Why aren't class-D amps, digital amplifiers?

[philthien]

This has come up a few times, when I refer to class-D amplifiers as digital amps and someone corrects me and says they're analog amps.

Aren't they technically digital amps? They output a signal that approximates an analog one, but they're doing it by switching on/off very quickly.

It seems to me they're analog amps about as much as the CD is an analog medium because we hear something that approximates analog coming from it.

Please discuss.

Extra points for anyone that would like to talk about early experiences w/ Class-D amplifiers. How long ago were you exposed? How did they sound then? Any notable older products that used them?

[Pete Schumacher ®]

A digital amp would be akin to a Power DAC, which a class D amp is not.

[dcibel]

Hypex whitepaper on the subject:
https://www.hypex.nl/img/upload/doc/...e_analogue.pdf

[philthien]

A white paper, wow! And pretty well thought out. So digital/analog are determined by what is done with the output. But what if the amplifier is generating tones (symbols) via a speaker, I guess in that special case, it would be a digital amp.

I better get used to calling them class-d.

[dst]

My main system right now uses some older DIY Hypex UCD180 amp modules. They make very good amps. My preference in amplification is tube. Overall, if you like the sound of a decent solid state amplifier then I think you would like Hypex. Very quiet background. Lots of bass slam. Not bright or edgy. I still prefer my EL34 tube amp (currently out of commission), but the class D are still good. Hypex has been around for more than a decade I think. They still sell UCD amps but also have the newer N-Core technology.

Class D uses transistors but not as linear gain devices (as in typical solid state amplification). Instead the transistors are used as on/off switches. Class D is essentially a switching amplifier, and sometimes called that. The audio is encoded (but not digitally) in an electronic pulse train by a modulator. Thus Pulse Width Modulation. It is released from the pulse train through a simple low pass filter (not a DAC). No sampling, error correction, or binary conversion exists in Class D as in digital (DACs/ADCs). This modulation and pulse train are very different than digital conversion. Digital audio takes tens of thousands of "samples" per second and converts them to binary (digital) form -- 1's and 0's. Class D is not binary, and therefore not digital.

Switch mode can also be used in power supplies, thus SMPS, a switch mode power supply. You see them all the time. Even in high end audio an SMPS can replace a linear power supply. Just as with a Class D amplifier, an SMPS is not a "digital" power supply. Electrical current and its properties are still in play, not 1's and 0's.

When you plug in your iPhone or laptop via its wall-wart power supply, that power supply is a small SMPS. It is still dealing with electrical current and taking 115 volt AC and turning it into 5 volt DC, but not in the traditional linear power supply manner (step down transformer, rectification, filtering).

I probably did not explain it so well, so you may want to read up on the subject. Another thought: Essentially you can make analogs in many ways, but they need to be in a binary format to be considered digital. A traditional photo from film is an analog of a visual moment in time. Light hit the film which later is "developed" via a chemical process that creates the photo, a pretty good capture of that moment. A digital camera skips the film and chemicals and instead captures the moment in binary (digital) form. The end product looks quite similar, but the digital process is much more efficient.

Call it what you will, Class D can make very good music.

[dst]

Another good place to read about all kinds of audio stuff, including digital, is at the PS Audio website. they post all kinds of things, articles, blogs, even a small magazine called "Copper." Great place for the latest in audio electronics, especially digital. they also have a cool DIY phono preamp kit available.

[dcibel]

A white paper, wow! And pretty well thought out. So digital/analog are determined by what is done with the output. But what if the amplifier is generating tones (symbols) via a speaker, I guess in that special case, it would be a digital amp.

I better get used to calling them class-d.

"Digital" is often a misused term. It should only apply to a representation of data/information. A switching signal such as a PWM, is ultimately a representation of an analog signal, and the device functions as analog at the input, analog at the output.

[djg]

Still no love for classdaudio dot com.

[TMM]

Pulse width modulation like Class-D amplifiers and Switching powersupplies use is not digital because although they switch between 'high' and 'low' pulses, unlike a digital data stream there is no timing clock which the 'high' and 'low' pulses must adhere to. Transitions between low and high can occur at any moment. The result is that because the amplifier can adjust the width of each pulse with infinite resolution, each pulse does not simply represent two discrete data values (1 and 0) but continuous analogue values (any number between 1 and 0 with infinite resolution e.g. 0.2349587394857983475....). Because of this there is theoretically no loss of information when an analogue signal is converted into a pulse width modulated signal as long as the switching frequency is higher than the signal frequency.

Digital audio data is different because the sample/pulse rate is fixed to a predetermined rate (therefore each sample occupies a set time slot) and each sample is quantized into one of a set of values. Take CD for instance, it uses 44100 samples per second each being 2^16bits = 65536 different values. Information is lost in the analogue-to-digital conversion process - e.g. the converter might have to decide whether a value of 12656 or 12657 best represents the analogue signal during a sampling time slot. It cannot represent a fractional value such as 12656.23423459.
The advantage of digital data is that when stored or transmitted, small errors in timing or amplitude can be corrected by simply reclocking the data because we know that each sample should occupy a set time slot and range of discrete values.

[PWR RYD]

Still no love for classdaudio dot com.

I think they are awesome sounding amp modules.

[billfitzmaurice]

At some point early on someone thought that the 'D' in Class D stood for 'digital', and the myth was born. It hasn't died yet.

[neildavis]

That digital vs analog argument is a lot like the blind men and the elephant parable: it depends on what part of the amplifier you are trying to describe. There are parts of class D amplifiers that use analog processing, and there are parts that may use digital signals and digital processing. It depends on what part of the elephant you are looking at and what you feel is the most important characteristic. Also, there are a lot of different "species" of class D amps. Some are self-oscillating, some use fixed-clock PWM, some use PDM, and some use Delta-Sigma conversion techniques. Some people seem to have an obsession to correct others who use the term "digital amplifier", but there are many types of class D amps and aspects of those amps where the term makes sense from an engineering perspective.

[philthien]

Still no love for classdaudio dot com.

There was recently a 4x300-watt unit with balanced and unbalanced inputs on eBay for a reasonable sum. If it were a little smaller or if I had a little more room, I'd likely own it by now.

[philthien]

At some point early on someone thought that the 'D' in Class D stood for 'digital', and the myth was born. It hasn't died yet.

Why were they originally called class-D, why not class-S (for switching)? Or were there already class-S amps?

[JRT]

Why were they originally called class-D, why not class-S (for switching)? Or were there already class-S amps?

Class D because D is the next letter after class C, sequentially.