Power supply PG value? huh?

[lilshawn]

Preface, I'm an electronics repair technician for a company responsible for several hundred pieces of amusement and arcade equipment. As part of my testing arsenal I use one of those cheapy power supply testers to quickly diagnose ATX power supplies often used in the computers contained in arcade games these days to determine if i have a motherboard issue or a power supply issue.

 

I'm not concerned about the accuracy of such a device... more that, all the rails that are supposed to be there... ARE there. (and trust me, the amount of supplies i have run across with dead/failing rails would fill a dumpster several times over.)  but one aspect about it, and others like it, kind of eludes me. the display of a "PG" value.

 

the only thing the manual for the unit I own says about it is...

 

"The unit of PG is ms "millisecond". When the test for PG fails, the backlight turns red, the system displays the letter "F" under the PG icon, and the buzzer makes a "beep" sound"

 

informative comma NOT!

 

normally I would just ignore it but, knowledge is power and power supply knowledge when dealing with them...all the time... several times a day... i decided looking into what it actually means might be beneficial.

 

I often see talk about this value gauging the health or reliability or whatever of a power supply by using a "tester" measuring... and looking at the supply's "PG value". now, normally I would be all over understanding  data... but, I don't definitively see, really anywhere, WHAT this actually means or what it's measuring when i search about it. I see a lot of the same "corsair" phrase being thrown around all over forums... (including here)

 

"PG Values are usually considered normal if it falls within the range of 100ms - 500ms, but can be lower than 100ms at times. 0ms PG values would be considered a failing PSU."

 

this seems to be the only tidbit about what PG means that is out there and it's being parroted over and over without any further explanation of why these values were "chosen?" or even... what it is.

 

But to me, this method of data presentation seems very counter intuitive. ms? how is less milliseconds bad and more better? ms until...what? what is being measured? can someone technically explain what is actually being quantitatively measured by this test and why a low number would be indicative of failure and a larger one be better? could an 999ms measurement be ideal? what about 3500ms? would a power supply with an infinity ms PG value last forever? i don't know... because nobody seems knows what it actually represents.

 

any of the Power supply testing crew able to explain?

 

A long time watcher, first time caller. Lifelong ADHD'r. 

 

Thanks a bunch!

[Average Nerd]

The PG (Power Good) Signal is being sent as a +5V signal to the motherboard after the PSU has completed its internal self test and the Power rails have stabilized.

The PG Value is the delay between the Power on signal being pulled to ground and the PSU sending the Power Good signal. If that signal takes too long, it can be a sign of a failing or otherwise malfunctioning PSU.

[lilshawn]

thank you for replying. the explanation does not make sense to me, that a commonly accepted "good" value would be 100ms - 500ms and a "bad" value would be zero.

 

here, allow me to explain my thought process behind that.

 

for example...

 

getting a zero value would mean that the computer got a signal back saying the supply was ready immediately when requested.

 

*****

computer- i need power you ready?

 

*0 ms of time passes (immediately)*

 

power supply - all tests pass i can do that i'm ready to go lets go here's the juice

 

computer - bruh, did you even test?

*****

 

and getting a value of anything other than zero would mean the supply took additional time before it was ready... which to me, seems like the supply would be failing because it took longer to say it was ready.

 

*****

computer- i need power you ready?

 

power supply - oh... yeah uhhhhhh. lemme check to see if i'm good to go

 

*500 ms of time passes*

 

power supply - all tests performed i can do that.

 

computer - uhm... you okay in there? that took quite a while?

*****

 

sorry, not to demean you or make fun of your explanation, but this is how that explanation plays out in my head and still does not make any sense to me. how is a bigger value acceptable and a smaller one...not.

 

maybe if the actual physical/electrical test itself was explained. but i'm still not sure how a test that takes longer to perform would be good. i seems this quantitative value is measuring...something... but to me, not how long it's taking to complete that test.

 

this is my primary issue with how the data is presented with the test... and i wonder if using something else or some other value would be more appropriate or beneficial... maybe even reversing the value where a bigger value was LESS desirable.

[Average Nerd]

13 minutes ago, lilshawn said:

getting a zero value would mean that the computer got a signal back saying the supply was ready immediately when requested.

 

*****

computer- i need power you ready?

 

*0 ms of time passes (immediately)*

 

power supply - all tests pass i can do that i'm ready to go lets go here's the juice

 

computer - bruh, did you even test?

As far as I know, if the value is >100ms, the conversation would be exactly like you described.

The PSU only starts to test itself after it has received the Power on signal, which takes a bit of time, and when it takes too long, the computer will ask: "Hey, you took way too long, I'll trigger an error message to inform the user or cancel the POST process."

[Poinkachu]

21 minutes ago, lilshawn said:

 

*0 ms of time passes (immediately)*

 

That's a trap

[lilshawn]

37 minutes ago, Poinkachu said:

That's a trap

OOOOkay. This piece of info has kinda made all this START to make sense. this means we are not only testing for the time but also testing for a non-zero state.

 

am I correct in concluding this fact as... "as long as the ms value is NOT zero and the ms value is no longer (say for example, over 500ms is the cutoff) than 500ms... then, power supply is actually good.

 

so, does anybody know how the state is reached in a zero ms time? i assume this "test signal" is simply just the power supply turning on and feeding 5v back, and this delay is the time the supply takes to do the test. if the supply immediately turns on "without performing a test" this is how I assume it would do so. maybe it's unable to stop itself from powering on and the rails might not actually be nice and stable?

 

i'm still confused because I would think taking longer to supply said 5v would seem like it has problems and the shortest (non-zero) time would be better. i mean it could be indicative of a supply having tighter tolerance for rail voltages and waiting for a smaller variance in the voltage before throwing the switch and outputting voltage. but could also be indicative of a supply having to wait a long time for the oscillations to stop due to failing capacitors or other filtering.

 

I think the phrase "PG Values are usually considered normal if it falls within the range of 100ms - 500ms, but can be lower than 100ms at times. 0ms PG values would be considered a failing PSU." reads weird and I think that's maybe where i'm getting confused.

 

it reads to me as though 100ms to 500ms is good, 1 to 100 is undesirable but could be encountered, and zero is bad news your supply is dying. (being that 1-100ms is between good and dead... i chose undesirable since that is the only thing in between those two extremes.)

 

if i were to revise this, Am i correct in stating something along the lines of...

 

"PG Values are usually considered normal if it falls within the range of 1ms - 500ms, but a value of 0ms can indicate evidence of an internal power supply failure. Power supplies with values over 500ms should be treated as highly suspect."

[Average Nerd]

3 minutes ago, lilshawn said:

so, does anybody know how the state is reached in a zero ms time?

Some cheap PSUs tie the Power Good signal to a 5V rail, so it just goes to 5V even if the PSU never self tested or waited for the rails to stabilize.

 

6 minutes ago, lilshawn said:

"PG Values are usually considered normal if it falls within the range of 1ms - 500ms, but a value of 0ms can indicate evidence of an internal power supply failure. Power supplies with values over 500ms should be treated as highly suspect."

PSUs with a PG value of <100ms can also use similar trickery as described above, and should also be treated with caution. Other than that, I don't see anything inherently wrong with this statement.

[seon123]

The Intel PSU design guide explains why they recommend at least 100ms for the PWR_OK delay

https://edc.intel.com/content/www/us/en/design/ipla/software-development-platforms/client/platforms/alder-lake-desktop/atx-version-3-0-multi-rail-desktop-platform-power-supply-design-guide/2.1a/2.0/

 

Check chapter 4.3, and Table 4-10. Note 2 for that table. 100ms minimum is recommended for older generation systems.

(It's a .pdf file and I cba copy pasting it properly in text)

 

It also takes some time for the voltages to rise to the required values, so there should be some non-zero delay before the PSU sends the signal, just because of that.

[lilshawn]

33 minutes ago, Average Nerd said:

PSUs with a PG value of <100ms can also use similar trickery as described above, and should also be treated with caution. Other than that, I don't see anything inherently wrong with this statement.

I mean, at this point... if we are buying a power supply cheap enough to implement a simple delay like this to HOPEFULLY have good rails by the time we output... without actually having to put the money into the actual circuit to test... it seems almost a wash at that point... but 2 cents on capacitors is 2 cents in your pocket, so i get it.

 

22 minutes ago, seon123 said:

The Intel PSU design guide explains why they recommend at least 100ms for the PWR_OK delay

https://edc.intel.com/content/www/us/en/design/ipla/software-development-platforms/client/platforms/alder-lake-desktop/atx-version-3-0-multi-rail-desktop-platform-power-supply-design-guide/2.1a/2.0/

 

Check chapter 4.3, and Table 4-10. Note 2 for that table. 100ms minimum is recommended for older generation systems.

(It's a .pdf file and I cba copy pasting it properly in text)

 

It also takes some time for the voltages to rise to the required values, so there should be some non-zero delay before the PSU sends the signal, just because of that.

great point. i hadn't thought to look to intel to have had a hand in the specification. but it makes sense since their CPU's would be relying on the power being smooth and accurate.

 

so...

 

assuming a good reputable power supply company's supply under test, the statement,

 

"PG Values are usually considered normal if it falls within the range of 100ms - 500ms, but a value of 0ms can indicate evidence of an internal power supply failure or trickery to simulate a good supply regardless if the supply is good or not. Power supplies with values over 500ms should be treated as highly suspect, with 1ms to 99ms being a gray area with regards to compliance with modern motherboards, but may be acceptable on older systems." 

 

seems pretty accurate now (your kilometerage may vary i am not an engineer warranty is 2 part if it breaks you keep both parts all rights reserved etc.etc.etc.)

 

but also.. in addition, if testing an unknown/unusual/noname/offbrand brand or manufacturer supply under test, the statement 

 

"These types of "value" PSUs may return with a PG value of anywhere from 1 - 500ms regardless of the actual power condition, and could be cause for concern due to the manufacturer trickery regarding the power good testing process, and should be treated with caution. As a matter of course, value power supplies should be changed regardless of PG value."

 

could be added and allies as well.

 

I think I understand this now, thank you all for your input and help! 

[seon123]

1 hour ago, lilshawn said:

-snip-

with 1ms to 99ms being a gray area with regards to compliance with modern motherboards, but may be acceptable on older systems." 

That says the opposite of Intel. According to them, it could be an issue especially on old systems. 

 

I would not worry about the PWR_OK delay at all, and instead worry mainly about ripple, when it comes to testing sketchy PSUs, as that is what actually kills components over time, outside of catastrophic failures. 

[lilshawn]

13 minutes ago, seon123 said:

worry mainly about ripple, when it comes to testing sketchy PSUs, as that is what actually kills components over time, outside of catastrophic failures. 

 

absolutely. if im suspect of a power supply that is still working, I will measure the DC rails with my meter on AC. This is a good indication if the output filter capacitors of the supply are shot. 3mvp-p? it's fine. 25mv? if it's got a good load on it, it's probably okay still. 100mv with no load? she's done.