Scrapheap stories: building my dual-mono ß22

Welcome to this first episode of scrapheap stories, or build log if you will, documenting how I am going from this junk box:



Into something like this:



EDIT: pictures of final assembly here.

As inspired by my Eufonikas:



Q: So what am I looking at?
A: End of 2019 I started the Bill Of Materials for my ß22 headphone amp, which I completed as pictured in plastic IKEA box early 2020. It is a fully functional 2-board, passive ground configuration powered by a single σ22 and controlled by a no-name stepped attenuator. I never got around to boxing it up. Not in a small part due to being impatient to listen to my handiworks, and knowing how much time & effort it took to previously nicely box up a Hypex UcD180HG HxR speaker amp. Following that I had some excursions into Singxer SA-1, MCTH & Eufonika H5 and H7 territories. It has been collecting dust in my garage for the two years since.

Q: Nice story so what changed?
A: My audiophile journey has led to a point where I am getting a much better understanding of what I like both musically and aesthetically. I am set to upgrade my BF2/64 to an Yggdrasil+ LIM, keep one of my Eufonikas amp, probably stacking it on top of the Yggdrasil. Great pairing for my high-Z Beyer T1.2 and Amiron Home. Then next to this OTL, what better differentiator then to have a DC-coupled, reference ß22 headamp with vanishingly lower output impedance, fully discrete topology, MOSFET output stage, and class A operation until your ears are done bleeding?

Q: Sounds like a Mjolnir 3.
A: I could not say as I have never heard a Mjolnir 3, but hey I get your point. It sure "sounds" like they are alike. I would not know but do know this: there is something golden about building your own.

Q: You say dual-mono but that's not in that IKEA box.
A: Correct. I am going to take what's in the IKEA box, add a second σ22 and toroid, add two ε12 boards for DC-protection, and add/remove some other parts here and there as well. Instead of spelling everything out in this topic start, I thought to spread it to the post when I get to it.

Q: I find your plans amazing, tell me more.
A: Thank you kind Sir/Madam. Be warned, this may be going to be something of a slow burn. With work, home improvement, kids, and other real life commitments, building time can at times be stretched thin. Unlike my previous impatience, I have two Eufonikas to listen to, so I am planning to take my time and get things right this time.

All kidding aside, thanks for reading, and let me know if you have any questions or remarks.

 

PART ONE: WOODWORKING

IANAC (I Am Not A Carpenter). I got some tools and even invested in a not-too-shabby manual miter saw to make some nice 45° cuts. Will use it for flooring my attic soon as well, so good to have in the shed. But yeah, IANAC. And when you want your headlamp to be a looker - it's a piece of furniture after all, visible to the eyes of the spouse - details matter. And I don't know about you but I find it exceedingly difficult, even with some not-too-shabby tools, to make consistent cuts, have all the angles sharp, no bends in the woodwork from the local shop, and so on, and so on.

Honestly I will be glad when I get over this stage and can say I've got a chassis to my satisfaction.

First, sawing the boards in 45° angles. I got a spruce board. It's cheap, unlike oak which has soared in price to ridiculous heights, and I also like its bright, natural color. Downside is that it's kind of soft of course, and can color yellow in sunlight, but OK.



Boxing it up. No idea how this tool is called in English - a framer?

Note that is was my second try of boards and boxes. I had stained my first batch in chene oak. I don't have a picture, and there's not debating about taste, but I didn't like it. I had expected it to be "whiter" than it turned out, but it was really on the grey side. Then I also found that you can't trust this "framer" to make nice 90° angles. You need that T-square to force it for the first few minutes.



Fitting the bottom plates on it. These are 33 x 28 cm covers from Modushop / Hifi2000. The top plates are alike, but have ventilation slits. I would have liked to have ventilation slits in the bottom as well, but having drafted the positioning of all the components, I could not find any that did not have slits interfering in the place where I am going to have to drill mounting holes.

I really like this contrast between the natural, bright color of the spruce and the black aluminium covers.



Finally, leaving them out to dry after two "wet-in-wet" layers of natural, transparent stain. It's hard to see on the picture that they're even stained, but they are, and have the slightest gloss on them.



Coming days I will see if I have to do another layer of stain. Next step in line is to screw on the top and bottom plates, and drill the holes for the connectors.

 

PART FOUR: POWER SUPPLY

Mods, I probably should have created this thread on the DIY subforum - at your option, would you move it?

Q: Wait, what happened to parts two and three?
A: Just two additional layers of stain. Boring.

--

Thursday was my birthday and best gift I give myself every year is a day off from work. And what better way to spend it, than in the garage tinkering on my amp? Made some good progress.

Woodwork came out OK. I wasn't sure whether I would do the chassis in portrait or landscape layout. The choice to go to landscape was made for me, because the shorter sides were slightly tilted to the sides. Picture below is still in portrait layout and you can't really make out the tilting, but it's there and enough that you would see when put up in the living room.



Here it is in landscape, although the perspective makes it hard to see. But yeah it's like 5 cm wider than it is deep.

Didn't take any pictures in-between, so here you see most of the mechanical work done:

• Power inlet is a Schaffner FN9244 filter, 1A for highest attenuation while being plenty for the actual load (each toroid has a 160 mA slow-blow fuse). Principally it's not really necessary, as the σ22 regulators will attenuate it anyway. But hey, this is DIY so we go all-out, regardless of whether you buy into John Swenson's treatise on ground-plane noise or not (personally, I'm no believer in audible benefits of decrapifiers beyond the galvanic isolation that they provide, but I digress.)

• Funky umbilical cords! I was doubting the kind of connector I should use between the two chassis. Design goal being to find something that one doesn't mistake for a line signal, only to jolt 60V into the sorry gear at the other end. Then I thought: hey, why not dump panel connectors altogether and go for some mating of cables that are fixed to the chassis? There's a better view at the end.

• About those umbilical cords, notice they carry four wires: three for power (V+, V- and ground) plus a separate ground return for the headphone channels. Better not to daisy chain those ground returns through the amp boards in the other chassis, but to give it the lowest impedance return possible free of any voltage differences.

• Plastic standoffs. Long time ago I read that this introduces every so slightly less noise on PCBs, providing separation from the bottom plate that's connected electrically - in this case to protective earth through that sanded mounting hole near the power filter. Remember, anodised aluminium insulates otherwise!

• Two more sanded mounting holes near the toroids are there to connect the transformer electrostatic screens. Cleaner and lower impedance this way, than through a cable to a common earthing point.

• On/off toggle switch at the very bottom of the picture. Better to see in the picture at the end.

• Toroids tucked away as far as possible from each other, to prevent any crosstalk through EM-coupling.

On to the σ22 boards. I had two functioning boards lying around: one ±30V from my β22 before, one ±5V from my Soekris dac1121 before, both v2.0. So one I had to change from ±5V to ±30V.

Then for both, I decided to upgrade them to v3.0. That version adds reverse-current protection by adding four diodes per board. Not strictly necessary on well-functioning β22 boards, but with a detachable load who knows what might happen in the future.

So reading the v3.0 schematics and following the PCB, I came up with this tack-soldering solution. For each board, two pairs of 1N4148 diodes hot-glued to the back:



Note that if you ever venture on this path, be mindful that the direction of the diodes is reversed between the positive and negative voltage rails!

Intermediate result, most components mounted:



Toroids are very nice audio-grade 2x30V 30VA ones by Toroidy from Poland. The β22 boards draw around 200 mA when biased at 160 mA in Class A, plus 20 mA for the ε12 DC-protection boards. So 30VA gives a good amount of headroom without being too large causing voltage to go up, only to be dumped and turned into heat by the regulators.

Before I had been using toroids from the Dutch Amplimo. Also very nice, but a lot more expensive especially when you spec an audio-grade one with an electrostatic shield. This shield should again attenuate VHF content - see rationale (or lack thereof) why you would need that, same as with the power filter, above.

Above you can just make out that I wrapped copper tape around each toroid - single turn around it, not through it! This "belly band" may somewhat decrease any flux leakage. Toroids should have little, but it's not none. It's a topic of debate how much this belly band helps, and whether it should be iron not copper, and thicker. Probably/maybe all true, and like all the rest of the filtering, if it doesn't help it will not hurt either.



Ahh, wiring coming together!

As you can see, I like using faston terminals for quick and clean installation. Other cables I crimp with ferrules for cleanliness, ensuring tight fit, and decent insulation between adjacent wires (don't want a loose strand wreaking havoc).

All wires are twisted for EMC - except the headphone ground return of course which we want to limit taking crosstalk from the power wires.

Bottom-center you see two fuse terminals. Better protection than to have a single one guarding two toroids - because who knows, worst case one could be providing reverse over-current and the other "normal" over-current and net the fuse would not trip!

Bottom-left I've temporarily terminated the two secondaries to see if I can find a "common" ground or midpoint - of course not "commoned" in the galvanic sense, but to make sure they are wired in the same direction. The σ22 transformer inputs take one "ground" and one "line" per secondary and I could find no documentation on how these toroids were wired. Measuring AC across them like this, I indeed could find the pairs in which they had the lowest potential difference.



All done! Notice how the AC and DC wiring is completely separated for lowest noise. I’m really happy with how it turned out.

 

PART FIVE: NEW BOARDS

Just a small update.

After USPS in all its wisdom shipped my order for PCBs and some components from AMB from Sunny Vale to San Francisco to Bell Gardens, back to San Francisco, then to Richmond, to City of Industry, back to Bell Gardens, back to City of Industry, AGAIN BACK TO BELL GARDENS, to Los Angeles, to CANADA, to Jamaica, NY, it FINALLY arrived in The Netherlands where it was held at customs.

So yeah after over four weeks after ordering I finally got everything in to solder the amp parts.



On the left a genuine 50K ALPS Blue Velvet potentiometer, soldered to a simple AMB ε27 PCB for neatness and ease of wiring. This PCB has a common ground between the left and right channels, which I hesitated about for this dual-mono build. Reading up some it became clear that bonding the two signal grounds is a standard and recommended away to short ground loops with the source.

Middle and right there are two AMB ε12 DC offset protection circuits. My Yggdrasil is DC-coupled and so is my AMB β22, so let's make sure we don't fry any nice cans.

It doesn't really show on the photo but man, these are small PCBs with quite a lot of components on them! Easy enough to solder but took me some evenings with the kids tucked to bed. For the relays I went for the slightly more expensive ones with bifurcated crossbars.

One ε12 can protect two channels, but only when the channels are run from the same power supply. Not the case with dual mono so I had to build two. You'll see that when I wire it all up, I won't need all of the terminal positions but I didn't think it was worth to over-optimise for just a couple of resistors and pins.



Working on the layout. I think it's nice to have the volume knob and XLR-4 output equidistant from the sides. I may want to switch them around - I'm right-handed so could be better not to have the headphone cable in the way - but maybe I'm overthinking. Shaft is there to keep signal wires as short as possible.

DC protection PCB on the right I rotated 90 degrees because the power connection is to the right of the terminal. So this way I can wire the power cables without them ever crossing the signal cables.

Bottom terminal strip is where I'll solder a Zobel network. I know my β22 needs it for at least my Amiron Home cans, or otherwise it'll get excited by the reactivity with a high-pitched whine. The Zobel doesn't hurt the amplifier performance, so a no-brainer for maximum compatibility with cans.

Finally, I didn't take a picture, but I changed R34 and R35 on my β22 boards from 1Ω to 0.47Ω. Originally I had used 1Ω because it provides some protection against short-circuits when TRS plugs are inserted or ejected (β22's are known to fry otherwise). Downside is that it reduces voltage swing nicely. Not necessary anymore with XLR4 output so back to full voltage swing this way. Let's hope no-one ever puts a TRS adapter on it.

Going on a business trip for a few days, so a few days of consideration before I start drilling holes.

 

PART SIX: ASSEMBLY AND OH NOES SMOKE !!

Came up with the layout. Put the volume control on the right. And decided not to rotate the right-hand ε12 DC mute protection, because I figured I will simply fly the power wires over it.



Found a pretty neat solution by aligning one mounting hole of the ε27 board, the one with the volume pot, with the top-left mounting hole of the β22. This way the shaft is nicely in the middle of the chassis and the volume control is tucked away to the right.

RCA inputs I didn't drill yet in the photo above, but I've put them slightly off-centre, right in between the empty space between the left amplifier board and the voltage pot.

Front view:



In the photo above, the top chassis almost seems to float above the bottom chassis! In reality, it's standing on four black shock absorbers that lift it a good bit up. This allows proper airflow from the power supply below, and won't hurt EMI either.

Assembling it but oh noes magic smoke on the DC mute protection board!



I already replaced the capacitor with the light blue one, but the voltage regulation section seems defective (here is me asking about it on the AMB forums). Then during troubleshooting I also blew up that Q2 resistor.

Have to wait for the replacement components I ordered, and will tread carefully in the meantime.

 

PART SEVEN: PUTING IT ALL TOGETHER

Just look at this beauty ❤️



Not bad eh, compared to the junk box I started with?

 

PART EIGHT: BEHIND THE SCENES

I hit "Post Reply" too soon on the previous, so let me fix that by splitting it up into this second post.

Starting with the blown capacitor on the ε12 DC offset protection circuit, I suspect that it was a defective Zener diode that exposed the capacitor to the full 60V when it was only rated for 25V. Then during troubleshooting I messed up with my probe and shorted one of the transistors. To make a long story short, I replaced the entire voltage regulation section and that fixed it.

On the home straight I did not take a lot of pictures. Here is "v1" of the internals:



The speed drill chipped the wood, but fortunately on the inside only.

The input grounds should be bonded, which they are at the ε27 volume pot mounting board.

All pairs of wires twisted, except for the orange signal outputs - the ground return to the power supply runs parallel to the power supply lines.

AMB recommends using a Zobel network to tame reactivity in the output load, which can sometimes make the amplifier oscillate and cause a high-pitched whine. I know it did that on my Beyerdynamic Amiron Home before - weirdly, on one channel only - and the Zobel doesn't degrade the amplifier performance, so that's a recommendation I followed.

Although the wires joining at the terminal strip aren't as neatly tucked away as in other places, I like how it turned out. Here is a detail of it:



First power-up of the complete assembly was a success: no magic smoke (always an exciting moment) and no hum!

Until... I found this weirdest issue that pretty much sounds like a ground loop but only with an official Beyerdynamic XLR-4 cable and only with the volume knob turned all the way down. Swap in a different, aftermarket cable... hum gone. Quite elusive. If you have any idea, I would welcome your assistance in troubleshooting on the AMB forum.

Second finding: biasing the quiescent current to 160 mA makes it run hot. Like, REALLY F-ING HOT. To the point that it's barely comfortable to touch the top cover.

AMB describes 120-160 mA as "deep into Class A territory"... well deep into overkill territory, that's for sure! I mean, my 32Ω Beyerdynamic T5.2 only requires 4.4 mA to get to 100 dB SPL so we can easily tune that down. Not too much though, because the idea is that MOSFETs are supposed to sound nicer when they run hot. Still, dialling down from 160 to 120 mA is 25% off. That has got to help with the thermals.

So after a couple of revisions trying to fix that, this is what it is now:



The acute viewer will spot that I bonded the bottom cover to the signal ground, with the lug on the bottom left connected to the signal ground on the left amplifier board. Remember, the signal grounds are bonded at the volume pot, and indeed the bottom cover now measures as short to the grounds of either RCA connector.

Then, the even more acute viewer will see that I replaced the plastic screws on the volume shaft mounting tab with metal ones, and sanded the cover under it. In effect bonding the volume shaft to signal ground as well.

While this bonding should be best practice, it did not fix that elusive hum issue with the official Beyerdynamic cable yet. Of course, I tried the cable on other amps (none of them dual-mono though) where it works just fine.

So how does it sound?

Too early to tell in full but good! The amp clearly has a lot authority. Good slam, clarity, and micro detail. This is paired to my Yggdrasil+ MIB that I recently got, so all of it requires more brain burn-in. I'll let you know!

In the meantime, about that elusive hum issue...

 

Nice!

For grounding, best practice would be to tie the chassis and the volume pot shaft to safety ground. Then tie signal ground (and power supply ground) to safety ground at a single point through a ground loop breaker.

 

The PSU is in the bottom chassis, where the bottom plate is tied to safety ground.
In the upper chassis, there is only DC and signal ground.
So, no need for a ground loop breaker, right?

It's like option 1 in the "Wiring & ground" instructions, but dual-mono.

 

I'd still be inclined to have a distinct safety ground for the separate amp chassis. It prevents any noise absorbed by the chassis from dumping straight into the signal ground.

 

very nicely done - planning and taking your time always pays off. Enjoy this iconic amp which will likely remain timeless ..dB

 

excellent build!

 

Not too difficult to hack up and see if it works. I'll let you know if it does.

 

Alas, the ground loop breaker made no difference. It remains super-weird that this is with the official Beyerdynamic cable only, and only with the volume knob turned way to almost mute.

 

have you tried disconnecting the XLR housing from ground ( or the reverse - connecting it to ground )

 

Not yet - I can try tomorrow. I did measure the XLR connector on the cable, and it's unconnected. Both on the cable that hums when mute, and the cable that is hum-free. So it would somewhat perplex me - but hey, clutching at straws.

 

I would really love to build the Beta 22. I'm very afraid that it will sound too much like the MJ3. I finished a M3 last month and it and the MJ3 sound very different which is nice.

This build is fantastic by the way. For whatever reason I kept reading "dual mono" and thinking "balanced" and then being completely confused by the pictures. Modushop is getting expensive for my DiY builds so I may adopt your approach to my next project, whatever that ends up being. A friend of mine is into Radu Tarta stuff and your Beta is reminiscent.

 

Never heard a Mjolnir 3 but I would not be surprised if it sounded similar. After all, the topologies are more alike than they differ. On the other hand, I've owned a Bifrost 2/64 with DAC8812 chips and now an Yggdrasil+ LIM with DAC8812 chips and they are certainly distinct.

Thanks a lot. Honestly I went for wood because of aesthetics, not money (just looked up Radu Tarta; yeah my inspiration was Eufonika and it's alike). The aluminium covers I ordered with Modushop and buying a full chassis would only have been marginally more expensive. Yes, milling the back/front panels add cost but so does wood and paint.

Yeah, dual-mono single-ended. I'm in the camp that balanced is perfect for interconnects and unnecessary for headphones, undesired even. At the cost of double the impedance (OK, in the case of ß22 2x near-zero ohm is still near-zero ohm) you get double the voltage swing. A single-ended ß22 pumps 38Vpp (roughly 13.4Vrms) in a 32Ω load at 1 kHz before onset of clipping. That's 5.6 Wrms... well above the threshold of pain for any pair of cans.

 

Oh wow. That did improve it - in unexpected ways. Now when I twist & coil this specific cable, I can find positions where it is hum-free where before I could not.

This official Beyer cable seems to act as an antenna. The fact that its wires don't seem to be twisted isn't helping. Maybe it's picking up EMI from the transformers. Can't exactly say, because moving the cable around or away from the toroids doesn't really seem to matter.

 

To add to the murk, sometimes connecting the housing to ground with a capacitor also works. Or not at all.