Fixing analog audio on the $2.58 HDMI-to-VGA adapter

Fixing analog audio on the $2.58 HDMI-to-VGA adapter

on

I recently purchased an ultra-cheap HDMI-to-VGA DAC to hook up my Nintendo Switch to my CRT monitor and external speakers. Unfortunately its audio out jack produced serious noise problems on my audio mixer, as well as my PC's mic jack. I spent several days diagnosing impedance issues, prototyping filtering for the delta-sigma audio DAC, and reworking PCB components to produce the best quality audio achievable with the hardware.

Prequel: Repairing my old HDMI DAC

Before buying the audio DAC, my initial plan for video was to plug my Switch dock into my ICY-Box HDMI-to-VGA DAC. While reworking my I2C bodge, I damaged several traces and lost components on the board, and had to reconstruct them by tracing the I2C lines in GIMP to learn their functionality. (The DAC's HDMI input is an I2C slave/peripheral, the SDA/SCL lines had high-impedance pull-up resistors to 5V which I omitted, and low-impedance resistors to the chip to pull the bus to ground.)

I took this as an opportunity to practice microsoldering with NH's recommended magnet wire and 3 second solder mask, before I used them for Switch modchip installation. The 36-gauge wire was helpful, as the enamel insulation was less prone to melting than Kynar, and the wire was thick enough to not snap easily while also thin enough to largely avoid ripping pads. The solder mask could cover exposed pads I didn't want to solder onto; I also used it to pin "finished" solder joints and wires in place while working nearby, though it proved to have poor adhesion to the PCB. Perhaps I needed to clean and roughen the surface more, or try YCS's 3 second mask instead of Relife?

• While tracing the PCB in my photos, I messed up several times by reading the PCB with the wrong orientation while flipping between the top and bottom. Next time I may use KiCad's reference photo functionality to model the PCB symbolically, rather than tracing traces with the GIMP pencil tool.

Hooking my Switch to my audio mixer

Surprisingly the hardest part of my setup was connecting my console to my external speakers. The ICY-Box DAC does not have audio out (as far as I can tell, the ITE IT6892FN chip does not output analog audio), while the Switch's headphone jack output annoying interference to my audio mixer and speakers. After hours of troubleshooting by unplugging and swapping out various cables and components, I eventually concluded that when the Switch is connected to a ground loop and not in sleep mode, its audio circuitry picks up 60hz hum and amplifies it into the output.

Asking online, I received a large number of suggestions, of varying levels of viability.

• Ground loop isolators use transformers to pass voltage offsets between circuit segments with separated grounds.

• I read that small ones tend to attenuate bass and other low-frequency signals, and large ones were expensive. Perhaps a cheap ground loop isolator is still worth considering?

• Converting my audio equipment to balanced differential audio can avoid ground loops; this is the approach taken by professional audio mixing consoles.

• However the equipment seemed expensive, and DI boxes often lower voltages to mic level (which seems counterproductive since it raisessusceptibility to noise).

• However the equipment seemed expensive, and DI boxes often lower voltages to mic level (which seems counterproductive since it

• Another option was HDMI audio extractors, but chaos (👋) told me every AliExpress HDMI-to-3.5mm or TOSLINK extractor it tried was poor quality, while audiophile products were expensive.

• In my research, most HDMI audio extractors are part of full bitstream receivers that also process the video signal. Perhaps it's just as expensive to onlyextract data islands and convert to analog or S/PDIF audio?

• In my research, most HDMI audio extractors are part of full bitstream receivers that also process the video signal. Perhaps it's just as expensive to

• Negative-feedback audio optoisolators have two receivers, and the LED is driven so both receivers' voltage matches the input voltage.

• I was dissuaded from using them by the 1BitSquared Discord server, but don't know why.

• I researched processing the ground and two audio signals as inputs to a differential amplifier. I abandoned the idea since I didn't know how to make it work with inputs outside the amp's supply rails, and it could have poor common-mode noise rejection.

The easiest workaround I found was to turn my Switch's volume jack to max (with the headphone volume limiter off) so game audio would be louder than ground loop hum, then turn down audio in my mixer. However I still had to plug and unplug the headphone jack every time I put my console in or out of the dock, and I worried about long-term damage to the console's audio amplifiers. I also tested an Apple USB-C audio dongle, but did not get audio out of it.

(I also experienced the same audio buzz with a "friend's" Switch 2 hooked up to my mixer, but at the time attributed it to my third-party HDMI dongle. As a sidenote, I will never buy a new Nintendo console after they sued Switch emulator developers, abusing their corporate legal budget to threaten independent developers with personal ruin over legally protected activity to pad their corporate coffers, and continue to issue GitHub takedowns against forks.)

The HDMI-to-VGA audio extractor

The /crt/ Discord server recommended an AliExpress VGA dongle with 3.5mm audio output, which cost $2.58 plus shipping. (The price fluctuates slightly over time, and is currently sitting at $2.71 excluding a temporary 4th of July sale.) After the dongle arrived a week later, I plugged the VGA output into my video switch, connected the HDMI input to my Switch's video output, connected the audio output to my audio mixer... and heard the worst crackling I've ever heard come out of my Switch.

Confused, I connected it to my computer to capture audio through my motherboard's mic jack. I discovered a strange effect: the noise was almost inaudible when I set the Realtek audio's Mic Boost to off, but was overwhelming the usual sound when I turned it to +30 dB (even when I turned the volume down so game audio was at the correct volume). At this point I suspected the audio dongle was either pumping out ultrasonic harmonics, or its amplifier was distorting when changing mic boost changed my mic jack's input impedance.

To test the impedance theory, I measured the resistance between my motherboard's audio input and ground with my oscilloscope. Annoyingly the input pins had a constant DC offset of over 2 volts, which went away when shorted to ground but reappeared moments after I removed the short. This would eliminate any chance of properly measuring impedance. Undeterred, I measured resistance both ways and found that they measured open circuit one way and 300 ohms the other. Importantly, this measurement was unchanged by switching my Mic Boost levels. At this point, I would have to check for signal problems using my oscilloscope.

Probing the audio with my oscilloscope

I plugged the audio output into a breakout cable, and probed the exposed wire leads with my oscilloscope. There I found a startling 500-600 mV of noise covering the audio, with harmonics up to 50 MHz.

So I had confirmed the problem with the audio was ultrasonic DAC switching noise in the MHz range being modulated down to audible distortion by nonlinear amplifiers.

At this point I opened up my dongle to see what chip was inside. A photo revealed the chip was a NX3303X; I could not find any datasheets, but one Chinese website had a block diagram and pinout for the chip.

Searching around for other users of this chip, I found c0pperdragon (of Lumacode fame)'s HDMI2SCART project. This PCB uses chips largely harvested from AliExpress DACs like mine, and also hooks their audio output directly to speakers (through a DC-blocking cap yes, but with no treble filtering).