• Welcome to ASR. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

DIY fanless network streamer: retired ASUS RT-AC88U running OpenWrt + MPD + myMPD (DSD over DoP)

andrewr8712

New Member
Joined
Nov 3, 2026
Messages
1
Likes
8
Hi all,

Long-time reader, first project write-up. I wanted a network streamer for a Topping DX5 II to replace a dedicated PC running Volumio in the living room. Instead of buying a dedicated streamer, I reused an ASUS RT-AC88U router (BCM4709C0, dual Cortex-A9 @ 1.4 GHz, 512 MB RAM, 128 MB NAND) and built a minimal OpenWrt image around MPD.

This should all be reproducible on other OpenWRT compatible hardware with similar hardware specs - not sure how low the spec can go before performance suffers.

I’m not selling anything and I’m not making subjective sound-quality claims. The goal was to remove the PC from the playback path and make a deterministic, fanless appliance. Image, source, build notes and raw benchmarks are here:

What it is​

  • OpenWrt 25.12.6, wired-only, no firewall/NAT/DHCP server — it’s a LAN client
  • MPD 0.23.16 built with -Ddsd=true, ALSA hw:0,0, mixer_type "none", no soxr/libsamplerate
  • myMPD 26.1.1 web UI, plus upmpdcli for UPnP/OpenHome and MPD-native phone apps
  • Music library on an SMB share; no local music storage
  • MPD DB and myMPD state live on the overlay (/etc/mpd, /etc/mympd), so they survive reboots
  • Custom image + overlay, not a general-purpose OpenWrt install

Measured so far​

Playback test: 9 m 49 s continuous, 40 samples at 15 s, library of 1,965 tracks including 108 DSD files.

SourceMPD reportsALSA hw_paramsCPU avg / max
16/44.1 FLAC44100:16:244,100 Hz4.4 % / 6.6 %
24/96 FLAC96000:24:296,000 Hz5.9 % / 9.1 %
24/192 FLAC192000:24:2192,000 Hz7.7 % / 13.3 %
DSD64dsd64:2176,400 Hz (DoP)7.6 % / 10.6 %
DSD128dsd128:2352,800 Hz (DoP)6.5 % / 9.4 %
DSD256dsd256:2705,600 Hz (DoP)11.6 % / 12.7 %
Overall CPU average 5.85 %, peak 13.3 %, zero USB disconnects during that run. Every rate matched the source exactly; there is no resampling in the chain. These figures were taken on the 24.10.8 build; DSD64 was re-verified on 25.12.6.

I also published raw CPU/memory/storage microbenchmarks (BENCHMARKS.md) so others can compare hardware, and a porting guide for other routers/DACs.

Why this instead of a dedicated PC?​

Engineering/operational reasons, not audio mysticism:

  • No PC audio stack — no PulseAudio/WASAPI mixer, no sample-rate conversion, no per-OS audio “enhancements”. ALSA talks directly to the DAC.
  • Fanless, no HDD, low heat — nothing mechanical in the listening room. I have not measured wall power yet, so I won’t quote a number until I do.
  • Appliance behaviour — read-only squashfs root + overlay; boots straight into the player; no background updates, AV scans or driver churn to break playback. Power loss just means it boots again.
  • Network library, no PC in the path — MPD reads the SMB share directly; no PC needs to be awake to play music.
  • Cheap reuse — a second-hand router instead of a dedicated streamer or a spare PC.
  • Standard control surfaces — browser (myMPD), UPnP/OpenHome, and MPD apps such as M.A.L.P.
I’m explicitly not claiming it sounds better than a PC. Whether keeping the PC out of the room or avoiding the OS mixer matters audibly is exactly the kind of thing that should be measured, not asserted.

Limitations / what it doesn’t do​

  • Wired only. The BCM4366 5 GHz radio has no usable FOSS AP/STA support under OpenWrt; 2.4 GHz AP starts but isn’t used. 5 GHz would need the proprietary DD-WRT/ASUS driver, which isn’t portable to OpenWrt’s kernel.
  • DSD via DoP only. DSD64/128/256 verified; DSD512 would need native DSD, not enabled.
  • No ALAC/WMA/APE, no myMPD Lua scripting in this build.
  • UPnP is a renderer, not a library server — for UPnP browsing you still need a media server on the network.
  • Tested on exactly one router + one DAC (Topping DX5 II). Other hardware needs the porting notes.
  • No null-test or bit-error measurement yet — “bit-transparent” here means no resampling/mixing/volume processing, with the negotiated rate/format verified at the DAC. A proper null test is on my list.

What I’d value from ASR​

  • Suggestions for a practical null-test / bit-error methodology on this setup (PCM and DoP).
  • Whether the CPU headroom figures look sane for this SoC, and what else is worth measuring.
  • Wall-power measurement advice — I’d like a real number rather than a guess.
  • Anyone running a similar low-power OpenWrt/MPD build on other hardware: what did you use for storage, and did you hit the same 5 GHz Broadcom limitation?
Happy to answer questions or provide logs/commands for anything above.
 
In your present setup MPD is both server and renderer, whereas john61ct's suggestion is to run just a renderer - which makes sense for a modest processor such as BCM4709C0. Of course this means you would need to run the matching server, Lyrion, somewhere else on the network. If your existing SMB share device allows Lyrion to be installed, this would certainly be a good overall solution.
 
IMO there is no reason for this setup to be bit imperfect, it's a normal linux computer running on ARM hardware with standard Openwrt distribution, nothing custom-made with custom software. Routers with USB ports re-flashed with OpenWrt have been used as audio players for many years.
 
Good idea for reusing one of those routers we all have in a cupboard somewhere, and free is cheaper than a Pi!

In the end, all sensibly-priced commercial streamers are probably using similar ARM-based compute modules.

I assume any device that's re-used must have a USB port that the software can use to connect to the DAC, which makes me think an ethernet connection might be useful input for a DAC to have as well, in which case the router's USB could be used for local storage of music.
 
an ethernet connection might be useful input for a DAC to have as well
IMO there is no core standard for network audio like UAC/UAC2 for USB. It would require a complex software on the DAC, typically an embedded linux computer. Expensive (hw, sw, licensing), requires software maintenance, ... IMO just USB input is the best interface for a DAC. A USB hub can be used on the host side if more USB devices need to be connected.
 
There's no free and non-proprietary standard I'm aware of (other than AES67?), but many fancier DACs may well have a SoC capable of hosting a network stack. It all adds cost though!
I like what you've done though and can't see why the DAC would not get bit perfect data (assuming you know the software doesn't deliberately munge it up).
 
Last edited:
Suggestions for a practical null-test / bit-error methodology on this setup (PCM and DoP).
Use a USB to coax/optical adapter, and another machine with coax/optical input to record the output. Assuming no volume control or other DSP you should get a bit-exact match once you trim start and end.
Whether the CPU headroom figures look sane for this SoC, and what else is worth measuring.
Seems fine - there's even room to do other things if you want.
Wall-power measurement advice — I’d like a real number rather than a guess.
Various wall power measurement devices are available. My old one only has a resolution of 1W, but you should be able to find better now. Some of the home automation plug in power switches include power measurement if you already have the rest of the automation side to interface with them. They may also talk to a phone app.
Anyone running a similar low-power OpenWrt/MPD build on other hardware: what did you use for storage, and did you hit the same 5 GHz Broadcom limitation?
Not OpenWRT/MPD, but other minimal linux on a few SBCs. Usually the audio streamed from LMS running on a different machine, but local USB connected storage should work fine if that's what you want. Beyond that it depends on the router hardware - some have extra internal storage options if you tweak them. I've been keeping an eye on OpenWRT support for a broadcom-based AP I have, and there's been some progress on the driver front that hasn't made it upstream yet, so even with broadcom it depends on exactly what hardware you have.
 
IMO there is no core standard for network audio like UAC/UAC2 for USB. It would require a complex software on the DAC, typically an embedded linux computer. Expensive (hw, sw, licensing), requires software maintenance, ... IMO just USB input is the best interface for a DAC. A USB hub can be used on the host side if more USB devices need to be connected.
I guess you mean no single dominant standard like UAC(2) for USB - there are a multitude of standards, some of which are open like AVB and AES67, but they're a lot more complicated to use than USB. The latter has an open implementation on ESP32-P4 that looks promising.
 
Network library, no PC in the path — MPD reads the SMB share directly; no PC needs to be awake to play music.
Where is the SMB share hosted, if no PC needs to be awake to play music?
 
Everything can be seen as type of PC these days, but a NAS can use far less power than an actual PC, will have WoL, no screen and can be put on a shelf out the way... But I see the sentiment, and I like the project above, even if I have no need for it!
 
I guess you mean no single dominant standard like UAC(2) for USB - there are a multitude of standards, some of which are open like AVB and AES67, but they're a lot more complicated to use than USB.
Yes, but which streamer supports these network protocols out of the box, to have a network DAC (renderer) attached, just like it works for USB audio?
The latter has an open implementation on ESP32-P4 that looks promising.
Especially the way it fine-tunes its internal audio-clock PLL to avoid the two-clocks merging issues. It definitely is a VERY interesting solution. However, the former applies - which streamer can use such renderer device out of the box?
 
Yes, but which streamer supports these network protocols out of the box, to have a network DAC (renderer) attached, just like it works for USB audio?
Indeed - apart from making your own as in this thread. And those two have requirements that aren't met by all wired network ports, so there's a further complication. PipeWire seems to be adding support for a number of network protocols though, which should make it easier to support them where the necessary network capabilities are available.
 
OpenWrt 25.12.6, wired-only, no firewall/NAT/DHCP server — it’s a LAN client
Couldn't you just install MPD on Asuswrt-Merlin and keep the full functionality of the router? It's been a while since i've used Merlin (I moved to a pfSense based setup several years ago) but this seems like a simpler option.

Someone got this working 12 years ago on an RT-N66U, they had to compile the firmware to include the audio modules, I don't know if that would still be nessesary. https://www.snbforums.com/threads/asus-rt-n66u-as-music-player.18879/
 
Couldn't you just install MPD on Asuswrt-Merlin and keep the full functionality of the router?...


Someone got this working 12 years ago on an RT-N66U, they had to compile the firmware to include the audio modules, I don't know if that would still be nessesary.
IIUC that firmware is based on ancient 2.6.26 kernel and no sound support is compiled in https://github.com/RMerl/asuswrt-me....36/arch/mips/configs/bcm947xx_defconfig#L907 . IIUC this commit https://github.com/RMerl/asuswrt-merlin/commit/a41bcaa6acc152b59b5dc039cf75741fc94a528c adds proprietary binary blobs supporting non-open functionalities of that SoC exactly for that old kernel.

While the sound support could be compiled in when rebuilding the firmware, support for usb-audio in kernel 2.6. was still rudimentary.
 
IIUC that firmware is based on ancient 2.6.26 kernel and no sound support is compiled in https://github.com/RMerl/asuswrt-me....36/arch/mips/configs/bcm947xx_defconfig#L907 . IIUC this commit https://github.com/RMerl/asuswrt-merlin/commit/a41bcaa6acc152b59b5dc039cf75741fc94a528c adds proprietary binary blobs supporting non-open functionalities of that SoC exactly for that old kernel.

While the sound support could be compiled in when rebuilding the firmware, support for usb-audio in kernel 2.6. was still rudimentary.
Yeah, I guess that's going to be an issue for older devices where support for Asuswrt-Merlin has run out. Looks like none of the older AC or N devices are supported anymore.
 
IIUC the new openwrt should be pretty good with the ethernet features of that router, while lagging in wifi support. For wired router it sounds perfectly fine. Kernel 5.15 of the latest openwrt should be OK too for usb audio.
 
Back
Top Bottom