FEDEVEL
Platform forum

, 03-28-2024, 07:56 PM

Hi everyone,
I have a little knowledge on 1GBPS ethernet and if someone answers my doubts it would be very helpful for me .

I have designed a board with Nivida Jeston NX Xavier (In built RTL8211FDI Phy IC) and the 966378617 magjack ethernet connector. Initially these boards were used for 100Mbps and those worked without any issues.
However recently we had a demand for 1000Mbps, for this we have used the same board since this board was designed for 10/100/1000Mbps. While testing these boards with 1000Mbps we have observed some packet loss. We don't have any idea why packet loss occurred at 1000Mbps.
The same board tested with different magjack connectors (Part no: 233799-8) and observed zero packet loss. Since we have already designed many boards with 966378617 magjack footprint we couldn'ta able solder the 233799-8 connector on this board.
I have attached both magjack internal scheme for your references.
My Question is can we achieve better performance (zero packet loss) by changing
RTL8211FDI driver code instead of changing the magjack connector since the footprint of both connectors are different.
Even if I get the same footprint, the de-soldeeing and soldering back will create other issues. So please let us know if there is any possibility of achieving good performance (zero packet loss) by changing PHY IC driver firmware.
Like increasing eye width, amplitude or levels and introducing correction factors.
Many Thanks in advance

Robert Feranec , 03-29-2024, 06:06 AM

did you try to probe the signals?

, 03-29-2024, 03:51 PM

No Robert Feranec sir, Since I don't have enough samples (Bandwidth) DSO. If I want a probe where to probe. Is it near the magjack connector or at the input of the PHY IC drive ?

, 03-29-2024, 03:53 PM

, 03-29-2024, 03:54 PM

We have the above DSO

QDrives , 03-29-2024, 05:02 PM

Your scope is 350MHz, Gigabit is 250MHz, so it should be possible.
Naturally you measure at the same location. I assume the board is the same? Then measure at the connector as it will be much easier than at the PHY.

, 03-29-2024, 06:50 PM

Thanks @QDrives ,
I will do that but how is this board working with another magjack?

QDrives , 03-30-2024, 09:25 PM

1) How did you get the other magjack on it?
2) The capacitances are a bit different. Perhaps that has some influence.

, 04-01-2024, 10:35 AM

I made a small PCB for this footprint and on the other side my current PCB footprint. It is like an add-on PCB .

QDrives , 04-01-2024, 08:01 PM

That adapter board may distort the signal a bit, in this case towards good.

, 04-03-2024, 12:38 PM

I have DP-65pro differential probe, with this can I measure Ethernet signals?

, 04-03-2024, 12:38 PM

I think it's a 65MHz bandwidth

, 04-03-2024, 12:45 PM

By keeping this at 50 mode I can measure signal (65*50=3250MHz). Attached datasheet

Mini , 04-03-2024, 01:56 PM

No. You can't multiply frequency like that. If you have 65 MHz bandwidth then max frequency you can measure is 65 MHz. It actually means your input voltage will be reduced 50x on your differential probe input so you can measure higher voltages. Otherwise your differential probe input has to withstand very high voltages. Same goes for oscilloscope, but it is more complicated. Like you can measure max 1300V DC or AC peak with this differential probe(has to be on 200x range). It is actually written in the manual. Why not read it.

Mini , 04-03-2024, 01:56 PM

https://d17bck4wpaw2mg.cloudfront.net/att/a/2/5/x/a25xz2/dp65manual.pdf

Mini , 04-03-2024, 01:57 PM

Robert Feranec , 04-03-2024, 02:22 PM

you may be able to use also standard probe (no differential), important is to compare especially the level of signals between these two transformers. that is what I would look for as the first thing

, 04-04-2024, 05:49 AM

MP770208 probe part number

, 04-04-2024, 05:49 AM

Currently we don't have the 300MHz probe. We are using only 60MHz probe for our internal measurements.

Robert Feranec , 04-04-2024, 07:15 AM

just try, maybe you will see a noticable difference

Sniper2 , 04-04-2024, 10:55 AM

It will

Sniper2 , 04-04-2024, 11:12 AM

Tho do do it properly u will need 1 2gb probe and 2 3gb scope

Robert Feranec , 04-04-2024, 11:46 AM

I mean if there is something very different between the two transformers, any probe and scope should show it. Of course it will not show the signal properly but if levels are very different than the transformers are different and that will be maybe the problem. If the signals are similar, then I would look for a proper probe.

QDrives , 04-04-2024, 02:27 PM

Those are not the probes that were supplied with the scope. Or was the scope supplied without probes?

, 04-04-2024, 05:53 PM

Yes, please let me know what types of probes are needed for thos

QDrives , 04-05-2024, 08:24 PM

You have a 350MHz scope so you want at least 350MHz bandwidth probes.