There are some discussions about 1 gbps vs 100mbps Ethernet connection speeds for streaming. Any experience or opinions on full duplex vs half duplex network connection on sound quality, noise, etc.?
Depends on the gear and who you talk to. if you go back 3-5 years and look at the You Tubers who have something to say on the subject and review the specialist network gear that came out - a lot of the focus was on the merits of 100Mb/s networks (4 wires) as opposed to 1Gb/s networks (8 wires). Today the likes of Network Acoustics in the UK are leading the way in making a very strong case for the merits of 1Gb/s networks with their specialist cables and switches and network filters. Vinshine Audio, Melco, Silent Angel, English Electric and others have similar offerings - often offering both 100Mb/s and 1GB/s ports in their Chassis. For me there is no reason these days to go with 100Mb/s unless your streamer or integrated amp supports only that. Not seen any discussion around full duplex or half duplex - that feels like a red herring, as opposed to the speed and technology to power 100Mb/s or 1GB/s networks
1 Like
Modern switches default to full duplex, even when running at 100mbps. One would have to intentionally misconfigure it to cause it to do half duplex, except in the case where a very old 10mbps device is connected as that was designed to run at half duplex. I have also not seen any mention of intentionally misconfiguring a switch at half duplex to achieve better sound quality. But this is a crazy hobby so I guess I wouldn’t be surprised of someone is recommending this.
1 Like
@bladeraptor Many thanks for your input. I am not disagreeing, but lots of people did, and still do, think that the whole topic of network settings, network speed, switches and cables being able to influence the sound is a red herring.
Lots of people used to think the whole network-settings/switches/cables thing was a red herring. But those usually aren’t the folks chasing the best streaming performance. You won’t find many of them on the Antipodes forum. Most people here have heard first-hand how much the network can influence the sound.
2 Likes
The main switch here, a Waversa Smarthub V3, can be configured for 1G, 100Mb operation, including full and half duplex operation for both inbound and outbound sockets. All settings have a different sound, and for me the optimal is 100Mb half duplex both inbound and outbound. Also running Waversa Ref+ lan filter and Vertere Pulse HB ethernet cables.
This is not to say these setting will be best for everyone, but the beauty of that switch is that it is all user configurable on-the-fly and you can figure the optimal solution yourself.
3 Likes
I would point them to the excellent series of networking experiments carried out by the Greek Audiophile and the liked of Hans Beekhuyzen on his channel where he goes into a lot of depth around the influence of networking on digital audio. Some of it does seem to come down to things not necessarily sounding better or worse but perhaps different. Alpha Audio in the Netherlands have actually built a rig to test the impact of things like power cables, supplies and network switches on the sound delivered by digital sources. Worth checking out some of their comparisons
2 Likes
Many thanks for all your input. I too have experienced firsthand differences in sound between network settings, and have read some of the comments out there, hence the original question. Will continue to experiment.
1 Like
Most end user traffic is going to be one way. So you normally have two unidirectional paths if you really want to get into the nuts and bolts.
Then you also get into the possibility of symmetric vs asymmetric pathing.
I would take with a large grain of salt sound improvements with 100 1/2 duplex. Modern switching is creating dedicated Layer 2 bridging with their ASICs and software stack.
Here is food for thought: Native Tidal app will buffer entire tracks. I have a 300Mpbs Internet. A typical length Redbook encoded track will que up from a Tidal CDN in roughly 2.5 seconds. After that I can pull out the Ethernet cable and I still have 5-6 minutes of the song playing back.
I’ve never been able to hear when either the cable is pulled or the system put into airplane mode thus disabling the wifi.
I could make the argument that the higher in speed you go the better:
- Lower jitter
- Faster data xfer which means the tx/rx phy’s spend more time idle(nothing on the line).
I’m a Cisco Certified Network Pro X3 (Enterprise, Wireless, Data Center), Fortigate NSE 4-7, Palo Alto PCNSE, HPE Aruba ACEP (Switch/Route, Mobility, Design), Cradle Point Engineer.
I also spent 7 years doing Broadcast Television and Radio ( live, mastering/post production).
1 Like
Thanks. Putting aside data transmission and data integrity, what about difference in noise between 100mbps/gigabits, and half/full duplex?
Not sure what people are really getting sometimes. Always seems like a moving target. The sound signature issue is multi-component: You have the sync clock and then you have the encoding schemes.
Then you have, what I think most people are driving at: Layer 1 issues that have to do with power: Either the AC switch mode or the DC converted.
So here is the thing with speeds: Speeds come along with higher clocks. The higher the clock, the farther away it gets from human hearing and the related harmonics.
100Mbps is 31.25 MHz
1Gbe is 62.6 MHz
10Gbe is 400Mhz
So not only is your frequency getting higher your data dwell time on the wire is getting shorter.
My local storage is a TrueNAS setup with SMB/NFS/iSCSI with 24TB in a RAID 5 with a Solar Flare 10GBE nic going out to an Aruba 3810 SFP+ switch.
It’s all spinner storage and I get sustained 600Mbps. (5x 1Gbps). I can literally, in JRiver, buffer an entire Album in less than two seconds and play back with the cable unplugged for 50 minutes.
My work lab is a 5 node Proxmox Cluster with High Endurance WD NVME PCIe 4.0 CEPH with a Cisco Spine Leaf using QSFP+.
I get sustained 4000Mbps. That means if I had a play back machine with a QSFP+ adapter I could queue up an entire redbook CD in about 1/6th of a second.
And then in both instances there is nothing on the line besides maintenance traffic, most likely discovery protocol(s). And those are on a standard hello and hold-down timers regardless of speed and duplex.
Technically the higher the speed you go the less dwell time you have on the line.
Lower speeds are simply counter-productive in my opinion. Because at 10Gbe, which is incredibly affordable, we are talking transfers in the mil-seconds for a track. After that what’s actually on the line that isn’t going to be there with 10 1/2 duplex?
LLDP for example a tx every 30s and hold-down of 120s. Doesn’t matter if it’s 10,100,1000,10,0000, or 40,0000Mbps.
1 Like
One other point to really consider:
The higher the data rate means your power and tolerances (jitter, phase delay, cross-talk, NXT) all have to be orders of magnitude better.
You actually get a higher quality, almost everything, because higher speeds require it.
My two cents: this is a great question for the guys at Uptone Audio. John Swenson has been professionally involved with the design and development of PHY transceivers. These are what transmit and receive Ethernet at the physical layer. My guess is that they will say that it depends on the PHY more than the rate itself. I suspect that the PHYs used in the EtherRegen Gen 2 will be lower noise than what’s found in most other products, though I haven’t followed closely enough to say this for certain. Alex had learned a ton from John and he is always willing to answer questions so he’s your best bet. Post your questions to one of the Uptone threads on Audiophile Style.
2 Likes
As speed goes up so do the power and tolerances requirement of the PHY.
Uptone Audio has been promising for, I believe 7-8 years, to provide objective audio band measurements of what their PHY does.
Then on the other hand their ‘white paper’ supports my position of just shooting the gap and, if you can, go to 10Gbe optical. Johns suggestions of ‘low jitter’ and ‘isolation’ pratically scream Fiber Optics and 10Gbe.
Alex accepted an offer on Whats Best Forum for another member to come out and admin a blind test of his USB reclocker. Alex initally accepted and then ghosted WBF for 1.5 years immediately after the member asked when they could come out and was ready to purchase airfare and accommodations.
Then there is the Etherregen Firmware controversy:
802.3az is ‘green ethernet’. There was bug in the initial ER that would cause connectivity problems. UpTone releases a firmware update to properly enable it.
Users reported that there was an audible difference. Alex told his own customer base that they were mishearing things. That it couldn’t make a difference.