Thanks for raising this. When I’ve said Maestro doesn’t touch the audio stream, I wasn’t trying to mislead anyone — I was using it as shorthand for why I think Maestro sounds the way it does, and I shortened too far. My bad, and I apologize.
Marcin (whom I greatly respect) and the poster in the dCs forum, Anupc, are right about what that Wireshark capture shows. So rather than dance around it, let’s take a look at what’s actually going on under the hood.
When Maestro controls a UPnP/OpenHome renderer like the dCs Bartok, yes — we’re pulling the Qobuz stream into the Maestro app and then handing it off to the renderer over your local network. That’s how UPnP works with cloud content. The renderer expects a local URL; Qobuz doesn’t hand those out to anyone but its own endpoints; so something has to sit in the middle. Every UPnP control point doing cloud audio does this — BubbleUPnP, Kazoo, mconnect, JRiver, Audirvana. We’re not special there.
A couple of things worth clarifying though.
Roon does the same thing — more so, actually. Roon Core pulls the stream and rewraps it inside its own RAAT protocol before shipping it to the endpoint. That’s a heavier hand on the audio than a TCP relay, not a lighter one. The “Maestro proxies, Roon doesn’t” framing isn’t quite right.
Lots of techie talk. Let me explain:
TCP — Transmission Control Protocol — is the basic plumbing of the internet. Vint Cerf and Bob Kahn designed it in the early 1970s (about the time PS Audio was founded in my living room) as part of the work that became the modern internet, and just about everything that needs to get reliably from point A to point B on a network — email, web pages, software downloads, your Qobuz stream — has been riding on it ever since. The concept is simple: take a chunk of data, chop it into numbered packets, ship them between two addresses, put them back in order at the far end, and ask for any missing ones to be re-sent.
The post office for digital information.
What Maestro is doing is reading bytes off one TCP connection (the one from Qobuz) and writing those same bytes onto another (the one to your Bartok). That’s the whole job. The envelope gets passed from one address to the next, and Maestro never opens it, never reads what’s inside, never repackages anything.
There’s a real difference between forwarding bytes and processing audio. When using Maestro, the FLAC frames arriving at the Bartok are bit-for-bit identical to what came off Qobuz’s server. No DSP, no resampling, no decode-and-re-encode. You can checksum the bytes at either end.
Maestro is a deliberately minimal transport layer, not a playback engine.
We are not reconstructing the stream inside our own protocol stack, we are not running it through an audio engine, and we are not “improving” the bits — we can’t improve what we don’t change. Whatever character Maestro has comes from doing as little as possible to what passes through it.
To my ears, Maestro sounds better, and I know a lot of you agree. After fifty years in this business I’ve stopped pretending the measurements tell the whole story. If they did, nobody would have spent the last thirty years arguing about clocks, cables, power, and isolation.
Have a listen for yourself. Maestro can be downloaded and played with for a month for free. If you want to not worry so much about the why it sounds better, and give it a try and see/hear for yourself, then it’s easy!
Have fun and enjoy.