Bought a Dlink today. Looks exactly like the Aqvox switch minus the expensive goo. Has two onboard switchers for 3.3v and 1.1v. Apparently none of the modders consider removing the switchers important.
Due to the high price of OCXO/TCXO i would like to first get some sonic incentive. I am also unconvinced that changing just the oscillator in the switch without also changing the oscillator on the ethernet card will have any effect.
Despite this reservation will make a simple experiment: a little board with a cleanish 3v3 supply and a Crystek 957 24.576MHz. Looking at modded switches it appears some are indeed using audio frequency oscillators instead of proper 25MHz.
It would be nice if the switch could also incorporate an Ethernet filter but i have no idea about the RC values. Suggestions?
It is also easy to cut off all but two pairs to a particular port and force it to 10Mbit operation. This will allow for easy comparison against the unmolested ports.
Due to the high price of OCXO/TCXO i would like to first get some sonic incentive. I am also unconvinced that changing just the oscillator in the switch without also changing the oscillator on the ethernet card will have any effect.
Despite this reservation will make a simple experiment: a little board with a cleanish 3v3 supply and a Crystek 957 24.576MHz. Looking at modded switches it appears some are indeed using audio frequency oscillators instead of proper 25MHz.
It would be nice if the switch could also incorporate an Ethernet filter but i have no idea about the RC values. Suggestions?
It is also easy to cut off all but two pairs to a particular port and force it to 10Mbit operation. This will allow for easy comparison against the unmolested ports.
Attachments
This is clearly wrong: the difference is too big to make any sense. So, this experiment, unless i buy a useless for audio 25MHz OCXO just won't happen.
Hello sa,
Sounds like the switchers have not been replaced, but I would say upgraded on the AQVOX. Looks like a diode has been implemented on D5 and D6. This is an non implemented option by Dlink, even if the PCB is ready for it.
I cannot read the ref of the lite-on semi switcher, but looks like a Schottky Diode has been installed between SW pin and GND pin of the Buck Converter. Don't know if the adjacent selfs and capacitors have been upgraded. Could you please provide the ref of the switchers? Check the LSP5526 data sheet for the diode option.
Not sure what are you asking me.
You are certainly right about the diodes, only there is no way to know if those were not on the boards Aqvox used in the first place. Too many board revisions to know for certain. And is this not some trivial minutiae?
The 3v3 reg is easy to replace with linear but i don't have anything suitable for 1v1 in my drawers.
Generally think this is clutching at straws.
I was refering to the PN of the switchers, which is confirmed to be LSP5526 from LITEON.
Regarding the diodes, if you look at your Dlink board D5 and D6 are empty places where optional diodes could be installed. Looks like AQVox soldered a component at this place (look at first page picture), I assume it is a shottcky diode as advised as optional for LSP5526 LITEON Synchronous Buck Converter datasheet.
Regarding the diodes, if you look at your Dlink board D5 and D6 are empty places where optional diodes could be installed. Looks like AQVox soldered a component at this place (look at first page picture), I assume it is a shottcky diode as advised as optional for LSP5526 LITEON Synchronous Buck Converter datasheet.
The regular Dlink do not have the component in D5 or D6.
It is for sure a modification made by Aqvox. If Dlink would implement this, they would have put a B130 SURFACE MOUNT SCHOTTKY BARRIER RECTIFIERS as advised by LITEON scheme. (done automatically by a SMD machine). If you look at the picture from AQvox, the component in D5, D6 has been added to the board as an add on.
I have no clue if the add on component used by AQVOx on the LSP5526 is a Diode or not. It is hidden with black resin. I’m talking avour diode since it is an optional scheme advised by Liteon to improve the device performance
"
Optional Schottky Diode
During the transition between high-side switch and low-side switch, the body diode of the low side power MOSFET conducts the inductor current. The forward voltage of this body diode is high. An optional Schottky diode may be paralleled between the SW pin and GND pin to improve overall efficiency. Table 3 lists example Schottky diodes and their Manufacturers. »
Does this help in leakage current management? I dont know, Would current leakage blockage be helped by a capacitor or additionnal self at this place in the circuit, I dont know neither. This is where some electronic experts could provide some good advise. Blocking current leakage should help in sonic performance.
Attached is the recommended circuit application with optional diode.
This is by comparing regular Dlink and AQvox picture boards that I have seen the physical differences. This is investigation phase
Attachments
Ethernet is a term that covers all the 802.xx working group protocols.
As a standard TCP does indeed guarantee that a packet will be delivered. UDP is like throwing a dart. Even if a vendor uses UDP for the lower overhead they can code their own reliability transport upper layer.
It's why, from what I have seen, all video/audio streaming services employ TCP.
Audiophile switches
Correct. This is part of my daily work at a Uni. A good piece for everyone here to read is here AES67 - Wikipedia in conjunction with Real-time Transport Protocol - Wikipedia. Just know that reading it does not make you an expert, even an arm-chair one. This is a complex protocal/system that even some of the best engineers in the world wrestle with.
As far as UTP cable is concerned, definitely stay away from the cheap Eastern brands. We have had people bring in their own cheap cables to work and endless issues abounded. The RJ45 connectors are sh1te quality and even the cable is not well made/twisted etc. We have insisted upon a single manufacturer only, from the cabinet panel/keystone, horizontal cable and outlet-box/keystone and patch-cables & flyleads. Very few problems, but normally due to ID10T codes. Users think cables are excellent tow ropes or bungi cords.
All of our switches are enterprise level what with FM/HR systems, servers (SANS research etc.) and a large Wi-Fi presence throughout our 5 campuses.
On the last bit (Wi-Fi,) we have had to ensure that they are on seperate switches and VLANs and the up-links to these are a minimum of 1Gb/s upto 8 APs, but otherwise 10Gb/s. Initially they were installed on the user VLANs but once the APs were inundated with users (20 000+ students) the throughput levels was impacting the QoS on the VoIP phones; i.e. the OS/firmware and ASICs and onboard RAM were overtaxed.
This however will definitely not be so with these 8-port switches "smart hubs" as they definitely are not in the league of 48-port ER $4500 switches.
I agree with an earlier comment to direct link the streamer and DAC, i.e. KISS. The less connections the better. And make sure that you use quality Cat 5/5e or Cat6 cable. I always recommend quality factory moulded terminations from a recognised brand such as Molex/Panduit/Krone/Meganet etc.
Lastly, if you have to use a switch, make sure it is earthed to the same point as the DAC - just makes sense (electically that is) - make sure only audio devices are connected (no Wi-Fi or general use PCs, even routers) and that only one stream is active at any given time. Those little ASICs have tiny buffers and limited processing capabilities.
Correct. This is part of my daily work at a Uni. A good piece for everyone here to read is here AES67 - Wikipedia in conjunction with Real-time Transport Protocol - Wikipedia. Just know that reading it does not make you an expert, even an arm-chair one. This is a complex protocal/system that even some of the best engineers in the world wrestle with.
As far as UTP cable is concerned, definitely stay away from the cheap Eastern brands. We have had people bring in their own cheap cables to work and endless issues abounded. The RJ45 connectors are sh1te quality and even the cable is not well made/twisted etc. We have insisted upon a single manufacturer only, from the cabinet panel/keystone, horizontal cable and outlet-box/keystone and patch-cables & flyleads. Very few problems, but normally due to ID10T codes. Users think cables are excellent tow ropes or bungi cords.
All of our switches are enterprise level what with FM/HR systems, servers (SANS research etc.) and a large Wi-Fi presence throughout our 5 campuses.
On the last bit (Wi-Fi,) we have had to ensure that they are on seperate switches and VLANs and the up-links to these are a minimum of 1Gb/s upto 8 APs, but otherwise 10Gb/s. Initially they were installed on the user VLANs but once the APs were inundated with users (20 000+ students) the throughput levels was impacting the QoS on the VoIP phones; i.e. the OS/firmware and ASICs and onboard RAM were overtaxed.
This however will definitely not be so with these 8-port switches "smart hubs" as they definitely are not in the league of 48-port ER $4500 switches.
I agree with an earlier comment to direct link the streamer and DAC, i.e. KISS. The less connections the better. And make sure that you use quality Cat 5/5e or Cat6 cable. I always recommend quality factory moulded terminations from a recognised brand such as Molex/Panduit/Krone/Meganet etc.
Lastly, if you have to use a switch, make sure it is earthed to the same point as the DAC - just makes sense (electically that is) - make sure only audio devices are connected (no Wi-Fi or general use PCs, even routers) and that only one stream is active at any given time. Those little ASICs have tiny buffers and limited processing capabilities.
Last edited:
This isn't a forum for VoiP. I've done everything from AES67, to AVP, to SiP as a 15 year CCNP.
Playback of audio from file isn't real-time.
For VoiP you need 150ms or less round trip. For AVB? 2ms.
From a switch configuration standpoint AES67 is cookie cutter once you have the configs worked out and you are smart about your layer 2 diameter and vlans. With the Nexus switches we just go full 10GBe SR. Also I like using IPv6 for both AES67 and AVB since we don't have to deal with any broadcasts.
I work in an environment where one of the preferred cable vendors keeps using no name patch cable. Very stiff. In brand new installs for 170 drops on average I keep seeing FCS, CRC, and other errors, and we aren't talking just a few, I mean thousands. One drop so bad it was 13,000+ FCS errors in a matter of days. I've pointed this out, in email, every time it happens. Nothing has changed. But somehow I get the call with the tone that it's my fault. That's ok because I go out and get to bill full rate, come back with the bad patch cable, the switch stats, and let my boss handle it.
My suggestion: Get Tripp-lite or Panduit. Affordable, built well, exceed spec. No need for audiophile ethernet cables either. They simply can't make a difference. Same goes for switches.
What genius thought it was a good idea not to put campus wifi on it's own vlan? As long as your backplane bandwidth is there you can have chassis with wired and wireless. We do that all the time and use multi-rate 100/1000/2500/5000/10000 POE for Wifi with no issues.
You can't get ground loop as I think your post is suggesting. UTP cables have no ground to make a loop with. You should actually put the switch on another circuit breaker.
I also have to disagree with the tiny buffers. Tiny for what? Audio? Audio is nothing for even the $80 Netgears of the world. You can get full wire speed on these devices at ~107MB/s. That's like ~120 concurrent 24/192 tracks.
If you are really worried just run MM fiber SR with LC connections. You can get 10GBe SFP+ SR MM LC modules at FS.com for $17. You can get a 16 port Mikrotik SFP+ switch/router for $349 go all fiber on your direct endpoints and only use copper where you can't do anything else. They have a 9 port version (8 SFP+ and 1GBe copper) for $245.
<You can't get ground loop as I think your post is suggesting. UTP cables have no ground to make a loop with. You should actually put the switch on another circuit breaker.>
Hi Jinjuku,
No, I am definitely not suggesting anything about a ground-loop, just that it is always a good policy to electrically ground the metal chassis of any equipment.
<What genius thought it was a good idea not to put campus wifi on it's own vlan?>
That was forced upon us techies by upper management despite our protests. They wanted immediate results, and could not wait for additional equipment to be ordered and implemented. But what do we know, we're only technicians and engineers, not very bright like the Doctor and Professor Academics at all - hear the sarcasm in my voice.
Hi Jinjuku,
No, I am definitely not suggesting anything about a ground-loop, just that it is always a good policy to electrically ground the metal chassis of any equipment.
<What genius thought it was a good idea not to put campus wifi on it's own vlan?>
That was forced upon us techies by upper management despite our protests. They wanted immediate results, and could not wait for additional equipment to be ordered and implemented. But what do we know, we're only technicians and engineers, not very bright like the Doctor and Professor Academics at all - hear the sarcasm in my voice.
Last edited:
Improve Ethernet data?
The data is transmitted to the streamer by the TCP protocol... not morse code. If a package (chunk of data) is corrupt, missing or received in the wrong order, it is resent or just arranged in the right order.
Every single packet is recieved by the buffer in the streaming device BIT PERFECT!
How can you possible improve a perfect copy?
Just because streaming something over the internet can be lossy with distortion and blocky picture, does not mean its the same when streaming at home.
These transmission protocols are using UDP not TCP.
And even if that switch is used for a multocast UDP stream with packet loss, how is it supposed to improve data that no longer exists?
Use $50.000 on cables and at least half that on other tweaks. Just dont waste money on things like that. You are making us all look incredible stupid.
If you are a HiFi enthusiast running media servers, streamers etc. this is one important part of if you want to learn. Learn how to DIY the right way.
What’s The Difference Between The OSI Seven-Layer Network Model And TCP/IP? | Electronic Design
Here is a more in depth article. Lots of other related stuff here worth a read:
What is OSI model (Open Systems Interconnection)? - Definition
Hey George: Are you saying switches are/are not important (to sound quality)?
pj
As a R&S and Voice CCIE, I can guarantee they are not. You either have sound or you don't.
Oh, Really ?
I was afraid you may say that. Ultimately, the depth of this topic goes much further than 'electronics', and what is currently understood (or not) to contribute to observable differences in SQ.
Before we engage in a discussion, let me say that my minimal (4/5 years) experience with computer audio mirrors that of my general (but thorough) 40 years of experience concerning all other sources/aspects of hi-if -as it relates to clear, easy-to-discern sound quality variations by nothing more than a simple change of a single component/cable.
And then, there are those (such as yourself) -of course as a "R&S/V-CCIE"- that can guarantee no differences (audible) result!
GUARANTEED -no less. How in the world can one "guarantee" such a thing when my own Ethernet Switch change/swap (the other day no less) yielded dramatic changes in perceived SQ/tonality (Sound Quality) ? Me, and thousands of others. There is clearly a divide here. Can you help explain?
pj
I was afraid you may say that. Ultimately, the depth of this topic goes much further than 'electronics', and what is currently understood (or not) to contribute to observable differences in SQ.
Before we engage in a discussion, let me say that my minimal (4/5 years) experience with computer audio mirrors that of my general (but thorough) 40 years of experience concerning all other sources/aspects of hi-if -as it relates to clear, easy-to-discern sound quality variations by nothing more than a simple change of a single component/cable.
And then, there are those (such as yourself) -of course as a "R&S/V-CCIE"- that can guarantee no differences (audible) result!
GUARANTEED -no less. How in the world can one "guarantee" such a thing when my own Ethernet Switch change/swap (the other day no less) yielded dramatic changes in perceived SQ/tonality (Sound Quality) ? Me, and thousands of others. There is clearly a divide here. Can you help explain?
pj
Something measuring well and sounding poor, or conversely, something measuring poorly and sounding beautiful, is nothing new in audio. Some things that should make a difference don't, and other things that (in theory) cant make a difference do.
What you perceive and what you enjoy is the ultimate arbiter of what makes you happy.
Audio happens in between your ears.
What you perceive and what you enjoy is the ultimate arbiter of what makes you happy.
Audio happens in between your ears.
Hands down, this is the most absurd thread I have ever seen on this forum. At first I assumed it was a joke. It saddens me a bit to know that there are people who are so determined to be wilfully ignorant.
The concept of an audiophile Ethernet switch is truly and absolutely ridiculous. Anyone who claims differently shows a complete lack of understanding.
As someone has already stated, this is snake oil of the highest order.
The concept of an audiophile Ethernet switch is truly and absolutely ridiculous. Anyone who claims differently shows a complete lack of understanding.
As someone has already stated, this is snake oil of the highest order.
- Status
- Not open for further replies.