DP83TD510E-EVM: Poor data rate in one direction with full duplex transmission

I've continued to repeat the test.  Sometimes I see results like this:

Both data rates start pretty high, then one drops.

Hi Cyrus,

I am talking with the team regarding this issue. Will have an update by Friday 1/21/21 at the latest.

Sincerely,

Gerome

Hi Cyrus,

Are the boards configured in the same way? Can you please share with me the strapping information?

In addition, have you tried swapping the boards out for new boards? Want to see if the problem is setup related or board related.

Sincerely,

Gerome

Hello Gerome,

Yes, the boards are configured the same way.  Here are some photos.

The DP83TD510E is RMII Master by hardware the configuration instructions in the EVM User's Guide.  The EVMs are USB powered.  I also shorted R21 on both EVMs to connect USB ground to board ground directly.

3.1.3 USB Power Supply Option
DP838TD510E block powered by USB.
• Plug in micro-USB to J4
• Place shunt at J9
• Place shunt at J14 in the "On-Board" position
• Place shunt at J15 in the "On-Board" position
• Place shunt at J16 in the "On-Board" position

3.2.1 DP83TD510E in RMII Master Mode, DP83822I in RMII Slave Mode (default)
• Populate R149
• Remove R143, R136, R91
• Set the correct hardware straps or perform the correct software write.
Hardware Strapping
• Populate R84, R103 with 2.49k (1%)
• Place shunt on LED_0, J20 in "Mode 0" position

I have not swapped the EVM boards for other boards.  We found this issue in some boards we designed doing the same thing as the EVM boards.  Our boards use the DP83TD510E and DP83822HFRHBR.  Our boards are configured to have the DP83822HFRHBR as RMII Master.  They showed the problem so I went back to the EVM boards to see if I could reproduce it and I could.  So it should not be a problem with the EVM boards.  It must be setup related.

Thanks,

Cyrus

Hi Cyrus,

Would it be possible to get a register dump of the DP83822 Reg 0x0 - 0x1F? Can you also provide Reg 0x467 of both DP83TD510E and DP83822?

If you are to connect directly between PC's, do you see the same issue?

Also from my understanding, your design has DP83822 as master and DP83TD510E as slave, while the EVM setup has this swapped where DP83822 is slave and DP83TD510E is master, correct?

Sincerely,

Gerome

Hi Gerome,

Here is the reg dump of the DP83822:

Register 0000 is: 3100
Register 0001 is: 7869
Register 0002 is: 2000
Register 0003 is: A240
Register 0004 is: 0061
Register 0005 is: CDE1
Register 0006 is: 000F
Register 0007 is: 2001
Register 0008 is: 4DA7
Register 0009 is: 0000
Register 000A is: 0100
Register 000B is: 1000
Register 000C is: 0000
Register 000D is: 401F
Register 000E is: 0000
Register 000F is: 0000
Register 0010 is: 4017
Register 0011 is: 0108
Register 0012 is: 2400
Register 0013 is: 2800
Register 0014 is: 0000
Register 0015 is: 0000
Register 0016 is: 0100
Register 0017 is: 00E1
Register 0019 is: 8025
Register 001A is: 0000
Register 001B is: 007D
Register 001C is: 05EE
Register 001D is: 0000
Register 001E is: 0002
Register 001F is: 0000
Register 0467 is: DFF3

Here is the DP83TD510E 0x467:

Register 0467 is: 0086

You are correct: in our design DP83822 is master and DP83TD510E is slave, while the EVM setup has this swapped where DP83822 is slave and DP83TD510E is master.

I will try direct connecting the PCs and let you know what happens.

Cyrus

I connected the PCs with an Ethernet cable directly.  That turned out weird too:

But I get over 100Mbps in both directions consistently.  It seems like the 10BASE-T1L 10Mbps ought to be handled easily.

Cyrus

Hi Cyrus,

There seems to be a slight inconsistency when reading register 0x1 vs 0x10 on DP83822. Could you please read register 0x1 multiple times over and confirm it is 0x7869?

In addition, could you please provide the value for DP83822 register 0x468?

Sincerely,

Gerome

Here you go.

Cyrus

Hi Cyrus,

Thank  you for the confirmation of register 0x1 and 0x468. I will talk with the team regarding this issue. I expect an update by Friday at the latest.

Sincerely,

Gerome

Hello Gerome,

When you say "As another configuration, in addition to power up and cable connection, both DP83TD510E's had to have reg 0x17 written as 0x4021" do you mean that to setup the test you had to write reg 0x17 as 0x4021, or do you mean you did another test writing reg 0x17 as 0x4021 and got the same results?

• Please tell me the details of your test so I can reproduce it as closely as possible.  For example:
• What data rate can be sustained in each direction simultaneously with full-duplex communication?
• How do you generate and monitor the communications?  Are you using PCs and a specific software app installed on each one?  Are there any special setup instructions?
• How long did the test run?
• Any other details you can share about the test.

Thanks,

Cyrus

Hi Cyrus,

This setup was made by plugging in both boards to two ports of Spirent Smartbits and generating traffic on both ports while checking received packets. I can get back to you regarding data rate early next week, but I ran an experiment that ran packets from one port to the other and vise versa. Each port as transmitting and receiving packets. This test was run with no errors (such as CRC, among others) for about 10 minutes. 

This was on a single PC to monitor and configure both ports. 

And yes, to get this EVM to work in this configuration, I had to write 0x17 = 0x4021 on both DP83TD510's (as these were both RMII master) in order to get the DP83822 online. 

Sincerely,

Gerome

Hi Cyrus,

The data rate on both ports were 1,219,504 and 1,219,390 bytes per second. This would translate to ~9.75mbps for both directions.

Sincerely,

Gerome

Hello Gerome,

OK, those results are great.  Can you tell me the Spirent Smartbits equipment information (model number of chassis, model numbers of cards)?  I can try to get the same equipment and replicate the test.

Thanks,

Cyrus

Hi Cyrus,

We are using Smartbits 2000 with 2 SX-7405 cards.

Sincerely,

Gerome

Hello Gerome,

I was able to create a similar setup and the data rate results matched yours.  Thanks for your help.

Cyrus