First of all we are talking about opinions. Yours, mine, others. There are no proof that measurements can measure every aspect and interpret them in the same way as human ears, so at the end measurements are a guideline not a proof.
Secondly, as you can see from some other posters, I welcome input, but when I hear an improvement over several days (which for me indicates no bias), I have to trust my ears. I have made some tweaks earlier that I abandoned because they decreased sound quality, and in those cases I discovered it on day 2 of listening (I guess that it means day 1 is "bias-day" for me 🙂 )
Thirdly, you and many others talked mainly about network layers, bit-errors, fault tolerance and similar things, which is unrelated to the tweak I am making. This tweak is about making an ethernet card that is as electronically quiet as possible.
Last edited:
Magnus, NDK SDA in common audio clock frequencies can be ordered from: NDK NZ2520SDA 3.3V 49.152Mhz 45.1584Mhz 24.576Mhz 22.5792Mhz Ultra low phase noise oscillator - DIYINHK
With the above clocks, some people use a dedicated voltage regulator for each one, and leave both of them running at all times for stability. The outputs then have to be switched to the dac by some external device. Some people use an FPGA or CPLD for the switching, others use a couple of NB3L553 (running from yet another voltage regulator) to buffer the clocks with enable/disable of the clock signals using the buffers: https://www.onsemi.com/pub/Collateral/NB3L553-D.PDF ...Don't know if any of those solutions are optimal, but that's some of what can be seen in various dacs.
With the above clocks, some people use a dedicated voltage regulator for each one, and leave both of them running at all times for stability. The outputs then have to be switched to the dac by some external device. Some people use an FPGA or CPLD for the switching, others use a couple of NB3L553 (running from yet another voltage regulator) to buffer the clocks with enable/disable of the clock signals using the buffers: https://www.onsemi.com/pub/Collateral/NB3L553-D.PDF ...Don't know if any of those solutions are optimal, but that's some of what can be seen in various dacs.
Last edited:
Ferrite materials have specified bandwidths and frequencies that they are effective in. They are neither generic, nor universal. Their use is application specific with knowledge of that sort of noise or interference spectrum you are trying to mitigate.
I'm not trying to discourage an experiment, I'm adding information that will assist in success.
This is the typical mythology propagated by audiophiles.
My statements are based on facts not opinions.
This is the very definition of confirmation bias. We have no data or measurements for which we can provide an explanation to what you are hearing.
You have changed the conversation to ethernet card noise when you began with -
Last edited:
Probably better to say 'minimum bias' rather than 'no bias.'
Very few if any people hear as much blind as they do sighted, but then again most blind testing of audibility is probably done incorrectly especially if ABX protocol is used. One approach is to train test subjects until they are comfortable with the protocol as determined by score improvements with listening tests asymptotically approaching something closer to sighted testing, but not quite. Otherwise, it may be necessary to weight scores to account for high numbers of false negative trial results. The false-negative problem can also be minimized by using alternate blind protocols, such as A/B. Still blind, just easier on working memory is one hypothesis.
Wow. Just...wow.
I never suggested measurements using expensive equipment. In fact, the only specific measurement I've suggested involves a sound card and free software. Apparently, that's still too expensive, or too much trouble, or there's a fear that reality will be in conflict with preconception, and a belief system might have to change. I would have assisted with operation of the software, but there is zero interest here so never mind.
I've provided factual information, not opinion. Everything I've said can be verified using...um...google, and your brain...and all. Most of "learning" is acquiring information. Sometimes it's offered, sometimes you have to search.
The OP has neither searched for the applicable technical information, nor accepted it when it was offered. Instead, he insists on performing a pointless experiment and evaluating the result with a highly error prone method. He rejects the actual technical principles because they don't match his concept without even considering them, much less researching them. This is not learning! But because his means of evaluating the result is loaded with bias and error, he will convince himself (effortlessly, I might add) of a result that isn't there at all.
Yeah, OK, I get it. I'm done.
The above is a claim commonly seen in confirmation bias. It expresses a belief that one's own beliefs are facts and someone else's beliefs are not.
I've been posting verifiable fact, not opinion. You could do the work and verify what I've said if you wanted to.
There is 100% proof that you can measure everything you can hear. As I've stated previously, to deaf ears apparently, the problem is in relating the degree to audibility. You can still measure a change, even if you're not sure if it's always audible. But I'm not arguing this anymore with someone who has his mind locked closed.
Again, the statements regarding bias I've posted are verifiable from many sources. Hearing an improvement over several days absolutely does not indicate no bias, it indicates bias is in full force. But you go live in your world outside of reality if you want. No objection here.
What do you think an Ethernet card does? This time, it's on you to go learn and find out. Clearly you don't know, and clearly don't want to know. How anyone can want to experiment but seal the mind shut is beyond me, but like I said, I'm done taking my time to offer helpful information. The horse is dead.
Obviously you haven't read the entire thread but my statements were based on the IEEE 803.2 standard, which is not an opinion, you should try reading it.
On the one hand low level noise generation and the coupling of it between different types of circuitry is one issue; on the other hand low level details of networking are fine for understanding basic data integrity. From what I have read of the conversation, it isn't that ones and zeros are changing when they shouldn't be. Its an issue of low level EMI/RFI produced in networking devices finding its way into and affecting the performance of mixed digital and analog circuitry that is supposed to work down to distortion levels below -120dB. That's hard enough to do well under ideal circumstances. RFI/EMI noise effects can be hard to measure in some cases too, although there are folks who would prefer to believe otherwise.
So mr fact guy, where is that proof? Anyone else but you know about this fact?
I have yet to see a measurement that indicates width, height and depth for sound stage, for example. Yet different DACs (many with very similar measurements) does give difference in sound stages.
Btw, I asked where the proof was that all we can hear can be measured at Audio Science Review (ASR) Forum (Amir and his deciples), I got banned for asking it 😀
I don't know, I'm not an Ethernet expert. I'm just of the opinion that any changes he's making should at least impact the signal transmission (and yes, ultimately the signal integrity).
You said that much better than I did (I am not an electronic engineer) but I been trying to make people understand that.
I think Rob Watts demonstrated (and wrote a paper about it) that to fully hear depth-effects in sound stage you would need a dynamic range of 300dB (Rob Watts is chief engineer at Chord Electronics). Not sure if its true, but it does indicate how low levels of distortions you need.
I disagree. You can hear everything that can be measured, but you can't measure everything you can hear (for now, you can up until the ear canal, anyway). Why? Because we have a nice little filter, both unreliable and unpredictable, called the brain...
With a dynamic range like that, you should be able to hear stuff happening on the other side of the world! The Loudest Sound Ever Heard | Discover Magazine
Last edited:
Can you explain how this might be? Ethernet has some level of inherent isolation, and then you have multiple stages of conversion, and while they might not be able to completely reject the interference, you should at least be able to measure the effects at each point.
Besides, even a very good speaker would still have an average THD of over -60 dB (among other types of distortions). With the ability of our brains to mask extraneous information while listening to music, proving any sort of audible effect from Ethernet interference should require some thoroughly-vetted evidence.
Last edited:
Rob Watts sometimes has a flair for the dramatic. He was apparently talking about the need to use 64-bit math for processing million tap filters without audible degradation. Audible effects in that case would probably occur do to things like rounding errors accumulating over a very large number of calculations. Accumulated errors, if audible, would be expected to add up to a lot more than a number like -300dB would suggest. Shame on Rob.
Its something like from -200dB to 100dB, but as said Rob can be a little dramatic. But even if its "only" 250dB is still gives a good indication of how well we can detect some audible changes.
I ordered the NZ2520SD 25Mhz (could not find NZ2520SDA for 25 Mhz). But only found them on a Japanese site so it will take some time (and the shipping cost was more than order value even though I ordered 3 of them) 🙂
- Status
- Not open for further replies.