[En-Nut-Discussion] Ethernet HW questions

[Jean Pierre Gauthier]

>"Ethernut 1.3G will use A13 and A14 to mimic the EEPROM. This enables the
software to
select the duplex mode and costs one serial resistor at BD7 only. The
software is a bit tricky and becomes simpler, if you use real port bits
instead of address lines." Harald

Hello,
Unfortunately I modified the Ethernut1.3F /CS to drive 64K RAM, excluding
CPU and RTL8019 space (8000-83FF). Thus upper 32K space drives RAM directly.
If the Ethernut 1.3G will use A13 and A14 to mimic the EEPROM, I suppose
that a collision will appair between RTL8019 and RAM during A13-A14 "ON"
accesses . Is that true?
Can the software easily modified to map A13/A14 to CPU IO (prefered
choice)or devalidate duplex mode option/ A13-A14 manipulations (last
choice)? 
Jean Pierre















-----Message d'origine-----
De : en-nut-discussion-bounces at egnite.de
[mailto:en-nut-discussion-bounces at egnite.de] De la part de Harald Kipp
Envoyé : mardi 3 août 2004 01:09
À : Ethernut User Chat (English)
Objet : Re: [En-Nut-Discussion] Ethernet HW questions

Hi Henrik,

your post was held for approval, because the sender's email wasn't
subscribed at http://www.egnite.de/mailman/listinfo.cgi/en-nut-discussion
No problem, but I do not check the requests every day.

At 11:52 02.08.2004 +1000, you wrote:

>A) Power-on reset generation
>I like to remove the DS1811 and replace it with a simple RC combination 
>like many other AVR designs do. Isn't the DS1811 anyway redundant 
>because of the ATmega's internal brown-out detection?
>
>B) RTL8019AS reset pin
>I like to connect the reset to an AVR port to save the need for an 
>inverter gate.

Right, the ATmega128 doesn't require a reset controller, but the RTL8019AS
does. Using a port bit is the most simple solution and is even supported by
the standard Nut/OS driver.



>C) RTL8091AS interrupt pin
>This pin can be connected directly to the ATmega128 to save an inverter 
>gate. I assume the inverter was only necessary for the older
>ATmega103 chip which didn't have programable interrupt levels.

Yes, but the gate is still on Ethernut 1.3F (and will be on Rev-G) for
software compatibility. A few of egnite's long term customers still use
ATmega103 firmware and don't want to change anything.



>D) RTL8091AS half-duplex mode
>I also like to connect pin 77/EEDO/BD7 to ground to force the RT into 
>half-duplex mode. This was suggested by RealTek's support. In this case 
>LED_BNC has to be used to indicate the Link status as the LED 
>configuration cannot be changed by the cfg registers. I have seen the 
>LED_BNC solution in some other reference designs supplied by RealTek 
>and also tested it using the PacketWhacker.
>When grounding pin 77/EEDO/BD7, the JP pin 65 can then be left open so 
>the RealTek operates in "RT jumperless mode".

Never tested this. Ethernut 1.3G will use A13 and A14 to mimic the EEPROM.
This enables the software to select the duplex mode and costs one serial
resistor at BD7 only. The software is a bit tricky and becomes simpler, if
you use real port bits instead of address lines.


>E) Realtek 270R pulldown resistors on pin 48,49 Aren't these rsistors 
>obsolete because AUI is not used?

Should be. But, if you got PCB space left, I'd still add the pads. You may
later decide about stuffing them or not.


>F) /RD, /WR, ALE signals
>These signals are floating in reset or if the ports are not configured. 
>Shouldn't they have pull-up resistors?

Do you see any harm during reset? As long as the memory interface is
enabled, they are configured as outputs.



>G) AVCC/AGND/AREF
>I won't need analogues on the board. What's the best way to deal with 
>these supply signals? Connect them to VCC and GND or leave them simply 
>open?

Under no circumstances I'd leave them open. Some internal parts  may draw
power from somewhere else. If you do not use analog I/O, you can connect
them to VCC/GND.


>H) RTL8091 address signals
>I like to use only the A0 to A4 and A8 (connected to ATmegas A15 as 
>chip select) address signals on the RealTek. A9 is pulled high all 
>other address signals low. This results in the RealTek occupying the 
>whole upper 32K address space.
>Doing this I need less tracks to route and I save the inverter gate for 
>A15.

Yes, if you do not use any RAM or memory mapped I/O you can simplify the
board's routing by not using the Realteks internal address decoder.

Harald



>Cheers,
>Henrik

_______________________________________________
En-Nut-Discussion mailing list
En-Nut-Discussion at egnite.de
http://www.egnite.de/mailman/listinfo.cgi/en-nut-discussion