Fixing Drumboy’s headphone output for line out

Drumboy only works with headphones

As a backer of the Drumboy & Synthgirl Kickstarter, I received my Drumboy unit last year. This rhythm machine has been renamed to Drumboy Mini recently after its manufacturer Randomwaves launched a new Kickstarter for Drumboy Pro.

As an indie product, Drumboy is definitely an exciting instrument with a refreshing creative vision on musical user interfaces. At the same time, there were quite a couple of frustrations regarding the unit, as indicated by a high volume of comments on the original Kickstarter page.

One issue was insurmountable for me: Drumboy’s output only works with headphones and is even picky at which headphones it works with. When connected to a mixer, Drumboy does not output any signal. Which means that combining this instrument with other instruments (notably Synthgirl) is impossible. That, at least to me, defeats the entire point of a rhythm machine.

After an email exchange with Randomwaves, we confirmed that this was a hardware issue. So although the project is open source, a software patch could not fix this.

Headphone ground differs from actual ground

Why Drumboy works with headphones

Because Drumboy’s hardware design is open source as well, pinpointing the issue was straightforward. The SGTL5000 chip that converts digital audio into an amplified analogue signal can be incorporated in different ways into schematics. The chip provides three pins for headphone output:

HP_VGND on pin 4

the virtual ground reference a connected pair of headphones

HP_L on pin 6

the signal for the left audio channel, relative to HP_VGND

HP_R on pin 2

the signal for the right audio channel, relative to HP_VGND

Importantly, HP_VGND is not the same as regular ground GND. While an audio signal fluctuates positively and negatively around a center voltage (usually considered ground), this chip is powered by actual ground (0V) and a positive voltage (let’s say 3.3V). Rather than centering the signal on 0V, it’s easier for the chip to amplify signals such as HP_L and HP_R around a virtual “middle” ground of 1.65V. Besides, headphones don’t care: they are not connected to anything else but Drumboy, so from their point of reference, HP_VGND might as well be 0V. What matters to play audio left and right aren’t the absolute values of HP_L or HP_R, but their relative difference from HP_VGND.

Why Drumboy fails with mixers and other line equipment

When connecting a device such as a mixer to the headphones jack as if it is a line out, this difference suddenly becomes an obstacle. As a result, no sound is emitted besides an initial plop. After this, Drumboy refuses to output any audio even when reconnecting headphones, until you power cycle the device.

This happens because the receiving device has the sleeve of its connector connected to actual ground GND, which—depending on how the devices are powered—likely differs from HP_VGND. So in connecting Drumboy to a mixer, we’ve shorted HP_VGND to GND, which the SGTL5000 fortunately detects and reacts to by switching itself off rather than burning out. After rebooting and connecting Drumboy to a device with a floating ground, such as headphones, the SGTL5000 reports for duty again.

In theory, when powering Drumboy with batteries rather than from the same power circuit as your mixer, Drumboy’s GND is allowed to float away from mains ground. In that case, HP_VGND could still be shorted to mains ground without causing an issue. However, I found that the SGTL5000 in practice proved too sensitive to handle such scenarios well. After trying many different setups, I wasn’t able to get Drumboy working with anything but specific headphone pairs.

Adjusting the schematic

To make the Drumboy headphone connector compatible with line out, we need to remove the 1.65V difference. To eliminate an unwanted DC voltage offset, we can insert a capacitor between the source and the destination, since capacitors do not let DC current through. The Drumboy designers presumably left this out because the required components’ size would challenge its thin form factor. The schematic indeed reveals our suspicion that the capless headphone design from the SGTL5000’s data sheet was used rather than cap-coupled headphone design: the connector uses HP_VGND rather than GND for its SLEEVE.

In order to patch this schematic, we need to change both 0Ω resistors into capacitors of 220μF, the value suggested in the data sheet. For impedance levels of line out and typical headphones, a 220μF capacitance results in a cutoff frequency below the human hearing threshold of 20Hz, and thus does not noticeably distort the audio signal.

This schematic has been implemented at the top left of the PCB, where the two resistors R20 and R21 are SMD components. The usage of these 0Ω resistors rather than a direct trace shows that the designers may have wanted to leave this option open. In principle, I’d need to remove R20 and R21 and replace them with 220μF capacitors. Since those are produced as polarized electrolytic capacitors, the negative lead needs to point toward the audio socket, so the positive side will be 1.65V higher.

Furthermore, I’ll need to steal a GND signal from somewhere; I’ve identified pin 5 of the optocoupler (the white chip sticking out of Drumboy’s bottom) as a viable candidate. In addition to lowering the signal voltage, we’ll need to reconnect the sleeve of the audio connector from HP_VGND to GND as per the data sheet.

Hacking the circuit board

Planning the fix

My initial plan was to:

1. Remove resistors R20 and R21.
2. Attach wires to the right pads of R20 and R21 to capture HP_L and HP_R.
3. Cut the existing PCB traces toward TIP (HP_L), RING (HP_R), and SLEEVE (HP_VGND).
4. Attach a wire to pin 5 (GND) of the H11L1 optocoupler.
5. Attach wires to TIP, RING, SLEEVE.
6. Test the capacitor setup on an external breadboard.
7. Incorporate the fix into Drumboy.

Disassembling Drumboy

After removing the screws, I was wrestling the cable that connects the main PCB to the cover with the tactile buttons. Push the tiny black handles away from the connector (I recommend on the cover side) to make the cable detach.

The underside reveals Drumboy’s main board with the components we recognize from the PCB diagram. Note in particular our targets R20 and R21.

Removing the resistors

I first tried to melt their contacts with a hot air gun, covering their neighbors with Kapton tape. The heat at which my tape began to melt wasn’t enough to move them. Fortunately, they turned out to be quite easy to remove with a soldering iron instead.

Attaching the signal wires

Since SMD pads are tiny, I decided to drill holes into the left-side pads, such that the wire could go through them and reach the right-side pads. This adds mechanical strength to the connection by ensuring the wire does not pull on the pad. The consequence is that we’ll have to attach the negative leads of our capacitors elsewhere, but those left pads connect to the jack socket, which has a bigger surface anyway. So by drilling through them, we only sacrifice small pad access to TIP and RING, which wasn’t the most convenient option anyway. We must remember to cut the traces of those pads to the audio socket, just in case the wire still touches them.

After drilling, I pulled yellow (HP_L) and red (HP_R) wires through the holes.

Finally, I soldered them to the right pads of R20 for HP_L and R21 for HP_R. (Arguably not my best soldering work.)

Then I cut the traces of the original signals to the socket at the top left. (Arguably not my best cutting work either.)

I tested with a multimeter that the signals from the chip reached the new wires, and that the audio socket was disconnected from them and the rest of the board.

Grabbing the ground and the socket

Next, I grabbed the ground from the optocoupler by soldering a black wire onto pin 5. (Arguably slightly better soldering work.)

Finally, I also attached a separate 3-wire cable to the audio socket, providing TIP (left channel, yellow), RING (right channel, red), and SLEEVE (ground, black).

Testing the sound

And now for the make-or-break moment: let’s see if it actually works. I reassembled my Drumboy unit with the wires sticking out, and brought the red and yellow signal wires to a breadboard. I then inserted the positive lead of two 220μF capacitors into those same columns, their negative leads going to fresh columns. To those negative leads, I then connected the yellow and red wires of the connector, and its ground to the negative power rail for good measure.

I connected the Drumboy audio socket to my headphones, and it still worked. Then I connected the same socket to my mixer, and after a power cycle (the SGTL5000 is sensitive to the temporary short that occurs when inserting a jack connector), it worked too!

Success. For the first time ever, Drumboy was playing through my mixer, and I could jam on top of its rhythms with my other synths.

Installing a permanent patch

Soldering the protoboard

I needed to find a way to install the capacitors into my Drumboy without sacrificing its small form factor. After some trial and error, I decided on a separate protoboard such that capacitors could stick out of the side of the narrower middle board, while still being covered by the wider top and bottom boards.

The protoboard attaches to one of the four screws that secure that middle board into place, replacing a spacer bolt. I added two 1kΩ resistors wired in parallel from the capacitor’s negative leads to ground, gently pulling down the signal in the capacitors to avoid popping sounds. They’re optional but recommended; although in the final version, I’ve moved them to the side because the space between the middle and bottom boards had become really tight with the extra protoboard.

The signal wires go through the front to the protoboard on the side.

Reassembling the unit

Finally, I reattached the bottom board and confirmed that the capacitors and resistors weren’t sticking out too far. By moving the resistors to the side, I was able to also screw on the fourth bolt in the top right position, holding the protoboard firm in between the middle and bottom boards.

Like the original, it looks like a tight sandwich, yet the capacitors fit within the original height. Alternatively, I could’ve cut holes in the bottom like the designers did for the optocoupler, but that turned out to be unnecessary.

The end result

The end result is a fully working Drumboy unit—whether you plug it into headphones or any other audio equipment!

What started as a risky endeavor, has left me with a working Drumboy.

Verdict

All in all, we have to applaud independent makers attempting new devices and musical instruments. They add an exciting dimension to creativity that established companies cannot always provide, because risk/reward balances and responsibilities are different. When buying from Kickstarter, small defects such as this one can be very annoying, but understandable. In the case of Drumboy, Randomwave committed to an original product with open-source software and hardware, and they’ve always reacted in a friendly and helpful manner to inquiries.

To me, a non-functional audio socket is—understandably—a deal-breaker for a musical device. Only working through headphones is a no-go, especially for a device meant to work together with others. This shortcoming was likely considered during the design, given the 0Ω placeholder resistors, but ultimately not implemented. It was important enough for me to risk wrecking my Drumboy entirely, because it was useless to me without a line out. Another solution would’ve been to use the line out pins from the chip; it’s still on my To Do list, although the SMD pins are so tiny that I should probably improve my soldering skills first.

With the release of the Drumboy Pro, some previous backers feel disgruntled that many of their fixes will only be addressed in a new version. I understand both parties: it’s frustrating to own a device that doesn’t work, but Randomwaves also needs space to learn. My major suggestion would be more hackability. When I opened up Drumboy, I found this message on the inside of the bottom board:

Reading this means you’re part of the hacker community

And I like the idea, but here’s what’s missing:

• The source code is openly available, which makes it open source in words, but not necessarily in spirit. The repository is a code dump with 2 commits; it didn’t capture a history and fails to provide a future. It satisfies the hacker bit, but not the community notion. I suggest placing the actual repository online, including revisions. Seeing and understanding how the code evolves makes contributing easier.

• The hardware design lacks pluggable testing and extension points. Although the 0Ω resistors turned out to be a godsend, it wasn’t exactly straightforward to grab the signal from those tiny pads. Something as simple as finding a ground signal became quite an endeavor. There was also little room to add or change things, even though the PCB itself has ample space.

Let’s hope Randomwaves integrates these suggestions into their new projects.

As for Drumboy Pro, I won’t be backing its Kickstarter this time, although it just passed its minimum goal as I’m writing this. Not for lack of confidence in Randomwaves, but because I don’t have a use case for it. The original Drumboy is a tiny sidekick to quickly start a beat; for more serious work, I switch to a full sequencer that can also drive my synths. The Pro edition has none of the coziness of the Drumboy but all of the distractions of a more powerful device; the Pro isn’t something I’d grab quickly to dial in a rhythmic pattern.

But, thanks to the capacitor fix, I will grab the Drumboy next time!