[SOLVED] LT3652 solar charger - low charging current and hot IC

I am designing and building device that will work outdoor self-powered using solar panel and battery. Solar panel rated 10W gives Im=0.65A (short-circuit) and has power peak at Um=16.8V (21V open-circuit). The VRLA battery from Yuasa is rated 4Ah 12V (13.65V floating). I followed data sheet and designed circuitry with C/10 loading (no timer) so for this battery it is 0.45A limited by Rsense=0R22 and using extra 6v2 zener diode for BOOST pin proper voltage range. I am using tantalum capacitors as Cout/Cin (4u7/20V as I could not find 10u) as well as boost capacitor (2u2 between SW and BOOST). Power inductor initially used are rated 2.5A; I tried 15uH (approximated from other designs) and 68uH (calculated from datasheet).

Using lab power supply I observed that charging starts correctly when voltage reaches Um voltage. Problem starts when LT3652 is charging. I measured ~0.2A as charging current with voltage drop close to 45mV on Rsense (so it is not problem with current limiter that applies at 100mV). I was trying different power inductors from 4u7 to 220uH giving different max charge current, all far from expected maximum. I tried different capacitors, electrolytic produced even lower current. When charging IC gets really hot and it should not be that much for such low current (2A rated chip).

As I am not sure about tantalum caps rating, is it possible for such problems with their regular ESR (not low-ESR)?
Would PCB layout lead to overheating and current drop? (I expect that design flaws would lead to noise emission, grounding problems with non-zero referential voltage etc).

regards,
-andy.

Could be a layout issue, causing oscillation at high frequency. A photo of the build would be helpful.

I have found couple flaws already. First, too small PCB pad for heat dissipation, just twice bigger than MSOP case are - I have designed very large pad now. Next, for high current path I had long tracks and ground incorrectly isolated; new PCB will have Cin/Cout/Df/Rsense/L as close as possible and only one point of contact between high current ground and remaining parts. Also I see in datasheet that ceramic 10uF as Cout is a must and low-ESR as Cin, while I have normal tantalum. So far I worked with DC/DC converters switching with frequency less than 100kHz less restrictive in terms of layout and component parameters. I am waiting for ordered ceramic caps and low-ESR tantalum caps. I will share news after I re-assembly it on new PCB.

So I have redesigned PCB, got my components and LT3652 is just warm as I made large pads. To my surprise it still was generating 250mA of current util (shamed) I discovered that battery is close to charged, thus resistive enough to not accept more current. This is because I have left the battery charging for too long. The LT3652 is programmed to keep constant voltage so the current drops as the battery gets closer to full charger state. I have discharged battery and now it accepts as much current as limited by Rsense.