[Wijnand]

Maybe he means the same. Jitter time based.

More important is the reason that it happens. I think when crystal can’t run on a stable frequency. They must capture the data in less than a complete clock cycle. Which means half filled glasses. So when all clocks would synchronize downstream. We would get good filled glasses. We can do this by trying crystal to listen better downstream than upstream. The ferrite core will work against the stream which means the data that is sent will have more influence than the answer will have.

[Wijnand]

Let’s think of 8 glasses which we can fill with water. Each glass represents a bit in the byte. Now when there’s water in the glass it’s true when the glass is empty it’s false. Let’s put water in 4 glasses. One we will fill with 20% water, one with 24%, one with 30% and one with 54%.

This means the outcome is 01010101

Now we do this again but now we will fill those glasses with 100% water

The outcome is 01010101

Will the sound be the same when we make it analog by pouring the water through resistors?

 

[Wijnand]

Yes of course! It’s very good you keep repeating that, because it’s 100% true. That for sure is not the problem, but when we look once again to reply 197472 where I have drawn this 8 bit dac. When a digital value is 100% correct it doesn’t mean the shape of the bit that actually is read analog is perfect as well. Let’s say the data shapes are 100% the sound will be 100% as well. When the data is less than 100% for example 25%. Then the digital value can still be 100%, but when we use this 25% shaped data and transfer it in an analog voltage you will hear the build of the 25%. When you look at the vector. Then you see that each line will be more horizontal (smaller corners) this means that each frequency will be actually lower than it should be. High frequencies will suffer much more of this than lower frequencies

 

 

[Wijnand]

Yes, but than it still should not become as worse as without ferrite cores at all. The ferrite core in the last section should still do that same trick (when it was about noise. It also should be audible at once and not 24 hours later). But as I mentioned the sound was dark after 24 hours. Just like it was without the cores

[Wijnand]

Also in the test of Hans Beekhuizen with the help of Alpha audio the Wavecrest showed a lower phase noise. This supports the above theorie as well Crystal are less floating around. Which means more stability at a certain frequency.

[Wijnand]

Well maybe then I should use my router’s Wifi directly on my DAC streamer. Then there is nothing in between.

But first I like to share a theorie of which I think the problem is really based. A problem that you will not have when using Wifi. I even think that when you deal with this problem then you won’t need Wifi for the reason you thought you needed Wifi 🙂

Actually I should draw it, but then I first need to post it on another forum and then connect it here. That’s what I did with the other picture’s I posted.

I will try it with words and to make it more simple let’s think ethernet runs on 25hz instead of 25Mhz. So the ethernet runs at 25hz.

1. We have a Router (25)

2. We have a PC (26)

3. We have a Switch (24)

4 we have a Dac / Streamer (23)

These ethernet devices need to run at the exact same speed to communicate with each other. Each device has a crystal with a certain source speed. Not one device has the exact same source speed. Let’s think that each device runs on the source speed that I have put behind it. Now we’re going to run data downstream from the router to the DAC.

As soon as the router starts talking to the PC the PC will synchronize automatically to the router’s source speed. So the PC will meet the 25Hz. The PC will start talking to the switch, but now the switch must synchronize to the speed of the PC. The switch will forward the data to the DAC. This happens. I have measured this.

Before I go on I must explain that the blockform that is sent will be slightly in another time frame. Because it must pass a galvanic isolator transformer on each output and input. So between two devices are two transformers. Lets say the blockwave is delayed by 4 nanoseconds. Then it can be stored in the receiving devise and sent to the next device. Due to the fact ethernet communication goes in both ways. This means when a device is answering or talking back. It must be flexible to do this on its own terms. When it talks back there will be (due to the transformers) a few nanoseconds delay as well. But a complete clock cycle is 40 nanoseconds. This means now the device that first sends the data must see to receive the answer. This means it must synchronize to the answer speed. Before it can send new data at its own speed.

For audio the only thing that matters is the downstream. The answers are not audio related. When we only would have downstream (one way) I think our problems would be solved. I think these crystals that float around cause the main trouble we hear. Noise comes in second place.

Well the trick is to stabilize the downstream. Make the power downstream bigger than the power for the answering power. This way the sending crystal becomes some kind of a master. I have done a lot of tests that point in the direction of theorie. Also the latest with the ferrite core which I will explain.

The ferrite core in the last section of ethernet does the complete trick in my system.

When I put ferrite cores in the other sections as well they seem to do almost nothing. Then I have lowered the data power in each section.

[Wijnand]

Yes you can buy it 🙂

Most ethernet is sold as CCA which is an alloy of aluminium with a copper layer. This is the cheaper and lighter variant. When it’s 100% copper they will always put that in the specs as well as in the case of CCA.

Just like you. I haven’t tested your situation as well. I will look into that for sure and maybe I will learn that there’s an even better way 🙂

The only thing that I can say is that this situation can compete with the very best streamers. From 2014 I bought my first high grade streamer and after this one I have had 5 others in my system. Not one outperformed the current situation.

I could install the rabbit ears on my DAC and on one of my PC streamers to test the WIFI solution, would that make sense?

[Wijnand]

Today I have built two new cables with Ferrite cores. This time I have used 100% copper instead of CCA. This brings that extra touch in music. That few % more color, not warmer, but better sound color. I have filled the cores with warm glue.

We had the discussion about more equipment, more power supplies would reduce sound quality. Well I added a D’Link AQvox between the Jcat and the Dac and powered it with a Super 10. Between the JCat and the switch I have a ferrite core cable and also between the switch and the DAC. In my opinion this is as good as it can get 🙂 Really better than without the switch. Only the clock switch brought even better detail, but in comparison to the current solution. It’s makes not so much of a difference!

The complete stream

1. Wan fiber converter on Farad Super 3 – 5Vdc with QSA Red/Black fuse

2. Router Draytek Vigor2962 modified to work with Farad Super 10 – 12Vdc (Stock fuse) (fancy fuses make the sound warmer, which is less good)

3. PC streamer with Jcat Netcard XE fed by a Farad Super 3 – 5Vdc (Stock fuse) (fancy fuses make the sound warmer, which is less good)

4. AQvox switch Fed by Farad Super 10 – 5Vdc (on  both sides of the AQvox a ferriet Cable) (Stock fuse) (fancy fuses make the sound warmer, which is less good)

While listening no extra connection to the ethernet system.

For sure it’s not the cheapest solution, but to my ears it works better than the solutions with very expensive streamers alone. Improving the complete stream matters. It outperforms my own music library on SSD.

I think it’s time to enjoy listening for a while 🙂

@Tobias thanks for pointing me in the direction of the Ferrite cores. They really did the trick 🙂

[Wijnand]

Multitest external 12-volt power supplies – Alpha Audio (alpha-audio.net)

[Wijnand]

Yes there’s something there, but I think it’s not about a switchmode or a linear powersupply. It’s about the type of the output stage. Actually a few years ago I spoke with the head engineer / owner of Farad power supplies about the Super 3. The super 3 had in those day’s a tiny elco capacitor in the output stage. It was a very fast one with very low esr. The first few hundred of the Farad Super 3 were built like that. In the measurements of Alpha Audio’s multitest powersupplies you can find that type of Farad. Later the Farad was modified and the little elco was replaced by an MKS capacitor which was easely 100 times faster. Shortly after this also a little ceramic capacitor was put on the voltage regulator, which made it incredibly fast and stable. I had 3 Farads of that first type and they sounded better on analog equipment than the newer version, but the newer version sounded far better on digital equipment. Those days there were thought to build an analog powersupply line and a digital line.

I started modifying my own old types and tried almost 20 different capacitors in the output stage. Everytime the sound that it produced on equipment became different, I learned a lot those day’s about output stages.

If you want a powersupply for analog equipment. Make sure it has an elco in output stage or a MKP capacitor. When you want a powersupply for digital equipment take one with a MKS or MKP capacitor.
The MKP can be used for both applications, but they normally are a lot bigger than a MKS.
The Super 10 has a MKP output stage, but a Super 6 is coming with an MKS output stage and that one will fill the gap between the Super 3 and the Super 10.

[Wijnand]

Haha 🙂 yes sorry, but stating the obvious makes it even more clear. Just for another test in the meantime I have switched back to the Jcat with the 9 winding ferrite to the dac. The sound is still very good, but a few % more ‘grey’ and maybe more relaxed, because the soundstage lost some detail.

The few % more grey you will only hear the first few minutes after switching from A to B

[Wijnand]

Yes of course! That’s not even the question. The clock does nothing with the timing for the network. This is all about noise 🙂 and maybe it helps with a better signal built.

Actually it probably can even get better, because I needed the powersupply of the Jcat, so the Jcat went at least 20% down in quality, still the switch clock with a dual high grade powersupply lifted the sound up quite a bit 🙂

The clock is completely galvanic isolated from the switch.

[Wijnand]

Well the clock is not even warm. In only another 5 seconds it is so clear this clock is something special. Every detail thight’s up and a super system is born! The sound stage was already very precise, but now its 2 times more precise details have more impact and became completely free. Low frequencies are brought with more information. I will wait 48 hours now to see if the ferrite cores can prevent the problems with dark sound that normally appeared in my system.

[Wijnand]

Thanks 🙂

Well I have one test down. I have built 2 short cables with ferrite cores. One with 9 windings and one with 14 windings. From the Jcat I went directly into the DAC using the cable with Ferrite and 9 windings. After this I have tested the Ferrite with the 14 windings. I could not hear a difference between the two. Well when I would listen harder then maybe I would hear a small difference, but I guess that wouldn’t make sense.

After this I have positioned a switch between the Jcat and the DAC. From the Jcat 14 windings to the switch and from the switch 9 windings to the DAC. For the switch I have used a Farad Super 10 in floating from ground mode. The sound that I heard directly from the Jcat was pretty much the same now with the extra switch. Maybe now there’s a bit more organic sound, but it also was a tiny bit out of balance due to energie in lower frequencies.

After this I replaced the Farad Super 10 with a switch mode powersupply which is in floating mode from ground by default. Well the listening session only took 5 seconds. That was not high-end. The ferrite cores could not solve and repair the noise of the powersupply.

I will now replace the switch with the ultra clock switch. I’m very curious what this will bring. Sit tight! 🙂

[Wijnand]

Hi Tobias, I really believe you have the knowledge about digital systems. I can’t argue that. I am very happy with your input. It helps me to prevent a certain tunnel vision on the subject.

Still my experience with many other people that also thought they had the knowledge including myself are proven to be wrong so many times. Once, way back in 2013 I spoke with a guy that claimed to have all the network knowledge. He thought I was wrong and he was trying to convince me over and over. We spoke and discussed for almost a year. In the end I brought him the conclusion he was wrong. From that point he was telling me what I told him already so many times. This all due to the fact I worked with the result and searched for the cause as long as it takes. This thing has already happened too many times, I’m pushing and working against the odds. In 2015 I started over here on the alpha audio forum where bits were still 1’s and 0’s. I won’t say I was the trigger to the understanding, but I will say I already knew 🙂

Please in the meantime believe me that I have an effect that can’t be explained easily. Actually it can, but when a signal is on a certain level other problems seem to appear, problems that should be solved when the (sound quality) level that has been shown for short periods of time, must become the new normal 🙂

Maybe you are right so I will try out your solution for sure. I’m closing in, but that there is more to find than filtering only is almost a fact for me. But I hope I’m wrong this time and things will fall in place and seem to be as simple as they should be. Stay tuned 🙂

[Author]

Posts