Raspberry Pi 3 GPIO Current

[netmech]

well, this is a apparently a sore subject on the internet, and i'll add new post to the history of it for a clarification on a piece of it

From the various locations on the Raspberry Pi site and other places, it seems clear that the 26pin headers have the 16mA per pin, an 50mA max for all 3.3V pins
However, it's not clear to me if that really does apply to the 40pin header, particularly on the RPi 3.

And, adding to that... I just picked up the 2nd edition of the Raspberry Pi Cookbook by Simon Monk, and it states on page 199
"Do not draw more than 16mA per output and keep the total for all outputs below 50mA in total for an older 26pin Raspberry Pi, and below 100mA on a 40pin Raspberry Pi"

That would make sense since the 3.xmA per pin * 30 3.3V pins is right about 100mA.
But i've not been able to find any document that says what the 3.3V rail on the RPi3 can actually supply.

Anybody have that?
Does Mr. Monk frequent these forums per chance?
Anybody else know of a document giving some more details on the differences between the supplies to the 26 and 40pin headers?


So to summarize the power draws that may be in place on the RPi3, using his suggested max values of 100mA for 3.3V

2.4A Max
- typical board power draw: 400mA to 1.3A (depending on CPU usage and such)
-- that was using wifi.. granting it's quite a range. how much does ethernet add?
- 50mA for HDMI
- 100mA for 3.3V pins
- 250mA for camera
--------------------------------
1850mA

leaving (out of 2.4A) ~650mA for the 5V rail or USB devices (out of a maximum 1.2A)
- or allow more for these if the above items like the camera aren't in use

right?
(using values from https://www.raspberrypi.org/help/faqs/#powerReqs)


(I cross posted on adafruit.. https://forums.adafruit.com/viewtopic.php?f=50&t=98371)

[rpdom]

The limit for the 3.3V supply outputs will be whatever the on-board 3.3V regulator can handle, minus what any 3.3V devices on the board take (assuming there is sufficient free power from the 5V supply). I think the regulator can handle up to 1000mA, but I don't know what the SoC and other peripherals need.

I think (with a decent PSU) you should be able to take at least 250mA from the supply pins without problems.

[joan]

There is an interesting article at https://raspberrypise.tumblr.com/post/1 ... power-rail which attempts to answer that very question.

I've only skimmed it, but the author seems to understand the subject.

[MarkHaysHarris777]

From the various locations on the Raspberry Pi site and other places, it seems clear that the 26pin headers have the 16mA per pin, an 50mA max for all 3.3V pins
However, it's not clear to me if that really does apply to the 40pin header, particularly on the RPi 3.

The article Joan referenced is fabulous , as it pertains to the 3v3 rail... but does not address the GPIO pins specifically. I understood (I thought) that the 54ma total limit for all GPIO pins and the 16ma limit for each pin was a limitation on the GPU not a limitation on the 3v3 rail per se. If you exceed 54ma (total current source|sink on all GPIO pins) or if you exceed 16ma per pin, damage will occur on the GPU. Would someone please affirm or deny this ? Thanks.

[joan]

From the Pads document (I don't have a link handy).

What is a safe current?
All the electronics of the pads are designed for 16mA. That is a safe value under which you
will not damage the device. Even if you set the drive strength to 2mA and then load it so
16mA comes out that will not damage the device. Other than that there is no guaranteed
maximum safe current.

My understanding is that 16 mA per GPIO is feasible with the new 3V3 power system as the rail should be able to supply 26*16 or 416 mA.

[netmech]

That article on the RPi3's 3v3 power supply is great!
Nice to see a solid test on it.
I think that does, for the most part, answer my question. From that, Simon Monk's suggestion on the 100mA limit would appear to be a safe best practice that still allows for pretty heavy CPU loads and such.

Everything really does come down to knowing how a specific set of peripherals and CPU load will stress the system.
I needed to increase my understanding so that I can teach my FIRST robotics team with it.
I think i'll make some pretty pictures summarizing this, and post them back for comment.

From the Pads document (I don't have a link handy).

What is a safe current?
All the electronics of the pads are designed for 16mA. That is a safe value under which you
will not damage the device. Even if you set the drive strength to 2mA and then load it so
16mA comes out that will not damage the device. Other than that there is no guaranteed
maximum safe current.

The PADs document
https://www.scribd.com/doc/101830961/GPIO-Pads-Control2

In the paragraph before the one above, Gert clearly does say in there that the limitation was the 3v3 supply, and not the GPU


Good comments, thanks all

[boyoh]

I'm a great believer that the Raspberry Pi was designed to learn Programming and Electronics
The learner programmers are leaping forward, but lagging in electronics,witch they find hard
To grasp. Knowing the capability of the Pi is a good thing, but not to be used to It's limits
Power wise, I would stick to using the IN/Puts and OUT/PUTs to switch buffer stages for
Higher voltages and currents, This way you would have a few ma switching a few amps
To do this you would have to delve into electronics, witch would be a good thing for the
learning curve

[netmech]

That's absolutely what I intend to do..

The problem is not knowing the limits just makes everything uncertain, and if I can't give a fairly clear answer as to what are solid boundaries to stay within, then it'll discourage people from creativity and experimenting.

For example.. If, as in some examples online, you're driving a few LEDs at 20ma, you're already over the 50mA limit. I don't want a student (or myself, ha!) to needlessly fear because they combined a couple simple tutorials to drive a couple LEDs and perform some logic as well. If the limit really was 50mA, I can't use all the GPIO pins even at the 3mA lower bound. That'd be a problem.

So, i need to be able to understand and then explain why the limits are what they are, and what a reasonable best practice is to stay out of trouble, and what the thought process is to evaluate a given setup. That's a key learning experience too.
Then we can show how to drive the LEDs at low current, and/or use buffer circuits, switching, I2C, etc.. to only utilize the Pi for logic/switching, and use externally powered circuits for higher voltage/current stuff. This builds context to allow a better understanding.

Also, i don't want to treat a new Pi3 like an original Pi, and miss out on functionality that may be well within it's capability just because I really can't tell the difference between the capabilities. That would be silly too..

So, yes.. i agree.
I just want to learn!

[MarkHaysHarris777]

In the paragraph before the one above, Gert clearly does say in there that the limitation was the 3v3 supply, and not the GPU

No, he in fact does NOT. He says that there is NO safe maximum rating on the GPIO pin above 16ma... which is a function of damaging the pad/ and it has nothing to do with the rail.

The problem for everyone's understanding is realizing (and this is especially for noobs) that there are two (2) issues... the total current that can be delivered (Joan's comment from the pads doc seems to be correct) which pertains to the 3v3 rail, and the safe current for each @ pin on the pad... 16ma. ( I cannot believe he used the word 'blow up' to describe the damage, but there it is ) I'm assuming that the pad will develop a short that causes high heat build-up and a complete loss of the PI. That does appear to be what happens when the pad gets damaged anywhere (even on one pin).

If one of your students throws a 5mm 20ma LED on an output and drives it to 22ma there is eventually going to be a problem, and they need to understand this ... I'm only giving my students 3mm 2ma low power LEDs in the first place. It is nice to know that the 3v3 rail is the root of the total current limitation and that the PI 3B has a higher limit ! ~very nice.

Thanks for finding the pads document and posting it.

[davidcoton]

That's absolutely what I intend to do..

The problem is not knowing the limits just makes everything uncertain, and if I can't give a fairly clear answer as to what are solid boundaries to stay within, then it'll discourage people from creativity and experimenting.

For example.. If, as in some examples online, you're driving a few LEDs at 20ma, you're already over the 50mA limit. I don't want a student (or myself, ha!) to needlessly fear because they combined a couple simple tutorials to drive a couple LEDs and perform some logic as well. If the limit really was 50mA, I can't use all the GPIO pins even at the 3mA lower bound. That'd be a problem.

So, i need to be able to understand and then explain why the limits are what they are, and what a reasonable best practice is to stay out of trouble, and what the thought process is to evaluate a given setup. That's a key learning experience too.
Then we can show how to drive the LEDs at low current, and/or use buffer circuits, switching, I2C, etc.. to only utilize the Pi for logic/switching, and use externally powered circuits for higher voltage/current stuff. This builds context to allow a better understanding.

Also, i don't want to treat a new Pi3 like an original Pi, and miss out on functionality that may be well within it's capability just because I really can't tell the difference between the capabilities. That would be silly too..

So, yes.. i agree.
I just want to learn!

It seems quite clear. the limit is 16mA per pin, and (less clear) 100mA total. So don't design a circuit using any pin at 20mA, or more than six pins at 15mA.
Teach proper electronic design -- so that your students use 2mA per pin (arbitrary figure) when that is enough, less when it can be, and more only if necessary. And so they keep an eye on the total power budget of the GPIO connections.

I understand that Pis are a hard-won resource, but even with guidelines in place expect some casualties -- either through ignorance/error or students deliberately "pushing the boundaries". I've seen both -- bored students experimenting outside the specified supply voltage on a poor innocent op-amp (there was a lot of magic smoke that got out), and a complete multi-thousand pound microprocessor development system that died when a circuit under development had a reverse polarity supply connected (fortunately for my career the damage was much less serious than the symptoms initially suggested).

[netmech]

@ MarkHaysHarris777
Sorry, i obviously mixed topics incorrectly with regards to the 20mA example.
My personal confusions were all around the total max current, not the per pin max. Gert noted that you couldn't drive all the GPIO pins at their maximum 16mA because the supply couldn't safely deliver it.



For reference, i had also run across this article the other day showing more testing of actual current draw of the system under load. He got it up to 720mA, which is inline with what one of the moderators on Adafruit noted that each core can use 175mA when going full bore. That seems to nicely support using 100mA as a best practice safe range for total 3.3V current draw on the pins.
http://raspi.tv/2016/how-much-power-doe ... ed-to-pi2b

[MarkHaysHarris777]

For reference, i had also run across this article the other day ...
http://raspi.tv/2016/how-much-power-doe ... ed-to-pi2b

@netmech, 'ey thanks! ... good read. The real-world comparison across models is helpful.

thanks again

[boyoh]

The Pi with good specifications ,and power handling. Wont perform any better then
The power supply feeding it. THis is left to the user to pick the recommended
Power supply.. This can be very confusing for the novice, not under standing
The specifications of what he is reading. This makes it more confusing
When the Pi ratings are not very clear. After all PLCs & micro controllers
Con troll external Buffer stages,.

[SonOfAMotherlessGoat]

The Pi with good specifications ,and power handling. Wont perform any better then The power supply feeding it.

And often the power supply feeding it doesn't do what is stamped on the tin. Wasn't it F T that had a list of labeled vs actual for supplies?

[Cancelor]

From the various locations on the Raspberry Pi site and other places, it seems clear that the 26pin headers have the 16mA per pin, an 50mA max for all 3.3V pins
However, it's not clear to me if that really does apply to the 40pin header, particularly on the RPi 3.

The article Joan referenced is fabulous , as it pertains to the 3v3 rail... but does not address the GPIO pins specifically. I understood (I thought) that the 54ma total limit for all GPIO pins and the 16ma limit for each pin was a limitation on the GPU not a limitation on the 3v3 rail per se. If you exceed 54ma (total current source|sink on all GPIO pins) or if you exceed 16ma per pin, damage will occur on the GPU. Would someone please affirm or deny this ? Thanks.

I'm no expert but this sounds right to me.

A header with 26 pins or a header with 40 pins is not what is important**. Both headers are a collection of connectors, some are connected to the 5v rail, some to the 3.3v rail, some to the ground plane and some are GPIO pins. I think the chip driving the pins is what makes the difference.

** I think it is confusing to call these GPIO headers or even to think of them as 26 or 40 GPIO pins.

[boyoh]

The more I read about the Pi power out/put capability I have come to the conclusion
That the Pi designer , also designed the Uk domestic Ring Main that, you could put
As many 13a switch sockets on the ring,as you want, controlled from a 35a mcb.
To do this he applied a Diversity Factor, hoping that not all the sockets would be
used at the same time

[Cancelor]

The more I read about the Pi power out/put capability I have come to the conclusion
That the Pi designer , also designed the Uk domestic Ring Main that, you could put
As many 13a switch sockets on the ring,as you want, controlled from a 35a mcb.
To do this he applied a Diversity Factor, hoping that not all the sockets would be
used at the same time

Made me LOL

[netmech]

The more I read about the Pi power out/put capability I have come to the conclusion
That the Pi designer , also designed the Uk domestic Ring Main that, you could put
As many 13a switch sockets on the ring,as you want, controlled from a 35a mcb.
To do this he applied a Diversity Factor, hoping that not all the sockets would be
used at the same time

That's awesome

A header with 26 pins or a header with 40 pins is not what is important**. Both headers are a collection of connectors, some are connected to the 5v rail, some to the 3.3v rail, some to the ground plane and some are GPIO pins. I think the chip driving the pins is what makes the difference.

Yes, however, I'm sure the 26 vs. 40 pin notation was used in the book to keep it simple for people not familiar with the different models. That quick rule of thumb is a lot easier than referencing the actual supply chip and fuse which change as the models progress, and is rated higher on the models that have a 40pin header. (Of course, in a few years you may very well have to be more specific on the models with a 40pin header...)

[Massi]

From the Pads document (I don't have a link handy).

What is a safe current?
All the electronics of the pads are designed for 16mA. That is a safe value under which you
will not damage the device. Even if you set the drive strength to 2mA and then load it so
16mA comes out that will not damage the device. Other than that there is no guaranteed
maximum safe current.

My understanding is that 16 mA per GPIO is feasible with the new 3V3 power system as the rail should be able to supply 26*16 or 416 mA.

it would be nice to have a "official" position from RPF guys about that.

[joan]

...
it would be nice to have a "official" position from RPF guys about that.

They probably are not in the position to know.

Remember Broadcom have not published electrical specifications for the GPIO.

[davidcoton]

The more I read about the Pi power out/put capability I have come to the conclusion
That the Pi designer , also designed the Uk domestic Ring Main that, you could put
As many 13a switch sockets on the ring,as you want, controlled from a 35a mcb.
To do this he applied a Diversity Factor, hoping that not all the sockets would be
used at the same time

Minor correction: the "standard" UK ring main is protected by a 32A MCB (or 30A fuse).
[off-topic]
The point is that, in most circumstances, the cable used is rated at 20A (there are exceptions, like when it runs through loft insulation). Since it is a ring (two cables to each socket) 32A gives a fair margin to protect the cable -- even when overloaded, the trip/fuse will open before the cable melts. The trip or fuse is intended to give fault (short circuit) protection. It is not intended to protect against minor overloads, as you say that depends on users just not plugging too many heaters in. There are surprisingly few problems with overloaded ring mains (overloaded sockets with extension leads are far more common).[\off-topic]

[boyoh]

The more I read about the Pi power out/put capability I have come to the conclusion
That the Pi designer , also designed the Uk domestic Ring Main that, you could put
As many 13a switch sockets on the ring,as you want, controlled from a 35a mcb.
To do this he applied a Diversity Factor, hoping that not all the sockets would be
used at the same time

Minor correction: the "standard" UK ring main is protected by a 32A MCB (or 30A fuse).
[off-topic]
The point is that, in most circumstances, the cable used is rated at 20A (there are exceptions, like when it runs through loft insulation). Since it is a ring (two cables to each socket) 32A gives a fair margin to protect the cable -- even when overloaded, the trip/fuse will open before the cable melts. The trip or fuse is intended to give fault (short circuit) protection. It is not intended to protect against minor overloads, as you say that depends on users just not plugging too many heaters in. There are surprisingly few problems with overloaded ring mains (overloaded sockets with extension leads are far more common).[\off-topic]

Thanks for the correction and picking and poking.
It was only meant to be a designee Comparison for
The post.Not meant to be a designee debate

A Ring Main that is fused with a REWIRABLE fuse
Can be open to dangerous abuse by some one that
Knows no better, He is not breaking any regulation
He is doing what he is al owed to do change a fuse.
Not all circuits are protected by MCBs
There is nothing stopping a idiot rewiring the fuse
With three strands of 30amp fuse wire on the
Ring Main

You are correct 32a mcb, my mistake

[Milliways]

well, this is a apparently a sore subject on the internet, and i'll add new post to the history of it for a clarification on a piece of it

From the various locations on the Raspberry Pi site and other places, it seems clear that the 26pin headers have the 16mA per pin, an 50mA max for all 3.3V pins
However, it's not clear to me if that really does apply to the 40pin header, particularly on the RPi 3.

And, adding to that... I just picked up the 2nd edition of the Raspberry Pi Cookbook by Simon Monk, and it states on page 199
"Do not draw more than 16mA per output and keep the total for all outputs below 50mA in total for an older 26pin Raspberry Pi, and below 100mA on a 40pin Raspberry Pi"

That would make sense since the 3.xmA per pin * 30 3.3V pins is right about 100mA.
But i've not been able to find any document that says what the 3.3V rail on the RPi3 can actually supply.

Anybody have that?
Does Mr. Monk frequent these forums per chance?
Anybody else know of a document giving some more details on the differences between the supplies to the 26 and 40pin headers?


So to summarize the power draws that may be in place on the RPi3, using his suggested max values of 100mA for 3.3V

2.4A Max
- typical board power draw: 400mA to 1.3A (depending on CPU usage and such)
-- that was using wifi.. granting it's quite a range. how much does ethernet add?
- 50mA for HDMI
- 100mA for 3.3V pins
- 250mA for camera
--------------------------------
1850mA

leaving (out of 2.4A) ~650mA for the 5V rail or USB devices (out of a maximum 1.2A)
- or allow more for these if the above items like the camera aren't in use

right?
(using values from https://www.raspberrypi.org/help/faqs/#powerReqs)


(I cross posted on adafruit.. https://forums.adafruit.com/viewtopic.php?f=50&t=98371)

You ask a number of questions which are really independent.

The GPIO CAN supply 16mA per output (but needs configuration {in 2mA steps} because the default is 8mA) but is subject to a per block limit and total SOC limit. These are documented in the published data and as the SOC is identical across all Pi up to B+ and GPIO part reported to be the same in Pi2 & Pi3 even though the processor part changed.

The 3.3V regulator on early Pi was a linear regulator and severely limited, but the switch mode regulator on later Pi is rated at 1A, and testing indicates it can provide close to this to peripherals (PI power source permitting).

[boyoh]

It is time the Raspberry Pi designers got to gether and ,designed a Pi with a in built power supply
Even the experienced have problems with the out/put power of even the latest modals
So the novice as no chance. The familiar saying " Have I Fried My Pi"

[joan]

...
The GPIO CAN supply 16mA per output (but needs configuration {in 2mA steps} because the default is 8mA) but is subject to a per block limit and total SOC limit. These are documented in the published data and as the SOC is identical across all Pi up to B+ and GPIO part reported to be the same in Pi2 & Pi3 even though the processor part changed.
...

Could you give your source for the 8mA default, and the block limit (what is a block?), and the total SOC limit?