Newbie question about power supply "sleep" mode

[bvz]

I am not sure if this is the best forum or if I should be in the newbie section...

I am trying to mod an old eMac (to run it headless) by placing it inside a small case (actually a Mac performa 6115CD case for any who are interested). I plan on using a micro ATX power supply. I have been able to find the appropriate info on making sure the ATX power supply works with the voltages that are required for the eMac (including having to down-convert 5.5V to 2.5V).

My question is as follows: Macs have a deep sleep mode wherein the computer more or less shuts down (drive, all fans including the PSU fans, etc) but enough power is kept on to keep the memory active. I don't know the first thing about modern PC's. Do they also have a similar deep sleep mode? If so, how is the PSU signaled to go to this sleep state? How is it told to wake back up?

Finally, while I have you all here can you recommend a decent yet still affordable micro ATX PSU (or any PSU that is less than 2.5 inches high)? It will be running the mainboard (including built in video), the built in USB (3 of them) and firewire (2 of them) ports, a single 250GB HD, and a DVD/CD RW. I don't think I need a really powerfull unit, but quiet and efficient would be nice. I am reading the Core Info on PSU's right now but any help is appreciated.

Thanks
Ben

Edited to add the DVD drive to the specs

[dukla2000]

Welcome to SPCR!

Sounds like an interesting project.

My question is as follows: Macs have a deep sleep mode wherein the computer more or less shuts down (drive, all fans including the PSU fans, etc) but enough power is kept on to keep the memory active. I don't know the first thing about modern PC's. Do they also have a similar deep sleep mode? If so, how is the PSU signaled to go to this sleep state? How is it told to wake back up?

AFAIK ATX psus are on or off. However there is one circuit, +5VSB (purple wire IIRC) that is on when the rest of the psu is off (as long as the AC circuit is connected and on). This line is very rarely rated over 2A though so not sure how much you can do with it. The psu is switched on by pulling PS_ON signal low (the green wire - you can power up an ATX psu without a mobo by connecting the green to any black).

The ATX specs do detail all this stuff though - there is a link to the ATX specs on the SPCR PSU Fundamentals page.

[bvz]

dukla,

Thanks for the response.

I am still researching what constitutes "sleep" on the Mac side. Maybe it turns the PSU completely off except for some line similar to the ATX +5VSB. Maybe it is something more complicated. This stuff is all really a bit over my head but I am having fun learning.

In addition to the excellent intoduction to power supplies on this site, I also came accross the following two references:


http://www.pcguide.com/ref/power/sup/index.htm

http://www.formfactors.org/developer...public_br2.pdf


The second one has some text in it that goes as follows:

The “Energy Star” efficiency requirements of the power supply depend on the intended system configuration. In the low-power / sleep state (S1 or S3) the system should consume power in accordance with the values listed in Table 10. Note: To help meet the “Energy Star” system requirements, it is recommended that the power supply have > 50% efficiency in standby mode.

and

For power supplies designed for low standby power, the following provides some general guidance. Requirements will vary with geographic region and target end user market. To help meet the Blue Angel*, RAL-UZ 78, US Presidential executive order 13221, future EPA requirements, and other low Power system requirements, it is recommended that the +5 VSB standby supply should be as efficient as possible.


These statements lead me to believe that the ATX spec either requires or at least suggests a low power mode (which may be called "standby mode"). Now, whether that low power mode is simply the always on +5VSB line or not I can't tell. So I guess it is more research for me!

I've never done any of this kind of stuff before, but if I continue (and as I learn) I will post my results here.

Cheers!

Ben

[Tephras]

The +5 VSB is a "standby supply output" as it is stated further down in the PSU design guide which is the document you are referring to. On the website where you found that document you can also find the ATX Specification in which the same thing is stated, it's also mentioned that a proper implementation of the +5 VSB is "required for an ATX12V compliant power supply."

[bvz]

Tephras,

This standby supply output, then, is the "sleep state" of the PSU? I guess that is the part I am not completely getting and what I am not able to derive from reading the documentation so far (I am really really new to all of this).

Here is my guess based on what I have read so far, but it is only a guess:

The PSU can exist in one of three states: Completely off (as in, not plugged in), off except for the single +5VSB line, and fully on. The motherboard then is responsible for determining the difference between the computer being "off" and the computer being in sleep mode. When asked to shut down via soft power, it shuts the PSU off and does not attempt to draw any current from the +5VSB line. When asked to go to sleep, it also shuts the PSU off but continues to draw from the +5VSB line enough current to keep the memory active.

Again, this is just my guess and I am eager to be corrected and/or pointed to more docs (but this whole process is becoming somewhat problematic since I spend all my time reading about PSU's instead of getting any work done! I never thought it possible, but I am completely fascinated by all of this!)

Ben

[dukla2000]

I am far from expert in this area, but my 6p:

The PSU can exist in one of three states: Completely off (as in, not plugged in), off except for the single +5VSB line, and fully on.

That is my understanding as well.

The motherboard then is responsible for determining the difference between the computer being "off" and the computer being in sleep mode.

Yes: as well as other states a PC can get into such as ON.

When asked to shut down via soft power, it shuts the PSU off and does not attempt to draw any current from the +5VSB line.

No: when a PC shuts down the mobo (well actually OS tells the BIOS tells the mobo) tells the PSU to switch off. The PSU shuts down all rails except the +5VSB. The mobo has circuits that use the +5VSB - typically there are jumpers so you can power the PS/2 keyboard/mouse ports from +5VSB, and more recently some/all of the USB ports. That way you can configure a PC to start based on keyboard/mouse actions in addition to the traditional On/Off button.

When asked to go to sleep, it also shuts the PSU off but continues to draw from the +5VSB line enough current to keep the memory active.

I am next to ignorant in this area, but think that when a PC goes to sleep (S1, S3, suspend to RAM etc) actually it keeps the PSU running in "full" On mode. Except by shutting down as much as possible the power draw on the PSU should be very low. But memory is never powered from +5VSB (AFAIK) hence my assumption the PSU must stay "full" on when a PC is "sleeping".

[dukla2000]

Sorry to double post but was just reviewing the thread and suspect:

Note: To help meet the “Energy Star” system requirements, it is recommended that the power supply have > 50% efficiency in standby mode.

In my understanding, the standby mode referred to here is the mode when the PSU is Off but connected to AC, so the +5VSB line is alive. The point being that the average user thinks his PC is off and should be drawing 0 Amps, whereas in fact it could be drawing up to 2A (or more) via +5VSB. If the +5VSB circuit was badly inefficient then the AC usage could become significant.

[bvz]

Yes: as well as other states a PC can get into such as ON.

Yes. I forgot about that one (kinda useful state for the pc to be in, eh?)

No: when a PC shuts down the mobo (well actually OS tells the BIOS tells the mobo) tells the PSU to switch off. The PSU shuts down all rails except the +5VSB. The mobo has circuits that use the +5VSB - typically there are jumpers so you can power the PS/2 keyboard/mouse ports from +5VSB, and more recently some/all of the USB ports. That way you can configure a PC to start based on keyboard/mouse actions in addition to the traditional On/Off button.

Ah, this makes sense. I hadn't really considered those events. Thanks.

I am next to ignorant in this area, but think that when a PC goes to sleep (S1, S3, suspend to RAM etc) actually it keeps the PSU running in "full" On mode. Except by shutting down as much as possible the power draw on the PSU should be very low.

Ok, this also begins to make some sense to me. As long as the mobo is not drawing much current, the PSU (assuming a decent efficiency in this state) will not be drawing much current from the grid. I keep forgetting that a 350W PSU does not in fact always consume 350W... it only consumes what is needed to supply whatever load is placed on it.

Now, this leads me to one final question... according to some of the docs I have read, there is in the optional specs of the SFX PSU the ability to turn the fan off and on. If the computer is asleep, we don't want to hear the fan running. Somehow the mobo needs to shut the fan off but how can it be sure that the PSU is cool enought to actually run without this fan? Hmmm. More research!

Thanks all for your help so far!

Ben

[Bean]

When asked to go to sleep, it also shuts the PSU off but continues to draw from the +5VSB line enough current to keep the memory active.

I am next to ignorant in this area, but think that when a PC goes to sleep (S1, S3, suspend to RAM etc) actually it keeps the PSU running in "full" On mode. Except by shutting down as much as possible the power draw on the PSU should be very low. But memory is never powered from +5VSB (AFAIK) hence my assumption the PSU must stay "full" on when a PC is "sleeping".

If so, how come the PSU fan is turned off in deep sleep mode. Can the PSU stay full on when a PC is sleeping yet somehow stop its PSU fan from running?

[bvz]

Ok, after a bit more research on the formfactors.org website, I have come across the following statement in the specs for SFX PSU's:

1.2.5. External Fan Control - Optional
With the implementation of Suspend To RAM (STR or S3 sleep state), the system can obtain a low power condition without the need for external fan shutoff. In STR mode, the main outputs including 12 V are not available from the power supply and all system fans and the power supply fan will be off. FANC and FANM functions described in Version 1.1 have been removed for Version 2.0. Some vendors may still offer the external Fan functions as an option.


I am still not clear whether S3 sleep state is something that is available for the mobo alone, or whether the PSU also actively engages in. Specifically, in the description above it indicates that in S3 mode that 12V outputs are not available from the PSU. This suggests to me that it isn't simply a case of the mobo drawing less power, but more that the PSU iteslf has been instructed to go into this mode. But my question remains: How is it told to go into this mode?

The search continues!

Ben

[Tephras]

Ok, after a bit more research on the formfactors.org website, I have come across the following statement in the specs for SFX PSU's:

1.2.5. External Fan Control - Optional
With the implementation of Suspend To RAM (STR or S3 sleep state), the system can obtain a low power condition without the need for external fan shutoff. In STR mode, the main outputs including 12 V are not available from the power supply and all system fans and the power supply fan will be off. FANC and FANM functions described in Version 1.1 have been removed for Version 2.0. Some vendors may still offer the external Fan functions as an option.

The FanC and FanM functions are mentioned in the ATX specification too, more specifically in section 4.2 and they are provided through an optional 6-pin power connector. I've never seen this 6-pin connector implemented but some PSUs has a two or three pin connector to provide the FanM function. The FanM tachometer signal "allows the system to monitor the power supply for fan speed or failures. Implementation of this signal would allow a system designer to gracefully power down the system in the case of a critical fan failure." and the FanC signal "allows the system to request control of the power supply fan from full speed to off. Implementation of this signal would allow a system designer to implement a request-fan-speed control or shut-down during low power states such as sleep or suspend."

I am still not clear whether S3 sleep state is something that is available for the mobo alone, or whether the PSU also actively engages in. Specifically, in the description above it indicates that in S3 mode that 12V outputs are not available from the PSU. This suggests to me that it isn't simply a case of the mobo drawing less power, but more that the PSU iteslf has been instructed to go into this mode.

The PSU doesn't do much on it's own, e.g. when you start your computer it's the PS_ON# signal from the motherboard that tells the PSU to turn on (and off) the main power rails. PS_ON#="is an active low signal that turns on all of the main power rails including 3.3VDC, 5VDC, -5VDC, 12VDC, and -12VDC power rails", "PS_ON#...allows a motherboard to remotely control the power supply in conjunction with features such as soft on/off, Wake on LAN*, or wake-on-modem."

But my question remains: How is it told to go into this mode?

I think you must leave formfactors.org and dig into the documents at www.acpi.info and Intels Instantly Available PC documents if you want to know more about how the sleep states are controlled since this is not all about hardware but rather the interaction through ACPI between BIOS, OS and hardware.

[dukla2000]

In STR mode, the main outputs including 12 V are not available from the power supply ...

This part seems slightly unclear: either it is the PSU OFF state (with +5VSB alive) or something else? I am not sure my OFF terminology is correct, but if it is something else then I have no idea how the signalling between mobo and psu achieves it. AFAIK there is only the PS_ON signal (which the mobo briefly grounds to start the PSU) and the PWR_OK which the psu raises a couple of milli-seconds later when all the rails are up and ready. And I always presumed that the reverse "shut down" process was triggered by the mobo (again) briefly grounding the PS_ON signal.

An even bigger hole in my knowledge: I presume from that statement that in STR (mobo) state the psu is OFF except for +5Vsb. I also presume the RAM needs some power to prevent it completely forgetting what has been saved there. And the only source for that I can imagine is +5VSB. I always thought +5VSB was a 'novelty' rail to keep the LED on my mouse alive when the PC was off etc: seems it may have a more serious role.

[bvz]

I think you must leave formfactors.org and dig into the documents at www.acpi.info and Intels Instantly Available PC documents if you want to know more about how the sleep states are controlled since this is not all about hardware but rather the interaction through ACPI between BIOS, OS and hardware.

Tephras, you are a genius. I am still sifting through the info, but I am now getting a lot closer by looking through these documents.

Specifically, the following document has really helped:

http://developer.intel.com/technology/I ... rev1_2.htm


It describes the PS '98 spec which included ACPI support. There are a few interesting details in there which I will summarize here. Keep in mind that I have only skimmed this and some other documents, so my summary may be (most likely is) a bit off.

It seems that in the spec there are two sources of power... there is the main power (5V, 3.3V) and then there is a dual power (5V, 3.3V). Ideally, this dual power runs off of a separate converter within the PSU. In addition to this, there is a 3.3Vstby (and, I think, an optional 5Vstby which is the old +5VSB that we know and love).

A table in the document mentioned above shows the following:
----------------------------------------------------------------------------------------------------------------
Voltage Plane: 3.3V
Consumer: 3.3 V PCI, 3.3 V components, processor cache, Expansion Bus Bridge

Voltage Plane: 5V
Consumer: processor VRM, PCI and other 5 V components on the motherboard

Voltage Plane: 3.3V dual
Consumer: Host bridge with memory controller, SDRAM components, memory clock driver, 3.3Vaux PCI connector auxiliary power pin

Voltage Plane: 5V dual
Consumer: USB devices (USB wakeup capabilities will be included in the future devices), 5V based PCI motherboard devices intended to be wakeup devices.

Voltage Plane: 3.3V always present standby
Consumer: This voltage is always present when the system is plugged in to the AC socket. It is intended to be connected only to the ACPI controller’s resume well, the power button and power indicator (LED).

The remainder of the platform voltages are what today’s desktop power supplies already comprehend.
----------------------------------------------------------------------------------------------------------------


This indicates that the memory and PCI bus and such which need to be powered while the rest of the machine is in the S3 suspend to ram state draw their power from the dual lines, and not the main lines.

Now, as to what state the PSU is in at any given time, that is still not completely clear to me.

There is another table in this document which (in part) has the following information:
----------------------------------------------------------------------------------------------------------------
Feature: Memory subsystem
Recommendations/comments: Powered by dual power circuit • SDRAM recommended • RDRAM devices target for the future designs • Memory devices are in self refresh state while in S3 • Voltage must be 3.0-3.6 V • Typical power per DIMM device 1-5 mA

Feature: USB wakeup
Recommendations/comments: Requires 500 mA per port • When system is in S3 state the standby converter supplies power to the USB port • The power for USB needs to be from dual power source or standby converter has to have capacity needed for wakeup devices • Use 5 V dual to enable USB wakeup

Feature: Signals to control power supply. SUSB# SUSC#
Recommendations/comments: Main power supply converter is turn off while in S3, S4 and S5 state • SUSB# signal from ACPI controller can be used to turn off the main converter. Ensure the proper polarity. • SUSC# signal is generated by the ACPI controller to indicate that system entered S4 or S5 state. It can be used to control power to the memory array.

Feature: Thermals
Recommendations/comments: Base designs have fan on is S0 and fan off in S3 • Use thermal sensor for active fan control • Use either linear voltage controlled fans or pulse width modulated fans to minimize noise while in S0 state • Use ACPI thermal policy for optimal performance
----------------------------------------------------------------------------------------------------------------


From this I gather that

a) the ram (and any other devices that need to still be partially active) are indeed powered by this new dual circuit.
b) the +5VSB is still there to optionally power the USB bus... but that the 5 V dual should be the one used if you want the USB device to wake the machine up.
c) that there exist some signals to control the power supply and let it know which power state to assume. How exactly these signals manifest themselves I still don't know
d) that the fans are not controlled by the PSU, but rather are turned off by the mobo in conjunction with the OS and BIOS.

These assumptions may be wrong, but I think we are starting to get closer...

Ben