Hi everyone,
here’s a link to an exciting Switch. The website isn’t online yet, but you can order the Switch directly from Farad or Mattijs de Vries and test it for three weeks. I’m still breaking it in—it needs at least 300 hours. The design is certainly impressive.
You can check it out for yourselves at the link below:
Mattijs is definitely one of the good guys in audio. I hope this product does well for him. The power supply matters a lot so this is very promising. Thanks for letting us know about it.
Yeah, Kenny, I agree.
He’s what people used to call—at least in Germany—a “cool sock”!
In English, I guess you’d translate that as “cool guy.”
I’ll make my final decision tomorrow.
The connection to the Zero-One-One shows 2.5 GB when paired with the Oladra; apparently, the two have come to an agreement on that…
This comment on the facebook page is interesting. “When the analog Hf signal is not perfectly square wave and there is noise present, the timing of this interpretation will be compromised, resulting in bad and digital sound.”
I can hear the network engineers thinking … if the signal is not square enough and the interpretation of it is compromised, it results in dropped packets and retransmissions.
However, noise riding on the transmission might compromise the ground plane of the receiving device. That might be a problem if the receiving device includes the DAC.
In my case the DAC and streamer are decoupled from the network via wifi, but when I change things way upstream in the network, those changes can be clearly audible.
I also recently had a very interesting experience. I had music playing and was sounding static like lots of tiny drop outs, but the track continued to play without stopping, pausing, stuttering etc. It could be said what I was hearing was extremely course distortion. It turns out the cause was an ethernet rj45 that was not fully/properly inserted p the cable was downstream of the server. This suggests roon keeps the track playing with whatever data it has, unless the data is so deficient it can’t.
So perhaps the impact on sound quality of networks is actually missing data that wasn’t retransmitted in time.
In my experience, sound quality improvements caused by changes to network seems to be because:
@dbastin2605, very interesting post. We’ve been going down a somewhat parallel rabbit hole with the little WR902AC used as a WiFi bridge over on WBF.
Your observation that changes far upstream remain audible even with WiFi between the network and streamer particularly caught my attention. That has very much been our experience too. Breaking the electrical connection clearly does something useful, but it doesn’t seem to make what happens before it irrelevant. Power, activity and the WiFi environment before the break can still matter, while the 902, its power and what follows it towards the streamer matter as well.
The Farad itself is interesting for a different reason. We’ve also experimented with Ethernet speed, including 100 and even 10 Mbps, while Farad is deliberately going in almost the opposite direction with 5 Gbps. I don’t think there is a simple “slower is better” conclusion from our experiments either, especially once good switches or filtering enter the picture.
@Alexander, did you eventually keep the Farad? I’m curious what attracted you to it in the first place? Farad’s power supply approach, the 5 Gbps design, or both?
And if you’ve heard something in a similar price range (or not) or what you replaced it with , I’d be interested in how they compare.
Last but not least, is it really for sale now? Or still in trial period? I can not seem to find a single unit for sale. I’ll check with Mattijs.
Small update:
Mattijs just confirmed to me that the switch is a final product and already available. Apparently quite a few have been sold directly; the ZOZ website and formal sales setup simply aren’t ready yet.
Hi @dbastin2605
I now also went through Matijs’ full explanation of the Zero One Zero switch. I’m less convinced by his explanation around 0/1 interpretation and 5Gb speed, but the design itself is much more interesting than that explanation suggests.
What stands out to me is the power architecture: no switching regulators internally, three separate low-noise linear/supercap supplies for the core, OCXO and I/O, plus substantial capacitance and filtering. That fits rather well with what I’ve been interested in. Noise can propagate through different paths, and once you reduce one source, others can become more apparent.
So perhaps the 5Gb choice is not the main story. I assume Matijs experimented with different speeds during development, but I don’t actually know that. Either way, there seems to be much more going on in this switch than simply running Ethernet faster.
In any case, his switch is another interesting example of how far optimizing power inside the network itself can be taken.
Hi Paul,
You may recall I uses Mikrotik mAP lite , powered by USB powerbank, with Akiko USB tuning stick all on Synergistic Research POD. You probably dont know I’ve optimised its operation, rf power, traffic rejection etc (easier because its a mini router), I also optimised traffic on my main router to minimise traffic to the server/CX and endpoint. The only traffic that reaches the endpoint is the essential traffic it needs, nothing more . The collective improvements doing these are significant rather than subtle.
This leads me to believe we should focus of minimising unnecessary traffic, which includes retransmissions caused by distorted ethernet signals.
Its probably that simple in concept but very hard to achieve. Performance of Fancy switches are significantly limited because they’re not managed switches. Furthermore, I’d guess most audiophiles spending big bucks on those fancy switches haven’t managed traffic to minimise the workload of the endpoint cpu and everything enroute to it.
My main router, mAP lite and CX cpus operates at about 2%.
Im working on triple cascade power supplies for the ONT, and router andbattery > LT3045s to EtherRegen. The triople cascade will probably be SMPS > CM and DM chokes > Linear drop down to 13v > LT3045s. And Lessloss DC power cables. Hence, very clean power, very low ripple, minimal disruption to ground plane = optimal operation of transmissions, minimal errors and retransmissions and workload on cpus enroute to endpoint = even lower disruption of ground planes and power ripple = further improvements of transmission accuracy etc.
Hi Dale,
Very interesting, and yes, we have ended up surprisingly close. I didn’t realise how far you had taken the mAP optimisation itself.
We actually ran into an interesting limitation of the WiFi client idea (your mAP, our WR902 client). In one already highly optimized network/system, the WR902 in client mode was ultimately outperformed by a very basic low power switch. That was important because it showed us that simply breaking the electrical connection isn’t necessarily enough. The result depends on what noise mechanisms you’re removing & what is already removed, but also what remains upstream, what the client and its power introduce, and what happens downstream.
That eventually pushed us much further into optimizing the WR902 itself: very, very clean power, 3.3V battery versus the internal conversion, a rewritten minimal OS, router setup and power, dedicated WiFi access point setup and power, WiFi band and channel, and then the last metre towards the streamer. The idea was also to minimise unnecessary activity at the client and make the dedicated AP-to-client link as clean and efficient as possible, with as little interference and contention as we could achieve. In other words, rather close to what you’ve been doing, although from a different starting point.
The two paths make the last part easier to see:
Router → switch → last Ethernet cable → streamer/DAC
Router → AP → WiFi → WR902 client → last Ethernet cable → switch or filter→ streamer/DAC
What became particularly interesting to me is that there doesn’t seem to be one universal network-noise solution.
The last metre may actually deserve special attention here, both for a network switch and a WiFi bridge/client. There are published measurements showing common-mode noise reduction in the last Ethernet connection to the DAC, alongside what we hear. So electrical isolation further upstream doesn’t make what happens close to the endpoint irrelevant.
That’s why I find Mattijs’ switch interesting beyond the 5Gb argument. He has gone unusually far on power, regulation, clock and I/O inside the switch itself. Your work, ours with the WR902, and this switch seem to be approaching the same problem from different places.
Hi, you did everything right. Right now, you can order the Switch directly from him via email. Yes, the Switch is up and running at my place now. It really does have a long break-in period—it needs at least 300 hours of actual use. The sound is very clean, and it plays quite neutrally. Ultimately, everyone can do as they please, but the Farad power cable does indeed pair exceptionally well with it. I’m selling my Melco S100/2, which is also very good. Right now, I’m still thinking about a suitable network cable. I’m going to test a cheaper one from O2A. We’ll see—it’s on its way 