Since LT3572 outputted only digital signal and I needed analog, I found some new chips. I was targeting class D piezo speaker amplifiers. Piezo speakers are used in places where there are not enough space for normal speakers. They are harder to drive and control (high voltage, capacitive), but have much more output power in smaller volume. Class D amplifiers use high frequency digital switching to make lower frequency analog signals.
LM48580 development board
Next in line were LM48580 and LM4960. Both had boost converters and worked from single cell lipo.
The LM48580 is in BGA package. Fortunately only 12 balls and all the middle ones can be shorted with neighbouring balls. I tried to solder it with reflow oven but school's stupid oven has ventilators in it that blew the chip away from the board. After trying to solder it for couple of times I finally got it on good enough. Large amount of solder paste and hot air was needed. And since it is BGA, every desoldering destroys your chips - so I destroyed about 3-4 chips in the progress.
While debugging I noticed that the boost converter isn't working. I got a little freaked out, but while explaining the problem to a friend I remembered that this is class H amplifier. Meaning that the boost converter output is as low as it can be to be more efficient. Since I was not driving it with signal the boost converter was in standby. After applying some signal it all started to work magically.
LM4960 testboard
As I learned - both of these chips are differential output chips - meaning that they will reverse the outputs to get two times as much peak to peak output to speaker. It can be a good thing for speakers but since my motor cannot be reverse biased I didn't end up using these chips.
Schematic of LM49590 testboard (PDF)