Home Streaming Audio Network Alpha Labs – Audiophile Clocks and Network Switches

Alpha Labs – Audiophile Clocks and Network Switches

Alpha Labs – Audiophile Clocks and Network Switches


Audiophile or ‘audiophool’? We’re going to talk about ‘audiophile’ clocks in switches. And we do that the Alpha Audio way. In other words: listening… and measuring. Brace yourself, because this is pretty nerdy….

Switches are quite a controversial topic. After all, how can a network affect the sound? We are talking about data packages … and they either arrive or they don’t. In short: it cannot affect playback.

Now that is 100% correct… Data either arrives or it doesn’t. So data corruption does not play a role at all. But what is in fact going on?

Now we are not going to explain exactly what is going on, since we have already written about that in this extensive article. We also did some extensive live testing of switches. These can all be found in the switches category.

Test setup

We grabbed a Dlink 108 and a TP Link 1008 switch to do this test. We have two of each switch: one we modified with a clock-input. The clock signal-wire goes on the clock-input, the shield on the ground. Sometimes we had to be a little creative with finding a decent ground connection. We always looked for a ground as close to the clock as possible.

To this external input we can connect a nice, audiophile 25 MHz clock. We also use this clock as a test-clock for the Wavecrest.

In both the measurements and the listening tests, we use both the standard power supply and a Sbooster 5 Volt power supply. This is to determine where the big differences come from: the clock or the power supply.

To do the listening tests we take the Volumio Rivo. We also used the AlphaPC with similar results in the listening tests.


  1. Hello Jaap and Wijnand, I really appreciate your work! I too have experimented with a lot of Ethernet switches with linear power supply upgrades and I still oscillate between using fiber optic (for greater accuracy and lower noise floor) and copper/silver (for more realism and body) . I was reading this article last week and it really got me wonder how are those Ethernet clocks synchronized so I started searching for it … and I think I got the answer.

    The are two Ethernet protocols that come very handy: NTP (with simpler version SNTP) and PTP (IEEE 1855). The first one, NTP = Network Time Protocol, has a precision in the 1ms range (one sample at 44.1kHz is ~0.0227ms !!!) while the PTP = Precision Time Protocol has a precision of ~1us (1/1000ms). The NTP can be used at home, you just need an internet connection and a managed device that can sync to one of many NTP servers out there (there is also a possibility to set up you own NTP server with a Raspberry Pi 4 and a GPS module) and then sync further the rest of the devices to that specific clock. PTP is a different kind of beast which implies some expensive gear and is mostly used for video and music streaming.

    Anyway, I’ve set up NTP on my router (client for retrieving the time from the internet server / server to broadcast that time inside my local network) and set up also my managed devices (main CISCO switch, two Synology NASes) to retrieve the time from the router. Hopefully the other switches (2x SOtM sNH-10G with Mutec REF-10 master clock, 1x MOTU AVB switch, 2x fiber optic media converters, Auralic G2.1 streamer) synced their clocks as well to the rest.

    I’ve noticed better defined instruments with improved texture and to be sure about this effect, I’ve set NTP on one of my friend’s router as well, without telling him what is going to happen to the sound … he noticed the same improvement.

    Looking at the clock numbers, the sync between my router and the internet NTP server occurs each 6-10 seconds and the local clock drift is about 1.5ms – 2ms. Between my router and main CISCO switch the clock drift seems to stabilize around 0.4ms – 0.2ms (this also brings down the latency and the jitter). As a note: after powering on at least one Ethernet switch (or other device that includes a clock) you have to wait some minutes until things synchronize in the network.

    You can read more about this:

    • Hello Valentin, thanks 🙂 first of all nice to hear you got an improvement. I will look into the links you have added in the next few day’s. The NTP is not meant to crystal clock sync, but time sync. This way devices can sync the time with a time server. That’s completely different from crystals that run on a frequency. It’s about frequency sync. For example a 24Mhz crystal can impact a 25Mhz crystal.

      How many devices due you sync and what is the time between to syncs? I could be something with the traffic you added in your network.

      • Hmmm, so it seems more complex. I had the idea that the crystal clock only counts, without knowing the real time (each time it is powered on, it starts all over again) so here the NTP comes handy as it provides correction every X seconds by setting a correct time and preventing the clocks from drifting too much.

        I have the main router sync with a NTP server (each 6-10 seconds) and then the router broadcasts that time in the local network. My main 16 port switch, 2 Synology NAS-es and 2 PCs reads the time from the router at a larger amount of time. I don’t know exactly, perhaps hours apart.

        The network traffic is pretty low, something about 80 bytes per second or even less, I’m not worried about that.

  2. My system came free of trouble out of the last test. I’m pretty sure of my previous thoughts now and I think it’s explainable.

    *The datastream is the source to sync on
    *The internet is not controllable so the router must sync on the internet frequency
    *The router is due to the internet frequency a source of trouble for the house network
    *Switches with normal crystals will sync to the network frequencies of which the router is the source.
    *Due to the galvanic isolators in switches, routers and other network devices there will be a small sinc timeframe between the crystals of the devices.
    *OCXO clocks will not sync to the datastream by default. They will try to take over the lead. The more they warm up and stabilize they will get more followers. The devices that are connectionwis near by an ocxo clock will sync to the OCXO. The devices that are nearby the router will sync to the router.
    *The main problem in networks for audio is that audio-equipment uses clocks as well. These clocks will interfere with the network, but will get in trouble with the inflexible router as well.
    *Due to the fact that there are several masters. The internet in combination with the router and the audio and OCXO clocks. The ethernet is a troubling factor.

    The solution.
    A separate internet subscription for the audio system. The router is separated with 4 switches before connecting to a streamer (PC) , then a switch with an OCXO and then the audio system.
    The switch near the router will sync mostly to the router and the switch near by the OCXO switch will sync with the OCXO the other 2 switches will float between.

    My system is perfectly stable now! Thanks for reading and I hope it will help you 🙂

    • Interesting observations Wijnand. Are you aware of this discussion?

      (…) Now it becomes more complicated, because there is a different clock in the endpoint—with its own phase- noise, switching other circuits in the endpoint—and creating ground-plane noise correlating to the phase- noise of its internal clock. Thus, the ground-plane noise in the endpoint is the combination of the ground- plane noise from the switch’s clock (the external clock) and from the phase-noise of the internal clock. The external clock phase-noise is thus overlaid on top of the noise from the internal clock. (…)
      – – –
      Not sure if I am allowed to post a link but try paste the following in your browser:
      – – –


      • Thanks Paul, I’m going to read more about this and will try to reflect it on my own situation for better understanding. A few months ago I also tested in my own situation with fiber connections between every device and still I had the same clock problems.

        At the moment I have a test in which the network is floating up to the Dac. The DC from the power supplies is completely free of ground. This seems to help. Thanks for the link. I’m sure it will help me understand more!
        I will start with the paper of March 2020


        • Hoi Wijnand,

          I enjoyed reading your observations, a few more things to consider maybe:

          Optimizing and segregating the audio network from your home network is crucial, alongside ensuring accurate clocking as you found out. One effective approach involves employing a secondary router such as the Mikrotik CRS305-1G-4S+IN. This device, featuring four SFP+ ports and a 1 Gbit copper ethernet port, operates with minimal power consumption, resulting in reduced noise—a quality highly regarded by audiophiles. Pairing it with a 12V LPS further enhances its performance. Or consider integrating a 4G router with a SIM card into your audio setup like the TP-LINK TL-MR6400 4G. Place it in the audio room and connect with a short and good ethernet cable to player/server, ideally use a good LPS.

          It’s advisable to connect your control device—be it a tablet or smartphone—solely to the dedicated audio Wi-Fi network, abstaining from sharing it with your home network.

          Today to me playing local files from SSD and no network is still the reference in SQ, but who only does that these days? With above measures and (audio) network optimization you can come a long way to enjoy Tidal and Qobuz as you do local SSD. Avoid ROON if you can because it connects constantly to the servers. Good news is that in the near future Roon aims to eliminate its dependence on internet connectivity to enhance performance with (local) files.

          For insights into switches and related expertise, John Swenson is an invaluable resource—I personally use his EtherRegen switch and JS-2 LPS. Wishing you “veel succes” in optimizing your setup.

          • Hi Paul, Thanks for the good advice!

            I have used the TP-Link 4G, but I hardwired it. I have used it with the router of the internet provider in bridge mode. I also bought a router with an SFP module. The SFP modules I tested produced additional noise even when it was only pushed into the slot.

            At the moment the sound is very good. Really on top streamer levels. The connection is on a dedicated internet line. I have 2 D’link DGS108 in series between the router and the streamer PC. The router and these D’links are powered by their stock power supplies. The internet connection is connected to a PC’s stock network card. In the PC is a Jcat Netcard XE which is fed by a Farad super 3 with a QSA red/black fuse. From the Jcat a 1 meter ethernet cable goes into a third D’link DGS108 which is fed by a Farad Super 10 with a QSA Red fuse and from there a 50cm cable into the Dac. It’s been playing now for 36 hours and the sound is still incredible, but when I replace only one D’link with, for example, a Netgear GS108E the sound will get darker for 30% immediately and go darker every hour until it seems like the tweeters are switched off.

            The difficulty with networks seems there’s a limit. I think it really matters how clean the powernet is, the worse the less trouble. The D’links seem to be the only switches that I can use in combination with the super power supplies. I disconnected 5 Farads and installed a few stock power supplies to get the noise that seems to be needed. When adding an additional powernet filter the sound will go darkening. Also when I connect a tablet on one of the 2 switches between the router and the PC the sound will go darkening.

            I’m not sure if an Ether regen can help, because devices like a PC can go dark probably due to crystal cync. When this happens nothing seems to be able to repair the signal.

            The next session will be a second streamer PC, because I once listened to streamers in series and that did something as well.

            “Dankjewel :-)”

  3. A second router seems a great success. Unfortunately there are a few protocols that don’t work on one of the routers (The one from KPN). For example we can’t use the TV guide on the television and a few other minor problems. The second router is a Draytek which works flawlessly with all protocols, still it was great for the test. The isolation factor is almost 100%. There’s now interference that I can point to, considering the clocks, between the networks behind the routers. So I could remove all isolation switches in the audio network, besides this due to the split routers the audio network opened up enormously. Internet streams are great… really better than from my SSD’s inside the streamer. Normally this is the other way around.

    After 24 hours clock stabilisation I could still hear a slight SQ deration, the clock still seems to gain grip on other clocks, but the end result is a lot better than in the past. I think the router clock normally must adapt or control all other clocks in the network. Due to the router split there are far less clocks in the audionetwork. The work around is switching off the powersupply output to the switch after listening.

    I decided to order a second internet subscription which will arrive tomorrow. This way I can extend the test, but for now a dedicated internet router especially for an audio network seems a very big plus when multiple devices are wired.

    About clock synchronisation. I think that it won’t be possible to use one master clock in a network to take the place of all dedicated clocks in switches etc. This is because every galvanic isolation transformer probably will take some time on the signal. It’s just a thought, but I think this would cause timing problems. I think the way switches work is to synchronise on the packets that flows in.

      • Hello EJ, Great to hear it helped you. Well I try to avoid them at the moment, because my vision of ethernet noise etc has changed dramatically. I think that unstable clocks have more impact on the sound than noise. The question is what ethernet noise actually is. Is it switching noise from for example power supplies or is it bad timing due to less stable clocks due to bad power supplies? Or both? I will give an example: why does a router have an impact on the sound when a stream is active from an internal drive? So the router is not active in the stream. Still it is clearly audible through a switch, a fiber or whatsoever. The only explanation that still stands in my opinion is that it’s affecting several clocks which affect each other even more 🙂

        About the ethernet isolators and switches as well. I think they can set the stream a little bit back in time microseconds. This way the clock behind the isolator will run a few microseconds or maybe nanoseconds out of phase with the other clocks. Adding an extra isolator will put the stream back in time more, but the stream won’t rebuild after the isolator plug. Thats why I would use a switch instead. When the stream is set back in time its impact on other clocks will be a few microseconds out of phase as well. The more switches run out of phase of each other the better the sound seems to be. Don’t get me wrong at this point because the clock does seem to synchronise on the stream, but the stream became a little bit out of phase due to the galvanic isolator (transformator EMK’s).

        When a clock signal becomes too strong many switches will follow and make the clock signal even stronger, which I think is bad, because then the stream will be clocked in and out at a less perfect moment. I mean a bit or word is active for a very short period of time 40 nanoseconds. Normally the clock will pad in and out on its own terms, but what happens when this happens on the terms of several clocks combined?

        I hope I did not go too deep :-), but keep in mind that everything is partly based on thoughts, experience and insight, but it’s not proven yet. 

  4. Such interesting research by Alpha, thanks for that, really appreciated! And what wonderful contributions by others in this thread, also much appreciated!
    The remark in the original article about the positive effect of grounding the switch triggered me. I am not a technical person, so I really don’t know how I could go about doing just that. Could someone explain this a bit more thorough?
    I do have a grounding pin on my Paul Pang Quad switch, but there nothing like that on my Audes power conditioner… Should I somehow ground to the grid instead, or would that compromise the benefit of isolating my audio components from the grid by this balanced transformer?
    Hope somebody could educate me a bit. Thanks anyway.

      • Yes, indeed. Thanks for you swift reply! Good to know as is well.
        One thing keeps me wondering, some manufacturers, like for instance Shunyata or Puritan, seem to discriminate between two additive ways of grounding that yield benefits. One concerning the signal path and the other concerning the chassis of the components. Do have you any experience with those two approaches and their proclaimed effects on sound quality?

    • To ground a switch, which does have a ground connection like the Dlink: you connect a wire to the earth of a regular three pole plug and attach the wire on the other side to the grounding connection on the Dlink. The plug goes into an outlet.

      If both your switch and your streamer are connected to the Audes, they share a common ground plane and additional grounding has no effect. I tried this with the mentioned Dlink switch.

      For your Paul Pang switch, see the comment by Jaap.

      • Thanks for your answer, it is very clear to me now. Also I have thank you for another thing. When I was looking for a solution to clean up the power for my audio set up, your report on the Audes ST-900 steered me into that direction. Before that I had a AQ Niagara 1200 in place but I never had the impression was functioning well (enough). The sound quality was fluctuating from day to day in such a way that I could only best explain by attributing it at least for the latest part to changes in the degree of pollution of the power. And wow, it really is a night and day difference the Audes makes. Sound quality has really clearly gone a few notches, and consistently so.

  5. It seems I went another step forward again. Last week I installed a Jcat card on my PC. This was the old version that I bought somewhere in 2018 I think. This Jcat was modified with a better Connor and Winfield clock. I didn’t expect it, because of all the work I have done in my network and the components, but I must say that the impact of this card and clock was clearly audible for the better. Unfortunately after 2 days the Jcat died, probably due to all the measurements I have done during these days. So I ordered the new version Jcat netcard XE on which I won’t do new measurements. This card also sounds very good already with its stock clock which is an OCXO as well. I have tried the Jcat directly on my Dac streamer for a few day’s and this morning I have put the D’Link switch with the Pink Faun clock back on the network. The SQ immediately gained so much precision, natural sound colors and musicality. When I would rate the steps (between 0 and 10) from the stock network card to the Jcat and finally to the complete setup with the Jcat and the D’link with Pink Faun clock. Starting at zero with the stock network card then the JCat would get the 4 and the complete setup the 10. The stock network card and the D’link with Pink Faun clock would get a 7 or 8. So the combination brought another few steps.

    The difficulty is the clocks. The clocks seem to interfere with several other network components until the sound is dull in a few days. The better the power supplies the worse it gets. This morning I installed a second router in parallel with the normal router. This way I could separate the network completely. The audio system and the PC are now connected to one router with only one isolation switch in between. See it as a dedicated solution. I have to sit back for a few days to see what this will bring in the end. But for now it took away a thick layer. Although the sound was already on a high level. Now the sound is less warm, cleaner, but also more natural. In the end the sound always seem to lose SQ due to interference, but again fingers crossed for a few days

  6. Today I have built in a Jcat netcard in my PC and I installed a spare Connor and Winfield on top of it. Without an isolation switch I connected the Jcat directly with the switch with Pink Faun ultra clock. After a few hours the sound was really great!! I decided to do a measurement on the Connor and Winfield. At first it run around 24.879 Mhz. After a while the sound was not so great anymore I measured again and it showed 25.0005 Mhz. Which is exactly the speed of the ultra clock. Clock synchronisation seems the problem of bad sound! Also a picture of both clocks combined. It shows 50.0010 Mhz.

    I will send the scope pictures!

    • I’ve looked into the ethernet standards. It seems devices should synchronise perfectly for packet transfer. This is probably not so easy with OCXO’s. The cercuitrisch are built to synchronise with each other. There can be a masterclock. There’s a lot of information about this on the internet. I have homework to do 😉

  7. Thanks Wijnand!
    I now realize that you are much more experienced than me!

    Having said that, a lot has been explained logically only since the pandemic.
    In my experience the ethernet cables don´t do much on their own, since they are just carrying the digital signal. They must be fed with a clean signal first! Then they matter since they then protect that already clean signal into the next gear.
    I say this because my experience, and many others on different forums, is that the ethernet cables does matter a lot, in a streamer/switch/PS context, but “only/mostly” when fed with a very clean signal.
    Maybe you already understood this back in 2018, then i am sorry for pointing the obvious out like this. 🙂

    Note that I am not saying that i am correct but this is where I am today, in my thought process.

    • Hello Tobias,

      No problem at all. I had to learn everything by trial and error and so I’m not smart enough to understand everything as most of the people. The problem is that things I don’t understand and still want to know, will keep my attention until I understand. This problem is the hardest I ever worked on. Literally I was off track for a few thousand hours, but now I’m close to understanding as well.

      Indeed, actually from 2010 when I bought a PS audio PWD1. I became aware that ethernet really matters. Those days I thought it was all about cabling. I became a member of the PS audio beta testers for the PWD2. I also added a T+A streamer dac for comparisons with the PWD2, after this the PS audio directstream and the NAD M12 were added. I spoke and wrote with many professionals. It went on and on and nobody seemed to understand and now we are here and almost there 😉

      • The funny thing is that Paul (PS Audio) said that switches can´t matter, just a few weeks back, since it is just one´s and zero´s…
        Then he went on, in the same video, saying how amazing their new Airlens technology is, since it isolates the noise towards the DAC…
        Not sure how he fails to see that that is exactly what the switch is also doing. 🙂
        It really is crazy how people, even on his level, get caught up in their normal “Digital is just one´s and zero´s” thought and fail to see the obvious that really is very simple and logic.
        This is the problem, in my opinion, that digital audio is developed by IT people that are too caught up in classical technology/digital ideas… So that they even fail to understand the simplest of logic, due to their IT knowledge/background.
        I´m in technology my self.

          • As stated in one of my previous comments I was a PS audio customer for at least 5 or 6 years. I bought 3 dac streamers , two power conditioners, powercords etc. because there was no other brand that could compete with PS audio. Especially the directstream and the P10 were products like no else 🙂

            There’s no forum that I know that discussed ethernet and switches more in those days than on the PS audio forum. There were so many success stories.

            I must admit the directstream was pretty good for the money in those days, but the powersection inside is really nothing special. The P5 and P10 were totally not my thing as well as the powercords AC12 or something. Maybe in the 110Vac countries these products probably sound good, but not in my system and also not in another high grade system that I knew.

            As far as I know Paul, we discussed a few articles on the forum and I once met him, when he demonstrated the PWD2. He is a really nice guy who stands after his products, but he was seldom into tweaks.

            With a German friend we added a galvanic I2S port on the directstream. This friend was ready to take the step first and it worked. The modification was engineered by the head engineer of Pink Faun and modified at Pink Faun. I wrote about it on the forums, but it was nothing special. Nowadays the directstream comes with a galvanic isolated I2S port as a standard 🙂

  8. This morning I woke up with the thought that I have better insight about jitter :-)I never really understood how jitter could influence music data. Of course I understand fluctuations, but fluctuations in a bitstream?You probably would have seen those blockwaves with longer and shorter blocks (that would be fluctuations of atto seconds on those points with the current clocks reaching 15 pico at 24.756 Mhz)Still this is the way it frequently is explained, for sure it is the way I understood these explanations.Let’s set a few things apart. An audio clock runs at 24.576 Mhz let’s take a Music file of 48 Khz and 24 bitThis means that 48 thousands samples of 24 bits will be clocked per second by the audio clock.When the audio clock speed is divided by the music file sample frequency you will get 512 samples 24.576.000 / 48000 = 512For a 192 Khz file you will get 128 samples.
    Those samples are musical information. 1 sample contains 16, 24 or 32 bits.For this example I took 48Khz 24 bit. This means the audio clock should move a sample of 24 bit every 512 taps to reach 48000 samples in 1 second.On this point the translation to analog takes place. I think this is where Jitter could be audible. Because what if a sample is moved after 490 taps or 530 taps instead of 512 (jitter?).The maximum theoretical analog frequency could be 48 Khz on this file type. I know we can’t hear this, but please keep this in mind.Or what if the clock daviates and the actual time length of every 512 taps is longer or shorter than it should be.
    To take this one level up. Still not knowing if i’m right for only one ‘bit’, but stay with me a few more sentences.Let’s take Roon that streams a Wav track from a PC to a streamer dac. If I’m right this already can be audio frames in a Wav container sent by ethernet.What if an USB clock daviates and misclock data to get on ethernet. Would it be corrected in this kind of audio streams?What probably could happen is that every few clock taps (512 tap) that there isn’t data at all. The analog output won’t move on this tap, because there’s no new audiodata, no new noise that would be moved.Think about this, you have to say the complete alphabet in 10 seconds. So you have to say one letter every 0,384 seconds. Your tap frequency will be 0,384. And you must keep the sound alive First you have every tap right and it will sound like ‘abcdefghijklmnopqrstuvwxyz’ The second time some taps are missing and this time it sounds like ‘aacdeeeeijklmnoooooouvwxyz’ 
    When this would be the case I would understand why I get darker sound when the clocks daviates. Because the maximum 48 Khz that we could not hear in the first place should probably only reach 46 Khz or something, because data is missing.Not data that is controlled on the ethernet, but data that wasn’t sent by missclocking in the first place. When data is missing every frequency will step down a bit, but we will first hear this in higher frequency ranges.This also will mean music will tighten up, details will be missing and the soundstage will be cleaner, but still everybody I know thinks it sounds better, including myself until now.
    This theory is not that easy to explain when I had used a 192 Khz Wav file, because there would be four times more data and the theoretical maximum analog frequency would be 192 Khz.Still Jitter should imo be explained in the analog conversion 

    • i think you are overcomplicating it.
      I think it is mostly about noise and vibration in the DAC/Clock and how to make sure to feed the DAC/clock with minimum amount of noise and vibration. One way of achieving lower noise is to start messing with additional clocking but I think that also adds a lot of complexity that isn´t really needed.
      My view is that if you focus really hard on the noise issue only, from all aspects, then your DAC clock will do its best with the incoming signal.
      Start with a very clean ethernet signal from a good (cheap) WiFi extender, like the TP Link TL-WR902AC v4, with a good power supply for it. From there on you can filter even more and use good shielded ethernet cables (they do matter, even more expensive ones, when the noise floor is low!) into your streamer.
      That is my two scents on this topic.

      • Hello Tobias,
        I think I’ve done this. Noise is controlled to levels you won’t see much. I started this search for understanding in 2018 when I wrote a complete report on the PS audio forum, complete with measurements on the analog output. I already had problems for two years then. I have worked with 3 different high grade streamers in which high grade clocks up to 6 were used in my system.

        Everything besides the streamer was powered by Farads from 2019 or something to solve the problem, but the troube I auditioned became worse and worse every time I tried to solve it by adding better equipment. At first the sound was always on a level that could not be described with words, but the problem is that the sound became dark and bad over time everytime. Last year I sold my streamer and partly bought new equiment and started over. I only use one clock and it seems I can get the same results as with a streamer, but now the SQ lasts longer, so I’m on the right way and almost there 🙂

        We don’t have Wifi in our home for other reasons.

        • Little correction! The problem actually seems solved 100%. The sound did not deviate a little bit in one day. In the last few weeks I already had a very stable sound, but I still could hear a little deviation after 8 hours listening. This seems solved as well now. The only thing that’s left is understanding 🙂

        • Did you feed that good streamer with the cleanest possible ether signal back then? I mean, did you really go all the way in terms of making sure the ethernet connection was the cleanest possible.
          If your streamer has a top quality poser supply the noise floor will never be lower than the incoming ethernet signal anyway, which will pollute the streamer, unless it is an amazing streamer that cleans up the signal itself.

          • Really, I did go all the way in every way one could think. My hobby is not driven by money but curiosity, I have used up to the Audioquest Diamond ethernet cables and several others. All sorts of high grade cable types in my system. From 2008 I’m aware ethernet cables do matter (due to a German magazine of 2008), but this seems overrated by far. When other aspects are ethernet cables should only slightly matter. I discussed ethernet from 2011 to 2014 on the PS audio forum, wrote about 500 articles about ethernet, but I have another vision these days 🙂 I compared computer hardware and software from 2012 to 2015. Built 3 high grade headphone amps, one really high grade pre-amp (with inductive volume control), several power supplies, modified high grade streamers, helped friends with their systems every time, but still I couldn’t solve my own problem 😉 for longer than a few hours. I did solve it at least a few hundred times, but it never lasted.

            Nothing of the above was profesional, but most was fun to learn.

  9. I have to correct myself on the point I made in the comment of December 28, 2023 At 11:41 am
    I wrote * A normal PC streamer can sound very good when a USB to network device such as from sitecom is used. This seems to be a better starting point than connecting to a stock network card.

    This was a mistake and I fell into my own trap 🙂
    Due to the Sitecom USB to network converter I heard a more tighter soundstage, less noise and almost free of artifacts. Very close to a live performance for as far as I know audio reproductions. Still I heard little clock daviations (because I know what this does to the sound). It seemed that the Sitecom was the problem, which makes sense when you know that the USB clock has another clock speed as the ethernet and DAC streamer.

    After I removed the Sitecom USB to ethernet converter. I had a gap to close, because I got a less stable soundstage, less atack and more artifacts. Still it was thé right move, because I could add a few better powercords, and I could move the PC one switch up in my switch stack. This closed the gap a described for 90%, but besides this the sound became a bit more natural as well.

    So I was wrong about that point, but it fits the theorie better now 🙂

  10. My conclusion
    * Crystals and clocks seem to be the most important factor in networks used for audio. The better the clock stability and power conditioning the more change the clocks will interfere with other clocks. This is not a problem when other clocks should run on the same rate, but clocks that should run on another rate should not be influenced and of course this goes as well for the other way around. 
    * Non audio related devices with several clock rates should be isolated as good as possible.  think about PC’s and Routers
    * Switches could be used for isolation. For isolation use only 1 port in and 1 port out for the best result. Don’t use ports next to each other, because inside the switch 2 ports (first port uneven and second even) are combined on the same transformer.
    * When there’s a need of a thirth cable in a switch, don’t connect this to an isolation switch.
    * Switches with 10mm crystals isolate better, than switches with 3.2mm crystals.
    * High grade power supplies can stabilize crystals to a level where clock synchronisation between several devices can occur
    * The amount of switches up to 10 (when high grade power supplies are used) is not an issue, because when 1 port in and 1 port out used. The benefit of clock isolation and the rebuild of the signal weight clearly up. There’s no downgrade in sound quality that I could point out.
    * When clock synchronisation between different clock rates occurs. The sound can tighten up, it will lose vividness and will sound darker.
    * A normal PC streamer can sound very good when a USB to network device such as from sitecom is used. This seems to be a better starting point than connecting to a stock network card.
    * In situations where a Streamer DAC is used make sure at least one isolation switch is between the Streamer Dac and the PC.
    * In all situations only connect one cable to a Router Lan port and from there the first connection should be to an isolation switch.  
    * It would be better to not use the very best power supplies on the switch which is connected to a PC or Router. It then will be more difficult to prevent clock (interference or synchronisation).

    When the above is right. You probably will get the very best sound out of your system. Even more when you already make use of a high grade streamer. At least is a very good starting point for dotting the i’s and crossing the t’s

    Good luck!

        • Thanks Wijnand! May I ask what type of ethernet cables you used during the tests? Sorry if you have already shared this information.
          I ask since i have noticed that the more you lower the noise floor, using switches, power supplies and LAN isolators, the more you need to protect that low noise floor with good (rather expensive) ethernet cables.
          My personal felling is that if you “only” focus on the noise floor, to the extreme, then extra clocks are not needed. Only a good last clock.

          • Hello Tobias,

            For sure! I make use of standard Cat5e s/Ftp and after the last switch to the Streamer Dac. This one ROLINE S/FTP (PiMF-) patchkabel Cat.8 (Klasse I), LSOH, soepel, geel, 2 m


            In the past I also have used several exotic cables up to extreme prices, but I promise normal cables can be great as well if your network is in good condition.

            I only use a OCXO clock on the very last switch. A switch with a dual powersupply. One for the clock and one for the switch. Really It’s a big deal and I can’t audiowise live without it anymore.

            Yes indeed the better and lower the noise floor the more you need to isolate the clock for interference with other clocks (I think this goes in both ways). It will gain more and more grip when it warms up and stabilizes. The more grip the worse the network quality for audio will be without isolation

          • Alpha Audio, I believe you have listened with the Networks Accoustics Muon Pro, right?
            That is an amazing filter ethernet filter.
            Isn´t that the proof that additional clocking is not needed (from a clocking perspective), since that involves no clocking and it just makes the most out of your DAC clock instead, by giving it the cleanest possible signal.

  11. Second session down, In this session I measured the crystal of a switch. Without the cables connected the crystal ran at 25.0040 Mhz. With the cables connected it directly ran at 25.0005 Mhz. Just like the Pink Faun Ultra clock. So they do seem to synchronize. Actually I shuld run this test during a few day’s and measure them on 2 channels to see if they reach an absolute synchronisation.

    I also received a second new D’link DGS108 today. This one was tested with its stock powersupply and was located between the PC and a Netgear GS108E. After 6 hours the sound was completely different to what I auditioned in the last few years. The sound was very open, a bit brighter, less tight. The soundstage was not as deep as before. Actually the sound was pretty relaxed and easy to listen to. I was asking myself if I liked the sound and yes I did like the sound pretty much, but could I accept the sound for a longer period no for sure!

    What was so different? I think the easiest way to explain is that I normally can position my loudspeakers half a mm by half a mm. I can move my loudspeakers 40 to 50 times to get them in perfect position, but now the loudspeakers were already good enough in their position. I think the precision is normally on a very high level and that’s not the case with this switch in the stack on its stock powersupply.

    I’m not sure if I already wrote that the 10mm crystals like in the D’Link DGS108 and the Netgear GS308 seem to isolate other clock frequencies better, probably due to a slightly less stable signal as with switches like de Netgear GS108E which has the 3.2mm crystal inside. So routers and computers seem to have less impact on the sound when these switches with the bigger crystals are used. This also is true when a less stable powersupply is used. The impact of routers and computers seem to go down. The overall sound of course won’t be better due to the less stable powersupply.

    I’m again really surprised how dynamic a network can be. My system already showed so many faces (really different systems). Which face should I choose (which network solution)? I guess it all depends on the goal. Feeding a high grade streamer-bridge or a streamer. In my case directly into a streamer dac. I like super precise, tight, with lots of detail and natural saturated sound colors, a big natural soundstage.

  12. One session down. A comparison between the PC network card vs the PC USB network output.
    At first the PC was connected to a Switchstack of a Netgear GS308 into a GS108E into a GS108E into a D’Link DGS with OCXO clock.

    The sound with the Network card is very open and detailed voices are surrounded with natural air and instruments for example the guitar strings are so natural.

    The sound with the USB Network is very powerful. Every detail has more impulse strength. Voices have less air, but surrounded by more black backgrounds. Guitar strings maybe are a bit to spiky

    On the second occasion I moved the PC one switch up. PC into GS108E into a GS108E into a D’Link DGS with OCXO clock.

    Both situations described above lost a bit of their qualities. It seems switches with the 10mm oval crystals. Seems to isolate the routers and computers better. Routers and computers use several clock frequencies that are different from the clocks from the network, for example USB is at 24 Mhz. It could be that the network is troubling these signals or the other way around.

    In all situations I measured the Pink Faun Ultra OCXO on top op de D’Link DGS108. This clock showed a near perfect waveform like a sinus and it measured an exact 25.0005 Mhz every time.

    In the third situation I restored the switch stack and skipped the D’Link DGS108 with OCXO.
    Now the USB network card sounded best, but it’s really clear the DGS108 with OCXO is missing.

    Tomorrow I will receive a new D’Link DGS108 for the next comparison without the Pink Faun OCXO. 

    • Hey Wijnand, I really like your work and your write-up´s, much appreciated!
      I didn´t really understand this test though, with a computer connected.
      In my mind, the whole idea with the switches (digital domain noise reduction) is partly obscured by having computer connected to the HiFi components? I mean, then you are going to have some noise going in that route which will mask much of what you are doing on the ethernet connection. Don´t you “need” to use a streamer only to really discern the differences clearly? Or, what am i missing?

      • Hello Tobias, Thanks 🙂

        Yes, that actually is the problem! With streamers I had the same issues even with really high grade streamers. I think every streamer is some kind of a computer. They use several protocols, like Ethernet, Wifi, USB, SPDIF, they have a chip set and a processor. In any situation a streamer will need a network connection. No matter if the streamer is used as a bridge from the network to the audio system. All the same issues will be there! The network with its trouble will still be there.

        That’s why I started after 7 years of “trouble” with a “normal” computer and kept it isolated and separated from the audio system. For an end-point in my opinion it will be easier to deal with a switch than with a complete streamer.

        The reason I write down my experience is that it really opened my eyes, because it really can provide that special touch of music. Something that was really difficult and unstable with a streamer. It’s difficult, but basically it’s the most logical way. Give audio an ethernet connection and a clean signal. The alternative is for me a streamer with all its parts, protocols, software that should be near perfect in all aspects.

          • Hello Martijn, Of course, that is a good thing to mention 🙂

            Network is the basis of streaming. Inside my Dac is a streamer as well which is partly used. Still every streamer needs ethernet or at least USB. My experience when the network is not great, a streamer fed by it won’t help that much.

            Starting at the source “the network” is what I’m focusing on. All the other things like streamers on several levels could not satisfy completely. There’s always something not completely right and I think this comes through the network. This source (the network) signal seems so incredibly dynamic for sound quality. That it can outperform or mess up the best streamers. Still a combination of a great network and streamer can be ultimate.

            For myself. My network has grown out of curiosity the last year to pretty much the same financial level as a great Streamer. I think this is more a technical journey than a journey into streaming. I mean I hope to provide some quick tricks to upgrade the network. Not something for a complete copy. Which on the other hand could be interesting 🙂

      • It all started in 2009 for me. When I bought a PS audio PWD1 with an ethernet bridge. This device replaced my Meridian G8.2 cd player. In only a few day’s it was clear there was something really different in comparison with the Meridian. Through the years a lot of computer hardware, ethernet cables, etc, was tested. In 2016 I bought a streamer with an OCXO clock on the mainbord and an I2S bridge. It was good but not good enough to call off the search. So in 2017 another streamer, a better model was bought. This streamer had 2 OCXO clocks, one on the mainbord and one on the USB bridge. The clock on the mainbord was a 25Mhz clock and the USB needed a 24Mhz clock. This streamer never sounded good to my ears. So I replaced the mainbord by an Asus Z something, but I remember it had a 24 Mhz chipset. So I could use two 24 Mhz clocks in this streamer. This streamer sounded much better then the original streamer (more open, better clarity). Later
        when a newer model again had a different mainbord. We did a comparison between the new model and my own modified streamer and indeed at that moment the new model sounded better, so I modified my streamer again with the newer hardware. The trouble started again. Something was clearly wrong. I think this was still 2017. The search went on for years. Sometimes my system sounded good, but most of the time it was pretty good.

        Two and a half years ago we moved to another home and the trouble became much more severe. The trouble was mainly that the system sounded really good for a while to really bad in a few hours or day’s. In this period I already had 5 OCXO clocks in my system. 3 of them were 25 Mhz and 2 were 24 Mhz. I really spent a few thousand hours on the search and almost 150 topics were tested.  This went from sticking antennas to the laminate floor to 20 new powercords. From drilling a new earth electrode…. to another internet provider. From building and modifying power supply’s to buying new and better equipment. Etc. etc. Almost a year ago I started with an OXCO clock on switches. I had two models, but the trouble went on. Also a year ago I bought a few new high grade powersupplies with exoctic fuses. I asked every professional I knew what could cause my troubles and I also questioned the switches several times, but nobody ever had such trouble 🙂

        Well in the end I started with the network again. I bought almost 20 new switches, a more professional router, 4 pairs of media converters, and 300 meters cable. Yes I wanted to attack the problem once again. And this time I really found a solution. Isolating the router and the streamer by switches. A few switches seemed capable of isolating better and other switches sounded better. Well with a combination I could have both. It was a switch stack of 6 switches powered by four high grade power supplies and an extra switch with an OCXO clock fed the DAC. Yes, it was good and stable again.

        Due to this topic I looked into the question again. I bought a new switch and connected an OCXO clock to it and powered the switch with 2 separate power supplies one for the switch and one for the clock. The sound with this new switch was really better in some aspects, but it lacked in another aspect. It was possible to solve the aspect where the system lacked a bit by adding another isolating switch. Still it was clear that this new solution with two separate power supplies was even more trouble on this particular aspect. Finally last night I found something that was reproducible. Different clock frequencies are interfering. In my streamer is an USB to ethernet adapter, because this one sounded better than my normal ethernet card. I will explain why I think this was better in a bit, but first when I removed this adapter from the ethernet the sound really became better. The other way around when I removed the switch with the OCXO clock and stuck the USB adapter in my streamer again the sound was also good. Different but that one aspect that lacked was gone completely in both occasions.

        Well now the calculations.
        A computer chipset runs mainly on a 25 Mhz clock
        Ethernet runs on 25 Mhz
        USB runs on 24 Mhz

        The audio Dac clock’s run on 24576 Mhz which is actualy meant for sample rates of 48 Khz, 96 Khz and 192 Khz

        There’s also an audio clock that runs on 22.5792Mhz
        This clock is actually meant for Music sample rates 44.1 Khz 88.2Khz and 176 Khz. But this clock is not used on the DA converter. Therefore 24576 Mhz is used.

        And of course there’s DSD
        Also a computer is build for 48 Khz mainly

        Lately I visited Alpha Audio and we measured an OCXO clock I have. This OCXO should run at 24576 Mhz, but in the first measurement it only runs on 24200 Mhz. It was troubled by the other OCXO that ran 24 Mhz. When we disconnected the other OXCO it showed 24576. So it was troubled by the other clock.

        Last night I did some measurements on the new switch with an OCXO clock and the signal was realy strong. It went through 4 switches before it could trouble the USB ethernet module, but it did. It became much worse when I removed a few switches that isolated the two from each other. After this I did some extra experiments and it came out that all clocks seem to interfere with each other. Especially the router and the streamer have produced strong clock signals that are not audio related. I have built a lot more intel through the years that makes sense today while I was still in the dark yesterday. I will do some more measuring and tests in the coming weeks, but there’s no doubt that the clock interference between computer hardware and digital audio hardware is the problem. I mean when a switch is slightly out of phase it is not a problem, but when it pulls the audio clock’s or maybe the clock’s in the computer streamer out of phase it could be a problem. I even think and I will measure that later that some clocks are completely synchronised due to the strong clock signal. And in that moments the sound my equipment produced was super tight solid and dynamic, but there’s was clearly one aspect wrong

        Routers, switches, computers etc should be isolated as good as possible. I will soon send my solution to Alpha Audio and this solution really deals with all the trouble I ever had with the network and the clocks. I’have heard the same trouble in many other sytems on a less severe level, but still there’s a lot to gain for many people

        Still clock do make a difference in switches and networks. It was never about the data, maybe some noise, but mainly interference

        • A little correction. Strong signal should better be read as a stable signal. I mean I have to check, but even though fiber the clocks seem to synchronise. It seems when one of the clocks is very precise the rest will be troubled. I think when all clocks are less presice there’s nothing to follow and the trouble will be gone. This is really unbelievable, but trouble showed everytime while it’s tested exesive in the last few months. Still I will check this again in the coming weeks 🙂

  13. Yesterday I received a new D’link DGS108. I replaced the GDS105 with Pink Faun ultra clock by the DGS108. In just a few seconds it was clear that the basic sound was already very different to the DGS105 basic sound. For a direct comparison I let the DGS108 break in for 8 hours. This probably is not enough to break it in for 100%, but it is close enough for a comparison.

    The D’Link DGS105 (without clock) is a pretty silent switch. It has some similarities as a Netgear GS108E. Still the Netgear is a better sounding switch

    The D’link DGS108 is a very open sounding switch. It is not as silent as the DGS105 or the Netgear GS108E. The DGS108 has some similarities as the Netgear GS308, but the DGS108 is the better sounding switch, although the soundstage is pretty troubled. I also connected the DGS108 to the ground of the powersupply, but after an hour I was missing the touch of music. It became pretty analytical… read dead.

    Today I opened up the DGS108 and the big 10mm oval crystal (trough hole) showed showed inside. In the DGS105 already has a 3.2mm (SMD) crystal.

    The above seems very interesting when one knows which crystals are used in the switches. The sound similarities could be due to the crystals. Because the Netgear GS108E also has the 3.2mm (SMD) crystal like de D’Link DGS105. And the D’link DGS108 has the same crystal type as Netgear GS308.

    Today I took the Pink Faun ultra clock and connected it with the new D’link DGS108. I also drilled a separate power connection through the case. The Pink Faun ultra clock is free of potential so a seperate powersupply can be used. I did this to eliminate the thought that the powersection could be troubled due to the clock.

    I connected two Farad power supplies to the switch and fired it up. The first few minutes the sound was very dull, but after 2 hours the switch revealed a complete other system.

    The D’link DGS108 powered by LPS without an external clock showed a very open but troubled soundstage. Voices and instruments had that brighter sound, which is pretty nice.

    The Netgear GS108E powered by LPS without an external clock showed a slightly less open soundstage, but the soundstage is less troubled with this switch. And the same for the D’link DGS105.

    The D’link DGS108 with dual LPS and with the external clock. Brought that super tight sound, very dynamic (like dynamite). The sound is a bit less bright (for which I lately found a solution). Everything is super precise. Really it takes a system to a complete other level. Not exaggerated for the tiniest bit 🙂

    When I look at the switches in the pictures of this topic. I see that the crystals are replaced by an external clock and connected to ground. Normally crystals sit in the middle of a circuit 10Kohm for and behind the crystal. I’m not sure if connecting to ground is the best solution. I must say I have tried that in the past as well and I didn’t hear a big difference, but still it measured 25Mhz to ground as well. This is different from VCXO, TCXO, OCXO, which do connect to ground.

  14. I think I should come back to my words that it is subtle, a switch with or without an external clock. It’s actually far from subtle, it’s massive. While playing the track of the album Unusual from Marian Hill track 2 differently. It’s very clear in just a few seconds without the clock. The track is nice, but with the clock I have to hold tight to my chair, because it feels like it’s trying to pulse me out of my chair. Wow what a control. In other tracks piano keys hit edgy like it should, without the clock the hits are softer. A foot drum stands very precisely in the soundstage, without the clock the foot drum is somewhere.

    Is it all better with the clock? No! with the clock the system needs more space between the loudspeakers. It’s only 5 cm in total, but that’s a lot because normally I’m always puzzling about a few millimeters. My systems drives everything to the edge in the listening room there’s not much ‘room’ left for tuning. I even have to use some stock cables to limit the system. Without the clock I have much room for tuning. The loudspeakers have a better possition. I can use high grade cabling etc.

    Just a thought, because you have reached the opposite with the test. I’m sure you have used very high grade cabling, but for as far as I know the cable from the clock to the switch should be as short as possible. This cable will act like an antenna to noise. Coax is possibly less good then a shielded piece of signal cable. Of course one of the wires is ground, but a shield around this ground wire can help.

    • I must say that i agree with Alpha Audio that the clock, and involved extra cabling, is probably just impacting the noise level. I am personally very skeptical that the actual clocking would do anything. But, this is just my amateur thoughts…

      • Hello Tobias, I think we are all pioneers on this topic. No amateur or professional seems to understand completely what is happening and what is important. The most important thing is that more and more people understand that digital was never really ones and zeros. Actually it’s just an analog signal that we interpret as a binary string. I think analog is the keyword to understand how it can be solved. Indeed a better clock has less or at least a different impact on the signal. For sure it’s not about timing. Still a “perfect” analog stream which is shaped perfectly for digital reading seems what is needed and maybe the clock can help in a certain way 🙂

  15. I have done a listening session with the Dlink 105 with and without an external clock. The most interesting thing is that in my situation the stock 105 sounds ‘lighter’ I would say 2 bits lighter:-). It’s very difficult to say which is the better for any situation, because it always depends on what the system needs. On my system the Dlink 105 with clock brings more attack, more weight, better sound colors, a more black background, but due to the attack and weight the picture maybe sounds a bit upfront, the soundstage is deeper as well, but this seems to depend on the track that’s playing. In other tracks the depth is not necessarily better. It’s subtle, like one to another power cord, so it can be heard in a few seconds. I have used a Farad powersupply with a QSA red fuse both on the stock Dlink and a Dlink with a Pink Faun ultra clock.

  16. Hello Jaap, Very nice! I should try the coax as well, this makes it far more easy to swap a clock. I have tried the Sbooster as well on several occasions and switches, with some switches it was hard to tell which was better the stock or the Sbooster. I think it depends on the powerinlet of the switch. Switches with an external clock, didn’t work well with the Sbooster in my opinion. The soundstage was very clear, but troubled, it actually was bad. I’m aware the Sbooster is already a low noise powersupply. It depends on the type of clock. For example a Connor and Winfield OCXO was not clearly better then a stock crystal. Although it always will be subtle, with the right clock and powersupply it can be very welcome. Thats why I would say; there’s room to explore and in some situations it really can make that difference.