16:04 <zyp> does anybody have figures on maximum power consumption for ecp5? the datasheet only specifies idle current and the lattice calculator seems to be useless without a specific design to calculate for, along with not supporting highend parts in the free version

16:04 <daveshah> Maximum power is largely a useless concept for FPGAs

16:05 <zyp> sure, but I still need a power budgets for sizing regulators

16:05 <zyp> -s

16:05 <daveshah> A device with every LUT and register running at 400MHz and almost all the routing used is going to consiume way more than a useful design

16:05 <daveshah> The best way is to calculate what power gives 65degree rise over ambient, any more than that will be thermally constrained, if a general purpose board

16:06 <daveshah> If you aren't pushing it hard, 1A on Vcc is more than enough

16:06 <mwk> ... you could even try to purposefully trigger metastability everywhere for extra crowbar currents!

16:06 <mwk> okay I'll shut up now

16:12 <zyp> daveshah, what if I am pushing it hard? :)

16:39 <awygle> i just copy the Versa power supply topology on all my boards lol

16:39 <zyp> cabga381 got a thermal resistance of 19°C/W at zero airflow, meaning it could dissipate 3.4W if you allow a rise of 65 degrees

16:40 <zyp> haven't run the numbers on the other rails yet, but 1.2W at the core rail doesn't seem like it'd come close to be thermally constrained

16:41 <daveshah> 1.2W was more of a typical unambitious design type level

16:41 <daveshah> 3W sounds about as high as it could get on a non synthetic design anyway

16:43 <zyp> https://bin.jvnv.net/file/afyrp.png <- I'm working on a design that'll hook up a ddr3 sodimm, also planning to break out all four serdes channels

16:43 <zyp> does that count as ambitious? :)

16:44 <daveshah> Yeah, 3W is probably a better target for that

16:45 <daveshah> On the thermals side, DDR3 will probably dissipate a nonneglible amount of IO power too on the ECP5 side

16:48 <zyp> I've already made it a point to make sure the sodimm won't get in the way of a heatsink on the ecp5 if needed