Pi Power - the power supply the Raspberry Pi *should* have come with!

[nsayer]

I've decided to try IndieGoGo to get enough seed money to build a batch of 1000 Pi Powers.

http://igg.me/at/PiPower

Pi Power is a 2A @ 5V power supply for the Raspberry Pi. It accepts 6-15 VDC in on a 2.1mm barrel connector, and sits on the GPIO header (yes, it works on the new B+ model too). It comes with a stacking header, so you can plug (and power) something else as well. The stacking header raises Pi Power high enough off the Pi so that you can still use the display connector underneath it. Pi Power doesn't change the footprint of the Pi, and the Pi will still fit in the bottom half of most cases without trouble. Cases like the Pimoroni Coupé are even better, since Pi Power can sit outside on top.

I've sent a unit to Dave for Mailbag Monday. It should get there in time for the next one.

With funding to build a batch of 1000 Pi Powers, it will be self-perpetuating, since the proceeds from the first batch will buy the second batch, and so on.

And if it fails, I will still sell them for $15 on Tindie (I am now, in fact).

[true]

So the $35 computer should have had a $15 PSU included?

[bwat]

So the $35 computer should have had a $15 PSU included?

Maybe nsayer is of the thinking that cheap and underdimensioned is still underdimensioned.

Meaningless exercise: according to [1] the power supply and fans should be around 1% to 3% of the cost of a circa 1990 workstation assuming 1000,000 units. A quick costing (actually price not cost!) of a workstation built around a raspberry pi with that PSU (monitor [$125], mouse [25], keyboard [25], external harddisk [75]) gives a cost of about $300. Since $15 is 3% of 500, we're not that far off target. Mind you modern monitors are about 1/10 of the price ($1000)! The PSU and fans are quoted in the book at $55. The CPU board is quoted at $200 to $800 so a bit of a difference there.


[1] J. L. Hennesey, D. A. Patterson, Computer Architecture: A Quantitative Approach. Morgan Kaufmann, 1990.

[rolycat]

So the $35 computer should have had a $15 PSU included?

Maybe nsayer is of the thinking that cheap and underdimensioned is still underdimensioned.

Maybe nsayer is of the thinking that he will sell more of his product if he can persuade people that it solves an actual problem.

[janekm]

It's a shame that the B+ now actually does come with a switching power supply, pretty much obviating the need for this... (Using micro-USB is sensible, as the plug-packs are widely & cheaply available, and the B+ now uses a more robustly mounted socket).

[nsayer]

Crappy USB power supplies are, indeed, ubiquitous. That's partly why powering such a high current, voltage-sensitive device like the R-Pi that way is such a lousy idea. Higher voltage, moderate current "wall warts" are even *more* ubiquitous, and regulating their output with a buck converter means that having a less-than-perfect one is much less of a problem.

Also, microUSB connectors can only be inserted in one orientation and the cables are much thicker than necessary for just supplying DC power. That increases the stress on the SMD connector. It's just not the best solution.

The B+ has switching converters for the 1.8 and 3.3 volt power rails, but it still requires 5V power - meaning the problem Pi Power aims to solve is still there.

Pi Power works extremely well with the B+. In fact, one problem with the B+ is that 26 pin IDC (ribbon) cables can't be plugged into the GPIO interface because pins 27 and 28 get in the way. Pi Power - in addition to powering everything - can be used as an adapter to make those cables fit. If you want to use all 40 pins of the GPIO interface, you can use a 14 pin stacking header next to Pi Power.

I don't think Pi Power is overpriced. A proper, high quality 2A USB power supply and a decent cable can cost as much as Pi Power and a surplus 800 mA @ 9V supply.

[nsayer]

So the $35 computer should have had a $15 PSU included?

If they'd have designed it in, it wouldn't have added $15 to the cost.

$15 is what it costs when 1000 units at a time is the wholesale lot. At higher volumes, the price drops further.

[Legit-Design]

So the $35 computer

Where to get one for $35? 

[rolycat]

Since it has twice as many USB ports one might think that the Pi Power would be especially useful on the B+.

My (possibly flawed) understanding is that the Pi doesn't use 5V directly but does forward it to the USB and HDMI ports and a couple of GPIO pins.
So the B+ should be relatively insensitive to the supply voltage if used for example as a headless web server, but will require a solid power source once those peripherals are in use. Is this correct?

[nsayer]

Even if it only really uses 3.3 and 1.8 volts (and I don't think that's entirely true), there's a 5 volt TVS diode across the +5 rail. If you supply too much voltage, that diode will conduct and possibly blow either itself or the input polyfuse. If the input voltage is low...? I'm not 100% sure that simply not powering the 5 volt peripherals you're not using would result in a functional Pi.

Besides, one of them is the USB/Ethernet controller. It's hard to imagine how you'd be communicating with your Pi without at least one of those two subsystems operational, and I think both require +5.

But, yeah, Pi Power is great for B+s. You can't quite have 500 mA * 4, but you certainly could have a couple of high-powered peripherals along with a keyboard and mouse without worrying.

[nsayer]

I'll just throw out there... Pi Power is also great for amateur radio folks. They quite frequently have high amperage 12 volt power busses. With Pi Power and (say) and a PowerPole to 2.1mm barrel plug cable, you're good to go.

[rolycat]

Besides, one of them is the USB/Ethernet controller. It's hard to imagine how you'd be communicating with your Pi without at least one of those two subsystems operational, and I think both require +5.

From the schematic for the model B and the LAN9512 datasheet it doesn't look as though that's true -  the 5V input is fed directly to the USB ports on the board, bypassing the chip.

I can't help feeling you might have more success promoting your solution on a less technically savvy forum - I imagine most Pi-using EEVBlog regulars are well aware of the dangers of cheap Chinese knock-offs and have already armed themselves with a meaty USB supply. And of course, not every forum user is a Pi fan...

[nsayer]

Oh, sure, but this isn't the *only* place I'm mentioning it.

[Kean]

I've used R-Pi's as controllers on several automation projects and something like this could be handy.  I certainly find the micro USB connection to be a hassle.  I also think the price is reasonable for a ready to use module.

Unfortunately a couple of things put me off backing the campaign:
1) IGG - at least you're using fixed funding, but I'm really avoiding IGG due to their poor response to scams
2) Limited input voltage range - I would actually need 24V input for most projects, and ideally over 30V (DC UPS can reach 28V)

Nice to see them on Tindie at the same price, but shipping is a killer at $25 for international delivery.  I guess it wouldn't be too bad if ordering 5 or more, and 15V max would still suit a lot of applications.

I currently have my own board based on the ST Micro L5973D, which I put in a small DIN rail enclosure.  That ends up costing a lot more than $15 (mostly in assembly time), but it is also configurable as dual output - e.g. if I need another regulated voltage to power sensors.  I can also swap in various 3 terminal style DC/DC converters from Recom, XP Power, SpeedyLab, etc depending on the current/voltage needs.

[nsayer]

Unfortunately, as designed presently, 15 VDC is the maximum input voltage.

Dave actually asked on Twitter why I didn't make it a SEPIC instead of a simple buck converter, and the answer to both is simply that I didn't consider use cases outside of a 6-15 vdc input range. At this point, I have a successful design and have ordered a small run of boards, so changes will have to be a follow-on product. A wider input voltage range would likely be a more expensive product, and people are already whining about it being $15.

As for IGG, I tried Kickstarter before, and that didn't work so well, so I thought I'd try someplace else.

My current shipping method is USPS small flat-rate boxes, which are $6 domestic and $25 int'l. That way I don't really have to try and figure out the weight and the box itself is free.

You are right, though.. The last few orders I sent clearly could have been sent cheaper in just a padded envelope. It's got me thinking.

[bktemp]

The 5V rail on the raspberry pi is directly connected to USB and to the internal stepdown of the main processor for the core voltage. Especially when using both usb ports (for wifi + bus powerd hub for mouse & keyboard) a stable 5V is mandatory. Most cheap micro usb cables and power supplies have a significant voltage drop, therefore a 5V regulator right at the pi is a very useful idea.
I have build a similar module using a MCP16322 5V 2A synchronous stepdown placed onto to gpio connector. This gives a 6-24V input range and allows powering the pi using two lipo cells.
I have also added a 3.3V stepdown and removed the linear regulator to increase the efficiency. This was a few month before the pi B+ came out... 

I have not calculated the bom cost, but the MCP16322 circuit could be a bit cheaper than your design.

[nsayer]

That's a nice chip. I guess the one thing I see that's slightly nicer about my design is that it has an over-current shutdown, and that chip (by itself) lacks one.

[nsayer]

Nice to see them on Tindie at the same price, but shipping is a killer at $25 for international delivery. 

I poked around a bit at the post office and an 8 ounce parcel to Australia would be just shy of $13 (plus the cost of a suitable box or padded envelope). Still kind of substantial for just *one*, seems to me, but definitely less.