Diference in Lankom device numbers?

  • Thread starter Thread starter comitter
  • Start date Start date
C

comitter

*ENW-9501+ mre¾na UTP 10/100 Digital Rev. A
Lankom 9809 LF-H42X

i

*ENW-9501+ mre¾na UTP 10/100 Digital Rev. A
Lankom 9820 LF-H42X

Both are ethernet 10/100 devices, what could be the difference in
Lankom numbers ?


Tenkyu.
 
*ENW-9501+ mre¾na UTP 10/100 Digital Rev. A
Lankom 9809 LF-H42X

i

*ENW-9501+ mre¾na UTP 10/100 Digital Rev. A
Lankom 9820 LF-H42X

Both are ethernet 10/100 devices, what could be the difference in
Lankom numbers ?


Tenkyu.
Click to expand...

Lankom makes Ethernet transformers. The operation
of the transformer makes no difference to the
purchaser of the card. (LF-H42X is listed here, but the
datasheet is not available.) Ethernet uses transformers,
to avoid a DC path between cabled devices. So the
transformer is for electrical isolation.

http://web.archive.org/web/20030802070114/www.lankom.com.tw/epage/products/data/10-100Base[1]4_2.asp

The Ethernet chip can change the performance of the
card, with some of the older RealTek chips lacking
ring buffers in the interface. I would be slightly
more concerned with what Ethernet chip is on each
card, than I would be about the choice of
transformer. Your card could have something like
a DEC 21140 for example. Study the characteristics
of the chip, to understand whether it makes a
difference or not.

(Picture of a DEC 21140 Ethernet chip...)
http://aceh.hp.infoseek.co.jp/jpg_2/dec40ae2_c.jpg

Paul
 
Paul said:
Lankom makes Ethernet transformers. The operation
of the transformer makes no difference to the
purchaser of the card. (LF-H42X is listed here, but the
datasheet is not available.) Ethernet uses transformers,
to avoid a DC path between cabled devices. So the
transformer is for electrical isolation.

http://web.archive.org/web/20030802070114/www.lankom.com.tw/epage/products/data/10-100Base[1]4_2.asp


The Ethernet chip can change the performance of the
card, with some of the older RealTek chips lacking
ring buffers in the interface. I would be slightly
more concerned with what Ethernet chip is on each
card, than I would be about the choice of
transformer. Your card could have something like
a DEC 21140 for example. Study the characteristics
of the chip, to understand whether it makes a
difference or not.

(Picture of a DEC 21140 Ethernet chip...)
http://aceh.hp.infoseek.co.jp/jpg_2/dec40ae2_c.jpg

Paul
Click to expand...

There are differences in chip model number 21143-PC, 21143-PD, both DIGITAL

Still looking at your writings...
 
Majki said:
There are differences in chip model number 21143-PC, 21143-PD, both
DIGITAL

Still looking at your writings...
Click to expand...

I guess the DEC networking chip designs, were bought by Intel. If
you download the "hardware reference manual" and go to page 107,
they describe a ring buffer type structure for control. That is part
of DMA transfer of packets to system memory. The FIFOs in the chip,
are for temporary storage of data, until the DMA can be completed.
This means the 21143 is probably a better chip than some of the
$10 RealTek based PCI cards.

http://www.intel.com/design/network/products/lan/docs/21143_docs.htm

The LanKom part, is used in the following fashion. The transformers
in the center of the diagram, would be the LanKom type device.

http://www.hardware-guru.net/board_design/Ethernet/image004.jpg

The fact there is no path for DC currents to flow (because of the
transformer), solves the problem of the network devices having
different electrical grounds on them. The turns ratio of the transformers
is selected according to the Ethernet chip being used, and the
amplitude of drive the chip is capable of. As a designer, you read
the chip datasheet, and the datasheet tells you whether a 1:1 or
1.41:1 transformer etc would be needed.

In that picture, the chip on the left, is the PHY, or physical layer
chip. A typical complete chain of command, looks like this.

PCI --- MAC chip -- MII ------- PHY chip ----- Ethernet ------- RJ45
BUS Interface transformers

In some designs, the MAC and PHY functions are inside the same chip
package. That would be a fully integrated design. In other cases,
the MAC is inside one chip, and the PHY is in another. Separating
the PHY can reduce the cost. The reason for that, is the MAC functions
can be built in pure CMOS, while the PHY functions might be built
in BiCMOS or Bipolar technologies. The TXP and TXN pads have to
provide a bit of power on their outputs, and historically, CMOS
wasn't good at doing that. But CMOS has been used since then,
to do some amazing things, such as sink or source 2 amps of current,
so anything is possible now.

Paul
 
Back
Top