reference design will enable some differentiation but 90% of the AQ is usually in the chipset used.

[Roxor]

Fun fact: PWM is a long-used technique among PC programmers to get some pretty good sound out of the PC Speaker. Just keep in mind this is _the PC Speaker_ we're talking about. A little 60-80mm cardboard-cone thing driven by a strictly on/off timer chip that normally only outputs square waves and these days really only gets used for the beep codes from motherboard diagnostics.

Still, with the timer chip running at around 1.2MHz, that's enough to construct a PWM signal with a clock of around 16-18kHz and around 6 bits resolution for the width. CPU-intensive and inferior quality to even an 8-bit sound card (or home-made parallel port resistor-ladder DAC), but still pretty impressive for what's being used.


If you've got good hearing (in terms of frequency response), you can hear the 16.5kHz PWM carrier whine in the background of the credits. If you can't, load it up in Audacity and switch to spectrogram view and you'll see a very strong line at 16.5kHz.

[randomgeocacher]

Back in school the definition of digital signal was a signal that had discreet values on *both* time and value, e.g. PCM. So if a signal has any axis that is analog (e.g. time is analog or amplitude is analog), it wasn’t considered digital. If I understood you correctly this PWM system has a digital amplitude but an analog time axis, I.e. not meeting school definition of digital.

[shaunhw]

2.48: 16 bits (unsigned) is max whole number = 65535 = all 16 bits on, or 1. IE:  (2 (power of) 16)    - 1.  One is subtracted because there are 65536 binary states, (number of combinations 16 bits can have) but 0 (all bits off or 0)  is the first state, so 65535 is the last state. For audio, bit 15 would usually indicate the polarity in some way, for example with the max number reduced to - 32768 to + 32767 (two's complement form)

[tacticalAV]

MOSFETs... it's really the post filter feedback and output filters that give class d amps their signature sounds. Pascal, ICE modules and are leading that market now. I just bought 2 Red Dragon Audio s500 stereo power amps at 1000 watts RMS at 8 ohms but with .003% thd.

[tonyhunt]

Great explanation Paul. I've learned something today. Cheers!

[jeffm]

For most Class D amps the pulse width is infinitely variable within a lower and upper range based on an analog input. This makes that type Analog in nature. Now some class D amps take a digital signal only (PWM) and convert that signal to discrete pulse widths with a finite resolution, these are effectively digital amps as numbers are involved internally. The output of course ultimately ends up Analog like everything else a person needs to Sense. Is my OLED TV digital or Analog, well it's both. Same can be said for Some Class D amps, while others are very much analog end to end.

[janinapalmer]

Sinclair electronics introduced the first PWM amplifier way back in 1963 a very simple little kit that gave 10 Watts output and ran stone cold ... transistors are at home with pure square waves ..

[Peter_S_]

Here's another way to think of it.... A class D amp absolutely uses a digital output stage but it's like a 1 bit DAC with an _extremely high clock frequency._ In the systems we usually think of as "digital" we use lots of bits at a relatively low sample rate and the result is that higher audio frequencies are not able to be captured with the same fidelity as lower audio frequencies. With a 1 bit output and a 2MHz sample clock, that frequency dependant error quality of digital is shifted upward so now your "low" frequencies comprise the entire audio band, allowing the whole audio band to be recreated with uniform, excellent fidelity. An inductor has no problem converting that high speed digital PWM into golden analog goodness. It's all about the speed of the data, not its mass.

[klazzera]

If class d amplifiers do have a clock and on/off state, then the state of that switch in a duration of an audio clip should be exactly represantable by a digital data set of a finite size. If the PWM generator didn't have a clock and the pulse width was modulated using input signal amplitude directly or after some short analog integration, the signal would have a resolution only limited by the analog noise. But when you bring the clock in and define the pulse widths in finite steps, the on/off states become digitally represantable thus digitized, just not recorded in numbers. I'm still in favor of class d amplifiers because they are very efficient in many means

[dhpbear]

Even if one considered the ON/OFF pulses as 'digital', the signal gets passed through an RC filter which *IS* analog :)

[marianneoelund]

Yes, Paul got this one right.
It is perhaps unfortunate that when the designation for Class D was chosen, the letter 'D' just happened to be the next one available after the class A, B and C amplifiers had been named.
Even more confusing, 'D' can stand for either "digital" or "discrete" which are two different concepts.
Now, just to start a war: Is DSD truly digital? Yes, it is usually generated and converted back to analog using true digital calculations, and the data is always stored on systems which are certainly digital, however - it can be coded and decoded with analog electronics only, and there is no defined bit weighting or word length which allows the bits in the stream to be interpreted as numerical values. It is rather like asking whether a set of tally marks on paper truly constitutes a number system.

Can you talk and explain your opinions for audio exciters (speakers, that you attach to surfaces to make sound).

[motorradmike]

Paul, does the new Sprout 100 convert an analog signal to digital using a Burr Brown analogue to digital converter in the preamp section to control the volume in the digital realm and the back to analog to the class D amp?

[kencohagen]

According to what I remember, when it comes to digital amplifiers they used several pulses per second to synthesize analog type A amplifiers. In fact Type A amplifiers are somewhat limited by there bias voltage. But that's a lengthy discussion and what I mean is that where a type A amp shows zero voltage can be biased higher or lower than actual zero volts. Tell me if I have this wrong Paul. An A/B amp uses two tubes or transistors to divide the voltage equally down the middle so that zero volts is zero volts period. The problem with this type of amp is that there may be a few miilivolts positive or negative before it actually starts to move the drivers. So there is some slack in the middle. Again, correct me if I'm wrong Paul, I'm a student as well as an closet audiophile! With digital amplifiers there's are many more positive and negative pulses than an analogue amplifier, based on the theory that Nyquist established. Nyquist said that if an audio signal sample was sampled something lie 8 times, or eight pulses per second, you couldn't hear the difference between a digital and analogue signal. Well that might be true for low frequencies, but not mid and high frequencies so manufacturers raised the sampling rate substantially. When they got it high enough it worked, and the benifit is a reduction in heat, or wasted energy required to drive a speaker. So, if done right, class D or other more advanced digital amplifiers can sound as good as an analog amp.

[InsideOfMyOwnMind]

If you were to consider class d to be digital then in the same sense AM and FM radio are also digital with the sampling rates being 1Mhz and 100Mhz respectively. Take it a step further and vinyl and tape are digital with the sampling rate being a function of the listener's attention span or the rate that the grains of metal oxide pass the head gap. Ah symantics.

[joshuacooper]

I love your videos, I've learned so much from you! I experience a video lag on every video you make. I don't experience this with other channels I watch. It makes me think it's something on your end, I could be wrong though. I just wanted you to know.