Pauper Switch - no audiophilia nervosa

Any updated Mike?

Not yet Adam,  its just playing  iRadio this evening with background music for Santa's very busy little helpers prep'ing/cooking & wrapping goodies.   It is a slightly different sound  but I (we) need time to dig deeper,  probably tomorrow.    But that said it is a new switch thats been changed for purposes other than SQ,  I just thought it might be interesting to check that element out in line with the "Posh" switch thread.    

Ok Mike - standing by for more info. Interesting to read that: 'slightly different sound' is already apparent.

A few things from last evening:  It was playing iRadio OK all day, but late last evening I noticed the NDX display screen was not correct, iRadio just displayed the preset channel number & not the preset title & it flashed that up for just a few seconds & not for the set 30 seconds. I suspected I had not powered up the switch & system sequence correctly & it had not 'learned' the connected equipment correctly. I powered off everything then re-powered the switch,  wireless 'router' (wait & let it finish) , NAS & NDX; problem solved. Also found some iRadio preset display names had changed slightly, but surely that has to be something outside the scope of the switch, new app maybe & I had not previously noticed it. Anyway a bit of editing on the app & its OK.

After two days on iRadio only,   I have a definite impression the sound is different, it is subtle but I'm pretty convinced.    Comparing iRadio to FM the difference now seems more pronounced; the dynamics & the top end detail appear stronger (as has always been the case with high kb/s iRadio) the new switch seems to show that up a bit more. That said, FM does still sound more natural on BBC R3 with its low/no compression,  not so with the more compressed stations.

Moving to NAS, as with iRadio, I hear more detail, it sounds like a touch lighter, sound stage is more obvious, hard to say what in the bass region, not more or deeper but it does carry more feeling of texture & timbre.   

To round it off, I would have liked to have gone with the “Posh” Cisco Catalyst, but I really don't have the room without major disruption. Its been interesting to hear this little Cisco “baby” has made a small difference, & it is different as it demonstrated when it was not correctly 'introduced' to the other equipment on the net, I suspect that's something in the Cisco QoS tag program. Bottom line is it works & another small form SOHO switch to add to the recommended list.

Mike - similar to my experience. I run a small streaming setup in my office to the obvious joy of my staff

Small CISCO proved to be very reliable and stable with good sound quality. 

I spent all of last evening listening to NAS albums,  (feeling rather smug 'cause all the Chrissie stuff is done, wrapped & packed)        This has been the first time I've actually listened  (the other sessions were more like 'testing').    It confirmed all of my previous impressions, its subtle,  but I hear the same on UPnP (NAS), iRadio & TV sound;   more detail with a lighter feel in the top end, sound stage is definitely more obvious & bass has more detail & texture,  especially most noticeable when comparing iRadio to FM.   

The iPad app screen is not any faster as I previously suspected,  besides its now a different app so nothing constant to compare.   Loading some new albums yesterday I tried it wireless from laptop to router (wired to NAS)  & all fully wired from laptop.  It adds a touch more speed but is most noticeable in that the speed (Windows graphic) is more steady & constant. 

Whats doing this ???  if you believe bits-r-bits then I'm hearing things & talking nonsense,   the only real difference is the chip memory going from 128KB to 128MB RAM & 128MB Flash.  Apart from that it has a different QoS process thats coupled to the new Cisco cable diagnostics & port management system that the Netgear didn't have,  but I don't understand how that can change sonics.    Whatever its a keeper.    

Meanwhile I'm off for the holidays,  being fed & watered for a few days.   

Mike, you mention QoS tags, what is being tagged and what are you thinking if? I have not seen any 802.1p flags used in home streaming, and by default one would typically need to set the switch up to trust and act on such flags as opposed to simply transparently passing through or manually assign prioritisation to specific physical ports... I really think 802.1p flags on the switch or even DSCP flags at the router have no real effect here and are not normally used on home networks except occasionally for VoIP.

Their effect would be observable if used  when a port or switch back plane becomes heavily  loaded with different conflicting services and applications, or one was using low latency multicast for VoIP etc.

interested on your thoughts what you think is happening..

S

Simon,  I really don't know,  all I know is it has a different/better sound,  its a simple little consumer grade switch,  I was not expecting a change & was just musing what's different.  The most obvious is the memory is bigger/different,    but what else ??   can't be QoS as you have pointed out in your post,  not in home streaming.   It did have a different reaction (compared to Netgear) when I did not 'introduce' it to the other system components correctly.  The only other difference(s) that I can see are the new (incl. new to Cisco) cable diagnostics & cable + port power management.         

Mike, I do think the majority of what we hear as change to sound (I'll avoid the word 'benefit') is down to physical layer implantation, sometimes called PHY in engineering docs. This is the part of the implementation that has the basic ADC and DACs as well as the synchronisation clocks. It is also the part will couple noise from the powersupply onto the physical Ethernet twisted pairs. I do believe a switch that is specifically designed to support precision time protocol is going to have a better physical layer, as the PTC interface occurs between the PHY and TCP/IP or digital stack. The 2960 switches do actively support PTP.. 

if I am honest to do the same type analysis like I have done on SQ differences between media servers for the switch  PHY interface jitter I would need a suitably capable spectrum analyzer.. and I can't currently get my mitts on one.... so this for the time being has to be speculation.

Simon

Simon,  you have sort of summarized where my musings are going.   The previous Netgear switch had a Broadcom BCM5397 chipset,  I have no idea what the new Cisco 110 series has,  but doubt its a Broadcom.  Whatever, therein lie the changes I suspect,  does the same job but does it differently.  

I changed PS of GS 108 with a posh PS ( ifi psu ) and as you'd guess there is no beneficial addition to sq. 50 bucks went outta window. What was I thinking.. 

HiFi Nerd Question Warning  (I must get out more, but it is p-p-ersisting rain)

Playing around with beta testing, I noticed my new Cisco SG110D switch has caused a different reaction on the NAS LED indicators. Curious I've swapped switches back to Netgear GS105 to make sure I was not imagining it.    If any knowledgeable forum'ites know why this is so, or have a theory,   I'm all ears.

Synology DS214 has four LED's  - Its the bottom two of the four that are showing a difference in the variable flash/blink rates. This is during the album play period,  not at the track change period when they all flash.       #3 'Green' when blinking means “Disk is being accessed”,  with Cisco the blinks are not regular but its about once every 10 seconds.       #4 'Green' when blinking means “Copying data”  & with Cisco it blinks very infrequently.       With the Netgear this LED pair do a lot more blinking,  #3 maybe 2 or 3 times every 10 seconds.   

I will get out more, I promise. 

Mike   without setting up a network trace one doesn't really know what is happening for sure. However there are two distinctive types of 'pattern' I have seen from the Naim streamers in how they manage the data transfer - and this appears dependent on the throughput of the data from media server or on online server such as Tidal.

Now it just might be for what ever reason there is a difference in the effective throughput with your two different switches - which is prompting the streamer to change how it manages the reception of the media transfer - and it it might be the more efficient higher speed transfer causes the caching/buffering of the NAS to change how its reads data of the disk and it in bursts but less frequently...

As I say only speculation without looking on the wire - but there are certainly two quite distinct media transfer styles... and yes I find the more efficient / bursty network transfer style tends to sound better

Simon

Simon,  thanks for posting, I was not expecting a definitive answer,  but ya never know.   Your post is along the same speculation track as myself.  

The only difference I can see in the data/spec sheet between the two switches is memory.  Netgear is "128KB on-chip packet buffering ... & ... Queue buffer memory 12 kbytes per port"   Cisco is "RAM 128MB  ........ Flash 128MB"   

Other (what might be)  relevant data is reported differently,   Netgear use Mbps,  Cisco Mpps.   One or two lines however do show packet forward rates  .......  (on 1000Mbs port)   Netgear 1,480,000 packets/sec     Cisco 7.4Mpps (7,400,000 packets/sec)

I agree with you Mike - you need to get out more 

Hi Mike -

The packet forwarding rates are essentially the switch backplane bandwidth - that is the switch throughput - on switches that are not to become a bottleneck - called non blocking - the backplane bandwidth (or packet forwarding rate)  should be greater or equal than twice the bandwidth of each port.

Now if you Netgear is gigabit and has 5 ports - then its backplane bandwidth required for it not to potentially block is:

5 (ports) x2 (send+recv) x1,000,000,000(port bandwidth in bps) /8 (turn into Bytes) / 1,526 (to turn into packets) gives a non blocking bandwidth of 819,135 pps for full size ethernet encapsulated packets. Now packets can be smaller sizes - and a lot of control packets and non data transfer packets will be smaller - so if the packet frame size averages say 500 Bytes as opposed to 1526 Bytes - the non blocking bandwidth rises to 2,500,000 pps - and so here the Netgear will discard data... but I would say 1,400 is going to be ample for the vast majority of time for home networks.

The Cisco SG110D has 8 ports - so its minimum non blocking bandwidth for 1526 Byte frames is 1,310,616 pps and for 500 Byte frames this rises to 4,000,000 pps, which is still well within its bandwidth and so we can see the Cisco is more comfortable forwarding smaller frames than the Netgear before data is discarded...

But even here I suspect most home networks just will get no where near here at all........

BTW flash memory on the Cisco switch is used for running its firmware which may be upgradable - just like the devices managed cousins.

Anyway interesting distraction - and yes its still raining here.....  - I would still need to compare the traces on Wireshark with a media transfer to see what the real world differences are - and I thinks its unlikely to be backplane bandwidth.

S

Thanks Simon,  something to get my head around when time permits.  

Minor NB:  My SG110N-05 & is the 5 port.    +++   Re my post ".......... data is reported differently.......... do show packet forward rates  .......  (on 1000Mbs port)   Netgear 1,480,000 packets/sec     Cisco 7.4Mpps (7,400,000 packets/sec)    

These numbers are identical when expressed as pps per port:   Netgear show pps per port.  Cisco show the all ports total pps for the switch  .......    7,400,000 / 5 = 1,480,000

ahh thanks for the port count correction on the Cisco device - hmm I wonder if the same underlying chipset is used..... anyway in this case they both appear equal to each other with smaller packets... so curious to why the different network dynamics - do you know how to use WireShark?

Simon,  I believe it's a different chipset,  Netgear's GS 'Prosafe' units (my old one) use Broadcom BCM5397.   Cisco are a bit less forthcoming;  there's no info anywhere that I can find.  I know they have used Broadcom in the past,  plus others (Intel etc).  It seems since the Cisco/Linksys split it looks like some new product designs are in-house.    Aside from the variable spec/data sheets,  the obvious big difference between them is on-chip memory 128KB to 128MB.

I've used WireShark in the past,  I had a bespoke set up on my field team laptops to interrogate my cmpy's satellite comms systems,  but its been a while since I've used it.