PEC_POWER supply comments needed

[aolofsson]

In terms of the power connector, we have received very mixed feedback so far.

One of the major decision to make is what to support in terms of daughter card power delivery. We proposed putting a few different power rails on the PEC_POWER connector, but we are now reconsidering this decision due to the potential risk and support complexity.

Here's the feedback we need:

1.) How limiting/expensive would it be if the Parallella only supplies 5V to the expansion card and required that the card itself generate other rails? Unless a lot of you stand up and testify that "Yes, having a 3.3V or 1.8V rail for an expansion card is a MUST because...", we are going to keep the support of these rails as TBD for now.

2.) How many mA would be needed on these rails to make them at all useful? We can't afford to over design the Parallella power solution and want to make sure it makes sense to bring out a rail even if it only can supply 100mA.

Thanks!

Andreas

[tnt]

Personally I think the only useful rails in addition to the 5V are the one that are used for some IO level, and even if they have limited current capability.

This way, if a board needs some IO level shifting between 2.5v and 3.3v/5v, it doesn't need to have a 2.5V LDO on board just for that.

Ideally having a 'strong' 3.3v rail (i.e. like > 500 mA) would be good since I'd expect a lot of the DB to use that as main voltage, and even if they decide to use 2.5v, they only need to drop 0.8v and not 2.5v which reduces the power dissipation and inefficiencies of using a cheap/simple LDO vs Step-down SMPS.

Cheers,

Sylvain

[Sundance]

Hi,

I suggest a “KiSS” = “Keep it Simple, Stupid” for the Parallella board to make it as cost-effective and simple – and come out as soon as possible! The Parallella is NOT a DC/DC Converter, but a DSP.

I have suggested in the past to use the pillars for Power and leave the lovely Samtec connectors for what they are designed to do = High-Speed date and low-voltage signals. Use the propose DC pins to route every spare pin from FPGA, Epiphany, MAC or HDMI, whatever got a pin not connected – as they might be useful.

If the board INPUT is 5V from either a PSU Brick or from USB, then the Daughterboard could become the Power Source, if required. If the Parallella provide any power-rails, then 100% chance they are not sufficient for some application/ideas and then a waste. Thinks as the 5V as the backbone, like the 110V/230V AC.

My gut feeling is to remove the 5VDC barrel and only use the USB for Power or the pillars. Will save space, cost, build/test-time – and one less item to break - or for the supplier to forget to supply

The following Link gives information about the future of USB Power Delivery - – and with Profile 5, then we can power a LOT of Parallella’s!


Flemming

[tnt]

Please, do _not_ remove the 5V input barel jack ...

Some people might be looking at running a bunch of parallela together, but from the backer stats, most of us will only have 1 or 2 and I'm pretty sure a significant portion of those people will want to run is independently and not plugged to a PC.
So sure, you "could" use a USB port to plug it to a wallwart charger but that's imho wasting a perfectly good usb port for ... power. And if connected to a PC, it can only provide 500 mA (if you stick to the specs) and you might want to deliver more than that to daugther cards.

Allowing the daughter card to be the power source is a good idea OTOH (basically just have the 5V connection be a direct bridge to the barrel jack).

As for the voltages, as I said above, having the voltages used for IO exposed to DB would be practical so that DB don't need separate LDOs just to power level shifters. But obviously if they have high current requirements, they should pump from the 5V and have onboard supply.

[psupine]

Can I suggest that there is some kind of wake up interrupt on the power connector?

I imagine that the ARM chip can power down and wake the Epiphany chip, and the ARM similarly has some low power sleep mode. (Sorry, I'm really light on detail here because I haven't done any homework on the idea). I am proposing that something that runs at an even lower power could sit quietly monitoring a sensor until it decides that circumstances demand waking the ARM (to wake Epiphany) to have a big hard think about what's going on.

To my mind, this would seem to be a power state control rather than a general interrupt, and so it logically belongs on the power connector. Just for what its worth, I don't feel strongly about it.

[wuher]

Removing the 5V jack might not be a good idea, standard USB port can only deliver 500 mA, and that might not even be enough to run ZYNQ and Epiphany with full capabilities.

[tnt]

[Sundance]

[trioflex]

[Sundance]