I'm going to put in a request with the Electrical Engineer most familar with this SOM to see what his opinion. I know most of our SOM's are placed into a temperature chamber and operated after soaking at the exterme high and extreme low temperatures.

I am not sure if the USB system is tested at temperature or not, but I know adding more and more peripherals internal to the processor can increase its operating temperature.

I noticed that you posted on TI's E2E forum. This is good.

In your post you make the statement "These also get far worst with lower voltage" and I was hoping you might be able to elaborate on this. There are minimum and maximum voltage ratings for the various power rails, so I was hoping you might be able to elaborate on what you're doing with voltage and what you're seeing.

Thank you,

adam

I am still waiting to meet with our Electrical Engineer, but I found some thermal testing documents in the process.

At one point we published a white paper on the DDR timings. This article is mostly focused on the low-temerature extreme, but it describes the DDR timings which have been tested at both extremes. http://support.logicpd.co...rtalid=0&EntryId=963

You'll need to be logged into our system to download the file.

These timings came from TI, but tested by Logic PD.

Whenever we see strange behavior I like to look at the timings to make sure everything is operating within normal specs. Since I don't know when your configuration was first setup, I want to have you check the timings with this document and comare the DDR timings.

For the purposes of the above testing, several devices were placed into a temperature chamber and:

allowed to soak, unpowered, at their end temperatures for approximately one hour before testing. When the soaking was complete, the assemblies were powered up. Power to the assemblies was cycled by plugging in and then unplugging their respective AC plugs. This method allowed power cycling without opening the thermal chamber to toggle the baseboard power switches.[\i]

adam

Quentin,

It appears as if the USB was not tested at extreme Temperature, but the processor and DDR were through text I sent you above.

The datasheet for the AM3517 itself uses the phrase 'Operating junction temperature' would would indicate to me it doesn't take into account airflow.

adam

Hi Adam,

Thanks for the replies......I think you have answered my questions, and I thought I had replied to this one you had already but I don't see my reply here....so here it is.

Our four corner tests (temperature and voltage) were done within the specification bounds, so all margins were within spec. When I talk about low voltage, I am talking about -5%. So the issues could start to be seen at low voltage and nominal (room?) temperature, but then the same issues really came out in a much more severe manner at the low voltage / low temperature corner,

I hope that answers the question.

-Quentin

 

 

Posted By adam.ford@logicpd.com on 18 Dec 2017 10:27 AM
.......
I noticed that you posted on TI's E2E forum. This is good.

In your post you make the statement "These also get far worst with lower voltage" and I was hoping you might be able to elaborate on this. There are minimum and maximum voltage ratings for the various power rails, so I was hoping you might be able to elaborate on what you're doing with voltage and what you're seeing.
 

 

Hi Adam,

Thanks for the info. This might be what we are seeing with the USB, and I don't think this is giving us a problem because higher level retries seem to get the transaction through, but I thought I would ask about it because we are having another problem with NAND FLASH, and I was thinking that we might be allowing the SOM to get too hot. (and that problem might be related)

I don't have any other measurements that show that it is in fact too hot, but I was thinking that these errors might be a sort of canary in the coal mine....but I am thinking right now that it is not.

I believe I have found at least some of our NAND FLASH issues to be related to the code not handling both scrubbing of PEB (physical erase blocks) with ECC issues nor "empty" PEB with ECC issues. I think I see that bugs have been found in the released 2.6.37 Linux code,,,,,and I need to figure out how to either get those fixes into 2.6.37 (perhaps by porting MTD, UBI and UBIFS from 4.4.x to 2.6.37) or to port our changes all the way to 4.4.x (not a topic for this thread though :-) )

Thanks,

-Quentin

 

Posted By adam.ford@logicpd.com on 19 Dec 2017 10:21 AM
Quentin,

It appears as if the USB was not tested at extreme Temperature, but the processor and DDR were through text I sent you above.

The datasheet for the AM3517 itself uses the phrase 'Operating junction temperature' would would indicate to me it doesn't take into account airflow.

adam