AUDIO over IP - REDNET 3 & 16 Review. AES67 Sets A New Standard for Computer Audio

[peteAllen]

Question... I want to try out the audio from my oppo bluray player through rednet. My d16 has a spdif input. How can i configure it to switch between my audio server pc/roon and the bluray audio signal.
Also wondering if I can plug the spdif into my audio server and pump it through hqplayer? Any ideas?
Is any of this worth trying?

 

[peteAllen]

Oh, another thing...
I still haven't been able to get the direct connection between pc and d16 configured as 1gb lan. It appears as 10mbps on the pc even though i have tried cat5e cat6 and cat7 cables. Is a switch really necessary to speed it up?
In practice this means that latency often goes up to 10ms meaning I cant stream anything higher than 24/96

 

[Golfnutz]

Oh, another thing...
I still haven't been able to get the direct connection between pc and d16 configured as 1gb lan. It appears as 10mbps on the pc even though i have tried cat5e cat6 and cat7 cables. Is a switch really necessary to speed it up?
In practice this means that latency often goes up to 10ms meaning I cant stream anything higher than 24/96

Shouldn't need a switch. I was connecting my PC to D16 secondary port for awhile before adding a switch. I never had an issue with speed.
 
Edit: Just removed the switch and connected PC to D16 secondary port, it's showing as 1Gb.

 

[rb2013]

  I agree, cover this controversial subject in another thread. It has in any case been talked over so much already. There is the tech specs camp who swear by up sampling but have not even heard a decent R-2R DAC. There are the R-2R camp that are converts from DS, there are others who swear by up sampling PCM to DSD. It goes on and on.
 
My point is, listen to a decent R-2R DAC then make your case. I get fed up with pushing water uphill with those that have no personal experiences (listening) only tech specs.
 
The upsampling aspect is enormous. I can see why the industry went that route as a way to add a filter to 'remove' digital artefacts from the raw 44.1 data stream. The problem as many know resulted in ringing, a digital sound and basically CD replay being worse than the previous format (Vinyl). Remember vinyl tech specs are not perfect yet out sounds better, go figure?
 
I am interested in the current 'chip less' designs appearing. In theory such a FPGA topology could go beyond a chip based design, be bit perfect and avoid the issues of either DS or R-2R designs.
And it could be cheaper than multiple resistor arrays as TotalDAC for example.
 
Also bear in mind, the way a DAC handles the data stream is one point, the design of the power supply and line stage is another possibly even more crucial aspect as after all a DAC is a pre-amplifier with a digital board added in effect. Put in a cheap torriod PS transformer and an Opp amp gain stage and it won't sound so hot.....

Did you not read my post?  You are confusing D-S vs R2R noise shaping by oversampling/upsampling with audio wave form interpolation upsampling like in HQ Player and Foobar/SoX.  Yes they are both 'upsampling' as a process - but the implementations are different.
 
Then you confuse DAC 'oversampling' with 'upsampling' filtering with a NOS implementation.
Many DACs have various filters to choose from.   Very few like the Zanden are NOS (Non-OverSampling).  They used a 16-bit R2R DAC chip called the TDA-1541A.  The infamous Zanden 5000 and 5000 MKII used the closer matched resistors 'Double Crown' version - which had lower distortion.  With this chip you could use it without Nyquist sharp cut off reconstruction filters.   Instead Yamada-San used a analog 'bridged-T fixed-impedance filter'.
 
Just because a DAC is Non-Oversampling - does not mean it does use digital filtering. And this digital Nyquist filtering can cause all kinds of issues.
 
The Audio Note DAC 5 uses the 18-bit R2R AD1865 - which was designed for 16X per channel Oversampling (see the AD1865 datasheet).  Audio Note uses a proprietary filter with just 1X Oversampling.
 
Why is DAC Oversampling important to Nyquist filtering?
 
Great explanation of why DAC's use Nyquist filtering, Aliasing, Brickwall filters and phase shift, Oversampling and Upsampling:
http://www.soundstage.com/gettingtechnical/gettingtechnical200311.htm

 

[astrostar59]

  The Audio Note DAC 5 uses the 18-bit R2R AD1865 - which was designed for 16X per channel Oversampling (see the AD1865 datasheet).  Audio Note uses a proprietary filter with just 1X Oversampling.
 

All Audio Note DACs use the AD1865 in non oversampling mode. Also they have the digital filtering removed i.e. no filter. Any filtering is accomplished in the analogue domain.
 
My point on oversampling/upsampling is, what is achieved by doing that on your PC. Oversampling on a DAC was originally created to allow a better filter to operate with less ringing. Oversampling isn't creating new information or better detail. If as you say it sounds better doing that in HQPlayer, this may be due to a level of softening of the data IMO. Remember if you upsample in anything outside of ratio x 2 or x 4, for example 192 it will be even worse, i.e. resample all the samples and basically blur the data. It is like upsampling an image in photoshop, the sharp squares (pixels in this case) of detail go soft as they bridge the gap to newer and more frequent squares. Try it and see.
 
If you view a track in a music program look at the sample data, then upsample it. It will present a completely new set of samples.
 
My view on this subject is simple, listen to a good non oversampling DAC and then come back. You may be surprised. If not, fine. I found the way after years of cost and searching, buying and selling. It is what this hobby is about. It is impossible for everyone to agree on that destination piece.
 
Having said that, and getting back on topic, I don't see many (if any) claiming USB is better than RedNet.

 

[mourip]

  Maybe the upsampling conversation could move to another or new thread?
 
Joel

+1

 

[astrostar59]

  +1

Agree, but no point IMO. So much has already been said, so many get mad and disagree. Best to forget it.

 

[rb2013]

  All Audio Note DACs use the AD1865 in non oversampling mode. Also they have the digital filtering removed i.e. no filter. Any filtering is accomplished in the analogue domain.
 
My point on oversampling/upsampling is, what is achieved by doing that on your PC. Oversampling on a DAC was originally created to allow a better filter to operate with less ringing. Oversampling isn't creating new information or better detail. If as you say it sounds better doing that in HQPlayer, this may be due to a level of softening of the data IMO. Remember if you upsample in anything outside of ratio x 2 or x 4, for example 192 it will be even worse, i.e. resample all the samples and basically blur the data. It is like upsampling an image in photoshop, the sharp squares (pixels in this case) of detail go soft as they bridge the gap to newer and more frequent squares. Try it and see.
 
If you view a track in a music program look at the sample data, then upsample it. It will present a completely new set of samples.
 
My view on this subject is simple, listen to a good non oversampling DAC and then come back. You may be surprised. If not, fine. I found the way after years of cost and searching, buying and selling. It is what this hobby is about. It is impossible for everyone to agree on that destination piece.
 
Having said that, and getting back on topic, I don't see many (if any) claiming USB is better than RedNet.

Not true - read the Audio Note website: "

1xoversampling™ direct from disc™ technology based on the AD1865N"  And they use a filter - but it's analog not digital.
 

This is not 'no Oversampling'. 
 
Oversampling and Upsampling are different.

In an oversampling system, the input sampling rate (Fs) is increased 8 times, and the output sampling rate is therefore 8 x Fs. This 8 x Fs clock is generated by an 8 x phase-locked loop based on Fs.  So the two clocks are strongly linked and any imperfection appearing on the Fs clock will be reflected in the 8 x Fs clock. With upsampling we use unrelated clocks to drive the input and the output respectively. This means that even if the input clock is imperfect, the output clock will remain as precise as it is."
 
Oversampling, he says, can allow the use of a "lighter output reconstruction filter with all the benefits it brings, but we are not isolated from clock imperfections that would pass on to the 8 x Fs clock." Upsampling, on the other hand, can help to overcome the clock imperfections.

 

Just like on your HDplex using the LT1083 with 180uv of noise.  How many times do I need to prove you wrong?

 

[REXNFX]

 
True, maybe I went on a bit there! Back on thread, playing Josh Gabriel super loud and the system (inc RedNet 3) sounds really amazing. Any last vestiges if treble glaze or digital edge is gone. I proves to me that a lot of grunge was feeding in from a USB chain, regardless of money spent trying to get around that. Basically, any DAC can benefit from the Rednet IMO. Now which DACs sound better, Ha Ha another Everest sized thread!!! And then we have the server optimisations, another super sized thread, and while we are at it throw in the power supply noise (mains). No time to read all this stuff......
 
Quick question (off topic a bit) how many music files are available at true high res (not botched up sampled as some on HDTracks) and where are these websites that sell them. In many cases I can't even find many RedBook files as downloads.... For dance gender BeatPort is excellent, as is Armada, but other genders it is still a void.

Qobuz

 

[jabbr]

http://www.findhdmusic.com/

 

[astrostar59]

 
Not true - read the Audio Note website: "

1xoversampling™ direct from disc™ technology based on the AD1865N"  And they use a filter - but it's analog not digital.
 

This is not 'no Oversampling'. 
 
Oversampling and Upsampling are different.
 

Believe me, Audio Note DACs from 1.1 up have NO OVERSAMPLING or UPSAMPLING. And as I said they use some filtering in the analogue domain.
 
Maybe you want to rewrite their company design facts and philosophy yourself.
 
Email Peter Q the MD and tell him yourself.... reading stuff on the web doesn't make you an expert, sorry.
 
Quote from Peter's article about the DAC 5:

Among the handful of companies whose CD players do not perform either over or upsampling, Audio Note is best known in the high end circle for its steadfast advocacy in SET amplification. From this company that offers audiophiles the $90,000, 27 Wpc Ongaku integrated amplifier, comes a $30,000 DAC with no oversampling, upsampling, or digital filter: the DAC 5 Special.
 

1xOversampling is a registered term Audio Note use for marketing, it means NO Oversampling.

 

[mourip]

  Believe me, Audio Note DACs from 1.1 up have NO OVERSAMPLING or UPSAMPLING. And as I said they use some filtering in the analogue domain.
 
Maybe you want to rewrite their company design facts and philosophy yourself.
 
Email Peter Q the MD and tell him yourself.... reading stuff on the web doesn't make you an expert, sorry.
 
Quote from Peter's article about the DAC 5:

Among the handful of companies whose CD players do not perform either over or upsampling, Audio Note is best known in the high end circle for its steadfast advocacy in SET amplification. From this company that offers audiophiles the $90,000, 27 Wpc Ongaku integrated amplifier, comes a $30,000 DAC with no oversampling, upsampling, or digital filter: the DAC 5 Special.
 

1xOversampling is a registered term Audio Note use for marketing, it means NO Oversampling.

 
"Agree, but no point IMO. So much has already been said, so many get mad and disagree. Best to forget it."
 
How about moving to PM or starting your own thread?

 

[Iving]

  The PC *does* make a difference.
 
(iii) The new machine is a silent (no moving parts) PC comprising:
- Gigabyte GA-X170-EXTREME ECC Thunderbolt™ 3 Certified C236 Chipset Motherboard;
- Intel Core i7 6700 Skylake CPU;
- Corsair 16GB Vengeance LPX DDR4 2400MHz RAM;
- Windows 10 Pro on Samsung SM951 128GB M.2 PCIe NVMe SSD;
- *.flac on Samsung 850 EVO 1TB 2.5" SATA SSD;
- (choice of) Dual LAN (Killer™ E2400 and Intel®);
- fanless "0dB" 400W PSU (being replaced shortly actually because of coil whine).
I am using the Killer vs. Intel NIC because by comparison it has demonstrably lower latency - see http://techreport.com/review/29144/revisiting-the-killer-nic-eight-years-on/2.
 
what we (audiophiles) call PRaT is in a different league altogether. It is as if an adept and eager conductor stands before your artist ensuring tempo; or, as if @rb2013 had returned with yet a newer device having served an internship at NIST (https://www.wired.com/2014/04/nist-atomic-clock/).  
The musical elevation because of enhanced PRaT is very significant.
 
Latency through the Killer NIC is rock fast at <1ms unwavering mean.
 
Thunderbolt 3 ready too ...

 
Quote:

  I have had to swap the PSU from Seasonic 400fl because of coil whine (found on >1 instance) to Aurum AU 500 Gold (very nicely silent)
and
the mobo from Gigabyte GA-X170-EXTREME ECC because of mysterious screeching around the CPU (6700) area [voltage regulation circuitry?] to ASUS Z170 Premium (more expensive but a way to keep Thunderbolt 3). The latter has two different Intel NICs but not the Killer E2400.
 
When I get the machine back I'll report on any changes. I have asserted that an up-to-date competent PC with specs such as these (vs. a basic laptop or a good W10 tablet with Docking Station) raises SQ by as much as going from USB to AOIP.

 
Quote:

  Update: The ASUS Z170 Premium was quiet but had "connectivity" problems with Thunderbolt 3 ("when a thunderbolt device is connected to the system the devices are not receiving power and therefore are not being recognised in Windows" x 2 instances). The engineers are still waiting for a reply from Asus about it. The final replacement is a Gigabyte Z170X Designare which I am told is behaving itself. It has 2 x Intel NICs and 2 x USB C / Thunderbolt 3 ports as well as M.2 and U.2 on board (Windows 10 Pro is on a dedicated M.2). The RAM is upgraded to Dominator 2666 MHz.
 
With a following wind I will have the machine within a day or two. I don't expect ever to use the "teaming" capacity of the two Intel NICs (although I dare say one day I will experiment), but I do hope that the Killer NIC wasn't a necessary ingredient in the stupendous SQ enjoyed previously. The PC I am told is super-quiet now. I am really looking forward to its homecoming.

 
Quote:

 
In case of possible interest to prospective PC buyers or upgraders, two main  points:
- The Gigabyte Z170X Designare exhibits the same mobo ringing phenomenon as the Extreme-ECC - but whereas the latter screeched, the Designare's self-pronouncement is much milder and very probably below a bothersome threshold. We are talking mainly about The High Performance Power Profile (in Control Panel) and, in particular, Minimum Processor State = 100%. The effect is mitigated significantly in the Balanced Profile (Min. Proc. State = 5%) - in which mode the Designare can hardly be heard at all. My wife can't hear the new mobo even in High Performance mode - but she has tinnitus.
- Using the primary NIC; i.e., the integrated I219-V (PCI-Express 3.0 version of the I218-V), the SQ is top notch. At first I wondered whether it was deader than the Killer E2400 (present on the Extreme-ECC), but my system hadn't warmed up. There is nothing missing. If anything, the speakers are more transparent and easier on the ear outside the sweetspot. The tone is liquid. My early undisciplined thinking was that "wall of sound" noises such as Jackson Browne's "Looking East" and Liz Phair's "Turning Japanese" were more distilled (components of the sound distinguishable). Bass thuds. It is very AOIP. Latency averages less than 900µs and peaks at <1msec over extended playing intervals.
It has taken 40 days to establish this PC. That has been a PIA but I am grateful to the building company for staying with me throughout the time it took to get it all fettled. Long and short - I don't doubt the value of the investment in the PC.

 

First, and on the basis that the PC *does* make a difference in AOIP playback, the final instalment in establishing the ideal for PC for me. Whilst I have a separate Windows 7 machine for ripping and library management, my listening pleasure is gained at my coffee table where I can operate at my convenience in fb2k my source PC with, of course, DVS etc - all on a dedicated LAN (i.e., just a single ethernet cable with no switches or anything else in the way) between my PC and D16 AES. I wanted an *audibly silent* PC with no moving parts - and with headroom spec including future-proofing with Thunderbolt etc in mind. The *audibly silent* bit has taken two months to achieve. At first, Gigabyte mobos were chosen for Thunderbolt 3 credentials, but the two I tried (GA-X170-EXTREME ECC and GA-Z170X-Designare) howled, screeched and whistled and I could not establish the cause - except that the problem corresponded exactly with Minimum Processor State = 100% - an aspect of the High Performance Power Profile - an ordinary Windows setting. Now that I have my final mobo - an ASUS Z170-WS - I have been able to resolve the issue - which featured in a similar way on this expensive mobo too - by *Disabling C States in the BIOS*. I never discovered this solution with the Gigabyte mobos and there are no equivalent settings in their BIOSs - although for all I know similar ones may exist. It seems that Enabled C States are antithetic to Minimum Processor State = 100% - erupting in (probably) voltage control-related whining. With C States Disabled, my ASUS mobo and, in turn, the whole machine are super-silent. Sitting in a proud brushed aluminium case, it is also rather handsome like its owner - but I digress. The machine now comprises in a Streacom F12C:
AURUM Xilenser 500W PSU;
Asus Z170-WS Motherboard with dual Intel Gigabit LAN: I219-LM and I210-AT;
Intel i7-6700 3.4 GHz CPU;
NOFAN CR-80EH Copper IcePipe CPU Cooler;
16Gb Corsair Dominator Platinum DDR4 2400MHz RAM;
Windows 10 Pro on Samsung SM951 128Gb M.2 PCIe NVMe SSD;
*.flac on Samsung 850 EVO 1Tb SSD.
The SQ is better than ever with enhanced transparency across the spectrum, deeper soundstage and lucid thumping bass (a reliable AOIP feature to my way of thinking).
 
Big shout out to scan.co.uk for persevering with me to this point. A particular aspect of their service - old-fashioned emphasis on customer satisfaction - has been exemplary. If you want an audiophile PC, you can rely on scan to walk with you to your goal. Now of course they have the added advantage of having worked with me!
 
Second, a summary of trial-and-error settings with emphasis on Network Adapters. One advantage of a hi-spec PC is that all buffers and other settings can be set to their most favourable free of dread regarding performance. Also, library management on the fly is as smooth as butter. Accordingly, all DVS, RedNet Control and Dante Controller settings are stretched/optimal. Without fiddling with Advanced Network Adapter Properties, I210-AT latency is some 10-20 μs lower than I219-LM - possibly because the latter is " Dual interconnect between MAC & PHY ". I thought perhaps the latter more "analogue" and the former more "refined". Latency swung it and I am using the I210-AT for the RedNet LAN and the I219-LM for the internet. I think disabling the online route could only help rather than hinder the audiophile and have done that having completed all necessary updates etc. I have spent a considerable number of hours (not days or weeks I hasten to add - I am not that much of a loser) reading up on and experimenting with Advanced Network Adapter settings. The chart below is the result. I guess I was trying to generate the lowest possible latency. The only setting that made an unequivocal difference is Interrupt Moderation Disabled. That is recommended anyway. I am not going to lecture on these settings as I am no technician. Aside from latency, I was interested in stability of latency and of course SQ. The final set of settings I thought generated noticeable mojo and, since that is everything to me, I left it there (for now).
 

DATE: 2016-11-04
SETTING
Adaptive Inter-Frame Spacing	Disabled​	N/A​
ARP Offload	N/A​	Disabled​
Enable PME	Disabled​	N/A​
Energy Efficient Ethernet	Off​	N/A​
Flow Control	Rx & Tx Enabled​	Rx & Tx Enabled​
Gigabit Master Slave Mode	Auto Detect​	Force Master Mode​
Interrupt Moderation	Disabled​	Disabled​
Interrupt Moderation Rate	Off​	Off​
IPv4 Checksum Offload	Rx & Tx Enabled​	Rx & Tx Enabled​
Jumbo Packet	Disabled​	9014 Bytes​
Large Send Offload V2 (IPv4)	Enabled​	Enabled​
Large Send Offload V2 (IPv6)	Enabled​	Enabled​
Legacy Switch Compatibility Mode	Enabled​	N/A​
Locally Administered Address	– Not Present​	– Not Present​
Log Link State Event	Disabled​	Disabled​
Maximum Number of RSS Queues	2 Queues​	4 Queues​
NS Offload	N/A​	Disabled​
Packet Priority & VLAN	Packet Priority & VLAN Enabled​	Packet Priority Enabled​
Protocol ARP Offload	Disabled​	N/A​
Protocol NS Offload	Disabled​	N/A​
Receive Buffers	2048​	2048​
Receive Side Scaling	Enabled​	Enabled​
Reduce Speed On Power Down	Disabled​	N/A​
Speed & Duplex	Auto Negotiation​	1.0 Gbps Full Duplex​
System Idle Power Saver	Disabled​	N/A​
TCP Checksum Offload (IPv4)	Rx & Tx Enabled​	Rx & Tx Enabled​
TCP Checksum Offload (IPv6)	Rx & Tx Enabled​	Rx & Tx Enabled​
Transmit Buffers	2048​	2048​
UDP Checksum Offload (IPv4)	Rx & Tx Enabled​	Rx & Tx Enabled​
UDP Checksum Offload (IPv6)	Rx & Tx Enabled​	Rx & Tx Enabled​
Wait for Link	Auto Detect​	Auto Detect​
Wait on Link Settings	Disabled​	N/A​
Wake on Magic Packet	Disabled​	N/A​
Wake on Pattern Match	Disabled​	N/A​
Average / Peak LATENCY (>1 hr)	INTERNET [DISABLED]​	790 μs / 1 msec​

[td=rowspan:2]

Intel I219-LM Gigabit LAN - Dual interconnect between MAC & PHY​

[/td] [td=rowspan:2]

Intel I210-AT Gigabit LAN controller​

[/td]

 

 
My system feels balanced right now (remainder = | Blue Jeans Cat 6 ethernet cable (40') | Focusrite RedNet D16 AES | van den Hul AES-EBU 110 Ohm Professional cable (0.8m) | Dangerous Convert-2 DAC [Word Clock Out to Focusrite RedNet D16 AES via Pro Co Premium Canare cable (3')] | Bespoke Achtung Audio Silver XLR/RCA "Pin 3 Floating" interconnects (1.2m) | Linn AV 5103 System Controller | Linn Silver interconnects (1.2m) | 2 x stereo Quad 909 power amps with identical DADA revisions | vertically bi-amping via Linn LK400 (c. 3m) | Snell Type A III) - and after a difficult year (much, much more difficult than I would ever have thought) I want to promote listening and enjoying over establishing and fettling. I have a lot of vinyl records as well as a large digital library, and I have always wanted to educate myself in the domain of classical music.

 

[jabbr]

 

Jumbo Packet

Disabled​

9014 Bytes​

 
 
Do you feel this setting has a positive impact on SQ? 
I thought that Focusrite only supports packet size of 1500 bytes and if I understand correctly the receiving end will have to do extra 'things' to handle these Jumbo Packets, either fragmenting too large packets or dropping packets and requesting resends in a smaller package size.
 
A reply from Cisco on MTU mismatch in a network:

There are a number of things which can affect this but the short answer is that by default the frames will not be dropped, they will be fragmented.  The payload is split into smaller pieces and forwarded, then reassembled at the far end. In the example you give, 9k frames will probably not be received on the switch interface.  They'll likely be fragmented by the PC before they're sent out on the network. Windows, Linux, OSX and other operating systems that support Ethernet interfaces have their own MTU at each layer of the network stack and/or protocol involved.  If a layer receives a datagram from the layer above it for processing that's larger than that layer can support, it will generally fragment the datagram into an acceptable size.  Unless you've explicitly configured the Ethernet MTU on the host for a larger size, the host's MTU for Ethernet is probably set for 1500 and it will send packets of a maximum of that size, regardless of the size of the chunks that are being sent from the FTP server running on the host.
I'm throwing a lot of "probably" and "likely" qualifications in there because, as I noted, there are a number of factors which can affect this.  But fragmentation as it goes through the network stack is typical behavior.

 

[Iving]

   
 
Do you feel this setting has a positive impact on SQ? 
I thought that Focusrite only supports packet size of 1500 bytes and if I understand correctly the receiving end will have to do extra 'things' to handle these Jumbo Packets, either fragmenting too large packets or dropping packets and requesting resends in a smaller package size.
 
A reply from Cisco on MTU mismatch in a network:

 
Thanks for this.
I had it on and off (Disabled) at various points.
I can't say that it had an impact on SQ.
But I might venture that it is a little bit possible that some permutation of settings as per table does.
Because mojo dropped in - I am happy to say that.
I didn't know about Focusrite and 1500 - that is interesting and I will assimilate it.
I understand that all devices on the LAN have to be Jumbo-compatible else it doesn't work.
Perhaps it is irrelevant then.
As your Cisco person says, there are lots of "probably" and "likely" in the mix.
Perhaps I got "lucky"?