Correct use of "sup" direction pins?

Ryan Pettigrew wrote on Fri, 21 November 2014 05:26

You are misunderstanding what I am saying here.

No, we understand what you are saying, just that it's wrong.

Quote:

However, it DOES treat "sup" pins different from "pwr" pins. The ERC

for the output of a regulator should be treated the same way as any

supply.

No.  This is what's wrong, as you have been repeatedly told.

You originally came here asking how to use SUP and PWR.  You got several

answers on how Eagle is intended to be used.  Now you are arguing about

those.  If you don't want others to tell you how it's done, don't ask them.

Take the advice or not.  We don't care, it's not our designs that will be

messed up.  And no, we don't want to repeatedly hear the same diatribe

about how you imagine PWR and SUP should be used.  Remember, you asked us,

we never asked you.  We don't care what you think.

That all out of the way, it seems the underlying problem is that you are

caught up in what the various pin directions are called instead of what

they do.  First, keep in mind all these pin directions only affect the ERC

check.  Second, don't get too hung op on the names.  Consider them

arbitrary labels and remember instead what these various settings do.

SUP is a special pin function that forces the name of the pin onto the net.

Forget the fact that "SUP" was originally intended to stand for "supply".

Think of it instead as the pin-name-onto-net pin function.  Now consider

where this function would be useful.  This is for symbols in the schematic

that stand for a specific net.  The most common would be a ground symbol.

For example, I have a down-pointing triangle with its single pin set to SUP

and named GND.  Any net I connect that to is automatically named GND.

That's handy for a ground symbol.  One drawback is that sometimes you have

different grounds.  For that reason I actually have a bunch of ground

symbols, each with diferent names, like GND-ANA, GND1, GND2, GND-PWR, etc.

Except for the generic GND symbol, these others have the name written below

the triangle so that it's obvious in the schematic which ground they

represent.  It took a little more work to make all these different ground

symbols up front, but now that they are there it's really easy to make a

particular ground net in the schematic.

Note that this behaviour doesn't apply to physical parts, whether they

drive power rails or not.  You don't want the net on the output of a 7805

to always be called "5V", or whatever you picked for the pin name in the

symbol.  Worse yet, you don't want two 7805 regulators in the same

schematic to automatically have their outputs connected.  You do want the

ERC to know that the pin is driving the net, so that it can catch you

accidentally connecting another pin that is also trying to drive the same

net.  To get that feature, you set the pin to OUT.

Again, don't get hung up on what the various pin directions are called.  Go

by what they do instead.

--

Web access to CadSoft support forums at www.eaglecentral.ca.  Where the CadSoft EAGLE community meets.

Hi, Ryan

I disagree with:

The only way to have the ERC evaluated correctly for a regulator is with a "sup" direction pin

For an example, on my current project, almost all supply are build with buck controller, so the regulator/controller is not the supply: none of its pin is connected to the Vout supply.... the output rail supply is the inductor output ?

And I have few unregulated rail supply : in this case the output is the diode bridge...

I also forget, one of my bridge is a MOS Fet bridge....

By the way I also use a buck regulator as inverter (here the GND pin of the regulator is the negative rail supply)....

Also in that same project one of 3 identical regulators don't feed any component on the board (the output is simply connected to a pin of a connector. I don't want that this output is considered as 'supply'. It's an output.

So it will be stupid to set as SUP, the inductor pin, or the Diode bridge pin, or the FET source etc, and the regulators out pin: sometimes often the out is not a rail supply.

How Could Eagle guess how we use the regulator (and other power sources) ?  The only thing we know is the current flow out from a pin, or it it's a input. And it's not always the case (battery,  etc...) That we call 'supply rail' is up to us. And It' will be very complicated to have a complete ERC that could deal with all of these situation. So I think the way eagle work is not a big problem.....and at least it's simple, so we know how ERC will react:

And with that 'we do whatever we need to do to be sure that our design will work, and we don't suppose that ERC will solve every things'.

Guillaume.

Guest wrote:

And without a regulator, or some other power conditioning

circuitry, there is genuinely no case for supply nets at all, except in

pointlessly trivial cases, for which they genuinely aren't needed.

What if the supply is off board? EAGLE shouldn't assume the there's a

power supply in the design. Even worse would be to force such a requirement.

Jorge, if the supply is off the board, then you don't need the regulator. If you DO need the regulator, then the off-the-board supply is not the supply rail you need regular connection to. You connect the regulator to it, and then you connect most of everything else to the regulator. Now, you may have something like an op-amp, which needs access to external power, but the supply lines that feed the input are more likely to be based on regulated voltages and voltage references; each of which need their own supply rails, even with the off-board supply. What is the supply that produces the regulated voltage? The regulator. What is the supply that generates the voltage reference? Usually a voltage reference IC, but maybe even a subcircuit. Eagle needs to treat these as supply pins in the ERC check, because that is the function they perform. Doing otherwise in library construction is ignoring the job they do, and the checks that need to be made.

Guest wrote:

 

Even in this case supply nets would be useful. You would connect them to

the connector that's bringing power into the board.

Jorge, I'm not advocating against supply nets; far from it. This is one of those "trivial cases" I mentioned before, that is currently well supported. But it is grossly insufficient for modern board design. Your typical mobile device starts with a lithium ion battery charging circuit; the power feeding the battery has to be heavily regulated, in order to not make the battery catch on fire. That regulation step is an additional supply rail; this is not currently well supported, since it's the battery charging IC that needs to be the supply source, from which downstream checks must be made. The charged battery is then itself a supply rail. This is currently well supported, and I have no argument with that. However, the charged battery then feeds either a buck, or a boost, or a buck and a boost converter, or perhaps a linear or LDO regulator instead of the buck regulator. These are two more supply rails, currently not well supported, because the output is from an IC again. In short, 3 out of the 4 supply rails in this design are not well supported by Eagle's suggested use cases of output pin directions.

The community needs more thorough guidelines on exactly how and when a "sup" direction should be used, it's impact on ERC evaluation, and and an honest statement on the correct way for an IC sourced supply net to be correctly evaluated by the ERC rules.

Guest wrote:

 

Noone who wants to tie to the regulator supply net wants that check to be

treated like being tied to the raw power net through some other

component; they want it to be treated as a completely independent

supply, with all the necessary checks involved. ERC should treat it like

a controlled voltage or current source, which it currently does not.

ERC has no concept of what the circuit is doing. It doesn't know if the

component is a regulator or not, it just sees the pin directions of the

component and what they're connected to. It will never know if you

working with a voltage source, a current source, a resistor, etc. It has

no intuition of your design's function.

You are misunderstanding what I am saying here. Yes, I know full well Eagle does not evaluate the circuit. However, it DOES treat "sup" pins different from "pwr" pins. The ERC for the output of a regulator should be treated the same way as any supply. The exact same rules should apply; there are no other pin output rules that are the exact duplicate of the "sup" direction pins, but merely under a different name. The only way to have the ERC evaluated correctly for a regulator is with a "sup" direction pin. I don't expect Eagle to understand what a regulator is. But I do expect it to have a means by which I can tell it that a pin is supplying power, and not merely using it. Alternatives to "sup" do not allow for that sort of checking.

Guillaume barrey wrote:

 

Hi Ryan,

As I already said, this is how work Eagle today.

Sure, we can think to a better management of supply (been sure that one (at least) and only one regulator (or anything else) provide the supply, check that all devices are powered, and why not, checking that the range of the supply voltage is acceptable for each device, etc so on....)

You are right, things could be better...

But you should put that in the suggestion forum.

 

And, I know that it could be frustrating, but I guess that this is not really high priority. And it could be quite complicated to found a solution that covers all situation redundant supply, several regulator in parallel, or PWM regulator with external FET (in that case the supply is an inductor output !)....

 

And, my point of view, on that is: "sure it would be nice to have a better management, but it's not a huge limitation, and I would prefer that Cadsoft works on other things that would be more helpful".

 

And by the way, you still can use SUP, PWR, NC, OUT.... or what ever you want, there is no obligation Each solution have its own + and its -, and none is perfect. You have to select the one that fits best your needs.

 

Guillaume.

Wrongful ERC results are a support issue. How this issue should be resolved is still an open question that has not been resolved in the software.

On 11/20/2014 2:14 PM, Ryan Pettigrew wrote:

Jorge, you are misunderstanding the situation completely. The decision

to use "sup" direction on the output of a regulator is to avoid both

false negatives in the ERC check, and enable viable use of supply nets.

Without the regulator output being a "sup" pin, the circuit has no

supply for "pwr" direction pins to connect to, thus generating a false

negative.

Hi Ryan,

I think you're over thinking this a bit. As far as EAGLE defines it, Sup

pins are only for supply symbols. These are your GND, +5V, etc. That's

the only context in which the sup direction should be used. The Sup pin

allows all instances of the symbol in the schematic to be automatically

connected.

One way to think about this is that supply pins can only connect to

either power sources or power sinks. Following this logic, you could

define the output of the regulator as a power pin. This would be in line

with how EAGLE defines the use of these types of pin directions and you

could avoid the generation of the ERC errors.

And without a regulator, or some other power conditioning

circuitry, there is genuinely no case for supply nets at all, except in

pointlessly trivial cases, for which they genuinely aren't needed.

What if the supply is off board? EAGLE shouldn't assume the there's a

power supply in the design. Even worse would be to force such a requirement.

Even in this case supply nets would be useful. You would connect them to

the connector that's bringing power into the board.

  No

one who wants to tie to the regulator supply net wants that check to be

treated like being tied to the raw power net through some other

component; they want it to be treated as a completely independent

supply, with all the necessary checks involved. ERC should treat it like

a controlled voltage or current source, which it currently does not.

ERC has no concept of what the circuit is doing. It doesn't know if the

component is a regulator or not, it just sees the pin directions of the

component and what they're connected to. It will never know if you

working with a voltage source, a current source, a resistor, etc. It has

no intuition of your design's function.

 

I'm not asking for my own purposes; I'm asking for the sake of properly

featured library construction. An important part of this is that the ERC

checks for the necessary things correctly, and correctly treats a

regulator output as an independent supply. Further, in order to have a

meaningful supply rail to tie to, it usually needs to be regulated.

EAGLE isn't checking the soundness of what your doing, and is unaware og

regulation or the subtle details of electronics design. If it was, it

could design the circuit on its own and we'd all be out of jobs.

 

I see no reason why either connecting a supply output, or a

regulator output, to an op-amp output, barring at least protection

diodes or some other protection circuitry, would ever be a good idea,

so I really don't see your point.

See above. I've seen new PCB designers make the extreme case I

mentioned. The point is that EAGLE can't tell the difference other than

by how the pin directions are defined, that's all the ERC can check. By

EAGLE's vernacular pin whose direction is set to out and is connected to

a pin whose direction is a sup, could be potentially dangerous. That's

all EAGLE knows.

Hi Ryan,

As I already said, this is how work Eagle today.

Sure, we can think to a better management of supply (been sure that one (at least) and only one regulator (or anything else) provide the supply, check that all devices are powered, and why not, checking that the range of the supply voltage is acceptable for each device, etc so on....)

You are right, things could be better...

But you should put that in the suggestion forum.

And, I know that it could be frustrating, but I guess that this is not really high priority. And it could be quite complicated to found a solution that covers all situation redundant supply, several regulator in parallel, or PWM regulator with external FET (in that case the supply is an inductor output !)....

And, my point of view, on that is: "sure it would be nice to have a better management, but it's not a huge limitation, and I would prefer that Cadsoft works on other things that would be more helpful".

And by the way, you still can use SUP, PWR, NC, OUT.... or what ever you want, there is no obligation Each solution have its own + and its -, and none is perfect. You have to select the one that fits best your needs.

Guillaume.

Jorge, you are misunderstanding the situation completely. The decision to use "sup" direction on the output of a regulator is to avoid both false negatives in the ERC check, and enable viable use of supply nets. Without the regulator output being a "sup" pin, the circuit has no supply for "pwr" direction pins to connect to, thus generating a false negative. And without a regulator, or some other power conditioning circuitry, there is genuinely no case for supply nets at all, except in pointlessly trivial cases, for which they genuinely aren't needed. No one who wants to tie to the regulator supply net wants that check to be treated like being tied to the raw power net through some other component; they want it to be treated as a completely independent supply, with all the necessary checks involved. ERC should treat it like a controlled voltage or current source, which it currently does not.

I'm not asking for my own purposes; I'm asking for the sake of properly featured library construction. An important part of this is that the ERC checks for the necessary things correctly, and correctly treats a regulator output as an independent supply. Further, in order to have a meaningful supply rail to tie to, it usually needs to be regulated.

I see no reason why either connecting a supply output, or a regulator output, to an op-amp output, barring at least protection diodes or some other protection circuitry, would ever be a good idea, so I really don't see your point.

Thank you for attempting a proper explanation. Please understand the problem better, so you can give a more relevant response.

On 11/20/2014 11:26 AM, Ryan Pettigrew wrote:

CadSoft Guest wrote:

 

 

 

On 20/11/14 05:55, Ryan Pettigrew wrote:

 

Because I still want a legitimate technical explanation of why this

is

 

not correct, and not just some rehash of the meaningless gibberish

in

 

the manual.

 

You've had one. More than once. You refused to believe it.

 

 

No, I haven't. Not one.

Hi Ryan,

As you're already aware EAGLE allows a pin direction to be assigned.

Among the options you have are Sup, I/O, Out, In, Pas, HiZ, etc. The

idea behind these options is to allow the ERC to be able to catch

incompatibilities between types of pins. If you don't care for this

feature you can set all of your pins to I/O, Pas, or HiZ which don't

receive any special checks.

With some background out of the way, let's get to your problem. If I'm

understanding correctly, you're concerned about why EAGLE flags a

warning when you connect a supply symbol (direction sup) to your

regulator output (I assume the pin direction is set to out). In your

particular use case this is intended and you would want EAGLE to not

flag an error, however not all outputs come from regulators.

What if instead you had an amplifier whose output was connected to a

supply pin. In this case you would definitely want EAGLE to complain

because that is an error that could potentially destroy your amplifier

or the source feeding it.

EAGLE can't tell the difference between these two cases(in both cases

you have a sup pin, connecting to a pin whose direction is out), so the

safest course of action is to always flag a warning whenever an output

pin and a sup pin connect.

Supply symbols are used to define the power rails and EAGLE expects them

to only connect to pwr pins as stated in the manual.

So what's the conclusion of all of this? You can set the out put of the

regulator to the pwr direction instead of out and that will take care of

the problem. If you don't want to change the regulator output's pin

direction then simply approve the warning. It works either way.

hth,

Jorge Garcia

CadSoft Guest wrote:

 

On 20/11/14 05:55, Ryan Pettigrew wrote:

Because I still want a legitimate technical explanation of why this is

not correct, and not just some rehash of the meaningless gibberish in

the manual.

You've had one. More than once. You refused to believe it.

 

No, I haven't. Not one.

On 20/11/14 05:55, Ryan Pettigrew wrote:

Because I still want a legitimate technical explanation of why this is

not correct, and not just some rehash of the meaningless gibberish in

the manual.

You've had one. More than once. You refused to believe it.