What is the suitable oscilloscope for PCIE and DDR 3 memory troubleshooting ?

[Kulunu]

Dear All,
I could figure out the reason for some of the custom hardware designs are not working due to discontinuity of the PCIE high speed signals. So this may cause some times board work fine but with the time (when it get heat) this PCIE link is going to fail. So in order to check these signals we need to have oscilloscope which has time-domain reflectometer (TDR). Could you please help me to find a suitable oscilloscope for this or else any laboratory to do these testings ?

As well as in order to measure signal integrity of the DDR 3 signals the oscilloscope should have 6GHz bandwidth. So I'm searching for oscilloscope which has above both capabilities. What are the low cost recommended oscilloscopes for this ?

I must be thankful to you if you can send me information on above equipment.

Have a nice day !!!

Regards,
Kulunu

[mohsin_qau]

Please check the specs of Agilent Part#DSO 9254A.

[robertferanec]

I can not really help with this question. This equipment is a way above my budget.

However, from my experience, unless you are designing something what has never been designed before (and have never been tested or validated) or something really extreme, then you should not need debugging these interfaces. From my experience, if you are designing a board based on a reference design, then if you see unreliable behavior on for example DDR3 memory, debugging is not really going to help - only what helps is complete redesign.

[Kulunu]

Hi Robert,

Dear Robert,

Thank you very much for your feedback.

This forum helped me to shear lot of information related to custom hardware design. First thank you for that. My custom hardware is based on nitrogen 6 max board and I could learn lot of thing from Open Rex open hardware memory layout. Hope you can remember I have asked several questions related to memory calibration and memory stress test. And finally our discussion ended with your suggestion of doing memory calibration in environmental chamber. But I couldn't find any equipment like that in my country and I was searching one to hire. But it is really difficult.

I have used nxp stress test tools and stressapp test in room temperature. But I didn't get any error though I ran those boards in week (7 days). But some boards get kernel panic when I run certain application with PCIE cameras and that application uses 3D gpu hardly. But the reason is some boards worked really well and didn't crash at all.

When I disable PCIE connection from dtb level some boards started to work. So I could figure there is a problem with pcie link (board to board connection) So I started to search deeper in to this and then I could see NXP has added pull up and pull down resistor to match clock as PCIE is using LVDS reference clock and need to to that small hack if LVDS clocks are using for PCIE. But still some of the boards are not working even disable PCIE from dtb level.

In order to isolate the problem again I need to debug this boards using oscilloscope or any other equipment. Like observing eye diagram of PCIE signals and memory. As well as in order to do design changes to memory layout or any other area I need to pinpoint the issue. Because I don't know which area should be redesign.

So what do you suggest for this ?

Waiting for your feedback.

Regards,
Kulunu.

[robertferanec]

This doesn't have to be PCIE layout or clock problem. It could be caused by many other factors e.g. heat (do you have a good heatsink on CPU?), power stability (any small drop on power can cause problems - where your PCIE power is coming from, is it strong enough?), SD card problem (are you running your system from SD card or SATA?), software issue (unstable driver or leaking software for the application), silicon issue (I think, there were some issues with PCIE on silicon level - it didn't work with all peripherals and I am not sure if I remember right, but software was doing some PCIE tweaking in registers), etc .....

I would start with the heat - I have seen boards crashing if CPU was too hot - and especially if you are running GPU, that is going to stress the CPU. Then I would double check the power - if the problem is only visible when CPU is stressed, than these are the two areas what I would try to check and be sure there is no problem with heat nor with power.

Also, still, try to run the environmental chamber tests - if there is a problem on the board, you may be able more easily to replicate it.

[Kulunu]

Dear Robert,

Thank you for your feedback.

Could you please give me more information about this ?

"(I think, there were some issues with PCIE on silicon level - it didn't work with all peripherals and I am not sure if I remember right, but software was doing some PCIE tweaking in registers), etc ....."

From where I can get more details ?

Regards.
Kulunu

[robertferanec]

I found out that when I was testing different PCIE cards in our iMX6 Rex Development board - some where not recognized, so I had a look into drivers and I saw there some settings with comments. But what files exactly, I do not remember.