cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 
bwk
Observer
Observer
7,018 Views
Registered: ‎09-30-2010

Aurora v2.9 Example on ML310 only working with parallel loopback

Jump to solution

Hi,

 

I generated the Aurora Core (v2.9) with the Core Generator and tried to implement the example design for the ML310 evaluation board (Virtex 2 Pro). I edited the example UCF to match the board pins. After that tried the parallel loopback and everything worked fine (got the CHANNEL_UP and LANE_UP signal). But when I switch to serial loopback I don't get the CHANNEL_UP and LANE_UP signal.

 

Any ideas?

 

The clock is connected correctly. I used some counters an blinking LEDs to test the clock.

 

Thanks, BwK

0 Kudos
1 Solution

Accepted Solutions
bwk
Observer
Observer
6,438 Views
Registered: ‎09-30-2010

Call me stupid. Now it seems to work properly. I only set TX_PREEMPHASIS to 3 on the first instance of GT_CUSTOM. Setting TX_PREEMPHASIS to 3 on the second instance solved the flickering LEDs problem. Now I can also use both reference clock rates (125.000 and 156.250 MHz).

 

Thanks a lot for help,

BwK

View solution in original post

0 Kudos
16 Replies
Anonymous
Not applicable
7,016 Views

Have you made sure that you are connecting the TX and the RX pins as recommended by the V2 Pro RocketIO user guide UG024 on page 91 and on? If this is not done correct you can get data contention.

0 Kudos
bwk
Observer
Observer
6,992 Views
Registered: ‎09-30-2010

They should be connected correctly. This example design is created by the Core Generator. I also tested this example on an ML321 board. On this board the example worked with parallel and serial loopback. And it also worked with a loopback connection via SMA cables.

 

Any ideas why it won't work with the ML310 board?

 

Thanks, BwK

0 Kudos
mcgett
Xilinx Employee
Xilinx Employee
6,971 Views
Registered: ‎01-03-2008

How do you implement the serial loopback test on the ML310?

------Have you tried typing your question into Google? If not you should before posting.
Too many results? Try adding site:www.xilinx.com
0 Kudos
bwk
Observer
Observer
6,967 Views
Registered: ‎09-30-2010

I set the LOOPBACK signal of the MGT to "10". This enables the serial loopback. Setting it to "01" enables the parallel loopback. And "00" is the normal mode without any internal loopback.

0 Kudos
mcgett
Xilinx Employee
Xilinx Employee
6,963 Views
Registered: ‎01-03-2008

Did the ML310 have anything external connected to the MGT that you were using?

------Have you tried typing your question into Google? If not you should before posting.
Too many results? Try adding site:www.xilinx.com
0 Kudos
bwk
Observer
Observer
6,957 Views
Registered: ‎09-30-2010

I only connect the MGT BREF Clock and some LEDs to the design.

0 Kudos
mcgett
Xilinx Employee
Xilinx Employee
6,954 Views
Registered: ‎01-03-2008

> I only connect the MGT BREF Clock and some LEDs to the design.

 

I meant physical connections on the ML310 board, not within the design.

 

Which MGT are you using?

------Have you tried typing your question into Google? If not you should before posting.
Too many results? Try adding site:www.xilinx.com
0 Kudos
bwk
Observer
Observer
6,946 Views
Registered: ‎09-30-2010

No, there are no physical connections.

 

I use MTG X0Y1, but I also tried some other.

0 Kudos
mcgett
Xilinx Employee
Xilinx Employee
6,941 Views
Registered: ‎01-03-2008

The MGT X0Y1 (and I think all of the MGTs) is connected to the XPM interface.  If you don't have a XPM module inserted then nothing would be attached to the MGTs.

 

You said that internal parallel loopback was working, but that the internal serial loopback is not.  The only two explanations that I can have for this.

 

1) The MGT PLL is not working correctly for the TX and RX path and this is not needed for the parallel loopback

   - This could be due to not having the right reference clock selected, PLL multiply/divide settings or a power down or reset state

2) The reflections caused by an unterminated TXP/TXN are creating bit errors for the RX path

------Have you tried typing your question into Google? If not you should before posting.
Too many results? Try adding site:www.xilinx.com
bwk
Observer
Observer
5,619 Views
Registered: ‎09-30-2010

I think I found out what the problem is. I tried again the serial loopback on the ML321 board. This time I disconnected the SMA cables and I didn't get the LANE_UP and CHANNEL_UP signals. After that I connected the SMA cables again and then I got the signals.

 

So it seems that I need a physical connection in serial loopback mode.

0 Kudos
bwk
Observer
Observer
5,611 Views
Registered: ‎09-30-2010

Now I have the serial loopback working on the ML310 board. I connected MGT X0Y1 with MGT X3Y0 and created the Aurora Core with 2 lanes (lane 1: MGT X0Y1; lane 2: MGT X3Y0). Now I get the CHANEL_UP and LANE_UP signals in serial loopback mode. But it still won't work in normal mode (without loopback).

0 Kudos
mcgett
Xilinx Employee
Xilinx Employee
5,602 Views
Registered: ‎01-03-2008

> But it still won't work in normal mode (without loopback).

 

You said earlier that don't have anything in your physical ML310 system to connect the two MGTs, so it can't work.

------Have you tried typing your question into Google? If not you should before posting.
Too many results? Try adding site:www.xilinx.com
0 Kudos
bwk
Observer
Observer
5,591 Views
Registered: ‎09-30-2010

I said there was nothing connected, because I thought the serial loopback works without any physical connection. But I have a connector that connects MGT X0Y1 with MGT X3Y0. With this connector the serial loopback works when it is plugged in. But the normal mode (without loopback) still won't work.

0 Kudos
mcgett
Xilinx Employee
Xilinx Employee
5,578 Views
Registered: ‎01-03-2008

Then the problem is likely with the module that you are plugging in to the XPM interface.  The other possibilities is that enter channel is too lossy for the default settings for TX output swing and pre-emphasis with the XPM interface/loopback and need to be adjusted.

 

I would start first with understand how the XPM module implements the loopback and is it correct.

------Have you tried typing your question into Google? If not you should before posting.
Too many results? Try adding site:www.xilinx.com
0 Kudos
bwk
Observer
Observer
5,558 Views
Registered: ‎09-30-2010

Now I have a "working" version. This time I generated a core with streaming interface and used a clock rate of 125.000 MHz as reference clock rate. I also set TX_PREEMPHASIS to 3 and now I get CHANNEL_UP and LINE_UP signals. Nevertheless the LEDs showing CHANNEL_UP and LINE_UP status are flickering. But I think (and hope) there are only some little adjustments to get rid of this.

0 Kudos
bwk
Observer
Observer
6,439 Views
Registered: ‎09-30-2010

Call me stupid. Now it seems to work properly. I only set TX_PREEMPHASIS to 3 on the first instance of GT_CUSTOM. Setting TX_PREEMPHASIS to 3 on the second instance solved the flickering LEDs problem. Now I can also use both reference clock rates (125.000 and 156.250 MHz).

 

Thanks a lot for help,

BwK

View solution in original post

0 Kudos