Forums

Gigabit Ethernet PHY without NDA?

Started by Matt Ettus December 17, 2008
I've been using a National Semi DP83865 Gigabit Ethernet PHY chip for
a while now.  It has 4 problems:

- Burns 1.25 Watts (and gets hot)
- Needs 2 voltages, 1.8 and 2.5 Volts, and 1.8 is used nowhere else in
my system
- Can't maintain a connection on 100 meters of Cat-5
- It is a very large chip

So I decided to look around for a replacement.  I have found
candidates from Marvell, Vitesse, and Broadcom.  All claim to use
about half the power, are much smaller, and don't need the 1.8V
supply.

The problem is that all of them require an NDA just to see the
datasheet.  Because I intend to publish the design (this is open
source hardware), I am not in a position where I can sign an NDA.  In
fact, there are FPGA development boards out there with published
schematics that have a page marked "this page redacted due to an NDA
on the Marvell 88E1111".  This seems nuts to me.  I mean, what could
possibly be NDA-worthy in the datasheet for a 4 year old part that
implements a technology (GbE) which is nearly a decade old?

Does anyone have any suggestions as to non-NDA encumbered GbE PHY
chips?  All I need is GMII or RGMII.

Thanks,
Matt
Matt Ettus <boyscout@gmail.com> wrote in news:3dbf05f8-b8cb-4036-800e-
940af73ce1e8@n33g2000pri.googlegroups.com:

> > I've been using a National Semi DP83865 Gigabit Ethernet PHY chip for > a while now. It has 4 problems: > > - Burns 1.25 Watts (and gets hot) > - Needs 2 voltages, 1.8 and 2.5 Volts, and 1.8 is used nowhere else in > my system > - Can't maintain a connection on 100 meters of Cat-5 > - It is a very large chip > > So I decided to look around for a replacement. I have found > candidates from Marvell, Vitesse, and Broadcom. All claim to use > about half the power, are much smaller, and don't need the 1.8V > supply. > > The problem is that all of them require an NDA just to see the > datasheet. Because I intend to publish the design (this is open > source hardware), I am not in a position where I can sign an NDA. In > fact, there are FPGA development boards out there with published > schematics that have a page marked "this page redacted due to an NDA > on the Marvell 88E1111". This seems nuts to me. I mean, what could > possibly be NDA-worthy in the datasheet for a 4 year old part that > implements a technology (GbE) which is nearly a decade old? > > Does anyone have any suggestions as to non-NDA encumbered GbE PHY > chips? All I need is GMII or RGMII.
The last time I looked at PHYs (a few years ago), the good ones required NDAs for the datasheets, and the ones that didn't have NDAs weren't worth using. I don't see how this clashes with open source hardware though. You need to document the design of the interface to the PHY but this isn't the same thing as violating the NDA for the PHY datasheet. The schematic merely shows the pinout. The pinout isn't covered by the NDA, as that information is already published in schematics that you can download. For example, I found this schematic containing an 88E1111 in less than a minute of searching: http://www.xilinx.com/support/documentation/boards_and_kits/ml50x_schemat ics.pdf Most NDAs contain a clause that invalidates them if the information becomes public through some other means. An NDA might be more of a problem for the software; device driver source typically contains information about many of the PHY internal registers. Disclaimer: I am not a lawyer, and you should read the terms of the NDA very closely. Regards, Allan
We have had the same problem and partially why we have not as yet got
a phy module at gigabit available. We didn't think it was worth
incorporating in development boards, or modules, if our customers
couldn't get the basic information. The other problem with these
companies is getting smallish numbers of chips is generally expensive
and difficult. I can only hope these companies stop being so snobbish
as they get rapid volume turnoff in the current recession in their
favoured consumer markets.

John Adair
Enterpoint Ltd.


On 17 Dec, 06:54, Matt Ettus <boysc...@gmail.com> wrote:
> I've been using a National Semi DP83865 Gigabit Ethernet PHY chip for > a while now. =A0It has 4 problems: > > - Burns 1.25 Watts (and gets hot) > - Needs 2 voltages, 1.8 and 2.5 Volts, and 1.8 is used nowhere else in > my system > - Can't maintain a connection on 100 meters of Cat-5 > - It is a very large chip > > So I decided to look around for a replacement. =A0I have found > candidates from Marvell, Vitesse, and Broadcom. =A0All claim to use > about half the power, are much smaller, and don't need the 1.8V > supply. > > The problem is that all of them require an NDA just to see the > datasheet. =A0Because I intend to publish the design (this is open > source hardware), I am not in a position where I can sign an NDA. =A0In > fact, there are FPGA development boards out there with published > schematics that have a page marked "this page redacted due to an NDA > on the Marvell 88E1111". =A0This seems nuts to me. =A0I mean, what could > possibly be NDA-worthy in the datasheet for a 4 year old part that > implements a technology (GbE) which is nearly a decade old? > > Does anyone have any suggestions as to non-NDA encumbered GbE PHY > chips? =A0All I need is GMII or RGMII. > > Thanks, > Matt
Matt Ettus <boyscout@gmail.com> wrote:

> >I've been using a National Semi DP83865 Gigabit Ethernet PHY chip for >a while now. It has 4 problems: > >- Burns 1.25 Watts (and gets hot) >- Needs 2 voltages, 1.8 and 2.5 Volts, and 1.8 is used nowhere else in >my system >- Can't maintain a connection on 100 meters of Cat-5 >- It is a very large chip > >So I decided to look around for a replacement. I have found >candidates from Marvell, Vitesse, and Broadcom. All claim to use >about half the power, are much smaller, and don't need the 1.8V >supply.
Perhaps Micrel has a nice & cheap phy. -- Failure does not prove something is impossible, failure simply indicates you are not using the right tools... "If it doesn't fit, use a bigger hammer!" --------------------------------------------------------------
Nico Coesel wrote:
> Perhaps Micrel has a nice & cheap phy.
Micrel has nothing gigabit on their website. -hpa
On Dec 17, 2:16 am, Allan Herriman <allanherri...@hotmail.com> wrote:
> Matt Ettus <boysc...@gmail.com> wrote in news:3dbf05f8-b8cb-4036-800e- > 940af73ce...@n33g2000pri.googlegroups.com: > > The problem is that all of them require an NDA just to see the > > datasheet. Because I intend to publish the design (this is open > > source hardware), I am not in a position where I can sign an NDA. In > > fact, there are FPGA development boards out there with published > > schematics that have a page marked "this page redacted due to an NDA > > on the Marvell 88E1111". This seems nuts to me. I mean, what could > > possibly be NDA-worthy in the datasheet for a 4 year old part that > > implements a technology (GbE) which is nearly a decade old? > > > Does anyone have any suggestions as to non-NDA encumbered GbE PHY > > chips? All I need is GMII or RGMII. > > The last time I looked at PHYs (a few years ago), the good ones required > NDAs for the datasheets, and the ones that didn't have NDAs weren't worth > using. > > I don't see how this clashes with open source hardware though. You need > to document the design of the interface to the PHY but this isn't the > same thing as violating the NDA for the PHY datasheet.
The fact that other users have been forced to redact pages of schematics tells me that this is a serious problem. Plus, all the source code is open as well, so who knows what secret registers that might reveal. Besides, I don't want to reward such bad behavior by these chip companies. Can you imagine how hard it would be to design anything if every chip you used had NDAs attached to it? I did find the ET1011C from LSI (formerly Agere). It is cheap, low power, available, and they give out the datasheets without an NDA. Anybody have any experience with it? Thanks, Matt