It’s only one of two key approaches. Push or pull. Push the controller hands control to the server, or pull the controller hands control to the streamer. Naim adopt the latter and match the features of the control between app and streamer
You need to look at the timing stats of the TCP flows using Wireshark or similar. Also note how TCP Zero Window is being used on Naim .
However upnp is all down at layer 3, upnp is at layer 6 and 7. So upnp doesn’t care.
Yes I was also streaming before upnp in the late 90s , but to stream with as little associated out of band network noise that we get with the later Naim products is a neat evolution.
Yes I notice a lot of rubbish on the Roon site regarding this… it’s annoying isn’t it.
There is no inherent limitation with upnp media transfers… they are very basic or straightforward transfers. Complications can arise when you need to sync multiple end points. I think that is where RAAT might have advantages.
2 Likes
Which I did when I was diagnosing some earlier problems with streaming higher resolution formats with the SqueezeLMS and Squeezebox 3, and then with the ND5XS I moved with, initially retaining LMS (with UPnP enabled) and then Asset UPnP server on a RPi (I wrote the installation guide).
I still don’t see how this wire-level interchange between Renderer and Server has anything to do with the UPnP Controller, and how using the Naim app ‘is designed to get the best SQ performance for Naim equipment’ and only the Naim app can utilize this between Streamer and Server. It just gives the instruction to the Renderer and backs out of the process.
Which Roon does very well. I have multiple endpoints, but 4 of them are grouped into a single Zone, effectively upstairs covering ChromeCast Audio based playback in Bedroom & Study/Office and then Chromecast embedded speakers (waterproof) in the Bathrooms. The Grouping ensures they are all in-sync depending what’s switched on.
So no other Naim devices involved, so no experiences of the mutli-room playback capabilities or issues experienced with Naim’s implementation.
But if they can do this, I would of thought the RAAT protocol can reliablably deliver a bit perfect stream from the Roon Server to the Naim streamer with a RoonReady implementation, and do this as well as UPnP.
I believe that is the case I was advised and approaches shared with the gen 2 streamers were in development. Now Naim don’t use certificates or encryption as far as I am aware, so in that case somebody could reverse engineer Naim’s approach - and fill the Naim streamer index buffer with server indexes and let the streamer play by it optimally pulling back the content - in a way that has been tuned to minimise noise on the streamer.
Digital EMF noise does cause interesting effects on the streamer - especially when using the inbuilt DACs… I know Naim engineering had a model where by changing code execution timing on their firmware they could simulate a largely controllable tone control on the resultant audio on their streamers… (I personally find this fascinating - perhaps not least because when I was humble under grad I really tripped up in my final year project with this sort of phenomenon, digital EMF causing interference).
Anyway as such I suspect Naim have optimised their streamer software to work optimally with the their app including with SQ… however yes I see how another app provider could potentially reverse engineer the Naim specific approach and drive the streamer the same way.
1 Like
Naturally - but I don’t think this is anything to do with being ‘bit perfect’… its about the side effects of the rendering from data algorithms in a closed system.
I think this is just UPnP implementation, and how the interchange and signalling between Renderer and Streamer is undertaken.
The Roon with SonoreUPnP Bridge approach seems to be that the cache is off-board the streamer, within the Roon server, and this then provides the traffic in a burst approach to the streamer, as it provides the next part of the track, which is then loaded into the streamer buffer.
This is why the bridged Roon solution can effectively work for the smaller buffers in the Gen 1 streamers, as it undertaking this off-board, leaving the streamer to do a lot less. Which I imagine is goodness, if undertaking management of the stream and processing draws current and therefore potential noise into the playback process.
My undergrad project, back in 1990, was a Wireless Security system using the permitted RF band for Car fobs ( 433.92MHz) and using FM to overlap signals to identify the PIR based sensors and status (alarmed, triggered and heartbeat).
Got the basic system working, just, between sensor and base station (could only afford one PIR, so only one sensor 
) at relatively short distance, had to black-box the keypad based alarming side, alarm setting/deactivation, etc.
But I think they were impressed with the RF standards documents enclosed, and a BOM for the unit, plus cost estimates. I was sponsored by a Electronics sub-contractor at the time.
And now, this stuff is churned out of China by the container.
But the RF side was pretty hairy and use of a double input FET, not covered in the course, to overlay the signal on the RF carrier.
Also having fun with a 10MHz OXCO external clock with -134 dBc/Hz at 10-Hz offset, to my EtherREGEN and the improvements this has bought to the streaming playback.
Curious on which clock you use?
I’ve thought about this (also EtherREGEN). What type of improvements are you experiencing? I have looked at a ready-made small board from audiophool (.nl) with these specs:
- OCXO clock 10MHz, 20MHz or 24.576 MHz
- Stability 5ppb
- Phase noise -140dBc/Hz at 1kHz
- Ultra low noise regulation circuit : noise less than 10uVrms
- Input : 5Vdc , max 1A
- Output-impedance: 0 Ohm
They are normally using it modifying Cisco Merakis, 2960’s and Melco N1ZH. Either DIY or ask them to mount it in some box for me. There is not enough DIY in HiFi these days 
Got an AfterDark one, at King level (-134 dBc/Hz at 10-Hz offset, -107 dBc/Hz at 1-Hz offset) with 75ohm output, with matching AD PSU from a Roon forum member, as there is a ‘For Sale’ section on that forum, unlike here.
Not the most expensive selected version, but not the cheapest either. Alex and John at Uptone state that once better than -125 dBc/Hz at 10-Hz offset there is benefit over the internal Crystek CCHD-575.
I have been happy with the ‘bare’ EtherREGEN since it landed back in Nov '19 with improvements made on the PSU from ZeroZero basic 15W to Farad3 with Level2 Silver cable, etc.
So this seemed the less risky approach over a DIY build using a OXCO board bought blind from China, or the other sub $1,000 clocks available.
Came with its original Clock and Power cable, so out of the box operation.
Been 21 days since power on, and the reported settling time is 30-days.
However straight away, there was a difference, for the better.
However I am now into PhoenixNET level spend between the EtherREGEN, Farad3 PSU, OXCO & PSU. And now considering alternate Silver Clock cables and DC power etc. 
and the show goes on.
1 Like
Hi, I reported here, as it happened Uptone Etherregen 50 or 75 Ohm - #40 by simon.pepper
So improved detailing, transients, almost ‘cleaner’
I was going to go down this approach, as there are DIY components readily available, have a spare 5v power supply, get a SMA to BNC cable and a case for the board etc.
How hard could it be?
And then this one, at King level, came up for Sale and bosh, deal done, for 50% of the cost, including shipping etc. Just had some Brexit tax to pay on goods that have already had VAT paid on them, but that’s just the stupidity of the UK being a 3rd country surrounded by a single market, but we won’t get into that ridiculous situation. 
1 Like
I believe it must be a nice improvement. Enjoy. I was tempted to do that too when I had the Etheregen, but the number of little boxes finally discouraged me. I went for PhoenixNet.
I may join you on the second ps you have for Nds. Soon or later. I will ask you how to connect it best when I will get it.
Uptone should produce the ER 2, upgraded. But still no news on it. Will you get it when it will be ready ?
1 Like
Don’t know on the ER2. I have no need for Gigabit on the ‘B’ port or a SFP cage, but it depends on the price, whether I can sell on the ER1 and what’s the delta between the ER1 & ER2
Happy with what I have for now. Most of the boxes are hidden behind & underneath the 2 Aavik racks.
1 Like
I just get the lp out of its sleeve, put it on the turntable press the power button and lower the arm.
I’d be interested to understand why using BubbleUPnP should sound better than using the Naim app? Is there a technical reason?
To be frank I have always been unimpressed with my EtherRegen… it was I am afraid apathy on my part that meant I didn’t return it, so I keep it in one of my switched segments from a 3750 - but quite honestly using an external DAC on my NDX2 acting as a digital streamer it’s transparent whether its in line or not for me. (that was not the case with my previous NDX … )
However since the next gen streamers and their improved network isolation, my enthusiasm, for investigating here has dwindled, not least lets be honest from an antctual networking perspective some of this stuff is cheaply or simply designed and manufactured dressed up with a good marketing budget for a niche market.
However would be curious to hear if you have any causation measures or quantitative measures you might have of the reduced phase noise power with serialisation clock using your external clock over the inbuilt one (I used to use a bandscope - a poor man’s frequency spectrum analyser - to look for line clock jitter and other HF common mode noise and harmonics on an ethernet lead generated by a switch) … or is it actually other things at play … like the grounding of common mode noise through the external clock cabling? - Or did you separately earth the Regen? The clock scaler/synthesizer is possibly quite a key element. Typical ethernet line speed clocks are 625 MHz, 312.5 MHz, 156.25 MHz, and 125 MHz . I don’t know if the Etherregen uses a quality scaler/synthesizer like the TI CDCM6100x to achieve these. It is the phase noise/jitter output of the clock synthesiser which will be important - rather than solely the input clock…
Also is this external clock used for the internal DC to DC switched mode power supplies (I suspect kind of unlikely) ?
If you have you only assessed through listening then no worries - but you strike me as a chap that likes to go poking around
1 Like
Yes, many of the switch offerings from EE, Ansuz, Silent Angel are just off-the-shelf switch boards repurposed. At least the Uptone products are a dedicated ‘ground-up’ design from a reputable designer (John Swensen)
Unfortunately no access to any Lab equipment anymore. When I worked for ICL Enterprise Systems (i.e. Mainframes) there was a good lab in the Materials Tech Group you could use. And getting decent equipment to performance measurement at this level, would be quite expensive.
Uptone are very transparent into how they are designing & implementing their products, and if you have questions and comments, then both Alex and John monitor their forum and post regularly.
The external clock input on the EtherREGEN v1 is just used for the Port ‘B’ the isolated dedicated output into a streamer/player.
Would love to hook things up to analyser and scopes and get down into the mechanics, but time & this level of equipment is not available. Listening is my common used measurement method now
1 Like
I would be interested to understand how a UPNP controller app has any bearing on anything, the music does not originate from it or go through it.
Naim is always using uniti core as a “transport” for the streamers but here I have a question, Melco has a dedicated eth connection for streaming signal. It is perfectly clocked with no layer 1 arp on it, a totally different approach from the core. Do you think it improves the quality or it is simply a way of complicating a simple connection?