WARNING: internal 10|100 PHY requires magjack hardware!
#21
(01-06-2019, 12:15 AM)t4_4t Wrote: Do you have enough experience to understand by looking at the circuit diagram?
For example, it is a circuit diagram as in the above post.
https://forum.pine64.org/showthread.php?...4#pid37014

If so, as an example, I can present a circuit diagram that I made and working well.

Hi t4_t4,

Thank you for your answer. Yes I have enough experience with circuits, even if it is now many years I don't touch the hardware at so low level. Wink

After reading some more info regarding the (very poorly documented) Sparkfun MagJack, and comparing with other projects where MagJacks were integrated, I've come out with this schema, which I haven't tested yet. Here I have also provided the connections to the MagJack LEDs, but I am not sure the PHY driver provided by ayufan will drive them properly.

I would be glad to have a look at your working solution, thank you.

[Image: image.png]
  Reply
#22
Unfortunately,
Just as many people on this thread already fail,
it probably will not work on the "circuit diagram" you presented.
Additional parts are required.

> but I am not sure the PHY driver provided by ayufan will drive them properly.

It is confirmation with the following image.
  bionic-minimal-rock64-0.7.9-1067-arm64.img.xz
  https://github.com/ayufan-rock64/linux-b...m64.img.xz

  o If "eht1" is changed to enable, "eth1" will work fine. ("eht1" is disabled by default)
In addition,
  o The green/yellow LEDs to function as "(Link-connetion)/(Send or Recive ?)" respectively.

---

 I am not a resident living in an English speaking country.
 In addition, my English ability is very very poor.
 Therefore, even if it is a short sentence, it takes a lot of effort to write it in English.
 Please wait few days to summarizing the "circuit diagram" and important point.
  Reply
#23
(01-06-2019, 10:09 PM)t4_4t Wrote: Unfortunately,
Just as many people on this thread already fail,
it probably will not work on the "circuit diagram" you presented.
Additional parts are required.

That's what I've read here from you in your past post, hence my request on the missing parts. As I wrote, not being an engineer but having enough practice with electronic components, I looked further on the net for information inherent the MagJacks, specifically circuits which have integrated  the one provided by SparkFun (because each MagJack it's a world on its own apparently, even if they provide similar function). If I am not wrong, the circuit may depend from the type of MagJack, and also by the type of PHY you are connecting to. That should be the reason some circuits include pull-up resistors to the TX/RX, lines because the PHY involved has no internal provision for them. The PHY integrated in the RockChip should not need them, as hinted also in the schema relative to the DAC/Ethernet shield, but again: I may be wrong.

Lately, I've looked at the following posts relative to the Mbed platform:

https://os.mbed.com/users/no2chem/notebo...t-testing/

https://os.mbed.com/users/rolf/notebook/ethernet/

The difference with my circuit is that they have used two separate capacitors to cut a possible RF interference (I used one, like in the Pine64 schema), and they have connected the central transformer pins to 3.3V and to the capacitors (I didn't, again following Pine64). The reason for those capacitors is to handle better longer Ethernet cables, though for short ones (below 1 m.) they should not make any difference. Other solutions I have seen had the central for the TX transformer connected to the VCC (via a 10K resistor eventually), still using the 100 nF cap, and the central for the RX transformed connected like in my schema.

Pretty confusing if you ask me Confused , I suppose this remarks the fact I am not an electrical engineer. Big Grin


(01-06-2019, 10:09 PM)t4_4t Wrote: > but I am not sure the PHY driver provided by ayufan will drive them properly.

It is confirmation with the following image.
  bionic-minimal-rock64-0.7.9-1067-arm64.img.xz
  https://github.com/ayufan-rock64/linux-b...m64.img.xz

  o If "eht1" is changed to enable, "eth1" will work fine. ("eht1" is disabled by default)
In addition,
  o The green/yellow LEDs to function as "(Link-connetion)/(Send or Recive ?)" respectively.

Good news! I am currently using for testing DietPi 6.19.7 with ayufan kernel 4.4.132-1075 (which was sort of a pain to update because the old DietPi distro I used for the initial install had a problem in providing a smooth way to update the kernel), so I should be set for the driver.

Thank you for your time and your efforts, I'll wait for your solution.
  Reply
#24
Your referenced page's Schematic : https://os.mbed.com/users/no2chem/notebo...t-testing/

As an example,(if the board is mbed LPC1768)
---
mbed LPC1768
https://os.mbed.com/platforms/mbed-LPC1768/

Schematics and Data Sheets
 mbed NXP LPC1768 Schematic
 http://mbed.org/media/uploads/chris/mbed-005.1.pdf

Look at page 4/5 of the above PDF.

The four terminals TD+/TD-/RD+/RD- have already been terminated at 50Ω(49.9Ω) in the board.

Therefore, In this example.
Even if it is directly connected with RJ45-JACK without additional parts, it will work fine.

I posted the circuit diagram, please refer to the following thread.

https://forum.pine64.org/showthread.php?...1#pid43561
  Reply
#25
Thanks a lot! Cool

So essentially we have been fooled with old picture and schema Huh , now it all makes much more sense.

I need some time to review all the info you've provided, and also to figure out where I am going to add the extra components that are needed. Space is tight in the case I am setting up, which is the black squared WDLabs used to sell for the Raspberries (see below for an example, the MagJack goes in the right bottom corner of the case).

[Image: box-open-100683007-primary.idge.jpg]
  Reply
#26
> 3.3v can be used instead of 1.8v ? If so, 1.8v-regulator do not need.
I understood that the point of your remark is like the above.
(My English understanding is as you are aware)

The following is the view on that premise.

https://forum.pine64.org/showthread.php?...9#pid39379
It is about half a year ago, I tried that condition (3.3v), that is the above post.
It certainly worked.

---

ROCK64 Schematic v 2.0 (Production Release)
http://files.pine64.org/doc/rock64/ROCK6...171019.pdf
Please refer to the pdf "Embed FEPHY" page (11/19) above.

The supply voltage to this FEPHY-block is 1.8v / 1.0v .
As far as the RK3328 data sheet is not available, everything is guessed.
Perhaps the PHY driver/receiver part is designed for use at 1.8v.
If so, termination to 3.3v is risky.
If there is something exceeding the absolute maximum voltage, it will lead to unrecoverable destruction.
I think that their design (terminated to 1.8v) is more reasonable.


Actually, At the termination to 3.3v, ROCK64(RAM=1G) is work.
However, ROCK64(RAM=4G/2G) did not work.
On the other hand, in the case of termination to 1.8v,
All ROCK64(RAM=1/2/4G) work fine.

Regarding the 1.8v-regulator, I described it based on this fact.
Again again, I think that their design (terminated to 1.8v) is reasonable.

*) However, "62Ω" is a mystery. I am the first time to see such a number.

---

It was quite a while ago, I tried changing the voltage at the termination.
and it's experiment data still remained.
According to it, it is sensitive.

x 1.61v ((27|27   ) / 22) : 13/22
o 1.73v ((22|27   ) / 22) : 12/22
o 1.83v ((22|22   ) / 22) : 11/22
o 2.01v ((22|22|47) / 22) :  9/22
x 2.16v ((22|22|22) / 22) :  7/22

----

> I have left that end dangling on my schema, although I was suspicious about it.

In conclusion, either way worked.

If you search, you can find examples of both.
Since it is not possible to obtain the data sheet of RK3328,
It can not be said that which is better.
So, they left their original.
  Reply
#27
t4_4t, first of all I want to thank you very much for the patience in answering my questions and the help provided, giving also the fact  writing in English is not an easy task for you (BTW, English is not my native language also)

Now...

(01-09-2019, 01:00 AM)t4_4t Wrote: > 3.3v can be used instead of 1.8v ? If so, 1.8v-regulator do not need.
I understood that the point of your remark is like the above.
(My English understanding is as you are aware)

You have understood very well. My biggest problem, apart from finding too late that all my best guesses were wrong (shame on me for having trusted too much the "official" documentation), is that I had planned to fit the second Ethernet port in this case, where there is just space for one of the (now defunct) WDLABS hard disk, and the MagJack in one of the corner, for which I had do to do a cut in one of the corners so that the MagJack can slide in, and to push the hard disk up by slightly less than 1 cm, so that the SATA connector doesn't clash with the MagJack.

If you look at the pictures I have attached, there's really no space left where I can fit the (unexpected) voltage regulator and, worst, its two electrolytic capacitors which tend to be bulky compared with the compactness of the other components, hence my reasoning in trying to use the 3.3V already available on the P5+ connector. I could push up further the HD, where not for the caps it could have been a viable solution, but really there isn't space. I'll take your recommendations and experiment results (thank you for sharing them) to look for a different solution as my board is 4GB RAM model. Perhaps a voltage divider would suffice, assuming the provided 3.3V is stable? Just guessing, I stopped working with this stuff a very, very long time ago and I've forgotten a lot.

(01-09-2019, 01:00 AM)t4_4t Wrote: Regarding the 1.8v-regulator, I described it based on this fact.
Again again, I think that their design (terminated to 1.8v) is reasonable.

*) However, "62Ω" is a mystery. I am the first time to see such a number.

Ohm's law tells us that with 62Ω and 1.8V you are applying on the pulled-up pin a 30 mA current, whatever it means. PHY TX/RX pins are working in the analogical domain, also, not digital.

(01-09-2019, 01:00 AM)t4_4t Wrote: It was quite a while ago, I tried changing the voltage at the termination.
and it's experiment data still remained.
According to it, it is sensitive.

x 1.61v ((27|27   ) / 22) : 13/22
o 1.73v ((22|27   ) / 22) : 12/22
o 1.83v ((22|22   ) / 22) : 11/22
o 2.01v ((22|22|47) / 22) :  9/22
x 2.16v ((22|22|22) / 22) :  7/22

What are the above? The resistor values for a voltage divider?

Thanks!


Attached Files Thumbnail(s)
       
  Reply
#28
(01-09-2019, 01:00 AM)t4_4t Wrote: The supply voltage to this FEPHY-block is 1.8v / 1.0v .
As far as the RK3328 data sheet is not available, everything is guessed.
Perhaps the PHY driver/receiver part is designed for use at 1.8v.
If so, termination to 3.3v is risky.

I am not able to find that voltage regulator Sad

For 1.2V yes, for 2.5V also, but for 1.8V NO, three shops told me already they had none. I am thinking about trying with the 1.2V model, I'll have to recalculate the resistors values. I have also a question regarding the caps for the regulator: why do they have different values? The example schema accompanying the voltage regulator suggest to use the same value for both caps. Why yours are different?

Thanks.
  Reply
#29
> I am thinking about trying with the 1.2V
Is it a 2.5v mistake ?

It will not work at 1.2v
According to the results I examined,
The range of working is between "1.73v <-> 2.01v"

> ? The example schema accompanying the voltage regulator suggest to use the same value for both caps.
> Why yours are different?

You can understand if you enlarge the photo.
The value is not so important, so I adjusted it to their design value.

---

The circuit diagram presented by me is a based on examining the delivered item.
(if unless it is necessary to change)

* Ex: "Unless necessary to change"
Their design of 1.8v part is polite and further "L/C" parts for quality improvement are added.
(Like "U2: LP2985-33" of "ROCK64_ES9023_Audio_100Mbps_Ethernet_Board.pdf")

But, these parts can not be obtained.
Because the chip "L/C" can not read its value.
Moreover, these parts are considered to have little influence on essence in this case(to Network).
For the above reasons,
I did not use these parts, so I removed it from the circuit diagram.

---

I recognize that your requirement is minimizing the space.

If I were you in the same situation, I will take the following method.

That is the same constitution as "mbed-005.1.pdf".
1.8v already exists on the rock64 board.
Connect 4 terminating resistors directly to that node.
In this case, neither a regulator taking up space nor a capacitor is required.

I already tried it actually.
Tried "ROCK64" model is only "RAM=4G". (which had less margin of Ethernet)
In addition, I do not plan to check with all the "ROCK64" that i own.

And it worked.
However, in order to operate stably,
Separate "C"=(0.1 uF) was required for "TD_TCT / RD_TCT" respectively.
But, it is not a big increase in size. (+ 0.1uF x 1)
It will be possible to place on the back of the "RJ45"

It is worth to try it if you own confidence in technology.

Otherwise, let's give up.

--- Important ---
Just because I succeeded, you may not always get the same result.
Always keep this in mind.




I exhausted my physical strength considerably for English translation.
I need a little rest.
  Reply
#30
(01-10-2019, 10:03 PM)4_4t Wrote: > I am thinking about trying with the 1.2V
Is it a 2.5v mistake ?

Yes, it's a mistake, I meant 2.5V, but I am by now aware that 2.5V is above the maximum voltage supported by the integrated PHY.

(01-10-2019, 10:03 PM)t4_4t Wrote: That is the same constitution as "mbed-005.1.pdf".
1.8v already exists on the rock64 board.
Connect 4 terminating resistors directly to that node.
In this case, neither a regulator taking up space nor a capacitor is required.

I already tried it actually.
Tried "ROCK64" model is only "RAM=4G". (which had less margin of Ethernet)
In addition, I do not plan to check with all the "ROCK64" that i own.

And it worked.
However, in order to operate stably,
Separate "C"=(0.1 uF) was required for "TD_TCT / RD_TCT" respectively.
But, it is not a big increase in size. (+ 0.1uF x 1)
It will be possible to place on the back of the "RJ45"

While I had the same idea of getting 1.8V straight from the board, since it's already there, I failed to find the right spot where to solder a cable. I tried to follow all the VCC_18 lines in the Rock64 schema (stemming from RK805-1), but I haven't found one point on the board where I would be able to get the needed voltage. To be honest, I have zero experience in soldering on SMD circuits, and I haven't a soldering station suitable for this task (I have a decent soldering tool, but good only for non-SMD components). It's also clear that your knowledge on the matter is way beyond mine Rolleyes

Your last sentence on where to put the double caps is not the problem for me (also it is suggested everywhere to have them as close as possible to the RJ45 port). It's the rest that is making me scraping my head, even more because I cannot locate the voltage regulator that you've suggested. It may be possible that this 1.8V voltage regulator is available in my zone (I'm not so sure, I'll have to ask), do you think it could be compatible? It can also work (and it seems that the manufacturer suggests it) with ceramic caps instead of electrolytic, which may help in finding a place for them.

MCP1702-1802E/TO

Thanks a lot again for all the useful info you've shared with us!
  Reply


Possibly Related Threads…
Thread Author Replies Views Last Post
  Any way to tell the rock64's hardware revision? kittyland 1 202 07-01-2021, 09:42 PM
Last Post: evilbunny
  Hardware random number generator? scalextrix 13 6,356 06-04-2018, 02:05 AM
Last Post: scalextrix
  Latest Rock64 board hardware revision Ryan 3 2,694 03-14-2018, 05:42 AM
Last Post: petec
Question How can I use the internal 100M ethernet? gabrielmartinslima 1 1,178 02-12-2018, 02:10 PM
Last Post: dkryder
Information Rock64 SSD internal PNY CS900 240GB w/Pine usb|sata adapter MarkHaysHarris777 1 2,091 09-30-2017, 06:48 AM
Last Post: dkryder

Forum Jump:


Users browsing this thread: 1 Guest(s)