Quickstart Board Issue - gets too hot - how to fix

cancel
Showing results for 
Show  only  | Search instead for 
Did you mean: 

Quickstart Board Issue - gets too hot - how to fix

2,967 Views
MXLuser
Contributor I

The first cut (Assembled / BOM per Rev C schematics) Quickstart boards get hotter than expected. The reason is a BOM error where a 10 ohm resistor was used instead of what I imagine was supposed to be 10K ohms. See attached images for the location of R199.

 

To determine if your board has this issue, measure R199. If it is 10 ohms, replace it with something in the 10K - 50K range. However, notice that it is a 0201 part, so find someone with a Metcal, good eyes, and steady hands.

 

Schematics affected: SPF-26565_C.pdf (shows R199 as 10 ohms). Assy numbers affected not known.

 

Freescale has been informed of this issue.

 

Attached also is image (R199_fix2.jpg) of a way to fix without specialized tools. Remove R199 by scraping it off with a soldering iron. Then attach a 10K 0603 or even an 0805 resistor from the polyfuse to the FET as shown. The value shown in the picture is 3322 or 33.2K - anything from 10K to 50K should be fine.

Labels (1)
0 Kudos
Reply
7 Replies

2,802 Views
clay_cowgill
Contributor I

I did a variety of experiments to track down heat on the board and I doubt that the increased core voltage is a big watt contributor.

 

The little diode hack on the back of the Quickstart board to power VDDOUT contributes a fair amount of heat (diode voltage drop * total board current).  After the R199 issue, it's the hottest spot on the system.

 

I was able to drop the PMIC temperature about ~7C by feeding VDDOUT from a 3.3V supply instead of the diode's ~4.2V output.  Since the diode was just a hack getting rid of it gives you a better idea of how custom hardware might actually perform.

 

Dropping the core from 1GHz to 800MHz only drops surface temps at the i.MX53 about ~3C on the package.  Cutting the DDR3 clock from 400MHz to 200MHz only brought the i.MX53 temps down another ~3.5C from that.  I can't see even a 20% core voltage reduction taking much more heat out than a 20% core clock rate reduction.  (I ran all my tests just with the stock ubuntu image with the 3D (pinball) demo and BBB movie running simultaneously to load it down a bunch.)

 

From my calculations, the core is only about ~29% of the total power budget of the Quickstart board.  So if they were to run the core at, say, 1.2V instead of 1.35V I can't imagine we'll see more than a ~5% reduction in total 'system' power.  It'll be cooler, but I doubt by much.  The power wasted in the PMIC by the diode supplying VDDOUT higher than necessary are bigger targets.  Of course fixing the R199 "whoops" saves about a watt too (~20%ish of the total power consumed on the board was due to R199 being too low of value).

 

-Clay

0 Kudos
Reply

2,802 Views
iafg
Contributor IV

NOTE - While the resistor may contribute to the board getting hot, the VPU errata FIX is running the processor at it's maximum voltage. See my notes here:

http://imxcommunity.org/group/imx53quickstartboard/forum/topics/linux-binary-demo-files-201103?xg_so...

Comment on VPU Errata and board getting hot.

0 Kudos
Reply

2,803 Views
PeterNewton
Contributor II

I had the 10 Ohm resistor replaced with a 10K Ohm, and the board is a lot cooler.  But there is a new problem.

 

When I boot the board with a USB device plugged in, the board powers itself off right when Linux initializes USB.  This happens right after the kernel prints:

 

usbcore: registered new interface driver usb storage.

 

This happens even with no USB device connected other than a keyboard in the upper USB port.

If, instead, I boot the board with no USB device attached at all, Linux boots normally.  I can then plug in a USB device (with Linux already booted), and it works. Has anyone else seen this?  Thanks.
0 Kudos
Reply

2,803 Views
PeterNewton
Contributor II

My board also has the 10 Ohm resistor.  How important is this fix?  Does the affected transistor run hotter than its specification permits?

 

Also, is there a best specific value for the replacement resistor?  The original post mentioned only a range from 10K to 50K Ohms.

0 Kudos
Reply

2,803 Views
paul_herbst
NXP Employee
NXP Employee

Hi Jonathan,  Thanks for finding this and posting to the community.  Freescale will be updating the next batch of boards and documents with this improvement.

0 Kudos
Reply

2,802 Views
MXLuser
Contributor I

You could return it, but these boards are hard to get, and it might take a while to get a replacement. If you can't find someone who can replace the part (can buy the replacement 0201 part from Digikey, also could probably fit a 0402 in there if need be), you could just live with it for now; just don't touch the part or transistor.

 

FSL is actually quite good at fixing broken EVKs, at least they were for me with the mx51EVK. I think they will come up with a way for QSB board customers to get their boards fixed.


X Xiao said:

Thanks for this info. Mine is 10 Ohms.

This is really hard to solder, can I exchange one or somehow FSL repair this for me? I consider this is a defect.

0 Kudos
Reply

2,802 Views
XXiao1z
Contributor III

Thanks for this info. Mine is 10 Ohms.

This is really hard to solder, can I exchange one or somehow FSL repair this for me? I consider this is a defect.

0 Kudos
Reply