Other Parts Discussed in Thread: MSP-EXP430F5529LP, DP83869EVM, DP83869, USB-2-MDIO
We have an in-house custom designed board that utilizes the TI DP83869HM as a media converter interface. The intent of this design is to interface an existing 10/100Base-T wired Ethernet interface (of a daughter card module) to an on-board SFP socket. The SFP socket then allows the use of SFP modules (wired, MMF, or SMF) to be installed in each board depending on link distance required for that unit’s field application (i.e., could vary from less than 100 ft per link to several miles). Initially we proved the design using wired SFPs which is the simplest case (i.e., does not require another external SFP to translate back to wired interface at opposite end of external link). Once we switched to fiber optic SFPs, we have had issues.
Our custom board design used the DP83869EVM as design reference for the DP83869HM portion of our custom board. It includes all the same configuration strap jumpers that are provided by the EVM board which are all implemented in the same way as shown in the EVM schematics. In our design, the DP83869HM is used as an unmanaged converter so we did not include the MSP430 launch pad or any other MCU on board. However, the MDIO and MDC pins go to an SMI connector same as on EVM board. This allows us to connect an external launch pad. I have purchased the TI MSP-EXP430F5529LP for this purpose.
I also have two DP83869EVMs that I can use as independent external media converters.
I have a simple strategy where I prove out any fiber optic SFP modules prior to testing them with our custom design. To do this I use the following basic test setup:
PC1 <====> [UWS1] <====> M/C1:SFP1 <::::x::::> SFP2 :M/C2 <====> [UWS2] <====> PC2
where:
PC1 & PC2 are PCs with 10/100/1000 NICs configured with separate fixed local IP addresses.
[UWS1] and [UWS2] are optional independent 10/100BASE-T Unmanaged Wired Switches.
M/C1 and M/C2 are external individual media converters (with single SFP socket and single RJ45 jack).
SFP1 and SFP2 are two fiber optic SFPs being tested (usually identical models but can also be mixed).
<====> represents a single wired Ethernet jumper.
<::::x::::> represents a dual fiber jumper with appropriate cross-over of fiber between the end connectors.
The above chain is everything in the basic SFP proving test setup. There are no other network connections to the PCs nor to the optional switches.
The two unmanaged switches allow me to force the link speeds on either side down to 100 Mbps. Without the switches, the PC NIC to M/C link speed will be 1000 Mbps. This is verified by link speed indicators on the M/Cs.
To pre-test a pair of FO SFPs, I set them up in the above configuration and verify I can communicate from PC1 to PC2 (simple ping tests both ways). I have 3 different types of packaged commercial M/Cs (in addition to the two EVMs) that I can use for the M/Cs. Once I have a particular pair of SFPs working in this proof setup, I then can remove SFP2 from M/C2 and plug it into our custom board's SFP socket to test if the SFP works in our design. In this way, I am making sure that everything from PC1 to SFP2 is working before testing with our design. The test setup with our custom board is then as follows:
PC1 <====> [UWS1] <====> M/C1:SFP1 <::::x::::> SFP2 : Custom Board : Daughter Card
The PHY interface on the Daughter Card is also configured with an independent fixed IP address. In this setup, if I can ping from PC1 through to the daughter card, then the test passes.
The problem is that I can get SFPs that will work between two EVMs as M/Cs and in this case I can then get the same SFPs to work with EVM as M/C1 and with SFP2 plugged into our custom board. Some of those same SFPs will also work using the commercial packaged M/Cs on both sides in the basic proving test, but I can never get a commercial M/C in M/C1 position to work with EVM as M/C2 or with our custom board.
I would have expected that if an SFP works with the commercial M/Cs on both sides and the same SFPs also work with EVMs as M/Cs on both sides, that I should then be able to get that same pair of SFPs to work with commercial M/C on the left and EVM M/C on the right. But I have not been able to get this to work. Can someone please shed some light on what I might be missing here that is preventing the commercial M/Cs from working with the EVM which in turn means they won’t work with our custom board? Or alternatively, anything that I can try on the EVMs to get it to work with a commercial M/C? If we can solve that, then we should be able to setup our custom board with the same configuration as EVM and get it to work with the commercial M/C too.
Thanks in advance for any help or guidance you can provide.