Hi all — I’m a hobbyist looking for help debugging a solar Li‑ion charger based on the LT3652 (schematic attached: schema_v1.pdf).
I’m charging two 18650 cells in parallel, but the charger consistently stops early: it terminates around 4.04–4.08 V instead of reaching the expected 4.20 V float voltage.
Setup
Input: small solar panel, ~18 V open‑circuit (VIN is the panel voltage under load)
Battery: 2× 18650 in parallel (treated as a single cell)
Charger: LT3652 with VFB divider set for ~4.2 V float
Icharge NEVER goes over 25mA (design current is 1000mA!!!)
Observed behavior
Charging current tapers and CHRG goes low when the battery reaches ~4.04–4.08 V
The battery never reaches 4.20 V
Same behavior across multiple PCB revisions and different divider values
VFB divider tests
Initially designed per datasheet guidelines (including input bias compensation)
Later increased equivalent resistance to better match recommendations (~250 kΩ)
Also tested other combinations (e.g. 330 kΩ / 1.2 MΩ)
Result unchanged: termination still occurs at ~4.0 V
Measurements and checks
Current/voltage monitored with INA219 (both battery and panel side)
Current drops to near zero at ~4.05 V
SHDN pin:
Always > 1.25 V → device is enabled (threshold ~1.2 V)
VIN_REG:
Measured > 2.7 V → no apparent input undervoltage regulation
VFB:
Around 3.3 V at termination (as expected for CV phase)
Input power observations
With ~16 V available from the panel, input current is only ~20–25 mA
Panel voltage stays close to open‑circuit, only slightly dropping under load
Even with a second panel, current remains very low (<50 mA peak)
This suggests the system may be operating in a very limited input‑power condition
However, VIN_REG / VREFIN remains above threshold, so input regulation does not appear to be the direct cause of early termination
Bench supply test
Powered VIN from a bench supply instead of the solar panel
Same behavior observed:
With a partially discharged battery (~3.7 V), input current is still limited to ~25 mA
VFB remains around 3.3 V
So the issue does not seem tied to the solar source
Strange observation on VFB
When probing VFB with a multimeter, input current increases (up to ~60 mA)
This seems too large to be explained by measurement loading alone
It makes me suspect:
Possible grounding issue
Noise or instability on the VFB node
The IC not correctly sensing the feedback voltage
(I’m also attaching a PCB photo for layout review.)
What I have NOT changed
SHDN is not driven by firmware and stays high
Safety timer is configured as in the schematic
Questions
Based on the schematic and measurements, what should I investigate next?
In particular:
Could PCB layout (switch node, ground routing, VFB trace, VIN_REG filtering) cause incorrect VFB sensing or early CV termination?
Are there known LT3652 pitfalls (sense resistor routing, diode orientation, inductor placement, grounding strategy, compensation network issues) that could explain termination around ~4.0 V?
Could the very low charge current (~25 mA even with bench supply) indicate a hidden limiting condition (e.g. ILIM, thermal, or internal fault behavior)?
Thanks in advance — any targeted debugging suggestions or measurement ideas would be greatly appreciated.
Are you aware that charging to a slightly lower voltage than 4.2 actually extends the lifetime of LiPo cells? The effect is dramatic.
I didn’t realize it had such a big impact. The problem is that I still can’t charge the battery! The charging current always stays around 25 mA, both with the solar panel (in full sun) and with the bench power supply. As a result, the energy storage is basically unusable, or at least very limited.
YES! I also applied some vias for enance thermal exchage with grounded bottom plane.
Here attached the second version of my design, but also this one has the same issues as the one reported in the previous schematic.
For the Resistor I used 330k // 1.2M 1% 0603 . But the main issue is the max absorbed current about 25mA, it grows up to 65mA while I put on Vfb terminal the probe.
Yes, even with a bench power supply (set voltage and current limit), I'm only measuring around 0.25 mA of absorbed current.
Is there any way to check whether the LT3652 has a proper ground connection using the exposed pins? (I soldered it using a homemade reflow oven.)
"Does my design seem to have any RF coupling or other problems
When I asked if it was connected to ground your answer was "YES!". So how did you check it then?
Using copper pours for ground/power connections can be quite problematic. Did you examine each and every ground path?
Sorry, I thought you were referring to the PCB connection itself (the F.Cu layer). In any case, given the LT3652 package, I’m not sure how to verify whether the soldering underneath was actually successful, even though I followed a standard reflow profile for the solder alloy I used.
Is that also with flat batteries.Most chargers reduce charge current between 4.0 and 4.2volt from full to zero.
My question was if your voltmeter was not reading too low.
I tried discharging the batteries — I did the same test with just one cell as well — down to about 3.6 V, and the behavior is exactly the same: it only charges at 25 mA, both in full sun and with a bench power supply. I checked it with a DMM, a small multimeter (DSO TC3), and the readings from the bench supply itself, and those three measurements are consistent with each other.
Thermal vias have 0.3 mm hole diameter. I was thinking the opposite: with heating from both top and bottom during reflow, the thermal wave should reach the vias before, or at least at the same time as, the surface reaches the peak temperature.
and
I can’t say for sure (except I tested each pad to be sure there arent sort circuit) and leave a GND plane under all the loop SW-BOOST and SENSE as shown in the img.
The components are quite similar to those on the LT evaluation board, except for the following:
Component
LT eval board
My design
Note
Bootstrap capacitor
1 uF, 50 V ceramic
1 uF, 50 V ceramic
Same value and type
Diode on SW path
Schottky diode
SS14
Same class, but lower current margin
Bootstrap diode
Schottky diode
CUS10S30
Vin<20V (Voc panel 18.5V -measured)
4.1. The resistance above Vfb, 330k // 1.2M, should match the equivalent resistance requested in the datasheet.
