One transistor too many

From Openmoko

(Difference between revisions)
Jump to: navigation, search
Line 1: Line 1:
 
Given-away fic1973's from phase 1 (red board) had one transistor too many. That one needs to be removed for correct sleep operation. See http://people.openmoko.org/roh/Gta01b_v4_front_Q1001.jpg .
 
Given-away fic1973's from phase 1 (red board) had one transistor too many. That one needs to be removed for correct sleep operation. See http://people.openmoko.org/roh/Gta01b_v4_front_Q1001.jpg .
  
Removing it is a nontrivial task, requiring experience and proper tools. Pavel (who wrote this page) did not yet dare to fix his fic1973. Be careful!
+
Removing it is a nontrivial task, requiring experience and proper tools. Pavel (who wrote this page) did not yet dare to fix his fic1973. You probably can't do this at home. Be careful!
  
 
The transistor pulls a 'reset' line on the gsm pmu. this pin on the
 
The transistor pulls a 'reset' line on the gsm pmu. this pin on the
Line 7: Line 7:
 
debounced. It is called q1001. When the pmu is shut down for suspend-to-ram the samsung somehow generates a
 
debounced. It is called q1001. When the pmu is shut down for suspend-to-ram the samsung somehow generates a
 
hard, short spike on all gpio, which would, due to no debouncing reset the gsm,
 
hard, short spike on all gpio, which would, due to no debouncing reset the gsm,
which wants to send data on the serial and wakes up the samsmung -> loop. since TI says the reset is only for testing, not for regular use, we removed
+
which wants to send data on the serial and wakes up the samsmung -> loop. Since TI says the reset is only for testing, not for regular use, we removed
 
q1001 from the placement map. Just the name was not bumped to gta01bv5 since that
 
q1001 from the placement map. Just the name was not bumped to gta01bv5 since that
 
would require changes to the gerber (labeling) and just not placing the transistor
 
would require changes to the gerber (labeling) and just not placing the transistor
does not. its basically a layout-revision.
+
does not. It is basically a layout-revision.

Revision as of 22:27, 25 November 2007

Given-away fic1973's from phase 1 (red board) had one transistor too many. That one needs to be removed for correct sleep operation. See http://people.openmoko.org/roh/Gta01b_v4_front_Q1001.jpg .

Removing it is a nontrivial task, requiring experience and proper tools. Pavel (who wrote this page) did not yet dare to fix his fic1973. You probably can't do this at home. Be careful!

The transistor pulls a 'reset' line on the gsm pmu. this pin on the pmu is not debounced, the other one for power (same wiring, other transistor) is debounced. It is called q1001. When the pmu is shut down for suspend-to-ram the samsung somehow generates a hard, short spike on all gpio, which would, due to no debouncing reset the gsm, which wants to send data on the serial and wakes up the samsmung -> loop. Since TI says the reset is only for testing, not for regular use, we removed q1001 from the placement map. Just the name was not bumped to gta01bv5 since that would require changes to the gerber (labeling) and just not placing the transistor does not. It is basically a layout-revision.

Personal tools

Given-away fic1973's from phase 1 (red board) had one transistor too many. That one needs to be removed for correct sleep operation. See http://people.openmoko.org/roh/Gta01b_v4_front_Q1001.jpg .

Removing it is a nontrivial task, requiring experience and proper tools. Pavel (who wrote this page) did not yet dare to fix his fic1973. Be careful!

The transistor pulls a 'reset' line on the gsm pmu. this pin on the pmu is not debounced, the other one for power (same wiring, other transistor) is debounced. It is called q1001. When the pmu is shut down for suspend-to-ram the samsung somehow generates a hard, short spike on all gpio, which would, due to no debouncing reset the gsm, which wants to send data on the serial and wakes up the samsmung -> loop. since TI says the reset is only for testing, not for regular use, we removed q1001 from the placement map. Just the name was not bumped to gta01bv5 since that would require changes to the gerber (labeling) and just not placing the transistor does not. its basically a layout-revision.