Maker Pro
Maker Pro

Excessive current draw at UPS restart

C

Chirrp

Jan 1, 1970
0
Hi all,
we have a 5000VA UPS from APC, 5 years old (incl.batteries) which still
holds for various minutes at full load so we are keeping it. Downstream
from that one we have about 8 servers which load the UPS at about 80%.

The UPS is connected to a 16A line (220V) with a C charachteristic
curve. Normally the whole thing draws about 10.5A from there.

When we have a blackout and then the electricity comes back, the 16A
breaker trips, then after a few minutes the UPS batteries are flat and
the whole thing goes down.

I have tried pulling up the breaker when the batteries were totally
flat, and it would stay up for about 3.5 seconds (this is approximate
because I didn't really measure that with a stop-watch), then broke
again. I waited 10 seconds then I pulled it up again and that time it
stayed up. Reports from my colleagues are similar.

Now looking at the C charachteristics curve for the breakers,
(it is the one at the bottom of page 2, isn't it:
http://www.altechcorp.com/PDFS/CBC.PDF )
tripping after 3.5 seconds would mean that the UPS was drawing between
2.2 and 6 times the rated current (16A)! That's quite a lot of amperes!
Especially considering that after this transient, the UPS draws just
about 11.5A while powering the servers and simultaneously recharging.

Do you have a reasonable explanation for what is happening here?
It seems there is like a very large capacity loading...
Might this be a thing like "memory effect" of the old batteries which we
have in the UPS?

Unfortunately our panelmount voltmeter measures a moving average so is
not reactive enough to let us see how many amperes we are drawing in
those 3.5 seconds.

Thank you
 
A

Andrew Gabriel

Jan 1, 1970
0
Hi all,
we have a 5000VA UPS from APC, 5 years old (incl.batteries) which still
holds for various minutes at full load so we are keeping it. Downstream
from that one we have about 8 servers which load the UPS at about 80%.

The UPS is connected to a 16A line (220V) with a C charachteristic
curve. Normally the whole thing draws about 10.5A from there.

Well, that just doesn't add up.

80% of 5000VA = 4000VA = 18.2A @ 220V

And that's without any UPS overhead.
 
C

Chirrp

Jan 1, 1970
0
Andrew said:
Well, that just doesn't add up.

80% of 5000VA = 4000VA = 18.2A @ 220V

And that's without any UPS overhead.

Hmm I am not sure... let's see...
I say 80% based on the LEDs on the front which tell the load level. Of 5
LEDs, 4 are lit.
Oh ok, I'm sorry, I was mistaken: reading the manual it says that 4 LEDs
mean that the load is between 67% and 85%.

If it is 65% the things mostly add up: firstly consider that 5000VA is
about 3500W in UPS terms (30% less). I'm not sure of the difference
between VA and W but it's there and in practice you have to disregard
the VA value and just consider it a 3500W UPS.

now:
3500/220*0.67 = 10.65A

So at 65% it does add up. After saying this it might really be a bit
higher than 65%, considering that the batteries are old and the UPS
might know that. Maybe this load is currently a 70% or even 80% on the
old batteries while it would have been an e.g. 62% load if the batteries
were new...
 
P

Palindrome

Jan 1, 1970
0
Chirrp said:
Hmm I am not sure... let's see...
I say 80% based on the LEDs on the front which tell the load level. Of 5
LEDs, 4 are lit.
Oh ok, I'm sorry, I was mistaken: reading the manual it says that 4 LEDs
mean that the load is between 67% and 85%.

If it is 65% the things mostly add up: firstly consider that 5000VA is
about 3500W in UPS terms (30% less). I'm not sure of the difference
between VA and W but it's there and in practice you have to disregard
the VA value and just consider it a 3500W UPS.

now:
3500/220*0.67 = 10.65A

So at 65% it does add up. After saying this it might really be a bit
higher than 65%, considering that the batteries are old and the UPS
might know that. Maybe this load is currently a 70% or even 80% on the
old batteries while it would have been an e.g. 62% load if the batteries
were new...


The answer is, as ever, to RTFM. As an example, this 5000VA 230v UPS
requires a 25A supply breaker:
http://sturgeon.apcc.com/techref.nsf/partnum/990-7032A/$FILE/990-7032A_REV03_eng.pdf

What the UPS "knows" about the battery age isn't relevant. What it does
on resumption of supply is relevant and that will be determined by the
inverter type. Which you haven't mentioned. If it is a True sine wave,
genuine online type, then it has to start by converting the input supply
to DC. That initial start up will be the same whatever the load is - it
has been designed to take over the full rating capacity of the UPS (and
a bit more). But the result *will* be a very large surge current. Other
UPS types, that simply switch over to the incoming supply using a relay,
are nowhere near as demanding, or as expensive.
 
A

Andrew Gabriel

Jan 1, 1970
0
Hmm I am not sure... let's see...
I say 80% based on the LEDs on the front which tell the load level. Of 5
LEDs, 4 are lit.
Oh ok, I'm sorry, I was mistaken: reading the manual it says that 4 LEDs
mean that the load is between 67% and 85%.

If it is 65% the things mostly add up: firstly consider that 5000VA is
about 3500W in UPS terms (30% less). I'm not sure of the difference
between VA and W but it's there and in practice you have to disregard
the VA value and just consider it a 3500W UPS.

Doesn't really work like that. You need to measure the current
with a clamp meter. You don't know what the power factor of the
load is, and it's irrelevant in this case anyway.

For the current draw and breaker ratings, you only care about the VA.
(For cooling the datacanter, you only care about the W.)

Even if it is at the bottom end, that's still

65% of 5000VA = 3250VA = 14.8A @ 220V

Add the UPS overhead and a starting surge, and it's not
surprsing you're tripping a 16A breaker.

The UPS instructions probably say what size circuit it should
be fed from, and it's going to be more than 16A (probably 32A).
 
C

Chirrp

Jan 1, 1970
0
Palindrome said:
The answer is, as ever, to RTFM. As an example, this 5000VA 230v UPS
requires a 25A supply breaker:
http://sturgeon.apcc.com/techref.nsf/partnum/990-7032A/$FILE/990-7032A_REV03_eng.pdf

Yes you are right, we noticed that, too late :-(

However as I wrote, after the first 10 seconds or so, the current draw
is very reasonable, like 1A higher than when not charging the batteries.
What the UPS "knows" about the battery age isn't relevant. What it does
on resumption of supply is relevant and that will be determined by the
inverter type. Which you haven't mentioned.

Because I don't know. It's not written in the manual! (the one which you
linked)
If it is a True sine wave,
genuine online type, then it has to start by converting the input supply
to DC. That initial start up will be the same whatever the load is - it
has been designed to take over the full rating capacity of the UPS (and
a bit more).

Oh thanks now at least I understand why it does that!

But still I have to say it really wasn't a good engineering AFAICS:
After the first 10 seconds the UPS is capable to determine the real load
and draw current only for that load +1A for charging the batteries.
Couldn't they design it so that it estimated the actual load value
BEFORE starting drawing current??

And then... no wait a sec... it still doesn't add up:
the C16A curve tripped at about 3.5 seconds which means between 2.2x and
6x the 16A so it is between 35A and 96A, still way higher than the 25A
required by the UPS as per its specs.
Should it have really drawn just 25A at startup it would have tripped
after 10secs--to--4minutes, which it certainly didn't.
Is the C curve in this PDF correct for any brand of breaker?
http://www.altechcorp.com/PDFS/CBC.PDF
our breaker is a Merlin Green one.

Thanks
 
C

Chirrp

Jan 1, 1970
0
Andrew said:
Doesn't really work like that. You need to measure the current
with a clamp meter. You don't know what the power factor of the
load is, and it's irrelevant in this case anyway.

For the current draw and breaker ratings, you only care about the VA.
(For cooling the datacanter, you only care about the W.)

Even if it is at the bottom end, that's still

65% of 5000VA = 3250VA = 14.8A @ 220V

How come the panelmount ampermeter tells me 10.5A then??
Are you 100% sure of what you say?
hmm we might have a malfunctioning UPS or ampermeter then.

More likely the UPS... with a higher load I have seen that ampermeter
measuring upto 13A I think... if the ampermeter was really scaled 40%
down the 16A breaker would have tripped... so I would say the ampermeter
is more likely to be correct, the UPS must be malfunctioning...
 
| Chirrp wrote:
|> Andrew Gabriel wrote:
|>> Well, that just doesn't add up.
|>>
|>> 80% of 5000VA = 4000VA = 18.2A @ 220V
|>>
|>> And that's without any UPS overhead.
|>
|> Hmm I am not sure... let's see...
|> I say 80% based on the LEDs on the front which tell the load level. Of 5
|> LEDs, 4 are lit.
|> Oh ok, I'm sorry, I was mistaken: reading the manual it says that 4 LEDs
|> mean that the load is between 67% and 85%.
|>
|> If it is 65% the things mostly add up: firstly consider that 5000VA is
|> about 3500W in UPS terms (30% less). I'm not sure of the difference
|> between VA and W but it's there and in practice you have to disregard
|> the VA value and just consider it a 3500W UPS.
|>
|> now:
|> 3500/220*0.67 = 10.65A
|>
|> So at 65% it does add up. After saying this it might really be a bit
|> higher than 65%, considering that the batteries are old and the UPS
|> might know that. Maybe this load is currently a 70% or even 80% on the
|> old batteries while it would have been an e.g. 62% load if the batteries
|> were new...
|
|
| The answer is, as ever, to RTFM. As an example, this 5000VA 230v UPS
| requires a 25A supply breaker:
| http://sturgeon.apcc.com/techref.nsf/partnum/990-7032A/$FILE/990-7032A_REV03_eng.pdf
|
| What the UPS "knows" about the battery age isn't relevant. What it does
| on resumption of supply is relevant and that will be determined by the
| inverter type. Which you haven't mentioned. If it is a True sine wave,
| genuine online type, then it has to start by converting the input supply
| to DC. That initial start up will be the same whatever the load is - it
| has been designed to take over the full rating capacity of the UPS (and
| a bit more). But the result *will* be a very large surge current. Other
| UPS types, that simply switch over to the incoming supply using a relay,
| are nowhere near as demanding, or as expensive.

The inverter should be seeing continuous DC when the AC comes back on. So
why would it need a surge of current at that time? I can understand it would
need this if the battery is already drained, or if the unit has switched off.
But in the brief power fail case (before the breaker trip that would cause
the battery to drain all the way), the inverter should see DC uninterrupted.

Maybe it's the _charger_ that is pulling the surge to fill up its caps?

I agree without even reading the specs, that this UPS needs at least a 25A
rated circuit. In the USA that would be a 30A breaker and a NEMA [L]6-30 plug.
 
| Couldn't they design it so that it estimated the actual load value
| BEFORE starting drawing current??

The possible reason I can see for a surge is the charger starting up and
filling its caps. In theory they could have made a slow start charger
that would fill the caps slow before actually having the charger output
any DC, whether the UPS is off (and hence slow to start) or on (slow to
switch back to AC as long as the battery doesn't fail in that short time
window).


| And then... no wait a sec... it still doesn't add up:
| the C16A curve tripped at about 3.5 seconds which means between 2.2x and
| 6x the 16A so it is between 35A and 96A, still way higher than the 25A
| required by the UPS as per its specs.
| Should it have really drawn just 25A at startup it would have tripped
| after 10secs--to--4minutes, which it certainly didn't.
| Is the C curve in this PDF correct for any brand of breaker?
| http://www.altechcorp.com/PDFS/CBC.PDF
| our breaker is a Merlin Green one.

Another possibility is the breaker has a loose terminal, heating it more than
would normally happen, and it's on the border line of a thermal trip. That
heat would be mostly retained across a short time frame of AC outage so it
could trip again soon.
 
P

Palindrome

Jan 1, 1970
0
Chirrp said:
Yes you are right, we noticed that, too late :-(

However as I wrote, after the first 10 seconds or so, the current draw
is very reasonable, like 1A higher than when not charging the batteries.


Because I don't know. It's not written in the manual! (the one which you
linked)

It's a Line Interactive, low distortion sine wave, IIRC.
Oh thanks now at least I understand why it does that!

But still I have to say it really wasn't a good engineering AFAICS:
After the first 10 seconds the UPS is capable to determine the real load
and draw current only for that load +1A for charging the batteries.
Couldn't they design it so that it estimated the actual load value
BEFORE starting drawing current??

Not really. It has been designed to work from a supply circuit with a
25A breaker. So the designer had no reason to limit transition current
to suit a 16A breaker. That would have either compromised or complicated
the design, or both. As soon as the input supply has been found to be
stable and within limits, it's in everyone's interest (other than yours)
for the transition from battery to be done in the least time possible.
And then... no wait a sec... it still doesn't add up:
the C16A curve tripped at about 3.5 seconds which means between 2.2x and
6x the 16A so it is between 35A and 96A, still way higher than the 25A
required by the UPS as per its specs.
Should it have really drawn just 25A at startup it would have tripped
after 10secs--to--4minutes, which it certainly didn't.
Is the C curve in this PDF correct for any brand of breaker?
http://www.altechcorp.com/PDFS/CBC.PDF
our breaker is a Merlin Green one.
No, because those time figures relate to ac currents to a constant
load. The load the UPS presents will not be constant. Initially, during
monitoring, a very low current will be drawn. Once the decision has been
made to switch to mains supply, everything has to happen very, very
quickly - because an inline UPS uses the same circuitry for running on
battery as it uses for running on mains. The initial (surge) current as
the supply switches to mains will be very high. If your supply impedance
is high (eg the lights tend to dim when a big load comes on) that will
make it even worse, as the UPS will compensate for the decreased voltage
by drawing even more current.

The designer merely had to make sure that his design wouldn't trip a 25A
breaker..
 
K

Ken

Jan 1, 1970
0
Hi all,
we have a 5000VA UPS from APC, 5 years old (incl.batteries) which still
holds for various minutes at full load so we are keeping it. Downstream
from that one we have about 8 servers which load the UPS at about 80%.

The UPS is connected to a 16A line (220V) with a C charachteristic
curve. Normally the whole thing draws about 10.5A from there.

When we have a blackout and then the electricity comes back, the 16A
breaker trips, then after a few minutes the UPS batteries are flat and
the whole thing goes down.

I have tried pulling up the breaker when the batteries were totally
flat, and it would stay up for about 3.5 seconds (this is approximate
because I didn't really measure that with a stop-watch), then broke
again. I waited 10 seconds then I pulled it up again and that time it
stayed up. Reports from my colleagues are similar.

Now looking at the C charachteristics curve for the breakers,
(it is the one at the bottom of page 2, isn't it:
http://www.altechcorp.com/PDFS/CBC.PDF )
tripping after 3.5 seconds would mean that the UPS was drawing between
2.2 and 6 times the rated current (16A)! That's quite a lot of amperes!
Especially considering that after this transient, the UPS draws just
about 11.5A while powering the servers and simultaneously recharging.

Do you have a reasonable explanation for what is happening here?
It seems there is like a very large capacity loading...
Might this be a thing like "memory effect" of the old batteries which we
have in the UPS?

Unfortunately our panelmount voltmeter measures a moving average so is
not reactive enough to let us see how many amperes we are drawing in
those 3.5 seconds.

Thank you

Bad breaker?
 
A

Andrew Gabriel

Jan 1, 1970
0
The possible reason I can see for a surge is the charger starting up and
filling its caps. In theory they could have made a slow start charger
that would fill the caps slow before actually having the charger output
any DC, whether the UPS is off (and hence slow to start) or on (slow to
switch back to AC as long as the battery doesn't fail in that short time
window).

IME of APC SmartUPS, it's actually very nice to the mains supply
when restored. It draws almost nothing for some period (probably
waiting to see it its going to stay on, and it syncs up its
inverter output to be in sync with the mains) before cutting
back to use the mains supply. Makes quite a useful contrast with
all the other devices on the supply which normally yank a good
surge without any consideration for anything else.
 
| In article <[email protected]>,
| [email protected] writes:
|> The possible reason I can see for a surge is the charger starting up and
|> filling its caps. In theory they could have made a slow start charger
|> that would fill the caps slow before actually having the charger output
|> any DC, whether the UPS is off (and hence slow to start) or on (slow to
|> switch back to AC as long as the battery doesn't fail in that short time
|> window).
|
| IME of APC SmartUPS, it's actually very nice to the mains supply
| when restored. It draws almost nothing for some period (probably
| waiting to see it its going to stay on, and it syncs up its
| inverter output to be in sync with the mains) before cutting
| back to use the mains supply. Makes quite a useful contrast with
| all the other devices on the supply which normally yank a good
| surge without any consideration for anything else.

Sounds like a reasonably good model. It could slowly charge the caps in the
charger during the examination of AC supply stability. That's assuming the
caps don't just run in parallel with the battery (I don't know if any UPS
does that or not).
 
D

daestrom

Jan 1, 1970
0
Chirrp said:
How come the panelmount ampermeter tells me 10.5A then??
Are you 100% sure of what you say?
hmm we might have a malfunctioning UPS or ampermeter then.

More likely the UPS... with a higher load I have seen that ampermeter
measuring upto 13A I think... if the ampermeter was really scaled 40%
down the 16A breaker would have tripped... so I would say the
ampermeter is more likely to be correct, the UPS must be
malfunctioning...

How sure are you that it is 220V and not 240V?? 220V is somewhat rare
nowadays.

10.5A * 240V = 2520 VA which is 50.4% of the 5000VA rating.

As far as 'malfunctioning', the only thing I would question is the LED
readout. It's only a crude bar-graph and most likely is not calibrated in
any real sense of the word.

As far as the panel ammeter, if it is a simple ammeter (D'Arsonval
movement), there's a good chance it isn't reading perfectly either. The
input stage of the UPS that converts the incoming AC to DC has a fair amount
of harmonics in it. D'Aarsonval ammeters respond to *average* current flow
and have their meter face marked of in RMS values assuming the current is a
simple sine-wave with little harmonic distortion. Since the input has
harmonics, the markings on the meter are going to be incorrect to read the
actual RMS value.

So many things... But since the manufacturer requires a 25A breaker and you
have only a 16A, from this distance it sounds like that is about the only
thing wrong.

daestrom
 
J

John G

Jan 1, 1970
0
Chirrp said:
Hi all,
we have a 5000VA UPS from APC, 5 years old (incl.batteries) which still
holds for various minutes at full load so we are keeping it. Downstream
from that one we have about 8 servers which load the UPS at about 80%.

The UPS is connected to a 16A line (220V) with a C charachteristic curve.
Normally the whole thing draws about 10.5A from there.

When we have a blackout and then the electricity comes back, the 16A
breaker trips, then after a few minutes the UPS batteries are flat and the
whole thing goes down.

I have tried pulling up the breaker when the batteries were totally flat,
and it would stay up for about 3.5 seconds (this is approximate because I
didn't really measure that with a stop-watch), then broke again. I waited
10 seconds then I pulled it up again and that time it stayed up. Reports
from my colleagues are similar.

Now looking at the C charachteristics curve for the breakers,
(it is the one at the bottom of page 2, isn't it:
http://www.altechcorp.com/PDFS/CBC.PDF )
tripping after 3.5 seconds would mean that the UPS was drawing between 2.2
and 6 times the rated current (16A)! That's quite a lot of amperes!
Especially considering that after this transient, the UPS draws just about
11.5A while powering the servers and simultaneously recharging.

Do you have a reasonable explanation for what is happening here?
It seems there is like a very large capacity loading...
Might this be a thing like "memory effect" of the old batteries which we
have in the UPS?

Unfortunately our panelmount voltmeter measures a moving average so is not
reactive enough to let us see how many amperes we are drawing in those 3.5
seconds.

Thank you
If this set up worked for the first x years and is now failing with old
batteries my guess would be that the battery voltage is below the design
point and the chargers tries tooo hard to recover as soon as the mains
returns.

Even if the charger is inteligent it may not handle batteries that are old
and hence have a terminal voltage outside the design paramerters after a
time driving the load.

John G.
 
Top