Charge not completing in ADP5063

[This is a followup and refinement on a previously posted question, ez.analog.com/.../bypassing-ldo-isolating-battery-when-fully-charged]

Short form:

• What must be true for the ADP6053 to enter "Charge Complete" state?
• What in my settings (below) is preventing it from going into Charge Complete state?

Details:

We are using the ADP5063 in a battery backup application. It is our hope that once the battery is fully charged, ISO_Sx will be set to VSYSTEM, but before we can answer that question, we need to understand why the battery never completes charging.

The setup:

• VINx = 5.0v (coming from a "stiff" power source, not from USB and not current limited)
• Battery is a single cell 18650 LiION battery
• ILIM (register 0x02) set to 2100 mA
• VTRM (in register 03) set to 4.2V
• EN_CHG (in register 0x07) set true
• VSYSTEM (in register 0x08) set to 5.0V
• all other register set to their defaults

What I observe is that ISO_Bx gets up to about 4.15V.  There is a 10 mOhm sense resistor at the low end of the battery that measures about 0.9 mV (and is slowly approaching 0.0v as I type this), meaning that the current through the battery is about 90 mA.   

The CHARGER_STATUS (in register 0x0B) bobbles between "fast charge CC" and "fast charge CV", HOWEVER it never gets to "charge complete" and CHDONE never gets set.

Additional notes:

None of the fault bits are set.  Watchdog is not enabled.

For Charge Complete, the documentation says:

The ADP5063 charger monitors the charging current while in fast charge constant voltage mode. If the current falls below IEND and remains below IEND for tEND, charging stops and the CHDONE flag is set. If the charging current falls below IEND for less than tEND and then rises above IEND again, the tEND timer resets.

In our case, IEND is 52.5 mA (the default).  And I clearly see less than 52.5 mA flowing through the battery for periods longer than TEND (7.5 miinutes).  However, ISO_Bx is at 4.15V, NOT the 4.2V as required by VTRM.  (Then why is the system switching between constant voltage and constant current mode?).  I tried lowering VTRM to 4.0V (by writing 0x64 to register 0x03).  Although the voltage dropped to 4.03v (now greater than VTRM), the Charge Complete bit did not set.

Two Hypotheses:

Favored Hypothesis: Noise on the ISO_Bx line.  The scope trace below suggests that ISO_Bx == VTRM briefly (so it goes into constant current mode) and then a moment later it reads ISO_Bx < VTRM (so it reverts to constant voltage mode).  In particular, since this happens every 12 uSec, the time requirements for TEND are never met.  My hardware colleague will make more precise measurements of this with his superior test gear.

Less likely Hypothesis: There is a possibility that something on the ISO_Bx line is sinking additional current.  If the Favored Hypothesis is debunked, my hardware colleague will cut some traces and make direct current measurements on the ISO_Bx line to make sure that there are no additional current sinks that could be fooling the ADP6053

Regardless, I'll update this note accordingly when I have more information.