i.MX6 - Ethernet PHY Interface

[durmus]

Hello all,

We're currently designing a custom board based on SABRE SDP. We chose to use KSZ9021RN as ethernet PHY.

We're having some difficulties about the voltage levels.

In our design:

i.MX6:

NVCC_RGMII = 2.5V and NVCC_ENET = 3.3V as in SABRE SDP.

KSZ9021RN:

AVDDH = 3.3V, DVDDH = 2.5V (similar SABRE SDP's configuration with AR8031).

Since the voltage domain of the signals RGMII_MDC, RGMII_MDIO, ENET_REF_CLK, RGMII_INT is NVCC_ENET,

it seems to me that a voltage level translation is needed for these signals. However, in SABRE SDP, a level translator

exists only for ENET_REF_CLK.

In datasheets of both AR8031 and KSZ9021RN, it says that an external pull-up resistor is required to 3.3V/2.5V digital VDD

for MDIO and INT. Since RGMII 2.5V is used in SABRE SDP, I believe it should be the latter (2.5V), yet they are tied to 3.3V.

What am I missing here? This issue confuses me a lot and I cannot be certain of the connections in our design.

Can anyone clarify this issue?

Thanks in advance.

[igorpadykov]

Hi Durmus

signals naming in schematic may be misleading.

RGMII_MDC, RGMII_MDIO, ENET_REF_CLK, RGMII_INT is NVCC_ENET are

on i.MX6 NVCC_ENET = 3.3V power domain.

Other - such as RGMII_RXDn,RGMII_RTXDn are on NVCC_RGMII = 2.5V.

Pads power domain allocation one can find in

Table 100  "21 x 21 mm Functional Contact Assignments"

IMX6DQCEC i.MX 6Dual/6Quad Applications Processors for Consumer Products - Data Sheet

Best regards

chip

-----------------------------------------------------------------------------------------------------------------------

Note: If this post answers your question, please click the Correct Answer button. Thank you!

-----------------------------------------------------------------------------------------------------------------------

[durmus]

Hi chipexpert,

Thank you for your reply. Actually, what I was referring to is the PHY side.

Ok RGMII_MDC, RGMII_MDIO, ENET_REF_CLK, RGMII_INT are NVCC_ENET = 3.3V

But aren't they on 2.5V on the PHY side?

Thanks,

Durmus

[igorpadykov]

Hi Durmus

could you attach PHY datasheet or part of it

which proves your statement ?

Best regards

chip

[durmus]

Hi chipexpert,

Below are screenshots from related sections of KSZ9021RN and AR8031 (phy used in SABRE SDP). Links to datasheets are also included.

KSZ9021RN

AR8031

It says pull-up to 3.3V / 2.5V and since digital level of PHY is 2.5V in SABRE SDP, I thought they should have been pulled-up to 2.5V instead of 3.3V.

Thanks.

Durmus

Datasheets:

http://http://www.micrel.com/_PDF/Ethernet/datasheets/ksz9021rl-rn_ds.pdf

AR8031 datasheet - Atheros Communications - Integrated 10/100/1000 Mbps Ethernet Transceiver

[igorpadykov]

Hi Durmus

first please look at attached part of AR8031 datasheet.

Second, regarding your statement

"It says pull-up to 3.3V / 2.5V and since digital level of PHY is 2.5V in SABRE SDP, I thought

they should have been pulled-up to 2.5V instead of 3.3V." -

Why do you think that "they should have been pulled-up to 2.5V instead of 3.3V." ?

Datasheet allows for both 3.3V and 2.5V.

Best regards

chip

[durmus]

Hi chiexpert,

Firstly, ok, my mistake. The link I provided is an old revision of datasheet. The one where I took the screenshot from has a date of August, 2011. Hence more recent, and must be more valid.

Secondly, I believe what they are referring to is not a range of 2.5V-3.3V, but the selection of voltage level of digital interface you are using (i.e. either 2.5V or 3.3V). Therefore, if you are using

2.5V for digital interface (which is the case), using 3.3V for pull-ups seem.

[igorpadykov]

Hi Durmus

these interfaces are all "digital".

Only i.MX6 RGMII_RXDn,RGMII_RTXDn are on NVCC_RGMII = 2.5V.

Other - at NVCC_ENET = 3.3V power domain

Best regards

chip