The β22 from AMB is one of the most highly regarded DIY headphone amplifiers you can meet around the web.It gets plenty of excellent reviews from plenty of people who have built it. Over the years I’ve had the opportunity to listen to quite a few headphone amplifiers, including DIY builds, and I ran across a β22 more than once. I’ve even had an opportunity to repair one for a friend after it got damaged due to an accidental short on the output. The β22 always sounded good to me, although I must admit that its one of these amplifier that didn’t give me that “wow” factor on our first encounter. In my book that can actually be a very good thing, as many of the amplifiers (and any other stereo component) that give a “wow” feeling at first, prove to be too fatiguing and unrealistic sounding in the long run. The β22 is one of these amplifiers that you appreciate more as you spend more time with it.
I’ve been thinking of building a β22 for a fairly long time, with the cost being one of the factors against it. Just like with any other DIY project, and I’ve seen quite a few, the builder has significant wiggle-room regarding quality and cost, as well as functionality. However, I wanted to build one that could serve multiple functions, perform well, and look well. I wanted something I could be proud of building and owning, and to be happy with it for years to come. Eventually, I’ve decided to pull the trigger on this build. In this post I’ll share the steps and some of the technical considerations that came into play during this build.
Unlike most headphone amplifiers, the β22 I was planning on building was intended to drive both headphones and loudspeakers. To get a sense as to how much power it should really have, I’ve measured the power output of my stereo amplifier (an overhauled vintage Kenwood KA7100 I might do a post about in the future), and my home-theater receiver. The output of both measured <5W/channel at what I would define as loud listening levels. This is despite using completely different speakers in significantly different size rooms. Therefore, I’ve decided I should have ~50W/channel output capacity to be able to handle high dynamic portions of the signal. Luckily, this is something you could pull out of a β22 amplifier if you go for a balanced build (4 boards, 2 per channel). So the overall list of demands was:
- Single-ended and balanced input connection
- Headphones (SE and BAL) and speaker driving
- Output protection of the speakers/headphones
- >50W/channel at 8ohms
For the PS I’ve opted for the σ22 from AMB as well. This was the latest version of the σ22, which had an improved ability to handle heavy loads (like a speaker amplifier). As I’ve shared on the AMB forum here, the basic capacitance multiplier is limited in its ability to handle large voltage droops over the bulk capacitors, which increases output noise and ripple under heavy loads. To reduce this sensitivity, a few components should be added. This was implemented in the version of the σ22 I was using for this build.
I wanted to have a few additional functions implemented, which led to a fairly long list of “parts” that needed to be built/designed/sourced to fill the list:
- 4xβ22 boards for a balanced stereo build
- 2xσ22 boards, one per channel
- 500VA toroidal transformer with 2 secondaries, each is center-tapped. One secondary per channel
- 2xα24 from AMB for a SE->BAL conversion for the RCA inputs
- 2xLT3042 based dual regulators from DIYinHK to power the α24’s
- ε24 switch driver circuit for soft start option
- Mill a board for input switching using relays
- Output switching and protection achieved by the DC protection circuit I’ve designed for this project
- A generic aluminium case was purchased which could be modified according to the needs of the project
There are more than a couple of boards in the list above that are designed by Ti from AMB. I’ll take the opportunity to thank him again, and say that he has very interesting stereo DIY projects, and he makes sure they are well documented to make them as accessible as possible to other DIY’ers.
As I’ve stated earlier, I’ve had repaired a β22 in the past. What I didn’t mention is how and to what degree it was damaged. This happened as a TRS (6.35mm) plug was inserted into the amplifier while it wasn’t muted. This has temporarily shorted the amplifier output, and caused significant damage. It damaged a β22 board, a σ22 board, and fried the power transformer. While this is something that shouldn’t have been so extreme if a proper fuse rating was used, this is still too substantial of a damage to just ignore this option. Therefore, I’ve decided to slightly modified the σ22/β22 boards I’m using by including a simple current limit. To make this perfectly clear, it isn’t meant to prevent thermal damage in case of excess current for prolonged duration. It is only there to prevent damage due to a temporary short in such an event.
Luckily this can be implemented fairly simple by adding a few resistors and a transistor for each high-current path as follows:
The circuit is straight forward and is easy to follow. The exact values of the resistors can be modified as desired to trim the value of the limit, and the values above are what I’ve used. In practice the devices will start conducting at <0.7V, especially if they get a little warm as the amplifier is operating. Therefore ~20% headroom should be left. The additional capacitance of the added BJT’s is sufficiently small to have negligible affect on performance.
The first thing was to assemble each board on its own and test the functionality to make sure there are no issues there. Most of these boards are known designs, so there is little to share about them. The only exception being the DC protection circuit, which doubles as the output selection board. However, it was already described in detail in its own post. Despite this I would like to linger on one point regarding this circuit in the context of this project.
The amplifier I was building is a balanced amplifier. However, it does support a SE output for headphones. Therefore, in the event of a failure that will cause the common-mode (CM) at the differential output to differ significantly from 0V (perhaps a blown rail, or a poorly connected power rail), the output protection circuit would be ineffective in protecting a SE headphone. This is fairly simple to understand since the protection circuit is observing the differential output voltage on the balanced output, as it was meant to do. However, the SE load is sensitive to the voltage difference between the positive output and ground. Therefore, to solve this issue I’ve decided to re-purpose the second detection channel on each of the protection boards (one per channel). The logic behind this is that one detection path is responsible for the differential output, while the other is responsible for the SE output. If either one of them fails, the output(s) are disconnected. To clarify, this was very simple to do simply because I went for a dual-mono build. While in a standard build a single protection board is used to monitor both L and R channels, in this build there is a separate board for each of the two channels. This meant that there is an unused channel on each of the boards, and it was used for this dual detection mode.
The enclosure I’ve purchased has heatsinks as the sidewalls, which meant all power devices are meant to be mounted to them. Since I wanted all trimmers to be accessible easily after assembly, I had to mount everything in proper orientation. For the σ22 this is straight forward, while for the β22 this called for some additional aluminium pieces.
Next step was to mill simple single layer boards to hold the stepped balanced volume attenuator, 2xboards to hold the input switching relays, and 3xboards to hold the low power transformers. There are 3 low power transformers overall. The first used to power the soft power switch driver board, the other 2 are used to power the LT3042 regulators boards that power the SE->BAL converters.
After this was completed, everything was “thrown” in place for another round of measurements before I could move on with the case work. After these measurements, the rear panel could be milled on the CNC machine. The rear panel holds the power input module (with switch/fuse/line filter), the input connectors, and the speaker output terminal. Additionally, it holds another aluminium plate that the α24 boards and the input relay boards are mounted to. This can be seen in the images below:
Next it was time to move to the front panel. It was at this point that Murphy jumped for a short visit. The Dremel I was using on my CNC machine as the spindle blew up while engraving the front panel after years of working flawlessly. This was a major issue as I’ve had no replacement for it. It was an older model that was no longer being sold, and my “spindle” mounting was sized according to it. As a temporary solution, I was able find a local forum member who had a similar tool and was willing to lend it to me. This of course brought work to a halt for a few days until I could go and pick-up the replacement. I was able to get the panel done eventually, although it didn’t turn out exactly as I’ve hoped it would.
Finally it was time to start assembly. I will skip (just like I did so far) the boring parts of sawing/drilling/tapping/sanding/painting things, and move on to everything being put together. If you have any questions about a step I’ve missed you are always welcome to contact me and I’ll be happy to help.
As you’d expect I had to measure it again after final assembly to make sure everything was connected. I’ve used this opportunity to check the output power before the onset of clipping with both channels loaded by a 8ohm resistors. The amplifier hit just over 60W per channel before clipping can be observed.
I have initially opted for a Zobel network for improved stability only on the speakers outputs. However, I’ve quickly came across oscillations when the AKG K1000 headphones were connected to the front output connector. Therefore I’ve revised my previous decision and added one to the headphones output as well.
This amplifier build was completed well over a year ago, and its still being used in my workbench. I use it both with headphones and driving my Proac bookshelf speakers, and it does a great job with both. The β22 really is a wonderful amplifier, and deserve all of the excellent reviews it receives. Its sound is very natural and well balanced. It is detailed yet nothing jumps out at you. It is probably one of the most transparent amplifiers I have ever listened to, and it can drive any headphone effortlessly.
Finally, I had to find some place I could place this huge amplifier. The lack of room on my work-bench was all the motivation I needed to build a “stereo-stand” to hold the amplifier, DAC (Gustard X20Pro), and the headphones in use at each point in time.