[Wijnand]

🙂 It would be a relief to really understand what is going on, no matter the financial part it already cost me. I started this thread due to the fact it did not turn out as simple as I and many others thought. Starting over and over from scratch a few hundred times (which is not exaggerated a bit) each time with another approach I could never find a logical explanation. Until I found out the crystal seemed to cause the problems. From this point I can target the test on this aspect. Finally and time after time I now can connect things with logical explanations and make steps forward. When I’m wrong I will admit that, but for now I’m still more certain each day. Still I will try the solution without the high grade power supplies once again as well. I also will try the Wifi solution to get a complete view.

The most important thing is that we can say for a fact what is going on. In all situations one could simplifie things to an ultimate level when proven knowledge can point us the way. 🙂

[Wijnand]

Good morning, Yes I found that picture as well, but I’m not sure it’s the latest version. This one is only 100Mbit I think because of the 4 wires that are used.

Due to the fact the phase noise chanched in the video of Hans Beekhuizen. I think something else is going on. I can’t believe it’s only the filtering yet 🙂 Please maybe read this twice 🙂

What has happened in my system is that the trouble I had with dark sound reappears after I put a filter in each section of my small network. I think the dark sound is caused by bad phase noise. Due to crystals that synchronize and unsynchronize. Because the negative effect did not take place when I only put a filter in the last section. I’m positive that the strength and or stability of the signal in each section is more of a concern than the filter itself. I think the crystal of the sending device must not be influenced by the receiving device when it’s talking back. When the signal between the sending device and the receiving device is on the same level I think they will float all around and there’s no stable factor. One must be the clear leader.

I think it causes this same problem when I put a high grade powersupply on each device. A powersupply which is ultra clean and sets a far more clean and stable signal in the path than a SPS will do. This is measured and this is proven. When the signal becomes on the same level in each section the sending can’t lead that easely. One could think that this would be the same problem with a switch mode powersupply, but I think that the instability of this signal causes another effect and no crystal seems to influence any other crystal at all.

When putting the ferrite filter only in the last section this section became of different strength. The inductor will weaken the signal a bit. The cleaning factor of the stream built by a high grade powersupply is much smaller than on a stream built by a SPS, but the weakening of the signal and the EMK’s will always be there. I think the downstream matter. The sending crystal becomes the master and the receiving crystal becomes the follower. The dac itself will be influenced less, because the crystal in the network does not float around that much and the phase noise level will benefit :-). I will dive deeper, but due to all the testing that I already did in the last few years this is my conclusion at the moment. For example in one test I have used 8 Pink Faun ethernet isolators. Each isolator is a filter as well all together I think they can bring a big filtering quality. Putting them all in series one by one weakened the signal that much that the sound became sharp as a serrated knife. It was perfectly readable, but the bits became very small

 

 

[Wijnand]

I see your point and still I almost try everything. Networks have surprised me too many times. I thought about this structure as well and for this purpose I have put a fiber converter several times next to my DAC. Still the copper connection Router -> Streamer -> Dac had the clear preference. I have tried 4 different fiber converter single mode and multi mode and fed them with different power supplies as well. This way there was only 1 powersupply at the side of the DAC, but the powersupplie on the other side had almost as much impact. Actually every ethernet connection has a galvanic isolation to the next device except for the shield. I guess with Wifi it will work better (I can imagine). I haven’t tried that, but the fun is that I have asked a friend of mine to try a powersupply at his transmitter side of the Wifi. It did make a difference and he was very skeptical 🙂
I just built a new ferrite cable with 9 windings. It’s between the Jcat and my Dac. At least until now It really does more than I thought. It’s on another level than what a powersupplie brings, but together they are great 🙂

 

 

 

[Wijnand]

Hi tobias, I must try out the Muon to get an idea myself, but first I will do some more experiments with the ferrite and I will dive in the Muon as well. I think we at least need a powersupply for a switch and a router, maybe a better power supply will lower the noise already for a starting point and less filtering is needed.

Do you mean additional clocking or replacing a clock by a more silent clock? I have done the latter, but I can’t use it because of the effects I get after several hours. I think these clocks can help in building a better signal, because a standard crystal has a pretty high phase noise level.

Just for the record. My relation with audio products is DIY only 😉

 

 

 

[Wijnand]

With the above theory you can see that higher values will drop the most. Lower values can’t drop that much. I think a percentage of each value must be combined to get the value that will be measured. Logically you can see the lines will be less steep and this means high frequencies will be decreased more than low frequencies. In fact there will be more low frequencies, because the higher frequencies will come on top of the low frequencies.

[Wijnand]

When we think about an 8 bit dac. Would the calculated value be the same as the measured value? I think with a perfect signal it would be, but it’s not about 1’s and 0’s but about analog build bits in a data container. When the shape of the bit is not perfect it still represents the digital value, but would it measure correct as well? I guess high frequencies will be lost first and low frequencies will gain body. See the attached scheme.

Think about what less shaped bits will do in a R2R dac. Every bit will find its resistor and all bits combined will output a voltage.

[Wijnand]

Hello Pieter,

Thank you for sharing this. What you describe I really recognize in what I often have heard in my situation. About the background haze this indeed can easily completely disappear. Still in my experience with this, is that it can mean two things. 1. The sound is so incredibly pure without any noise. 2. The sound is missing details on the same level as where the background hase is some kind of detail too.

Low frequencies can increase strength by a theorie that I will try to explain soon.

Relaxed sound can be a lack of detail as well. Just like a LP is most often more relaxed due to the lesser amount of detail than most digital files.

Less stress in music. Do you mean instruments and voices had less explosive impact? Or was it more free of noise?

Don’t get me wrong I’m not judging and I’m really open minded and trying to understand what the Muon brought in your system. Maybe you can react to the thoughts I described?

[Wijnand]

Thanks at least for trying. It really is appreciated. If we can understand what happens I’m sure we can make another step.

[Wijnand]

It took me a very long time to figure it out, but the sound of my system is really good nowadays and I dare to say that it is on a level which is not regular at all (at this level, there’s always a better system) :-). Even more with the ferrites in the last section. They are not a problem there (yet). 🙂 For sure I will dive deeper into this materie.

The downside is that I have heard my system can do even better than it sounds right now. It is really tempting to set that last step and to keep on looking. Which causes trouble in the end over and over.

[Wijnand]

Yes for sure 100%. But what is the reference of the time that’s needed to hear what we hear in our listening room. I mean when we listen to concerts miles away we can hear large time fluctuations due to wind and other weather influences like pressure etc. The fluctuations can be a few hundred milliseconds up to half a second. So what do we really need for a steady high-end reproduction? Of course it must be far better than that 🙂 We sit about 0,01Second away from our loudspeakers. When an audio clock will tap around each 40 nanoseconds at a precision of 15 picoseconds per second. And it will be even more precise when we look into it on a longer term, because it outputs data every 512 clocktaps and so we have a build of 48000 points. Still this timing can be near perfect, but because the signal isn’t built with the complete voltages I hear a dark sound. The image on the other hand is amazing.

[Wijnand]

Of course 🙂 Keep in mind it’s still just a theory

I came up with this theorie, because logically data bits that are not shaped well will limit high frequency response first. When a small amount of output voltage is missing on each clocktap due to badly shaped data, an analog audio waveform will have shorter spikes and because high frequencies are built with more spikes than low frequencies we will hear dark sound.

[Wijnand]

I think that’s a great illustration. I do that as well, writing it down and getting a better understanding my self 🙂  Hans for sure has far more knowledge on this matter.

May I add that we maybe should talk about datawords instead of bits for even better understanding. Let’s look into the R2R dac for another moment. When we draw vertically a container for 24 bits and we will draw the single bits vertically under each other in this container. Then we will see the shapes of the actual 1’s and 0’s. Not one bit will have the exact shape of another, they are all different. Together they will go through the resistors and will be combined at the output. They represent a voltage together. When we put another dataword with exactly the same data in the containers, but again not one bit will have exactly the same shape as in the first data container, because the shapes of the bits are actually built analog. Still they represent the same digital data value(bit perfect). Still due to the analog shapes they will create a slight other output voltage.

Personally I think that this is what we hear and not the picosecond of time difference. I think our brain must deal with all sorts of reflections which it will correct to a certain level. Besides this an audioclock of 24.576 Mhz will only clock out once every 512 clock taps for a 48Khz file.

[Wijnand]

In addition to the above. About driving a crystal of the receiving switch or device. Every ethernet device has a galvanic isolator transformer on each port. So receiving data is slightly delayed. This will also be the case for the crystal that will clock in the data (At least according to a measurement interpretation). At the moment the switch or ethernet device will change talking to another device it should fast synchronize to this other device to make sure the data can be received in good shape.

Please correct me, because I most certainly are off track now 🙂

[Wijnand]

I figured 🙂

The more we know the more it seems we can’t understand. Speaking of myself of course. I’ve been discussing ethernet for over a decade now and I’m still learning how little I know. For me it helps a lot when people with knowledge think with me. So thanks!

Wrong position. This was a reply on 197141

 

[Wijnand]

With the sinewave I meant the data waveform. In fact it’s not a sine wave. (see reply 197275 and 197277) On this waveform you can see another tiny waveform on top of it. That tiny waveform on top is not data, it’s noise, but there will also be noise that is so fast that my scope can’t picture it. Normally the noise that you can’t see will be smoothed first, but the tiny waveform that you can see on top of the actual waveform will be smoothed a bit as well by the ferrite (inductor). The more windings we make the more noise we cancel, until we cancel the main waveform as well.

What we can’t do with an inductor is repairing the waveform build itself.

With a high grade powersupply we can reduce noise in the first place by not creating it. Also the waveform builds are a lot better. I guess that this waveform is also driving the crystal of the receiving switch. The stronger the drive the less the flexibility of the receiving switch to talk with other devices. Still each device should capture the waveform that is received and sent by one crystal. I think this will cause a less perfect shape of the waveform itself.

I’m not sure if my thought about this is correct, but I think it can be explained best by a R2R Dac. With a R2R dac a dataword will find a resistor for each bit in the word. This way the dataword will be converted into an analog signal. I think everyone could understand that the width and shape of each bit going through the resistors to create an analog signal matters a lot.

I also think this matter on a conventional dac chip. And indeed still everything can be bit perfect at the same time 🙂

[Author]

Posts