Any tips or warnings for low-voltage high-power boost conversion?

[karpouzi9]

Thanks.

[David Hess]

At such high currents, parasitic inductance and resistance, including ESR of capacitors, becomes increasingly important, which places severe demands on circuit layout.

Using multiple modules in parallel is definitely the way to go, and proper phasing between modules will lower ripple current requirements.

[f4eru]

take voltage on the scale of 1s lithium-ion (3-4V) and boost it up to something useful for existing hardware, e.g. 12-15V. If you have, please share it!

The simple answer is : NO. Don't.
Take the load on a series of cells instead of paralell, if your load is more comfortable at higher voltage.
Voltage balancing and co. is much simpler than an impractical DC/DC converter.

[Peabody]

Isn't the efficiency of a stack of cells effectively 100%?  As opposed to whatever you lose in the boost process.

[Siwastaja]

If you need regulated output voltage (like 12V +/- 2%) over the whole state-of-charge, then you are going to need a switching converter anyway, so boosting from 1s (3.0 - 4.2V) to 12V is not a colossally bad idea, unless the current is very large. Bucking from 5s to obtain 12V would allow smaller battery current.

Note that both buck and boost topologies require an extra switch to disconnect the battery (boost lets current-unlimited Vin directly to the output through the diode; buck lets voltage at the output flow through and charge the battery through the body diode of the MOSFET).

For unregulated rough voltage ranges, just use desired number of cells in series.

[Siwastaja]

I think you are overestimating the risk of cells failing in series configuration. The only viable solution is to use high quality cells that fail rarely enough. If you use cells that are prone to fail, I suspect they might have safety issues in parallel configuration, even if you fuse them separately. Series configuration at least makes the product dead (by BMS) when a cell fails, which is a good thing safety wise. Availability wise, obviously not - but that can be only solved by cell quality improvement.