The Great Class D Controversy

Class D, or switching amplifiers, are something of a third rail for audiophiles. Some audio enthusiasts claim they are the best thing to happen to audio and are taking over the amplifier market. And maybe they are, but maybe for the wrong reasons. Here is an overview of the technology so you can draw your own conclusions. 

Like all technologies, things are constantly evolving, and some of us (me at least) have moved from being outright skeptics to at least agnostic and willing to re-evaluate as things progress. For an easy-to-read explanation of the basics of D-type amps, see this post in the Audio Science Review forum by member DonH56. Also, see this post in the same forum with a good list of links to Class D information.

They have been around for decades, and much like early transistor amps (see my article here), were pretty awful in their initial stages of development. B. D. Bedford was issued US patent # 1,874,159 in 1932 for a switching amplifier using tubes. The idea of using pulse-width modulation techniques is generally credited to Alec Reeves, a British scientist who first described the concept in the late 1930s and received over 50 patents relating to communications. The basic idea is to modulate the audio signal with a high-frequency carrier of varying pulse width, controlled by electronic switches. Since the switches are either on or off, very little power is wasted in the modulation process. Practical circuits did not emerge until good transistors became available in the 1960s.  Below are some proof points that Class D for HiFi has been around for a long time!

Above is the Sinclair X-10 amplifier DIY module which was introduced in 1964 in the UK, claiming to be the world's first pulse width modulated (later to be called Class D) HiFi amplifier. It relied on the speaker to act as a filter choke for the high-frequency switching noise. It was sold as a module aimed at the DIY/experimenter market since you had to build a power supply and enclosure to use it.

Advertising was primarily in Wireless World (UK) magazine and many were sold. And many blew up (mainly due to germanium transistor failures), prompting many complaints. So it was generally considered a failure. Like lots of other products introduced by Clive Sinclair, it was ahead of its time.

And we have an article that appeared in the September 1967 edition of Radio Electronics magazine describing the circuit of the improved version of the X-10, the X-20.

It took several years and improvements in semiconductors to make major advances in Class D. This is the Sony TA-N88, not introduced until 1977. It was rated at 160 WPC into 8 Ohms. Sony used their VFET technology to handle the switching function. Unfortunately, it was plagued with heat and reliability problems, so really did not take advantage of Class D technology.

Like many vintage HiFi devices, this has become a "collectible" despite the lack of repair parts, particularly the original VFETS, which are long out of production.

So why are they gaining in popularity?


They are much more efficient than Class A or A/B. This means less power dissipation at idle and full power, a smaller form factor, and lighter weight. A Class D amp can theoretically approach 100% power supply efficiency, although practical circuits are more like 90-95% due to switching and filter losses. Some manufacturers have used them to build multi-channel amplifiers for home theater in relatively compact chassis. Also, they have been used in micro-power applications like hearing aids where battery life is critical. They have been also popular in sound reinforcement applications - like PA and concert applications and for high-power automobile amplification. Also as musical instrument amplifiers where portability is important, but these are not exactly the same as audiophile applications where critical listening and euphonic sound are of interest to audiophiles. 

 A Hypex OEM module

However, some dedicated engineering talent has been applied to the technology and high-end amplifiers based on Class D have emerged in the past few years. One leading example is Bruno Putzys, who designed what many consider the first truly audiophile-quality amplifiers while working for Hypex. Hypex calls these nCore modules and they have been incorporated into many high-end manufacturers’ amplifiers. Bruno went on to his own startup partnership called Purifi. They produce a small amp module called the IET400A capable of 400 Watts into 4 Ohms with 95% efficiency and remarkably low distortion.


If you are wondering why mainline OEM manufacturers like NAD, Bell Canto, and Marantz incorporate modules like these instead of developing their own? The answer is simple: It is REALLY REALLY hard to get a high-end Class D amplifier to approach the ideal of a “straight wire with gain”. The investment required is too great for the potential returns in the high-end market. Bruno himself has mentioned this in interviews. The best-in-class techniques and mathematical analysis required are known to only a few engineers.


For a glimpse into some of the technical challenges implementing Class D, see this presentation Bruno gave way back in 2008 titled "A Universal Grammar of Class D Amplification". A more recent scholarly article is here, titled "The Class D Audio Power Amplifier: A Review" published by a group of Chinese researchers.

TI TPA3223 Chip

Texas Instruments (TI) and other semiconductor manufacturers have invested heavily in developing high-power Class D integrated circuits which have found their way into many Chinese DIY modules and finished low-cost amplifiers. These typically do not perform as well as the Hypex or Purifi modules but are far cheaper. TI sells an evaluation board and has extensive documentation on how to use its chips. This is one of their latest chips - the TPA3223. You can see how large it is and the required heatsink interface on top. It is rated at 170 WPC into 4 Ohms at 1% THD with a 45-volt supply.


Click the block diagram for a larger image. You can see the complexity involved and the level of integration. The chip includes fairly elaborate protection circuits and can also operate in bridged, mono mode, doubling the power.

Click to view a larger version

Here is an example of a low-cost (< $80 US) Chinese Class D amp based on a TI chip. It is optimistically advertised at 300 WPC into 4 Ohms with an upgraded power supply. The power supplies for these are typically large "brick" type external supplies similar to laptop SMPS supplies. These amps keep getting cheaper and incrementally better, but are they audiophile quality? In certain circumstances, they may be "good enough" for casual listening.

At these price points, the average audiophile might be tempted to try one out without much financial risk, or simply return it after one's curiosity is satisfied. On the other hand, YouTuber Andrew Robinson noted that several of these low-cost TI chip-based amps, while sounding OK in near-field (desktop) situations did not do so well in large rooms with good speakers, exhibiting some not-so-great sonics at volume, including spurious high-frequency noise and poor speaker control. His conclusion was that once you spent the extra money on an upgraded power supply (to get more power), you might as well invest in a decent A/B Class amp.

The Achilles heel of Class D is the output filter. All Class D amps require at least one choke and capacitor output filter to remove as much of the high-frequency switching component (typically 400 to 800 KHz) as possible, leaving (mostly) only the audio signal. One problem with magnetic core chokes is they exhibit hysteresis. This hysteresis is non-linear and introduces some distortion in the audio component of the signal. Attempts have been made to replace the magnetic cores (typically ferrite) with air core chokes, but these are very large and radiate lots of EMI (see more on EMI below). Toroidal chokes contain much of the field within their cores but don’t solve the hysteresis problem.


Attempts have also been made to reduce the size of the chokes needed by increasing switching frequencies closer to the 1 MHz level, but this is pushing the limits of conventional silicon switching semiconductors (usually MOSFETs). Several implementations have successfully used GaN (Gallium Nitride) transistors as the switching elements. Since these are fast, high-current devices (originally developed for the power control industry for inverters and motor controllers) the switching frequencies can be higher. One drawback however is the higher cost.


Feedback is important in reducing the negative effects of the output filter. This is usually global (that is AFTER the filter) and must be carefully designed to avoid instability. Negative feedback in Class D is really different than Class A or AB amps and is not seen as an audible drawback since it does not result in distortions like TIM. The best Class D amps use multiple feedback loops for control and stability.

Typical Class D EMI Noise spectrum - courstesy of Texas Instruments

Then there is the issue of EMI. If you are a ham radio operator or listen to AM radio, you probably don’t want a high-power Class D amp (or SMPS) around your gear. The harmonic content of both radiated and conducted EMI can extend to hundreds of MHz. While any commercial switching device needs to meet FCC regulations for EMI, the typical test for amps involves speaker leads around 12 inches long and twisted pair wires. Who does that? A few years ago Maxim (now Analog Devices) was producing a low-power Class D amp chip which they touted as low EMI due to a proprietary switching technique. The EMI was just under the FCC limits with 12-inch twisted pair speaker leads! Long speaker leads effectively become antennas for conducted and radiated EMI. EMI can also be an issue for low-level high-gain phono stages where the EMI can overload or be rectified by the input devices, resulting in noise. Most well-designed phono inputs will have some RF filtering (e.g. using ferrite beads), but this may not be adequate near a particularly noisy Class D amp.

We now live in a world where we are constantly surrounded by EMI - WiFi, cellular signals, blue tooth, wall wart power bricks and most digital devices in our homes (including TVs). You could argue that we don't need more of this, since Class D amplifiers add to the "buzz" all around us. On the other hand, audio listening is probably the least worst (sic) contributor, so just go with it.


A note about measuring Class D amplifiers related to this high frequency switching noise - something that some reviewers have overlooked - here is a quote from TI about measuring Class D amps:


"Most audio analyzers will not give correct readings of Class-D amplifiers’ performance due to their sensitivity to out of band noise present at the amplifier output. AES-17 + AUX-0025 pre-analyzer filters are recommended to use for Class-D amplifier measurements. In absence of such filters, a 30-kHz low-pass filter (10 Ω + 47 nF) can be used to reduce the out of band noise remaining on the amplifier outputs."


This supports the widely held belief that if it is outside the audible spectrum (usually defined as above 20 KHz), it won't do any harm, so just ignore it. I am not so sure, since there have been numerous studies to indicate that humans CAN be affected by or perceive ultrasonics. See this article for some information on the subject. Assuming you have a tweeter in your speaker system that goes above this limit, it may affect what you perceive in the reproduced sound. The flip side of this is James Boyk's famous paper There's Life Above 20 Kilohertz! A Survey of Musical Instrument Spectra to 102.4 KHz. A copy is readable here. Assuming it can be perceived, there is useful musical information above 20KHz, so why eliminate it with a filter or contaminate it with artificially introduced noise? [The same argument was made with the introduction of the CD Redbook standard, with brick wall filters at 22 KHz].


Another potential issue is caused by ultrasonic beat frequencies. When two (or more) waveforms of frequency f1 and f2 are combined in a non-linear system, the result is two new tones which are the sum and difference of those two frequencies, (f1 + f2) and (f1 - f2). The (f1 - f2) difference tone can "fold back" into the audible spectrum if the tones are less than 20KHz apart, causing audible distortion. This has been widely investigated in amplifiers with IM distortion measurements, but only with tones of 20KHz or less, not ultrasonics. Little research has been done on this subject with Class D amplifiers.


There are many other design considerations for Class D amps that can't possibly be covered here in an overview, but online searches can lead to 1000's of references if you want to delve deeper.


Since this website is about DIY, where does that leave us? It is extremely difficult for a DIYer to design and build a good Class D from scratch (for the same reason that DIY SMPS projects are few and far between). Multi-layer PCBs with careful layout and attention to high current paths are required as well as knowledge of high-frequency, high-power switching circuits, resonant circuits, and filters. Another issue is grounding - much of what we know from linear amp designs like star grounding and lifting for minimal ground loops goes out the window with HF designs. Also, an all-discrete component Class D will be much more complex than any of the others, although most of the components will be fairly low-cost items. This leaves the module approach. If you are on a budget you’ll need to look at TI chip-based boards or possibly a source like DIY Class D. They have DIY modules starting around 89 Euro according to their website. These appear to be based on Bruno's earlier designs. Another source of parts for constructing a Class D amp is Neurochrome. They make front-end boards and chassis to house Hypex or Purify modules along with a Hypex SMPS. I estimate you will spend at least $1,000 or more to get all that together.


Another supplier of modules is Icepower. Their modules appear to be aimed at a lower price point than Hypex or Purifi. For example, their 300A2 module outputs 300 Wpc and is listed on their webshop at a very reasonable $171. They also appear to have developed proprietary control chips. It will be interesting to see reviews of their technology.


Some recently available "high-end" DIY kits I have heard about:


Another consideration is the power supply. Most DIY attempts using modules end up using high-power SMPS supplies, although some high-end OEM manufacturers use linear supplies. You can’t really get something for nothing, so a beefy linear supply will end up adding bulk and weight to your amp. Of course, an SMPS will compound the EMI issue. I can find little information on the effects of power supply noise and ripple on Class D amp performance, but I would assume there are some measurable effects.


You’ll probably discover that building a high-end Class D amp with modules will cost as much or more as a Class A or Class AB amp build, but at least it will be smaller and cooler!