1 Topic by julian.accts (edited by julian.accts 2026-02-08 12:32:22)

  • Registered: 2026-02-08
  • Posts: 2

Topic: M-1620 Pro Latency

Hello,

I’m looking for a way to connect AES50 to AVB/Milan with the lowest possible latency. I’m coming from Dante and have very little experience with AVB.

I need to send 16 channels at 48 kHz from a Behringer WING into an L-Acoustics LA7.16i.
Current setup: Behringer WING with Dante card (AES67 mode) → LA7.16i (AES67 mode).

However, since the transmitter only supports a 1 ms packet time and the receiver uses a buffer of three packets, I end up with a total network latency of about 3 ms. That’s what I’m trying to reduce to under 1 ms, ideally even lower.

For this reason, I’m investigating the following possible solution:
Midas DL32 ADAT out → ADAT in on M-1620 Pro → AVB out on M-1620 Pro → L-Acoustics LA7.16i

Since latency is the main concern, I’m wondering what the total end-to-end latency of this setup would be. I believe the minimum network latency for AVB would be around 0.25 ms—is that correct?
If I understand correctly, this refers only to the transmission latency within the AVB network between the AVB/AoIP interfaces. On top of that, the internal latency of the devices hosting these interfaces would also need to be added.

How much latency would be introduced internally by the M-1620 Pro from ADAT input to AVB output due to internal routing or DSP? For example bit-depth conversio. I believe the ADAT inputs would be 24 bit, while the AVB streams would need to be 32bit PCM as supported by the listener. Since I’d like to avoid any SRC or unnecessary processing, my idea is to clock the system from the mixing desk via the DL32 (AES50 in, ADAT out). The M-1620 Pro would receive its clock from the ADAT input and then act as the clock master for the AVB/Milan network.

Thank you very much for your help.



German version

Hallo,

ich suche nach einer Möglichkeit, AES50 mit AVB/Milan bei möglichst geringer Latenz zu verbinden. Ich komme aus der Dante-Welt und habe bisher nur sehr wenig Erfahrung mit AVB.

Ich muss 16 Kanäle bei 48 kHz von einem Behringer WING in einen L-Acoustics LA7.16i übertragen.
Aktuelles Setup: Behringer WING mit Dante-Karte (AES67-Modus) → LA7.16i (AES67-Modus).

Da der Sender jedoch nur eine Paketzeit von 1 ms unterstützt und der Empfänger einen Puffer von drei Paketen verwendet, ergibt sich eine gesamte Netzwerklatenz von etwa 3 ms. Genau diese möchte ich auf unter 1 ms, idealerweise sogar noch weiter, reduzieren.

Aus diesem Grund untersuche ich folgende mögliche Lösung:
Midas DL32 ADAT Out → ADAT In am M-1620 Pro → AVB Out am M-1620 Pro → L-Acoustics LA7.16i

Da Latenz der entscheidende Faktor ist, frage ich mich, wie hoch die gesamte End-to-End-Latenz dieses Setups wäre. Ich gehe davon aus, dass die minimale Netzwerklatenz bei AVB bei etwa 0,25 ms liegt – ist das korrekt?
Wenn ich das richtig verstehe, bezieht sich dieser Wert nur auf die Übertragungslatenz innerhalb des AVB-Netzwerks zwischen den AVB/AoIP-Interfaces. Zusätzlich müsste also noch die interne Latenz der Geräte berücksichtigt werden, die diese Interfaces hosten.

Wie viel Latenz würde im M-1620 Pro intern von ADAT-Eingang zu AVB-Ausgang durch internes Routing oder DSP hinzugefügt? Da ich jegliche SRCs oder unnötige Signalverarbeitung vermeiden möchte, ist meine Idee, das System vom Mischpult aus über den DL32 zu takten (AES50 In, ADAT Out). Der M-1620 Pro würde seinen Clock vom ADAT-Eingang beziehen und anschließend als Clock-Master für das AVB/Milan-Netzwerk fungieren.

Vielen Dank für eure Hilfe.

  • From: Germany
  • Registered: 2007-04-10
  • Posts: 10,961

Re: M-1620 Pro Latency

Ich hab gerade keinen genauen Wert zur Hand, aber es dürfte sich im Bereich einiger weniger Samples bewegen... Wie genau muss die Angabe sein?

Regards
Daniel Fuchs
RME

RME Support's Website

  • Max
  • Max
  • Moderator
  • Offline
  • Registered: 2007-05-09
  • Posts: 358

Re: M-1620 Pro Latency

What is your target latency and why?

However, since the transmitter only supports a 1 ms packet time and the receiver uses a buffer of three packets, I end up with a total network latency of about 3 ms.

I don't understand how you get to the 3 ms conclusion. Could you provide a source of information for a "three packet buffer"?

AVB presentation time offset could probably be set even lower than 0.25ms, it depends on the number of hops. If you connect the M-1620 Pro directly and manage to set up streaming without a switch (unlikely that you'd want to do that), you could reduce the presentation time offset to 140us. Otherwise 280us (one switch).

The time it takes for for an ADAT signal to make it into the AVB stream should be within a few samples (less than 100us).

4 Reply by julian.accts (edited by julian.accts 2026-02-09 10:17:36)

  • Registered: 2026-02-08
  • Posts: 2

Re: M-1620 Pro Latency

RME Support wrote:

Ich hab gerade keinen genauen Wert zur Hand, aber es dürfte sich im Bereich einiger weniger Samples bewegen... Wie genau muss die Angabe sein?

Wenn wir uns im Bereich von 3-5smpls bewegen, ist alles OK.

Max wrote:

I don't understand how you get to the 3 ms conclusion. Could you provide a source of information for a "three packet buffer"?

I was talking about my current solution (AES67 Mode). As you can read here https://prdstglaxxwe001.blob.core.windo … deV1.8.pdf on page 11 and 12, the amplifier in AES67 mode accepts packets of 1ms time (= 48smpls at 48kHz) and 0.333ms (16smpls). And always applies a buffer of 3 samples to compensate for network transmission delays.


Max wrote:

AVB presentation time offset could probably be set even lower than 0.25ms, it depends on the number of hops. If you connect the M-1620 Pro directly and manage to set up streaming without a switch (unlikely that you'd want to do that), you could reduce the presentation time offset to 140us. Otherwise 280us (one switch).

The time it takes for for an ADAT signal to make it into the AVB stream should be within a few samples (less than 100us).


Currently the only AVB Milan device I have for testing is the amplifier. It only shows as a listener in Milan Manager. Since the latency is defined by the talker (and I don't have one here right now), I can't experiment with latency settings. On the web, I read about minimum latencies of 0.25ms. The manual of the l'acoustics amp talks about a maximum latency of 2ms, but not about a minimum. Unfortunately the web is not very clear about this things. So I don't know what the minimum value will be that I can configure in the Milan Manager software. (While with Dante there is a clear rule, that every device can set to fixed values between 0.25/0.5/1/2ms according to a safe value, based on the number of hops.)

I'm not planing in involving any switches, but direct cable between the M-1620 Pro and the amp.

5 Reply by Max (edited by Max 2026-02-10 03:00:10)

  • Max
  • Max
  • Moderator
  • Offline
  • Registered: 2007-05-09
  • Posts: 358

Re: M-1620 Pro Latency

I see. In MILAN, 6 samples make a frame (=125us observation interval), and it's likely going to be processed faster. But this should be answered by L'Acoustics. (edited - earlier version mentioned the frame transmission time, which is not relevant here).

In order to understand PTO and MTT, think of it this way:

- both devices have the accurate time
- talker adds an offset to the time (by default, 2 ms) and puts this information into outgoing frame as playout time
- listener receives early, but keeps frame in buffer until playout time

The offset can be configured to anything you like (it is a static setting for each stream output of the talker), so you are free to use whatever (adding 10 us shifts the phase at the output by 180deg...). You can configure that either in Hive/MILAN Manager or in the web ui of the M-1620 Pro.

However there is also the MTT (maximum transit time) which is measured before the stream is established. If the presentation time offset is smaller than the maximum transit time, you get a warning in Hive. Makes sense because why would you attempt to play things from the past.

https://docs.rme-audio.com/m1620/330-1c … t_streams/ (click on "show remote" to see how you can set the latency on the M-1620 Pro)

If you run a direct cable between the two products, there is no way that a controller can start the stream (I assume, unless the L'Ac device has a built-in switch or implements a controller within its listener). However, listeners will automatically re-connect after a network failure, so you could set it up once and then remove the switch later. I have not tested this setup, it's really a bunch of questions for L'Acoustics. Let us know the outcome of your experiments, please :-)!