SolidGear power supply

Anyone have an opinion about SolidGear?
I have a box with an Asus board that suddenly
rebooted to "prevent damage from power
supply surge". From discussions online I
can't tell whether the PS is faulty (it's fairly
new) or whether the Asus surge sensor is
faulty and should be disabled.

Mayayana wrote:
Anyone have an opinion about SolidGear?
I have a box with an Asus board that suddenly
rebooted to "prevent damage from power
supply surge". From discussions online I
can't tell whether the PS is faulty (it's fairly
new) or whether the Asus surge sensor is
faulty and should be disabled.

http://www.overclock.net/products/so...y/reviews/6552

I would wait and see if it does it again.
A new p/s is a bit cheaper and easier to install than a new MB.

Mayayana wrote:
Anyone have an opinion about SolidGear?
I have a box with an Asus board that suddenly
rebooted to "prevent damage from power
supply surge". From discussions online I
can't tell whether the PS is faulty (it's fairly
new) or whether the Asus surge sensor is
faulty and should be disabled.



"Anti-surge says unstable power supply"

http://vip.asus.com/forum/view.aspx?...e=en-us&page=2

Tis a strange function.

Do you think that is properly documented somewhere ?

The hardware monitor in the SuperI/O, is a single ADC,
with an eight or nine channel mux on the front. It doesn't even
read all the voltages at the same instant in time. It
might be using SAR (successive approximation register)
concept. That's really no better than using your
$20 Harbor Freight multimeter, and expecting to catch
something "peaking". When I look at the screen of mine, it
looks like I'm getting maybe two readings per second on each
voltage.

If they used window comparators and did "V1 = V = V2"
checking in the analog domain, that could be used to
set an interrupt. But, you'd have no readings to look
at later. All you'd know, is at some point, something
went out of bounds. And to build that, costs money.
Some hand-held PSU testers use a method like that
(have a single LED indicating "in-spec" per rail).

Asus do have some supervisor chips, for operating multi-phase
Vcore and the like. Would there be a function in there ?
What voltage does it monitor ? Vcore ? Or everything ?

*******

I would probably disable the BIOS setting, if it was
me. Unless I could be convinced it was a quality implementation
and could actually measure something important.

And what good does rebooting do, exactly ? Besides
**** you off ? Shutting down would make some sense.
But on the other hand, the act of shutting down (depending
on timing), might cause a surge behavior to become worse.

*******

The Asus representative here, leads me to believe it is
the "naive" SuperI/O hardware implementation of the function,
complete with too-low sampling rate to be effective.

https://rog.asus.com/forum/showthrea...tem-Faulty-PSU

"If there is any polling going on with multiple tools
on the system, the Super IO will mis-read, and this
can generate a shutdown. If that happens to be the case,
disable Anti-Surge, then reinstall the APC software."

That would be my assumption about how it works. Just an
attempt to read the Hardware Monitor ADC, and conclude
something from it. Back when the SuperI/O was on the SMBUS,
readout errors were an epidemic, because one utility could
"stomp" on the serial bus access of the other utility.
Modern systems use the 33MHz LPC bus, and as far as I
know, once a transaction kicks off, it completes. Readout
errors should not happen. Assuming everything involved
here has the appropriate "interlocks" so that the operations
can complete properly. For example, if your read the Hardware
Monitor, it should not upset any reading currently in progress.
There should be no timing windows where that can happen,
if the Hardware Monitor is worth having and using.

At one time, one of the developers of hardware readout software,
attempted to get all the parties to agree to a semaphore scheme,
so there would not be bus stomping. He got no-where fast.

*******

It turns out my new motherboard has this "anti-surge" feature,
and I didn't even notice. The setting is down the screen, and
I'd have to scroll down to see it. It is enabled by default.
This picture is selected chunks of a 12Mpixel digital camera shot.

https://s27.postimg.org/gm31udhhv/anti_surge.jpg

So what do I notice. The setting is right below the voltage
the Hardware monitor measures. The text claims

If enabled, system will have uvp or ovp protect function

One would hope the function was "and", not "or".

Implying they are using the formfactors +/-5% allowance on
the rails. It's unclear what kind of number would be useful
on VCore. The tolerance on that should be quite good, whereas
the best place to measure it is a problem.

Now, sitting in the BIOS, how do my readings vary ? There
is some oscillation. This is really quite reasonable. Only
the 12V rail is ridiculous. It is *not* sitting there precisely
at 12.000V. I don't know what is going on there. It's like it
was nailed down or something. Like, a dummy reading. At least
the others, the voltages are "realistic", like they were coming
from a cheap power supply or a cheap instrument.

Vcore 1.102, 1.104, 1.106
3.3V 3.328, 3.344
5V 5.080, 5.120
12V 12.000 --- hilarious, doesn't move

Each rail has scaling. The ADC is 8 bit, meaning 1 of 256 values.
The metering reads from 0 to 4.096V or 0.016V per step.

The 12V input would be around 16V full scale. The front end
resistive divider (external to the ADC) would attenuate the
signal by multiplying it by 0.25. So when you see 12.000 on
the screen, if the setting were to move, it would be by 0.064V.
It's allowed 5% or 0.600V. That means we have ten measurement
steps roughly on the ADC, between being perfectly centered,
and hitting "OVP".

The 3.3V on the other hand, you should be able to see that
one is unscaled. The converter reads 4.096V full scale, and
immediately you can see the two readings that it oscillates
between, are 0.016V apart. Again, you get about ten measurement
steps before you hit the 5% mark.

The power supply itself is far from perfect. There is a static
offset error on the voltage. The reason they give 5%, is because
it might need 3% on a shared feedback design, just for the realizable
error.

I would estimate there isn't all that much wiggle room, if you
set the alarm at +/-5%. And maybe it needs an even larger allowance
before going to "alarm state".

Have a stare at your hardware monitor screen in the BIOS,
and see how much "wiggling" it is doing. The BIOS is not
much of a load for the PSU, so the wiggling should be
quite small.

If a power supply 12V rail is "really defective", you will
hear some cooling fan speed variation. There is enough of
a speed change, you can hear it wandering in terms of rotational
speed. That's how I've detected PSU problems on two supplies
in the past.

And by all means, check that the main PSU connector is seated.
It has a locking latch for a reason, so thermal expansion cannot
cause it to "back out" on its own. Make sure it is latched.

And how is ATX12V monitored, compared to the 12V wire on the
main connector ? Well, the Hardware Monitor was never designed
to be a monitor that monitors every wire bundle. It may be
able to warn you the main connector is loose, or the ATX12V is
loose, but it cannot monitor all of them. There aren't enough
channels for that.

Paul

On Sun, 5 Feb 2017 20:39:25 -0500, Mayayana wrote:

Anyone have an opinion about SolidGear?
I have a box with an Asus board that suddenly
rebooted to "prevent damage from power
supply surge". From discussions online I
can't tell whether the PS is faulty (it's fairly
new) or whether the Asus surge sensor is
faulty and should be disabled.

I've disabled it on my Asus boards. Whenever there is a power cut they
always claim to have detected a power surge when the power comes back.

I can't remember why I didn't like that now though. I think they
started up themselves after a power cut and I've got them set to not
boot on power on until I press the power button. I've got them
connected to a KVM and it the booting PC is not selected it screws up
the display settings.

--
Faster, cheaper, quieter than HS2
and built in 5 years;
UKUltraspeed http://www.500kmh.com/

Thanks to everyone. After reading these posts I
looked into it further. Turns out my current machine also
has a SolidGear, with the same MB. And surge protection
is enabled. I never noticed. But while HWMonitor shows
almost no movement on my machine if I'm not
doing anything in particular, the new box
shows cpu vcore going between .8 and 1.3. It also
shows 3.3v and 12v values varying by at least
a few hundredths. The +12v reading, at 8+,
is varying by about .15. So there seems to be almost
a regular oscillation. The readings are also in
different ranges on the two machines, with the
same PS and MB, but different CPU. My current
machine shows almost constant 1.73 for cpu vcore
and shows +12v around 11.
I have no experience with this so I have no idea
what normal range is, but that seems like an awfully
lot of variation.

Paul in Houston TX wrote:
Mayayana wrote:
Anyone have an opinion about SolidGear?
I have a box with an Asus board that suddenly
rebooted to "prevent damage from power
supply surge". From discussions online I
can't tell whether the PS is faulty (it's fairly
new) or whether the Asus surge sensor is
faulty and should be disabled.

http://www.overclock.net/products/so...y/reviews/6552


I would wait and see if it does it again.
A new p/s is a bit cheaper and easier to install than a new MB.


There is a lot of contracting-out, in the PSU industry.

Some companies, actually do design their own stuff.
Fortron/Sparkle, Seasonic, Enermax, they would make
their own.

Even when companies make their own, they are not above
contracting out themselves, and buying a really cheap
supply to fill the lowest tier. Always be suspicious.

Companies like Antec, contract the stuff. They've
bought lots of them from ChannelWell, but later switched
to Delta for some of the units. The quality and internal
design, varies with the source. The Delta probably has
the noisy fan.

So when you see a "brand name", next you have to decide
what "tier" the supply is in.

Some of the 80+, initially they would all have been
coming from Seasonic. Somebody probably has a patent
on double-forward-conversion, and then other parties
might have to pay a licensing fee.

On an 80+ supply, conversion can be in two stages.
The massive 12V main rail. And a separate circuit card
has the +3.3V and +12V converters. They run off +12V.
So there are two converters in a row. The 80+ supply
gets most of its efficiency, when the user does not
load the 3.3V or 5V that heavily. The power drawn from
the "double path", is less efficient than the power
from the "single path". The hope is, your CPU and GPU
are loading that 12VDC rail.

400VDC ------------- 5VSB
+ convert
|
110AC -------+-- 12VDC -------- 3.3V, 5V board
convert | convert | |
v v v

The 5VSB standby voltage is used when sleeping. The fan goes off.
The front-end rectifiers continue to run. The active PFC
continues to run, which boosts the voltage on the main filter
cap a bit. The +5VSB converter runs off the same 400V rail as
the main supply, but uses an economical converter suited
to the tiny load it supports. The wasted power can be down
in the 0.5W to 1W range for standby. (In other words, if you
select "Shutdown" in Windows, and the +5VSB converter is running,
and RAM contents are not preserved, the PSU power wastage
drops to the 0.5W level.)

So when we do your surge analysis of that whacking
pile of converters, there should be very little cross-loading,
and generally the outputs should be well behaved. And
not be tripping any "anti surge" feature :-)

The old supplies without PFC, were simpler.

350VDC ------------- 5VSB
+ convert
|
110AC -------+-- 12VDC --- Single feedback loop.
convert 5VDC --- Loading one heavily, affects
3.3VDC --- the voltage on the others.

The main converter switches off when you sleep, hibernate,
or shutdown. The 5VSB provides power for RAM, while sleeping,
and for stuff like Wake On LAN. If there is a "load step"
on the outputs, as I understand it, it might take 1-1.5
milliseconds for the loop to respond. The supply can have
5000uF of filter caps (without affecting stability), so
they have an impact on the dynamics. If the motherboard had
too many filter caps, the PSU might become unstable, and
there are limits the motherboard makers know about. So while
some of you might be thinking "let's put a Farad across that
thing", actually that will make it less stable.

Paul

Mayayana wrote:
Thanks to everyone. After reading these posts I
looked into it further. Turns out my current machine also
has a SolidGear, with the same MB. And surge protection
is enabled. I never noticed. But while HWMonitor shows
almost no movement on my machine if I'm not
doing anything in particular, the new box
shows cpu vcore going between .8 and 1.3. It also
shows 3.3v and 12v values varying by at least
a few hundredths. The +12v reading, at 8+,
is varying by about .15. So there seems to be almost
a regular oscillation. The readings are also in
different ranges on the two machines, with the
same PS and MB, but different CPU. My current
machine shows almost constant 1.73 for cpu vcore
and shows +12v around 11.
I have no experience with this so I have no idea
what normal range is, but that seems like an awfully
lot of variation.

First, don't panic.

The purpose of the Hardware Monitor, is as a Warning.

The scaling math isn't always correct on the Hardware Monitor.
In many cases, it seems the 12V signal has the wrong scaling
resistor values recorded for it.

At one time, the Speedfan guy or the MBM5 guy, were collecting
info on a mobo by mobo basis. The scaling resistors were
figured out manually.

Now, the scale factors are stored in BIOS ACPI somewhere, and
this means a reduction in labor to get the correct values.

In any case, you can use a regular multimeter on "volts" to verify
the actual voltages.

One of the rules is to clip the black lead onto an I/O screw on
the back of the computer. This is easier said than done, because
most multimeters do not have good accessory kits. I bought
separate alligator clips from the RadioShack long ago, but
even then, my gear isn't all that good.

So I can use one of my extender wires, with alligator clips
on either end, to connect the multimeter black probe to
an I/O screw in the I/O plate area.

That leaves just the red probe to wave around, while you work.
Set the multimeter on 20V fullscale, and probe a few things.

A 1x4 Molex gives access to 5V and 12V. If the 12V is wavering,
that connector is all you need to check it.

A PSU with "real" SATA power, has an orange wire for
3.3V, but there's no safe way to probe that.

The main PSU connector, most times you can barely see
it, for the jumble of stuff in the computer case. In
theory, you can slip the red probe of the multimeter,
between the plastic connector shell, and where the wire
goes into the main connector. There is exposed metal for
each pin, and you can take a reading off that, even when
the connector is seated on the motherboard.

It's easier to probe the main connector, if the motherboard
is out on your test bench in the open.

I have two meters here, that have RS232 ports. And more
importantly, at least one of them, I was able to find a
schematic, and discover the meter has proper opto-isolation,
so you can probe circuits at voltage, without a flow path
forming through the multimeter and into the PC. The purpose
of having a multimeter with an RS232 port, is if you need
to log readings over a period of time.

So my best setup is:

1) Separate 9VDC wall wart with 9V battery terminals
on the end. The wall wart has no safety ground, and floats.
2) Multimeter with RS232. I power the multimeter with (1).
3) Connect the RS232 from the multimeter to the computer
doing the logging.
4) Record voltages to my hearts content, safe in the
knowledge that the meter won't run the battery down.

Note - meters have auto-power-off normally. If the meter
senses the RS232 is being used, it's supposed to disable that,
but you never know what will happen until the meter powers
off on you.

But that's only if you need a history or a log file.

If you pay $1000+ for a multimeter, some of those
models can make up to 1000 readings a second. Which might
be close enough, to do real "surge detection" :-)

If you need higher speed analysis than that, you need
an oscilloscope.

Even a few meter readings is better than nothing. As it
will expose any rubbish coming from the computer. An example
being, the 12.000V reading coming out of my new computer. Something
I need to check now, as it looks "fake".

Some computer hardware monitors read 15V on the 12V output,
and when you check with the multimeter, the power supply
is making 12.1 or so. A harbor freight meter isn't going to
be all that accurate, but at least it won't be reading 15V.

As long as you don't use the two multimeter probes in
close proximity, so they can short together, you won't be
seeing a "light show". My buddy at work one evening, shorted
out a 5V supply, and there was a burst of white light,
that threw his shadow on the ceiling. That's about as
good a reason as any, not to short out one of those
really powerful supplies. An IC leg burned off almost
instantly, and that's where the white light came from.
The 5V supply he was working on, had a 100A rating.
Like a kind of arc welder. I'm just glad he didn't bump
his head, getting away from the thing (he was partially
inside the equipment at the time).

Paul

Paul wrote:
Paul in Houston TX wrote:

There is a lot of contracting-out, in the PSU industry.
(snip)

Quite interesting. Thanks Paul.