Unable to start debugger on XL216-512 board Topic is solved

Technical discussions around xCORE processors (e.g. xcore-200 & xcore.ai).
User avatar
mon2
XCore Legend
Posts: 1913
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Post by mon2 »

Pin 102 of your cpu wiring is not correct as shown. This is a +1v0 rail via the 4r7 ohm resistor and cannot mate with +3v3 rail.


View Solution
User avatar
BEBDigitalAudio
Experienced Member
Posts: 82
Joined: Thu Nov 11, 2010 7:45 pm

Post by BEBDigitalAudio »

mon2 wrote:Pin 102 of your cpu wiring is not correct as shown. This is a +1v0 rail via the 4r7 ohm resistor and cannot mate with +3v3 rail.
It does not mate with 3.3V. PWR-FLAG is just a CAD symbol to say that PLL_AVDD is a power supply pin that is powered by 1V0 via the 4R7 resistor. If you don't put this symbol on the schematics, the software issues a warning telling that PLL_AVDD pin is not powered (since R6 "isolates" the PLL_AVDD from the 1V0 source).

I think you understand PWR_FLAG symbols are connected together, which is not the case (each PWR_FLAG is independent)

Benoit
User avatar
mon2
XCore Legend
Posts: 1913
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Post by mon2 »

Understood.

Are you using NCP303LSN09 or assembled with NCP303LSN10? The 1v0 may be borderline for use so do confirm that reset is released and remains released during normal use.
User avatar
mon2
XCore Legend
Posts: 1913
Joined: Thu Jun 10, 2010 11:43 am
Contact:

Post by mon2 »

Otherwise, the design looks to be solid. In reviewing some past notes from other developers who faced the same issue, the #1 root cause we found was related to the assembly of the XMOS CPU.

One developer emailed for weeks only to conclude in the end that the OTP_VCC pin was not mated to +3v3 as the root cause. If you have a hot air gun, would not hurt to gently go around the TQFP pins to confirm the visible solder joints are solid. After this, would your assembly house have an x-ray machine to inspect the bottom side of the TQFP to confirm the metal belly is soldered ok? We still see some assembly issues with our local ISO certified CM even after very high volume (million++ PCBs) of production for our company and 20+ years of SMD experience. So it is worth a close inspection - a good digital microscope will help. On TQFP parts, our experience has been linked to the TQFP pins being not coplanar (often in the corner pins of the package) so the pin is sitting on top of an "air pad". Or sometimes, 0402 parts are slightly tombstoned so you can see that one side of a cap is not soldered.

Did both of your proto PCBs behave the same with respect to the XTAG-3 tool and xrun use? How is your external flash mated to this design?

Your TI part has a PG output that was not used and it could have replaced the 3v0 supervisor as a small cost reduction.

The power up delay between the 3v3 and 1v0 rails is critical but from the applied cap values, you should be ok.

Do you have a reference XMOS kit based on the XCORE-200 series? If yes, could you code up a flash device for a simple LED blinky -> confirm that it works on the official XMOS kit and then remove that flash device with the LED blinky and apply onto your custom PCB. Then, does your custom PCB continue to blink the same port pin as the original? That is what I would do if I was there (aside from having a nice tea with a tasty pastry :)


update...

this developer (phaseimage) used 100nf (ie. 0.1uf) vs. 1nf on your PCB for the delay cap for the same voltage supervisor and has a working design:

http://isogrid.org/blob/CrowdSwitchHW_Design_0v5.zip

Code: Select all

http://isogrid.org/blob/CrowdSwitchHW_Design_0v5.zip
Can you try the same to see if the results improve?
Attachments
xl216_partial.png
(77.68 KiB) Not downloaded yet
xl216_partial.png
(77.68 KiB) Not downloaded yet
User avatar
BEBDigitalAudio
Experienced Member
Posts: 82
Joined: Thu Nov 11, 2010 7:45 pm

Post by BEBDigitalAudio »

mon2 wrote:Understood.

Are you using NCP303LSN09 or assembled with NCP303LSN10? The 1v0 may be borderline for use so do confirm that reset is released and remains released during normal use.
Yep, I already made the change on the PCB. The first version was equipped with a LSN10, but I replaced it with a LSN09 (we have older designs based on XS1-L16 chips, and we use the LSN09. The designer made a small error by choosing the LSN10, but we have not yet updated the schematics)

Beside this, I will follow your suggestions. It costs nothing to check the soldering, I am even thinking on putting some soldering flux all around before passing over with the air gun.

We will compare also with the second prototype, to see if we get the same results or if we were only unlucky with the first tested prototypes. I will also raise the value of the delay capacitor. Having a longer reset delay between 3.3V and 1.0V is not bad, as far as I know.

And last thing : yes, we have some XCore200 Explorer kits here, and we know our code is running correctly on them (but the part is not the same : XE on the Explorer, XL on our board). Anyway, I will switch back to basics and rather than running the whole code, I will play with a single port only.

About the tea and the pastries : make me a sign when you come one day in the Lille area in France, I promise you some excellent pastries to thank you for your help °-)

(And now, time for a break until next week...)
User avatar
BEBDigitalAudio
Experienced Member
Posts: 82
Joined: Thu Nov 11, 2010 7:45 pm

Post by BEBDigitalAudio »

Finally, since everything else seemed to be ok, we have decided to come back to the basics and redo manually *all* solder joints around the XL216 (visual inspection was OK).

And tadaaaaa !! The debugger has loaded its first binary in the chip this morning 8-)

In conclusion : even the best sub contractor can do mistakes when dealing with a TQFP128 package...
Post Reply