Ravenna Linux Driver

This is not directly related to the title, but it does sideline this discussion in my opinion and gives you something to think about. Is that day near…



Wow. That’s big news, thanks.

Hi Mika
Please forgive my ignorance, does this mean that audio over ip stageboxes such as those made by Presonus or Behringer can now be used with Ardour on Linux, or is an intermediate piece of hardware required, or have I got it completely wrong?
I had a look at the article you linked, but it was a bit beyond my understanding :exploding_head:

This is very off topic of SSL purchasing Harrision, but yes, you can use audio over IP with Ardour, but it is still somewhat cumbersome to configure.
The Ravenna driver maintained by Merging is a good start, but Merging is only interested in embedded use, so does not track the mainline kernel. To make it usable with distributions which ship an up to date kernel you need the patches maintained by Andrea Bondavalli (who also maintains an open source user space daemon needed for configuring the driver; the daemon shipped by Merging is closed source).
Andrea Bondavalli github repositories

Maybe someone wiser than me could tell us more. I linked it after finding it on https://linuxmusicians.com/. Mainly, because it shows that there is good development. But this is apparently from 2019. There is good development going on in Linux and SSL knows it. So I think this will be a good thing.

Split the Ravenna topic into it’s own thread.


Short answer NO.

There are specific AoIP boxes that may work, but those you mentioned use either a weird bastardization of AVB (Presonus) that I don’t believe supports AES67 (Or Ravenna) or AES50 (Behringer/Midas) that is not compatible at all with AES67.

Ravenna boxes such as those made by Merging Technologies would work, and theoretically boxes with protocols compatible with AES67 might work (So a lot of Dante devices and some specific AVB devices) but the only way to know for sure is to try them, which is a financial investment that may not pay off.


Sorry for confusing the original topic.
I have seen previous mentions of this topic on the forum and was interested to find out if any progress had been made for Linux users.
Thank you all for your knowledgeable input.
So waiting some more would be wise before buying anything.

I was taking the original poster at face value that he was actually asking about audio-over-IP, and not audio-over-other-things interfaces.
I made the apparently unjustified assumption that the reference was to Dante, which as you point out mostly do support AES67 compatibility now. That Merging driver does not support audio over layer 2 protocols (e.g. AVB on Ethernet like Mark of the Unicorn and others support), nor does it support non-Ethernet protocols that happen to use Ethernet cables (AES50/SuperMac, Ultranet personal monitor systems).

Short answer is yes. There are not many Ravenna devices around, and the Dante devices in AES67 mode until now still required you to use Dante Controller to get the connections made. Some people have reported success running Dante Controller under WINE, but I have not been able to get it to run.
Recently someone created a python based CLI utility for configuring Dante devices by reverse engineering the wire protocol. So far I have only had time to verify that it will display device information, I have not tried making and re-arranging connections with that tool yet. On my list of things to do after this week.
“Network Audio Controller” python program for Dante control

Note that software AoIP is pretty resource intensive, the Dante devices only support 1ms buffer time, which means running with 48 sample periods.

1 Like

Yep I figured which is why I wanted to clarify because those options were specifically mentioned so I didn’t want to give a wrong impression is all.

Also for clarification, and I could be wrong on this as I don’t know AVB as well as I do Dante, but I believe AVB is considered a Layer 3 compatible protocol, but it does have requirements that make up part of the spec that are Layer 2 specific. But again I could be wrong on that, as I haven’t looked into the details of the protocol in the same way (I have a few Meyer galaxies that use AVB(Milan) and some MOTU interfaces, but really don’t use AVB honestly).

This is good to know about thank you, I wasn’t aware of it. I was figuring the control protocol wouldn’t be to difficult to reverse but I never had time to look at it in Wireshark. That being said there are a couple of things in the targetted list that are not possible in Dante the way the protocol is designed, so I suspect it won’t be quite as nice as they want, but I am still going to keep my eye on it. The other thing is that I know some time back I remember people routing the data stream to be able to listen to the audio through gstreamer using fairly standard audio protocols, but without control it wasn’t doing them much good, so I am curious if this acts as a gateway or not.

Pretty sure that it can support over 1mS actually in the hardware, but I currently have Hamilton in at one venue and another is in tech for Rent and don’t want to jump on my control machine in either lcoation and freak out the A1 right now to check:) Obviously the software can go up to 10mS IIRC, and my memory is that the latency negotiation takes the slower of the two values between transmitter and receiver as the value for that link.


Yeah, and with Dante Virtual Soundcard my understanding is that it will even do things like run 1ms latency between Dante hardware devices, but allow just the virtual soundcard to run at 10ms latency.
But they aren’t going to let their AES67 mode do all the cool Dante specific things, so they limit to just the bare minimum required features, like 1ms buffer size only, and 48k rate only, even though AES67 allows devices to support more than just the bare minimum features that are required to ensure compatibility.

Yea I haven’t played much with the AES67 compatibility yet to know what will or won’t happen there sadly. Wouldn’t surprise me honestly, at least for hardware interface, but on the software side I would say it is unlikely they will enforce 1mS latency there, but I could be wrong.

Either way 1mS latency is still a good amount for hardware, as that should cover probably at least 90% of users, if not closer to 99%. To give some reference, I recently had some issues that caused me to raise my latencies to between .5mS and 1mS on my Dante network here, but previously I ran most devices on .25mS-.5mS for years without issues and probably could go back to that honestly. That is a standard network with between 3-4 hops in many cases for one of the larger road house theaters in the US, so more complex than the vast majority of users.