Here is my current little DIY project, not very advanced 
I took a cheap TP-Link PCIe network card, cut the 3.3v to PCIe/motherboard and replaced it with external power through a linear voltage board. The result so far is very good, sounds better than my battery driven switch.
I haven't cut the ground to PCIe though, I'm afraid it might prevent the card from functioning, but it would be preferable if it was possible.
However, to complete it I would like to replace the clock as well, so does anyone know of a low-phase noise quality clock for 25Mhz that can be ordered online and in quantity of 1?
Btw, if you don't believe that tweaks like this matter for sound, feel free to do that but please don't post about it here. Lets keep this an informative DIY post and not yet another flame-war about "bits are bits". Maybe the mods can help keep it clean?
I took a cheap TP-Link PCIe network card, cut the 3.3v to PCIe/motherboard and replaced it with external power through a linear voltage board. The result so far is very good, sounds better than my battery driven switch.
I haven't cut the ground to PCIe though, I'm afraid it might prevent the card from functioning, but it would be preferable if it was possible.
However, to complete it I would like to replace the clock as well, so does anyone know of a low-phase noise quality clock for 25Mhz that can be ordered online and in quantity of 1?
Btw, if you don't believe that tweaks like this matter for sound, feel free to do that but please don't post about it here. Lets keep this an informative DIY post and not yet another flame-war about "bits are bits". Maybe the mods can help keep it clean?
Attachments
Unless I am mistaken, the clock is also used for low-level generation of the ethernet stream (all digital transmissions needs a clock). I would guess in this case they use a 25Mhz and multiply it to get to the frequencies needed for Cat5, Cat6 and Cat6a (100Mhz, 250Mhz and 500Mhz).
And the more accurate signal that the card generates, the less overhead to interpret it on the streamer end.
But anyway, unless the clock is a OCXO its not expensive, so might as well try and see if it makes any difference.
Its possible, but I prefer to get power externally, lots of electronic noise inside the computer. I use 4 AA batteries now for testing, sounds very good.
If I had a HDPlex or other linear power supply powering the computer, I might do what you said, but I have some cheap switched vanilla power supply.
The clock doesn't need to be accurate, the data recipient synchronises it's clock with the senders (self clocking via a preamble at the start of data transmission), this happens at layer 1 (physical layer), there is no overhead for the streamer.
Yes, but unless the clocks themselves are synced (which they are not) you will get jitter (not the kind of jitter we talk about usually in audio though) which will deteriorate the signal.
But as I said earlier, its not an expensive mod, a very good clock like Crystek CCHD-575 costs about $15, and if nothing else it will be interesting to see if it changes the sound. But I can't find the Crystek clock in 25Mhz which is why I started this thread.
The individual xtals on the PCIe boards do not need to be sycn'd, the data stream and the associated clocking mechanism is synchronised via the preamble from the sender, that's the beauty of the idea, which makes your idea redundant.
There is always jitter in ethernet/wireless networks, but there are buffers which take care of the out of order data packets and reassembles them in the correct order. There is no deterioration of the signal, ethernet is a robust method of data transmission.
How does this modified accurate clock only effect audio data packets ?
There is always jitter in ethernet/wireless networks, but there are buffers which take care of the out of order data packets and reassembles them in the correct order. There is no deterioration of the signal, ethernet is a robust method of data transmission.
How does this modified accurate clock only effect audio data packets ?
It will of course affect every package. From what I have read, the idea is to make it easier for the receiver of the data to interpret and handle a "better" signal, which will lead to less noise internally (in this case, the streamer).
But lets leave that discussion for now, you might be correct. I will still try to get a better clock and test, just for the fun of it if nothing else
Where you soldered the ground wire to the nic, i made the solder island connect with the crystal, great mod, so you could try that. But I wouldn't then also connect your power supply ground wire to that point.
On topic: I think Abracon has 25MHz clocks, iirc around $/€50 on digikey, haven't tried those though.
On topic: I think Abracon has 25MHz clocks, iirc around $/€50 on digikey, haven't tried those though.
In doing what Chris had said, the 12v could be stepped down to 9V, then to 5V, then stepped down again to 3.3V. I have an sd card player that does that using 2 lm78xx regs, an ams1117, and several 2200uf 16V capacitors.
Also a small package size/.01uf, higher quality ceramic as close as possible to the clock helps in other digital applications.
Also a small package size/.01uf, higher quality ceramic as close as possible to the clock helps in other digital applications.
The ground connected to the "shell" of the crystal? Could certainly try.
Currently I have ground from both external power and motherboard connected at the same time, I would like to remove the ground connection to the PCIe/motherboard to get more isolation, but afraid the card might stop working. What do you think?
There are lots of 25Mhz clocks out there, some with "gigabit ethernet etc" as suitable application, but I currently have a problem: the clock can't just replace the crystal, it has to replace the surrounding clock-supporting electronics as well, and Im not certain exactly where to "cut".
This post declares that you have no clue of what you are doing. Things don't work like you want them to. I think it's important that this is stated so that other don't follow this idea and waste money and potentially ruin perfectly working hardware.
//
I know exactly what I am doing, I am experimenting. And I will report whatever I find here, but even with the same clock the card sounds better than my old ethernet switch run on battery (and yes, I know you don't understand how that can be).
But surely you have something better to do that troll every thread I post in?
Have a look at any CD player PCB, locate the microprocessor chip and then its xtal. You'll notice that, on a properly designed PCB's, there's a fair amount of effort invested in segregating the ground fill around this microprocessor chip and its xtal, from the rest of the PCB ground fill. Furthermore, the power supply rails that supply the current to microprocessor chip and its xtal, are heavily decoupled from the rest of the PCB rails. Expensive CD players have completely different rails supplying the microprocessor sections, from the rest of the electronics located on that same PCB.
By decreasing the noise of any uPC switching / any xtal located on the PCB that shares the ground fill with the rest of the digital audio playback (including digital audio on Ethernet) system, you will improve the overall performance of that PCB. Any decrease in noise in the digital replay chain, will be immensely appreciated by human ears.
I am not an electrical engineer, but I think thats what I at least partially accomplished by cutting 3.3v to PCIe/motherboard and replacing it with linear voltage regulators and battery power. I would like to cut ground to PCIe/motherboard as well but I am not sure the card would work after.
But you seem to know a lot about these things, so feel free to suggest further improvements
The recipient device doesn't care, it simply locks onto the senders data speed. It can jitter all over the place and it won't matter. The device internal noise is insignificant in the scheme of things, the physical interface is a balanced transmission line and is resistant to noise.
If this xtal accuracy is such a problem then every device with an ethernet interface would fail to connect to anything else.
You have an incorrect concept of how Ethernet works, and since you're most likely using TCP/IP over Ethernet, the incorrect concept extends to that as well.
Without a lengthy dissertation of both technologies, suffice it to say that what you're attempting is a meaningless errand because of the way things work. Basic Ethernet includes error checking, and TCP/IP was intended to be highly tolerant of received data problems, as part of the end goal of a highly robust data transmission protocol.
There are many layers of understanding to be had here. I'd probably start looking for articles about Ethernet, gain that understanding first, then delve into TCP/IP. It makes no sense to expound on both in this thread. But lets apply a bit of logic here. Ethernet, the current variant, is designed to transmit up to 400 gigabits per second over several kinds of media, the most common being Cat5 wire. The system includes dividing data into frames, and error checking those frames, which in turn may result in the higher protocol, TCP/IP, triggering retransmission. Ethernet over Cat5 has a practical single-run distance 100 meters. TCP/IP is applied for robust, error free data transmission without limit to distance. Global is the norm. It's intentionally tolerant to any form of network data corruption that can occur along the way, through dozens if not thousands of devices scattered all over the world and above it. It makes no sense that a system like that would require, much less benefit from any low phase noise clock anywhere in the system, as that's not specifically part of the specification.
Simply put, you're solving a problem that does not exist.
Odd concept of fun, doing something that is entirely futile. But I'm also guessing that the test method that will be applied is fully sighted, long form listening.But lets leave that discussion for now, you might be correct. I will still try to get a better clock and test, just for the fun of it if nothing else
The questions to answer first would be: 1. What phase-noise is present in the current clock? And more importantly, 2. With a full understanding of Ethernet and TCP/IP, is typical phase-noise of a clock even a factor at all?
The answers, if you don't want to bother are: 1. Low enough and 2. No.
Again, if you really want to improve your audio, look into what's happening in the speakers and room. There are huge problems to solve there that will be very audible to anyone.
- Status
- Not open for further replies.