Ethernet Gigabit Switch for Audio

£249 ! - At least it's black with green lights 

Metal cased Netgear GS xxx have green lights and you get loads of change too in comparison.

I know, but there are folks in this forum who have, say, AudioQuest Vodka ethernet cables, so what's a couple hundred quid on a "proper" audio ethernet switch for them...?

Also, the pro audio community is a tough customer - if they don't hear benefits, they don't buy the products, period.

As the devil's advocate I might point out that the main benefit advertised by MOTU for this product is low latency. Important when recording, but meaningless on playback, which is what we are talking about here...

If I had heard any difference with the AQ Vodka over my ugly but /10 less Supra,  then I might be mad enough to fork out that kind of money for a simple unmanaged switch that's x10 more than my GS105. 

That's fascinating, Mike, but some folks here have been interested in ethernet switches, and do have expensive ethernet cabling, so this thread is for them, not for you. Sorry

Gajdzin,

And there was me thinking that as you're talking about it, you'd also be spending the money to test it!

Call me thick, but why is a low-latency Ethernet switch important when recording?

I'm tempted, I'm tempted... But I'm re-building my recording studio now, some new Beyerdynamic overhead microphones and custom mic cables (36 of them!), very expensive stuff. Also thinking of upgrading main studio interface from MOTU to RME and that's REALLY expensive

The whole audio path should have the lowest possible latency. And these days an ethernet connection can be in the audio path, because some studio boxes started using ethernet instead of e.g. FireWire. Cheaper to implement and more bandwidth this way, which does matter if you record e.g. 64 tracks at 24 bit / 192kHz.

OK, I get the bandwidth bit, but why does latency matter?

Long explanation and outside the scope of this thread Recommended reading: http://www.ime.usp.br/~kon/papers/icmc04-latency.pdf.

If you don't feel like a deep dive into the topic, here it is in a nutshell: imagine you are listening to your band playing in your headphones (delayed, because of the inevitable latency of the playback track) while you record your instrument. Your playing is shifted by the latency amount. Then imagine that you hear your own playing in the headphones shifted by X milliseconds from what you actually play... It all affects the performance and the end results. That's why for top studio gear you pay a LOT for low latency.

Although, in my favorite role of Advocatus Diaboli, I always point out: tracks recorded with latency can always be shifted in ProTools by the latency amount, while for the monitoring you can always do hardware monitoring of the musician, which has a latency close to zero Still, because of the various setups in the studio, low latency IS the holy grail of modern studios.

But the latency difference between two gigabit switches is going to be a few nanosecond clock cycles.  This is totally irrelevant in the context of millisecond delays in the rest of the system, or even the performer moving their guitar by a fraction of a millimeter.  However, I can understand why labeling a switch as "low latency" might help to sell it.

That could be very true - marketing talk. When you say "low latency", studio crows listens... But in case of a network device it could be meaningless at a practical level. This switch also promises "Network-wide clock and sync accurate to nanoseconds." If and how it matters to a recording setup - I have no idea. I know I'm clock-synchronizing 4 devices in my rack using 75Ohm coax wordclock cable with T-connectors and terminating resistors on the ends (like in the dawn of computer networking age) and it does matter to me that the clocks stay synchronized, else I get clicks on recorded audio tracks when the synchronization "misses" one master clock tick. What's the benefits of a nanosecond accuracy of an ethernet switch - again, no idea, my studio setup is using the lowly FireWire, not ethernet...

All I wanted to address with this post is that I have seen all over hi-fi fora the repeated question: "is there an audiophile ethernet swich, since we have autiophile ethernet cables" and I think I discovered the first such device that has been engineered for audio applications. If somebody has a bad case of upgraditis and has just run his last meter of AudioQuest Vodka to his Naim streamer, he might be interested in this as a the next logical step

It is not latency that is important but the fact that high quality, high bit-rate switches contain of necessity longer buffers. This means that they can handle real-time traffic better without re-sends or contention problems.

Cheap inexpensive switches use small buffers (cheaper) and rely on TCP/IP re-send when there are any problems - this is very bad for real-time data.

In my previous job I had to design IP Ultra-HiDef Videoconference systems than really gobbled-up bandwidth - only Gigabit professional switches allowed them to work without errors. Even smaller Web-Cam based systems did not like some of the 100 Meg switches.

So a good switch is important - but even more important is to architecture your system so that you have a dedicated switch fed from the main house router and that switch is only used for your HiFi streaming traffic. If you segment your system like this then even a smaller lower quality switch seems to perform very well, as you don't trip it up and it is happy doing a simple job.

Just some thoughts.

DB.

How does this manifest with relation to hifi?

But we're not talking Ultra HiDef Videoconferencing - we are talking about audio data.  Even a 192/24 bitstream is only about 9Mbits/s, so there is stacks of headroom in a gigabit network.

+1, my switch cost less than 20 quid and I put music and hidef video through it all the time wiyhout any sign of a problem.  You can over think this stuff sometimes

You are right. I was referring to where such switches actually do make a difference - and we found it was the additional IO buffer capacity that helped.

For Audio, as long as you segment the network then lower specification switches are fine, as I said. If you don't segment it and just plug everything into a single switch (or just use the base router) then it needs to be much better, as it is not then just dealing with Audio but bursts of other data.

When I tried an NDS at home I found an inexpensive 100 Meg router worked fine when it was only dealing with a NAS, UnitiServ and the NDS; the switch was in turn connected to the router and all my other non-HiFi computer stuff worked to the router. Seamless performance with no problems at all for the week I had it on demo.

DB.

True, but overthinking these things can be fun and that's what hi-end audio enthusiast fora like this one have been created for

Can't argue with that Gajdzin 

I'm sure a $1 set of RCA cables will also deliver the music from your Naim CD to your Naim amplifier without any sign of a problem. Yet most of us prefer a little better cables, up to and including the Hi-Line 

If the MOTU switch interests anyone here it's not because the standard, inexpensive gigabit switches cause any obvious problems - it's because we (or at least some of us) are trying to squeeze every last bit of sonic performance from our system.

I'll take your word for it my friend.  I can accept speaker cables and interconnects, am willing to try ethernet cables but I draw the line at a switch. If it can reliably shift hires 3d video then music isn't a problem.  If build quality and electrical shielding was the argument I could at least see how that might be possible but that's about it.  But if you enjoy the discussion,  believe you can hear a difference or just like to experiment,  good luck to you, you may well be right

Actually getting pedantic and technical, low latency switches are not always desirable unless the physical data signal is of a guaranteed high quality. Most quality  managed switches allow you to sacrifice latency for improved reliability. Essentially you can tell your switch to receive more of the frame before onward processing and switching. This prevents bad or errored framed travelling across your network to be later discarded. However to receive more of the frame requires time, and therefore increased latency.

Speaking from a professional perspective, the best switch for the job will entirely depend on the network configuration and applications you are using. For a simple home network with little or no DSCP/QoS, none specific multicast groups, no monitoring or SNMP requirements, no bonding/EtherChannel, no Spanning Tree or other topology management features and simple flat LANS (no VLANS) then I suspect a bog standard Netgear consumer device will be fine.

The quality of audio on connected equipment I am sure will most be driven by conducted RFI and noise, and so chokes and good quality SMPS/Linear PSU will no doubt make an audible difference.

Simon

Another thing forums were created for was Votes, and we haven't seen one of those for a while. So.....all those who think Gajdzin should get his wallet out, buy one of these MOTU thingies, and tell us if it sounds any good, vote now 

If voting for oneself is allowed, I vote NO

Unless the forum can engage in crowdsourcing and make a me a beta tester