Verfasste Forenbeiträge
-
Beiträge
-
„the differential Ethernet signal has three states: +1 V, 0 V or −1 V.“
That is correct for 100Mbps Ethernet. For Gigabit Ethernet the same 2V swing (+1 to -1) used, but there are 5 levels, so separated by 0.5V each.
Hi Eric:
Thank you for the very kind welcoming!
It is clear that we are on the same page about most all of these things. 😉
As for your thoughts on interference of sorts in the transmitted light signal with fiber optic:
a) I will ask John his thoughts on the possibilities. (I am NOT the engineer here and never pretend to be.) I do not think he has any direct measurement experience with various fiber cables, though certainly we have discussed the valid technical reasons for differences between single-mode and multi-mode.
b) The disturbances (PM and AM) induced by the parts and power networks associated with copper>optical and optical>copper conversions seem to be alluded to by you as well. And on that we do agree and have direct (measured and listening) experience with. This is surely the reason behind the widely reported differences users hear with different SFP module models and types.
One interesting ideas that may bring fruit later on:
1) Everyone is aware that SFP transceivers are offered with different transmission distance maximum ratings–from 5Km to 50Km. The greater distance rated units sometimes have better electronics–or sometimes the same as shorter rated units. But for certain they vary by transmit wattage. But when used for very short distances there is some risk of saturation/overload of the receiver end–hence you sometimes see the use of fiber optic attenuators.
Yet we have found that a series of high quality Finisar SFP modules–with different distance ratings–contain the exact same parts. The variation is due to flash programming of the transmitters to output higher wattage for the long distance units. We are looking into the possibility of reflashing the EEPROM on a particular model–to output very low wattage–thereby making it more suitable/less saturation/better performance for the short distances used in the home. I do not yet know if this project (low-priority for us now) will be successful.
Next time I may speak about an entirely different, non-optical, isolated transmission option that may be far more appealing in the near future. Besides your WiFi that is. 😉
Best,
–Alex C.
Hi Eric:
My first post on your nice forum. 😉 Please forgive that I don’t understand how to properly use the WordPress quotation function. But in reference to what you said a few posts above:
With regards to optical–and to everything really as I will repeat below–one can look at the fact that the chips are used in optical SFP transceivers themselves generating many picoseconds of jitter. It is right there in the data sheets of those chips.
And the WHY of why that matters is that phase-noise modulation converts to amplitude noise modulation (and back and forth throughout all transmission)
In my own forum I recently posted this–reiteration of what we have been saying all along:
BOTH common-mode noise (caused by the leakage currents that travel all around our systems–and even on Ethernet cables) AND the phase-noise/jitter that gets embedded in the single (from bad clocking, leakage, and the chips themselves as they operate) are sources of transmitted ground-plane noise.
That’s exactly the reason that the measurements Hans showed–made by Jaap Veenstra (Alpha Audio magazine)–at the master clock input pin of the DAC (it was a cheap streamer/DAC) show modestly reduced phase-noise/jitter when the Muon transformer box was added.
It is a good test–with the very expensive Wavecrest SIA-4000C–and it would be nice for them to run the same with other Ethernet devices, including EtherREGEN. Then perhaps Hans would no longer disavow his past experience.
Someone asked:
„Regarding the measurements in particular did it seem odd to you that a passive device could improve phase noise?“
Not at all odd. See above. It is all about the pernicious propagation of ground-plane noise–which ultimately reaches the DAC. (This happens will all interfaces, be they USB, Ethernet, I2S, etc.) We have written about it in a variety of ways in our papers, discussing it as both clock phase-noise overlay and in other terms.
But a key mechanism, one we have only touched on a bit (in our Sine v. Square clock paper) and which deserves its own full treatise is the repeating back-and-forth conversion of AM>PM. (No, I am not talking about morning and afternoon.
It is amplitude modulation (noise of all sorts: common-mode, harmonics, leakage, etc.) that converts to phase modulation in the chips–and when the chip clocks back out its data, the jitter/phase-modulation that got embedded in that data then again causes amplitude modulation–in the form of ground-plane noise.
Of course we address the above in all sorts of ways with the EtherREGEN, with differential clocking and differential isolation being the most obvious and unique techniques used. But there are many deeply technical methods which go into Johns Swenson’s board layout–including 6-layer board, clock buffer placement, power networks, and highly critical small parts selection.
All that and more refinements coming with EtherREGEN Gen2…
======================
I hope it’s okay with you that I post the above.
My point being that optical connections, while they provide full and true galvanic isolation (MUCH more than just the transformer magnets behind every RJ45 port), they will still:
a) convey already imbedded upstream jitter (which will convert into AM noise);
b) generate their own jitter/phase-noise modulation (from their own chips) which will also convert again to AM (common-mode noise).
Best,
–Alex C.
UpTone Audio LLC
