Hi, I would like to share my recent work: a cheap USB Audio Class 1.0 (UAC) input to UAC 2.0 output converter.

 

For folks that may not understand why I have this crazy idea, here is the background:

For some shitty reasons, PS5 does not support outputting audio through certain USB sound card, and that is because PS5 only supports very old USB sound card (UAC 1.0), and usually high end speakers or sound cards would use newer, better UAC 2.0 protocol.

So, this shitty thing happens to me, when I thought my EDIFIER S880 speaker would be a plus for my gaming experience. EDIFIER S880 has a UAC 2.0 in input source, so my PS5 does not support that. So, what I can do is to use another 3.5 jack to connect the speaker to my PS5.

Unfortunately, there is a staticky buzz sound throughout this channel. What's even worse is that EDIFIER S880 has 6 input source selection, but I can only switch to the next source, wait for 2 seconds, and repeat this process 5 times every time I want to switch the audio from my computer to the PS5.

 

After lots of trials and flashing my custom kernel, I finally did it!

The idea is to use a board that has 1 USB otg port to act as a UAC 1.0 sound card, and has another 1 USB host port to connect to the UAC 2.0 speakers, then run programs on that board to redirect sound from USB otg port to USB host port. It is even a plus that the board consumes less power, so we do not need an extra or special power supply for it; this becomes crucial, especially when sharing the same USB connection with the USB otg port (SBC boards are more power-hungry these days!). The standard USB 2.0 protocol only allows 5V 0.5A to the device connected.

I actually built one with an Orange Pi One board, but it turns out that the CPU is not fast enough, and there are sound glitches sometimes.

 

With some research, I found this board that perfectly fits my needs: Radxa zero 3W. Here is the advantage:

- It is cheap, and the one I used is the nearly minimum SKU (1 GB ram with 8 GB onboard eMMC, they also have no onboard eMMC SKU).

- The otg port could handle USB PD protocol, meaning no special hacks are needed, you could get at most 30W of power, if needed (and turns out the normal USB 2.0 port on my PS5 works perfectly, so I guess it consumes less than 5V 1A = 5W)

- It is tiny.

 

However, things don't work right out of the box. At the beginning, I flashed the official, latest Radxa OS, modprobe the g_audio. It is running as a UAC 2.0 device, so it is not the one I want. As a result, I grabbed their kernel source, changed the config to UAC 1.0 gadget, and flashed onto the board, but there seems to be some issues in their kernel fork for UAC 1.0 driver: there was no /dev/snd/controlC0, so the g_audio failed to run.

As a result, I turned to use the Armbian build. This time, the /dev/snd/pcmC0D0c is missing, so g_audio failed to run again.

I then compiled my own kernel, flashed it onto the board. The Armbian build is with kernel 6.1, and the kernel I built is kernel 6.12, then the board failed to boot to the kernel due to U-Boot thought there were some errors.

Luckily, I found that building the whole Armbian image with kernel 6.12 can boot without issues, so I built a custom kernel 6.12 with UAC 1.0 enabled image, modeprobe the g_audio module, and it finally worked!

 

After that, I crafted a simple golang SystemV daemon for bootstrapping and terminating the alsaloop program for redirecting the sound, and used a udev rule to notify that daemon of the attachment/detachment of the USB speaker out event. The converter is finally working!

 

Here is the pre-built image for folks who want a quick test. It is based on the commit b27c86e620dcd9f55daadf52ccc85643dba2a381 from the armbian build repo with the following config modification:

diff --git a/config/kernel/linux-rockchip64-current.config b/config/kernel/linux-rockchip64-current.config
index d053d0997..6e360bba7 100644
--- a/config/kernel/linux-rockchip64-current.config
+++ b/config/kernel/linux-rockchip64-current.config
@@ -3507,3 +3507,8 @@ CONFIG_RATIONAL_KUNIT_TEST=m
 CONFIG_MEMCPY_KUNIT_TEST=m
 CONFIG_TEST_MEMCAT_P=m
 CONFIG_MEMTEST=y
+
+CONFIG_USB_CONFIGFS_F_UAC1=y
+CONFIG_USB_CONFIGFS_F_UAC2=n
+CONFIG_USB_F_UAC2=n
+CONFIG_GADGET_UAC1=y

 

Currently, the "current" build does not detect the Wi-Fi interface, but as it does not affect the audio converting feature, I may not put too much effort into this.

Edited June 29, 20251 yr by 李松錡

awesome, well done!  and thank you for sharing.

HI I arrived here after buying a Hiby R3II and trying to use it on my PS5. Stupid Gemini sayed that it has a software toggle to UAC 1 but it doesn´t. So I´m seeing what to do and this option popped up. Do you think it may work with the PS5 front USB? I looked for the board and is around 35 USD in Aliexpress shipped to Europe. Is that regular price? thanks a lot. 

Yes, absolutely works with the front USB.

I do not know the regular price, that depends on the spec you choose (how much RAM, is eMMC included). I got the similar price for that board including shipping, but do note that it was shipped from China to Taiwan.

 

Hmm, that's true for Hiby R3II, checking from the manual, it actually says:

 

Quote

When the player works on USB audio, you can think of the player as a USB HD decoding receiver or as a USB sound card. The player works on USB audio supports the USB Audio Class 2.0 (UAC2) protocol, but does not support USB Audio Class 1.0 (UAC1), so the host side also needs to support USB Audio Class 2.0 (UAC2).

 

Edited January 16Jan 16 by 李松錡

I am thinking about getting a Nanopi R3S, which could be a better option, for it has a dedicated designed case for it. The CPU on it is the same one, so computation wise should be feasible.

Hi, I learnt how to customize the image, and also found a simpler way to achieve the bootstrapping alsaloop. Here is the forked repository and a pre-built image for you to try. Supposedly, you just need to flash the image, and it should work out of the box. As the radxa zero 3w does not have the Ethernet port, so you will not accenditally connect it to some networking, I think it is fine to leave the root password as the default.

 

However, if you plan to connect the device to some networking, remember to set up the root password. The Wifi is also working with the latest kernel release (my wifi module is broken, but I do see dmesg is trying to interact with that module).

Thanks a lot, I´m thinking about buying a board purely for your image, would you advise me to wait for the nanopi r3s release? I have an old rpi 3  collecting dust, do you think I may use that as well?

thanks

Edited January 23Jan 23 by Taladro

Oh, I do not have a nanopi r3s, and I do not plan to buy it. The image is for the radxa zero 3w, but as I provided the source of building it, you can customize it to build with nanopi, too. The reason I mentioned nanopi r3s is because it has a dedicated case for it. I ended up using raspberry pi zero case for it, with some custom modifications, which is also nice.

 

Raspberry Pi 1/2/3, zero/zero w/zero w2 won't work, because they only have 1 USB controller. The trick here is to have at least 2 USB controllers: 1 in host mode to connect to the UAC2.0 speakers, and 1 in gadget mode as a UAC1.0 USB device. Raspberry Pi 4/5 should work, but you need to make sure the power supply is strong enough, which is a big issue here, because with the official image, you can only connect the USB C port for power and as a USB gadget port, yet PS5 probably won't support that much power for RPi 4/5, so you ended up needing a USB power and data splitter for that, and connect the data port to PS5, then a separate power supply for the power port (simliar things happened in PiKVM).

Edited January 23Jan 23 by 李松錡

Thanks a lot, I´m in doubt about buying the radxa and use it with my hiby r3II or just buy a Fosi Audio DS1 or Fiio Ka13. I have bought nice headphones and I´m sad I can´t use them on my ps5 correctly. If just for that matter what would you do? Thanks again 

I think it depends on what you want. If you want to have some fun, then building this one is interesting actually, and there is more extensibility, and since radxa zero GPIO pin is compatible with Raspberry pi's, you can also get a DAC Mini Hat raspberry pi module, so there is more you can play or customize.

 

But, if you just want to solve the problem and don't want to have these hassle, buying an existing product is a better option, so you have the warranty, customer support, and everything should work out of the box. I don't have Fosi Audio DS1 or Fiio Ka13, so every information I get is from the google search and cannot guarantee the correctness. From what I saw, Fiio Ka13 seems to be unable to work with PS5, and Fosi Audio DS1 can. This project only handles digital audio data, so I do not need to pay attention to the audio electronic properties  when outputting (because that's the job of that USB speaker), and I do not know the sound quality of Fosi Audio DS1.

 

One last word, if your original plan is to use this converter to connect to Hiby R3II, then to your headphone, that should work, but that is kind of redundant; there is more latency introduced, and the robustness could be wore because you have a longer output chain (from the engineering perspective). It would be easier to just get a high quality UAC 1.0 Amp.

Edited January 28Jan 28 by 李松錡

Hello,

Your project caught my attention ! I'm trying to solve a similar issue but for a slightly different use case: I want to use a Wave XLR MK.2 (UAC 2.0 audio
interface with a mic input) on my PS5 through a UAC1<->UAC2 bridge like yours.
 

Two questions if you don't mind:
 

1. Your setup redirects audio in one direction (PS5 -> UAC2 speakers). My case needs it both ways: game audio out to my headset AND mic capture back to the PS5 for party chat. Do you know if the g_audio/UAC1 gadget config supports a capture endpoint at the same time as playback, or would that need extra ConfigFS/kernel work on top of what you already did?
 

2. On the latency side. What I care about is the added delay on the game audio path itself (PS5 -> your bridge -> UAC2 output), since I play competitive FPS and rely on audio cues like footsteps and gunshots for reaction time. Did you notice or measure any perceptible delay there with your alsaloop setup?

Also,did you end up sticking with the Radxa Zero 3W, or would you pick something else if you started this project today?

Thanks a lot for sharing your work, it's a great starting point.

Hi SENGA,

 

For 1., there should be no extra setup for the kernel config change. `g_audio/UAC1` gadget config should be enough (no need to recompile the kernel). The thing you need to change is the alsaloop daemon (this line). I did not verify it, but AI gives a example of looping the speaker and microphone in 2 different threads like this:

 

alsaloop -v \
  -T 1 -C hw:Loopback,1,0 -P hw:CARD=B,DEV=0 -t 20000 -A 3 \
  -T 2 -C hw:CARD=B,DEV=0 -P hw:Loopback,1,1 -t 20000 -A 3

 

. It looks legitimate, you can give it a try.

 

For 2., I tried playing Astro Bot the last stage, and I did not notice any inconsistency or audio lagging. I did not have the equipment to measure the latency, so I cannot give you the exact number. But do make sure the device is properly cooled (a nice heatsink and do not seal it in a box should be good), otherwise the CPU is throttled, and there will be sound breaking.

 

So as the long text in the description, the reason I ended up using Radxa Zero 3W is because:

 

1. It has at least two USB controllers. That is the requirement, because you need one for the USB gadget to connect to PS5, and another one for the USB host to connect to the USB speaker for output.
2. The CPU is powerful enough. Before I used orange Pi One, but the single core performance is not fast enough, so I heard sound breaking. I guess maybe compare the single core score with rk3566 (the one on Radxa Zero 3W) might be a reference to see if a CPU is capable of handling the task. Higher single core score is better.

With these 2 reasons, Radxa Zero 3W seems to be the bottom line you can find on market for such project. You can get a even more powerful board as you like.

Edited September 18Sep 18 by 李松錡

Hmm, after I learned more about the UAC and other things, I realized that there might be cheaper solutions (and I did actually really hands on it), but they are not Armbian related. Anyhow, I'll just share here if anyone is interested:

 

TL;DR

A much cheaper solution is to have a UAC 1.0 to I2S component connect to the PS5, then have another MCU acting as an UAC 2.0 host, do the DSP conversion, then send these audio packets to the speaker.

 

• The UAC 1.0 to I2S component could be any existing modules, such as PCM2706. You can also use a RP2040/RP2350 to build a UAC 1.0 device to I2S component. I actually partially done that work, see the long store below.
• The UAC 2.0 host MCU has some restrictions. The main restrictions are:
  • The MCU must has the USB controller component that can be a USB host, and also supports at least USB Full Speed (FS, 12 Mb/s)
  • The USB controller must support isochronous OUT transmission. That is the last step I failed in the attempt, see the long story below.
  • STM32F401, ESP32-S3 are good candidates.

 

Long story: I actually tried to use waveshre RP2350 USB A board to build a UAC 1.0 to 2.0 converter, because it looks very neat, and already has 2 USB receptacles on the board. I managed to use the PIO as a UAC 1.0 device, getting all the audio packets from my macbook (which requires another fix here). However, I failed at properly playing the sound from the Edifier speaker. Considering it is a timing and latency critical task, I actually utilized the dual core on the RP2350. There is one dedicated core for handling the DSP, and the other core is dedicated for the USB tasks (UAC 1.0 device and UAC 2.0 host). In the end, I did successfully forward the sound from macbook to the speaker, but the sound is incorrect and not listenable. The sound is always skipping.

 

In the end, I tried to generate a 750Hz sine tone from the RP2350 itself to completely rule out the possibility of sound distortion from UAC 1.0 communication. The sine tone still has the same issue, and turns out it is because the USB controller on the RP2350 does not include the logic unit for handling the isochronous transmission. I tried to work around it, but seems to be impossible.

Edited October 4Oct 4 by 李松錡