ethernet portal circuit

[bahstrike]

i am ALMOST done with this, still need a voltage regulator or maybe some level shifters depending on what is chosen for SPI master..

the key to keeping it DIY-able  (imo, but im a lousy engineer) is finding a jack with built-in magnetics.  i have tried to use (even make) transformers in the past but it's always been too picky for me.

View attachment 40884

 

[Alex Tsekenis]

That is in the Microchip datasheet for that MAC/PHY chip with SPI interface.

Winding your own transformers for Ethernet is silly. Wurth Electronic (Midcom), Coilcraft and many others have hundrends of standard models to suit ANY application and wiring layout.

Good luck with the TCP/IP stack.

 

[bahstrike]

this is my first attempt at ethernet  (didnt even try a MAX232 until a few months ago)

i have ignored transformers for my past year-ish venture into digital electronics-  i did some crap in high school with them, but very little.  the 3rd or 4th one i wound actually brought something like 9v up to 90v.. it was supposed to be wound for 300v

programming is my forte, so im really just looking forward to getting the ethernet circuitry "out of the way", while still understanding it, and quickly doing an IP implementation.

from there, i could use it in any of several various projects 

 

[Alex Tsekenis]

If you are into software, consider an RJ45 module with built in magnetics, MAC/PHY and ARM microcontroller. Alternatively, the circuit that you have been putting together can be bought ready made here:

http://uk.farnell.com/microchip/ac164123/board-pictail-plus-ethernet/dp/1439831

 

[bahstrike]

is this many decoupling caps overkill?   i created one for each pair of VDD/VSS pins.

datasheet didnt seem to specify, so im betting this is more an engineer's logical hunch;  would it be more reasonable to use half as many, with 2 sets of power pins per 1 cap?

or just 1 is fine for the whole ic?


i guess it is best to use all of them so that variances in power consumption from one module (eg. TX)  do not interfere with power supply of another module (eg. OSC)

 

[Alex Tsekenis]

The idea with this IC is to use one decoupling capacitor per VDD line and then connect them all together to VDD using a star connection.

 

[bahstrike]

thanks Alex, I will keep them all.

What is a 'star connection' ?

I tried to google it, did not find a real explanation but did find a datasheet which points at a single junction where all GND of an IC met,  and it was labeled Star Connection.

Is it basically just all VDD traces to the caps do not join VDD 'just anywhere', but all at the same single point?  Likewise I suppose the same for VSS, although it shouldn't matter if a ground plane is used (?).

 

[Alex Tsekenis]

Hello,

I avoided calling it a star ground as this is not for ground. It is like a 'star ground'. Google that.
The idea is that all subcircuits that draw a small amount of current are not affected by the circuits that draw large currents. If they were sharing the same connection (not star), then a circuit that draws large current will create a voltage drop on the copper tracks. This voltage drop will also be seen by the low current circuits, possibly affecting them. If however all circuits went to a single point and at that point there was a large energy storage and low impedance from the power supply then sensitive subcircuits will be less affected, if at all.

The same idea applied for a star positive supply, although it is not refered like that.

Digital circuits are more tolerant to this but this IC has an analogue front end (the PHY part) that draws a lot of standby current. Audio amplifiers use this extensively.

Power planes radically reduce this problem but nothing can eliminate it completely as the impedance to the power source will never be 0.

 

[bahstrike]

updated circuit-  includes audio amplifier

shucks i gave my idea away  XD


i have drawn the path and labeled the audio's form, starting at the network connection to the device's speaker

View attachment 40887