UPS Eaton Nova AVR 625 - Not so fast...

[hgg]

Hi everybody,

I have a problem with my computer UPS.  Its the Eaton Nova AVR 625.
https://powerquality.eaton.com/SEA/Products-services/Backup-Power-UPS/Nova-AVR-UPS/Nova-AVR-specs.asp?CX=35

It was working alright for the last couple of years but now has the following problem.
It works fine in a blackout when mains power is completely off but it cannot keep up
during very fast brownouts.  If power goes out for a fraction of a second, then the PC
switches off and reboots.  I have replaced the battery but the same thing happens.

It is sold as an online line interactive UPS but I am not sure about that. 
Can you guess that from the pictures?

The mainboard:


The main transformer:


A couple of fuses under the heatsink:


One of the CEP83A3 N-Channel Mosfets:


One of the two LM324 opamps:


The main MC68HC908 8bit Microcontroller:


The USB control board:


...and finally some musical relays...:


Do you think that this problem is caused by the relays?  (..being Chinese quality parts..)
There are no big capacitors on the board. I have two of these UPSs and they both have
started doing the same thing.   

Any ideas?

Thank you for you help.
George.

[mij59]

May there is a problem with the power supply of your computer, have you tried the ups with an other computer ?

[hgg]

Both the PC and the connected LCD display switch off together.
I have two of these UPSs connected to different computers and they both have started doing the same thing.

[firewalker]

Check all the electrolytic caps.

Alexander.

[hgg]

I will measure them later when I have some time, but at least visually they all looked fine.

[mij59]

Please post a picture of the copper side of the mainboard, first check for bad solder joints, connectors etc.
On the usb control board is a bad solder joint, but don't think that is the fault.

[hgg]

I suppose you mean the one above the label C16.

[mij59]

I suppose you mean the one above the label C16.

Yes looks like the pins of a voltage regulator.

[hgg]

I've just checked all the electrolytics. 
The capacitance values are close (lower values) and they have the following ESR values:

a) 1000uF   16V   0.034 Ohm
b)     47uF   50V   0.714 Ohm
c)   120uF   16V   0.572 Ohm
d)     47uF   50V   0.651 Ohm
e)     47uF   50V   0.600 Ohm
f)      22uF   50V   0.602 Ohm
g)    10uF    50V   2.020 Ohm
h)     47uF   50V   0.946 Ohm

Do you think they need replacement ?

I have also checked all the solder joints and they seem fine.  Even the one on the USB board.
It just has less solder.

Why does the UPS works during a blackout but not during a brownout?

Below is a photo of the underside of the board:

[krish2487]

From the ppics of the transformer, If the farther side has more than 3 wires of the same gauge ( cant see clearly in the pics, it almost certainly is a line interactive UPS. The multiple wires are to indicate different taps to switch between to regulate output voltage.


caps (g) and (h) look suspiciously bad. All the more of they were not a single cap and multiple in parallel. I d replace them for a good measure.


The brownout can happen due to several reasons.


1. The line brownout transitions are extremely fast and narrow in duration (high slew rate) - confusing the processor to take a decision.


2. The brownout voltage itself in on the cusp inbetween two voltage taps and varying by a small amount, leading the processor to go nuts trying to decide between which taps.


3. The processor has gone nuts. (unlikely because as you reported the UPS works fine in black outs)


4. Most often the stepdown and sense analog circuitry will have a couple of op amps, it is also possible that one of the op amps and /or associated circuitry is the cause (This seems more likely, since I dont see a sense tap on the transformer, meaning non isolated voltage sensing.) It is possible that during one of the brownout or spikes one or more of the components in the circuit path were affected.


I d start by checking and replacing the op amps.


PS: R182 and R67 SMD resistors in the previous pic , towards the top seem burnt. Is that the voltage divider section for AC??.

[mij59]

A lot of the solder joints on the main board are not that good, there's a small darker band on the solder indicating the connection is bad, e.g. below C49.
Resolder the pins of connectors, power transistor etc.

[hgg]

krish2487 hi,
Yes, the transformer has 3 wires, so its a line interactive UPS.
It seemed strange to me the fact that there are no big smoothing capacitors in the circuit,
but then I looked at the output of the UPS...

With no load:


With a 15" LCD monitor connected:


Output Ripple with light loading...:


I am not sure if this output is any good for any of your PC components, even for 5 minutes...
I am starting to think that a pure sinewave UPS is a better idea.

4. Most often the stepdown and sense analog circuitry will have a couple of op amps, it is also possible that one of the op amps and /or associated circuitry is the cause (This seems more likely, since I dont see a sense tap on the transformer, meaning non isolated voltage sensing.) It is possible that during one of the brownout or spikes one or more of the components in the circuit path were affected.

In the first picture immediately right of the heatsink you can see a Cosmo 1010 optocoupler.
Two more are on the other side of USB control board.  (The unit has RJ45 protection as well.)
So maybe its protected from such spikes.

I don't like the thought of replacing the opamps.  They are heavily populated with tiny SMD devils
in a very small space....



What can I check on them without removing them?
I will replace the caps just to be sure and I have just removed a relay to check its switching speed.

PS: R182 and R67 SMD resistors in the previous pic , towards the top seem burnt. Is that the voltage divider section for AC??.

Sorry, this was not a good photo. 
The red stuff is the glue that they put to hold in place the SMD parts.

A lot of the solder joints on the main board are not that good, there's a small darker band on the solder indicating the connection is bad, e.g. below C49.

Same here.  Its the flux residue.

[mij59]

A lot of the solder joints on the main board are not that good, there's a small darker band on the solder indicating the connection is bad, e.g. below C49.

Same here.  Its the flux residue.

Its no the flux residue thats creating the darker ring, the ring is caused by missing solder.
Take a look at the 3 pins to right of the flux residue, the solder should have flown up on the pin, the solder doesn't stick very well to the pin. 

[hgg]

Although the soldering is not the best, I've tested the connections and they are ok.
I will reflow all the pins that are in a similar condition, just to be on the safe side...

I have also tested two of the relays and they have a switching time of 4.8ms which is
inside the maximum 10ms specified in the datasheet.  Maybe they should be faster (?)



I am going to buy some new capacitors and relays and test them.

[mij59]

I have also tested two of the relays and they have a switching time of 4.8ms which is
inside the maximum 10ms specified in the datasheet.  Maybe they should be faster (?)



I am going to buy some new capacitors and relays and test them.

Test the NO and NC contacts of the relay with an Ohm meter.
Please test how long it takes for the ups to supply the computer when mains fails.

[krish2487]

Meh, the MSW UPS are just fine for computer load.
High harmonic content and all but should not affect the performance.


The ringing too is just fine, should not be a problem. Its more likely the scope probe introducing a ringing, more than the circuit and load.


Neither is the relay switching time. 4.8 ms is ~ 1/2 of 10 ms which itself is a half cycle duration for 50 Hz.
It is pretty conservative switching time. I wouldnt worry about the relay switching time.


Which leads me to another question, is the brownout behavior consistent whenever it happens??  or is it more of a random appearance??
If it is random then I would not worry about it.
Freescale should have an application note about the UPS. It wont give you exact circuit (This is roughly a 2 decade + old circuit ) but it should provide you with a more informed idea of where exactly you should start board level debugging.


I seriously doubt you will get any kind of schematics/manuals for that.

[hgg]

Test the NO and NC contacts of the relay with an Ohm meter.

I measured two of the relays and the NO & NC resistance is 0.01 Ohms, so I guess they are fine.
I will measure the other 3 as well, but it looks like that the problem is not from the relays.

Please test how long it takes for the ups to supply the computer when mains fails.

I will try that as well, but the UPS works fine with blackouts.

Which leads me to another question, is the brownout behavior consistent whenever it happens??  or is it more of a random appearance??

No, its more like a random appearance.  I guess it depends on how fast the brownout is.

If it is random then I would not worry about it.

What do you mean you wouldn't worry?    The PC reboots off while working...

[oldway]

I really wonder how it would be possible it should be a line interactive UPS.
When mains fails, it is like a short circuit because all the loads are still on.
You need to limit the short circuit current feeded by inverter by a big inductance in serie with the mains.
This inductance is generaly produced by a magnetic shunt in the transformer.
I did not see such a magnetic shunt in the transformer.
Output waveform of inverter must also be sinus in an interactive UPS.

For me, that's not an interactive UPS, but probably an off line UPS with inverter running all the time.
Switching between mains and inverter is done by relais.
Load is normaly feeded by mains and switched on inverter in case of mains failure.

Check this, because you can't make a diagnose without be sure first of the principle of working of your UPS.

[krish2487]

@oldway
you are right. I have been blind.
It is not a Line interactive UPS.
It is a push pushpull output stage. Hence the square wave output.

[hgg]

So, my suspicions were correct.

Do you think that it deserves some more debugging time, or shall I throw it in the bin and keep the parts...?

I did not see such a magnetic shunt in the transformer.

Can you find an internet photo of a similar configuration so that I understand better what you mean?
Thanks.

[mij59]

Do you think that it deserves some more debugging time, or shall I throw it in the bin and keep the parts...?

First resolder the main board, may be this solves the problem.
Just wondering how do your check there's a brownout ? may be the computer is shut down by a virus.

[oldway]

So, my suspicions were correct.

Do you think that it deserves some more debugging time, or shall I throw it in the bin and keep the parts...?

I did not see such a magnetic shunt in the transformer.

Can you find an internet photo of a similar configuration so that I understand better what you mean?
Thanks.

@krish2487: Do not forgot that the transfer time is not only the switching time of the relay but you have to add the time of detection of mains failure.
The first thing I would do is to check if the inverter is sincronized with mains or not.

It was working alright for the last couple of years but now has the following problem.
It works fine in a blackout when mains power is completely off but it cannot keep up
during very fast brownouts.  If power goes out for a fraction of a second, then the PC
switches off and reboots.  I have replaced the battery but the same thing happens.

You are not sure that what happens with mains voltage is the same than what was happening in the past.
You can't be sure of that something has changed in your UPS.
What is clear for me is that it is a low cost, low technology UPS and you can't wait for good protection with such an UPS.
The only secure technology is on line UPS.

[hgg]

Just wondering how do your check there's a brownout ?

From an incandescent light bulb.

oldway thank you for the pictures. 
From what I can see there is no magnetic shunt on this transformer.

The first thing I would do is to check if the inverter is sincronized with mains or not.

Any hint on how to do that?...

[oldway]

From what I can see there is no magnetic shunt on this transformer.

You must have two separate windings side by side with magnetics shunts between them as on the picture.

 

Any hint on how to do that?...

Use a low voltage transformer to pickup mains voltage waveform for one channel of your scope and measure the inverter waveform on the second one as you did...or even more easy, sincronise your scope  on line and see if inverter waveform is stable on the screen.

[hgg]

You must have two separate windings side by side with magnetics shunts between them as on the picture.

I meant there is no shunt in my transformer.

Use a low voltage transformer to pickup mains voltage waveform for one channel of your scope and measure the inverter waveform on the second one as you did...or even more easy, sincronise your scope  on line and see if inverter waveform is stable on the screen.

Hmmm, nice.  I might do that.  Thanks!

So, do you think that this definitely is not a line interactive ups?
Its sold as one.  I guess it interacts with the mains to a degree...   

[oldway]

In an a line interactive UPS, inverter works in // with line, with only an inductance between inverter and line.
Inverter works as battery charger by controling phase angle between mains and inverter.
Inverter waveform MUST be sinus in a line interactive UPS.
That's definitively not an interactive UPS.

[bktemp]

In an a line interactive UPS, inverter works in // with line, with only an inductance between inverter and line.
Inverter works as battery charger by controling phase angle between mains and inverter.
Inverter waveform MUST be sinus in a line interactive UPS.
That's definitively not an interactive UPS.

No, that is wrong.

Line interactive means the ups can adjust the output voltage by using multiple taps on the primary winding working as an autotransformer. As long as a voltage is present, the ups tries to switch the tap to keep the output voltage regulated. Only when the voltage fails completely or there are not enough taps too keep the voltage regulated it switches over to the battery powered inverter.

See here:
https://en.wikipedia.org/wiki/Uninterruptible_power_supply#Line-interactive

[oldway]

If your only information you have is Wikipedia, you will not go anywhere 
http://www.baservices.nl/downloads/APC%20whitepaper%20soorten%20ups.pdf
There may have a taps changer added to regulate AC output voltage because output voltage of an on line interactive UPS without taps changer is the same as the line voltage.
This does not change nothing of what I wrote.

Wikipedia is wrong: there is no battery charger (rectification is made by the free wheeling diodes of the inverter) and current control is make by varying phase between inverter and mains.
There must be an inductance between mains and inverter as both have low impedance, battery has also low internal resistance and a very little phase difference would give a very large charging current.
Furthermore, there is no output switching between mains and inverter in an interactive UPS, only a way to switch the mains off if there is a line failure.
And you can't let work in // an inverter with non sinus wave with sinus line voltage.

[bktemp]

http://www.baservices.nl/downloads/APC%20whitepaper%20soorten%20ups.pdf
There may have a taps changer added to regulate AC output voltage because output voltage of an on line interactive UPS without taps changer is the same as the line voltage.
This does not change nothing of what I wrote.

Can you show me any line interactive ups without the tap changer? I have never seen one.

Wikipedia is wrong: there is no battery charger (rectification is made by the free wheeling diodes of the inverter) and current control is make by varying phase between inverter and mains.

Wikipedia just shows a very simplified block diagram.
APC uses the main transformer for charging to reduce the size and cost, but it is not mandatory for a line interactive ups!

There must be an inductance between mains and inverter as both have low impedance, battery has also low internal resistance and a very little phase difference would give a very large charging current.

I have several line interactive ups from APC and none of them has a visible magnet shunt.
To be able to generate a high enough output voltage even under full load, the transformer produces a much lower voltage than the battery voltage. So there is no current flowing into the batteries even if the transformer is connected to mains.
To charge the batteries they use the stray inductance in the transformer as a boost converter to increase the voltage to the batteries.

And you can't let work in // an inverter with non sinus wave with sinus line voltage.

You are wrong again, here is only one example:
http://www.apc.com/products/resource/include/techspec_index.cfm?base_sku=BR550GI&total_watts=600

Topology: Line Interactive
Waveform Type: Stepped approximation to a sinewave

[oldway]

Can you show me any line interactive ups without the tap changer? I have never seen one.

The firm SAB NIFE where I worked as a development engineer was making interactive ups from 5 to 20kVA's with no taps changer.

Note that most off-line and on-line UPS does not have a voltage stabilizer in the mains.
I worked also by AEG, one of the greatest industrial UPS manufacturer in Europe, and I have NEVER see an installed on line UPS with voltage stabilisator in the main circuit.

Wikipedia just shows a very simplified block diagram.
APC uses the main transformer for charging to reduce the size and cost, but it is not mandatory for a line interactive ups!

You simply did not understood the principles of a line interactive UPS.
There is a good reason for this: those informations are not available on Internet.
Customers does not give schematics and diagrams of their UPS on Internet....You have to work in these firms to have such informations.

I have several line interactive ups from APC and none of them has a visible magnet shunt.

Transformer with separate windings side by side give an short circuit voltage of about 20%
Magnetic shunts are used to increase this value above 20%.
It's possible that the design is such that 20% was enough.

To be able to generate a high enough output voltage even under full load, the transformer produces a much lower voltage than the battery voltage. So there is no current flowing into the batteries even if the transformer is connected to mains.

That's right, charging current is controlled by changing phase difference between mains voltage and inverter.

To charge the batteries they use the stray inductance in the transformer as a boost converter to increase the voltage to the batteries.

Wrong: what you say is impossible....energy from stray inductance of transformer is absorbed by RC suppressor of the semiconductors and by the load.
The PWM inverter is bidirectionnal and can supply energy to the battery from the main if not perfectly in phase with the main. That's what is used to charge the battery without the need of a battery charger.

You are wrong again, here is only one example:
http://www.apc.com/products/resource/include/techspec_index.cfm?base_sku=BR550GI&total_watts=600

Without schematics and diagrams to see what really has be done, it is not possible to answer this.
One of the principles of a line interactive UPS is just to be able to correct a distorted mains sine waveform, this why this name interactive, as it can to be interactive with mains voltage and correct it before feeding the load...Your statement is almost against this principle of working of a line interactive UPS...

[bktemp]

Note that most off-line and on-line UPS does not have a voltage stabilizer in the mains.
I worked also by AEG, one of the greatest industrial UPS manufacturer in Europe, and I have NEVER see an installed on line UPS with voltage stabilisator in the main circuit.

Of course not. An online ups works completely different and therefore does not need an additional stabiliser, because it creates a completely new sine wave.

To charge the batteries they use the stray inductance in the transformer as a boost converter to increase the voltage to the batteries.

Wrong: what you say is impossible....energy from stray inductance of transformer is absorbed by RC suppressor of the semiconductors and by the load.
The PWM inverter is bidirectionnal and can supply energy to the battery from the main if not perfectly in phase with the main. That's what is used to charge the battery without the need of a battery charger.

Someone achieved the impossible:
http://www.google.com/patents/US5302858

[hgg]

Two of the relays had a NC 0.25 Ohm resistance instead of the 0.01 Ohm of the other ones,
so I've replaced them.  I have also resoldered most of the through hole components.
I did not replace any capacitors because they were fine.

The result:  190VAC without any load instead of the 240VAC...   
Luckily, no explosions.   
UPS is going exactly where it was supposed to be.  The trash bin.

I am going to buy a new one.
Anyway, thank you very much for your help!
George.

[mij59]

Two of the relays had a NC 0.25 Ohm resistance instead of the 0.01 Ohm of the other ones,
so I've replaced them.  I have also resoldered most of the through hole components.
I did not replace any capacitors because they were fine.

The result:  190VAC without any load instead of the 240VAC...   
Luckily, no explosions.   
UPS is going exactly where it was supposed to be.  The trash bin.

I am going to buy a new one.
Anyway, thank you very much for your help!
George.

Most computer power supplies can handle even lower input voltages, so it should not be a problem, may be the multimeter reading is off due to the square wave form.
Well some you win, some you lose.

[oldway]

Of course not. An online ups works completely different and therefore does not need an additional stabiliser, because it creates a completely new sine wave

And what about the situation when load is feeded by mains voltage ?
If customer agree with mains voltage variations in this case, why sould he not agree with the same variations of a line interactive UPS without taps changer ?

Someone achieved the impossible:
http://www.google.com/patents/US5302858

A patent never proove it works 
There are thousands of patents who never worked...

@hgg: buy an on line UPS with sine output, that's the best technology even if it is more expensive and less efficient.

[hgg]

Most computer power supplies can handle even lower input voltages, so it should not be a problem, may be the multimeter reading is off due to the square wave form.

You made me think again.
Before the repair I measured the output voltage with the UT61E and showed 250VAC,
but after the repair I used the UT-136C which is not true RMS.  That's why the 190VAC.   

Now it measures 240VAC spot on. 
I think I will try it with a load.

@hgg: buy an on line UPS with sine output, that's the best technology even if it is more expensive and less efficient.

Indeed, that's exactly what I was planning to do.
I will try first to see if this one works any better and move from there.
A variac would have been handy for testing the brownout behavior but I don't have any.

[bktemp]

Someone achieved the impossible:
http://www.google.com/patents/US5302858

A patent never proove it works 
There are thousands of patents who never worked...

Watch this video:

[hgg]

That is what I was watching yesterday... 

I've plugged it in again and started to play with the mains cable in order to simulate
fast mains transitions and after some tries the PC rebooted.  The problem continues.

I would have to buy a new one.
Anyway, at least I've tried and thanks again for your help.

It would be interesting to see a repair video from Dave, but shipping it to Australia...   

[bktemp]

I've plugged it in again and started to play with the mains cable in order to simulate
fast mains transitions and after some tries the PC rebooted.  The problem continues.

Did you try to measure both the input and the output voltage on the ups simultaneously as oldway has suggested? It should be easy to trigger on those fast transients.
I am still not fully convinced the ups is faulty.

[hgg]

No, I haven't.  The less you probe mains, the better... 

If the UPS was ok, the PC should have stayed on.
Maybe the power supply of the PC does not like the crappy waveform which is basically a simple square wave.
Not even a staircase approximation.  The same thing started to happen with a different PC (new PSU) connected
to an identical UPS in a different workplace.

(I think that a modification to make it function like a Schmitt trigger would have solved the problem.
For example if the controller senses a drop in voltage and has to connect the inverter to the output,
it should also add a small delay like 5 seconds for example, even if the mains voltage has recovered.)

[bktemp]

No need to probe mains directly. You can use 2 transformers to get the voltage down. It must not be precise, but at the moment we have absolutely no information about the timing when switching to battery mode. If it takes less than 10ms, then the ups is moste likely fine and can be used for something else.
Most power supplies are rated for at least 20ms hold up time under full load, so when not fully loaded a pc should run for quite a long time (50-500ms) without power before shutting down.

[hgg]

I know I have one transformer lying around.  I will have a look to see if I find another one
and measure the switching delay.  I would have to measure the time from mains going off
till the UPS output switches on.

The problem though never happens when you go ON-OFF, but when ON-OFF-ON-OFF very fast.

[oldway]

There a lot of differents schematics of power stage of UPS but the correct one of the true line interactive UPS is this one.
There is no switching of the output, inverter and mains works in // with an inductor between them and inverter interact on line voltage to correct distortion of waveform.
There is one output transformer (T1) with magnetic shunts for coupling inverter and mains.
Greatest problem with this technology is that the inverter is highly overloaded when mains fails because it has to feed the load and the short circuit of the mains (current limited by inductance).
Line must be opened as fast as possible (by thyristors) to limit this current.

I have made a lot of experiences and tests of such a line interactive UPS using a 10KVA inverter with modifications for varying phase between inverter and mains and measuring resulting charging current.

I don't need to look at videos of Dave, I have 30 years of experience in power electronics....
With all the respect I have for Dave that I appreciate very much, Dave is not God ! (nor I... )

NB: one of the avantages of line interactive UPS is that there is no switching, nor transfer of the load.
When mains fails, there should nothing occurre to the load as inverter is already working in // with the mains.

@hgg: you have a storage  DSO, why not measure the transfer time of your UPS instead of looking if you computer is rebooting or not ?
Your approach of the problem is absolutely not profissional.
When you have such a problem, you have to know exactly what happen...What is the transfer time ? Is this always the same or not

[bktemp]

There a lot of differents schematics of power stage of UPS but the correct one of the true line interactive UPS is this one.

That matches the schematics for all APC UPS I have seen. But I can not see anything in there that confirms what you are saying. I see an autotransformer for line interactive with the full bridge on the secondary and a relais to disconnect mains for battery mode. It also matches the pictures of the ups on this thread.

There is no switching of the output, inverter and mains works in // with an inductor between them and inverter interact on line voltage to correct distortion of waveform.
There is one output transformer (T1) with magnetic shunts for coupling inverter and mains.
Greatest problem with this technology is that the inverter is highly overloaded when mains fails because it has to feed the load and the short circuit of the mains (current limited by inductance).
Line must be opened as fast as possible (by thyristors) to limit this current.

That may be the case for a large 10kVA UPS, but not for the common home or office use ones. I have several APC one here from a few 100VA upto a few kVA, they all are line interactive ones and use the battery charging as described in the patent. Look at Dave's video, they have no magnetic shunt (at least not visible, but since primary and secondary are wound on top of each other it is almost impossible to add a shunt somewhere inbetween). And they do not have a thyristor. They have large relais to disconnect the transformer when mains fails.

NB: one of the avantages of line interactive UPS is that there is no switching, nor transfer of the load.
When mains fails, there should nothing occurre to the load as inverter is already working in // with the mains.

Can you give me any reference for that? All sources I have found for line interactive describe it the other way: Line interactive is an offline/standy ups + voltage adjustment via the autotransformer. No waveform correction (except for emc filtering at the input/output).

[oldway]

Try only to explain to me how an UPS as you describe can "interact" actively with mains ? 
To add a voltage regulator in the line is not actively interact...its only stabilizing the mains voltage with a voltage regulator.
I never heard an AC voltage regulator named "line interactive" device...
Only in the power stage as I described, the inverter interact with mains because it is coupled through an inductor with main.
They use the name interactive for a non interactive schematic...That is wrong.

Can you give me any reference for that? All sources I have found for line interactive describe it the other way: Line interactive is an offline/standy ups + voltage adjustment via the autotransformer. No waveform correction (except for emc filtering at the input/output).

Try to do the same for your statements...As I said, you will not find schematics (only a very few, if any) on Internet...Manufacturers of UPS don't give any schematics, nor diagrams of their products.

Even I worked with great UPS manufacturers and I still have a lot of schematics, I can't publish them on Internet.

Look at the topology of an interactive UPS: no battery charger, no transfer switch of the output.

[hgg]

oldway can you suggest a quality line interactive UPS with a sine wave output.

I will buy the Cyberpower CP900EPFCLCD
http://www.cyberpower-eu.com/products/ups_systems/pfc-sinewave/cp900epfclcd.htm

...except if you know of a better solution.
I don't want to spend more than 200 euros.

[oldway]

I know I have one transformer lying around.  I will have a look to see if I find another one
and measure the switching delay.  I would have to measure the time from mains going off
till the UPS output switches on.

The problem though never happens when you go ON-OFF, but when ON-OFF-ON-OFF very fast.

Why is your computer rebooting when there is a mains failure and a transfer to the inverter ?
A SMPS has an hold up time, this the time the power supply of your computer can still provide a viable output without an input supply.
With a high quality SMPS, you should have specifications of this hold up time.
This old up time can be reduced if the electrolytics capacitors have loose their capacitance.
The UPS has a transfer time, that's the time taken by the UPS to switch from mains to the inverter.

If your computer is rebooting, that means that the transfer time is greater than the old up time.

In such a situation, this UPS is worse than nothing as it increases the short mains failures.
For this reason, it is far better to choose a TRUE line interactive UPS (without any transfer time, as there is no transfer), or an on line UPS who has only a transfer in case of inverter failure or battery low voltage protection.

I had bad experience with true interactive UPS and I don't like this solution. I often had to replace exploded power blocs of such UPS...Reliability is not that good. (can't work with gen set)

I recomend to go for on line UPS, that's the best technology and the most reliable.

[oldway]

oldway can you suggest a quality line interactive UPS with a sine wave output.

I will buy the Cyberpower CP900EPFCLCD
http://www.cyberpower-eu.com/products/ups_systems/pfc-sinewave/cp900epfclcd.htm

...except if you know of a better solution.
I don't want to spend more than 200 euros.

If I understood right, load is feeded by utility voltage (mains) and transfered on inverter in case of mains failure.
Transfer Time (Typical)    4ms
That's short but again you have to be sure that the old up time of the SMPS of your computer is greater than 4ms.
Far better to choose an on line UPS where there is no transfer at all.

[hgg]

I think that the switching lower time limit depends on how fast is the switching time of the relays
they are using.  If you "upgrade" the UPS with solid state relays maybe it will become much better.

Online is indeed better, but it gets very hot and it has fans making noise all the time.
My system is completely silent and I don't want any noise from the UPS.  Besides that,
they are very expensive...   I don't want to pay more than 200 euros for a UPS.

I am curious about the transfer time of my UPS and tomorrow I will buy a transformer
just to test it.

[oldway]

I think that the switching lower time limit depends on how fast is the switching time of the relays
they are using.  If you "upgrade" the UPS with solid state relays maybe it will become much better.

I don't think so...4ms is allready very good, you need a time to detect the mains failure and another to switch...1/4 of cycle, that's what we succeed with SCR's.
About price and noise, I think that if you want to buy an UPS, it is because you really need one...Than price and noise are not the most important factors : you have to look for highest effective protection and security.

And don't forget to replace the batteries when needed...

[oldway]

If you want a low cost solution, you could use your actual UPS only as an inverter , without the transfer relays (inverter diretly connected to the output), no mains connection and add a battery charger able to furnish the full DC curent of the inverter, more recharging current for the batteries.

[hgg]

Below are some of the EATON UPS power transitions without any load connected:












I did the same test to the Cyberpower by unplugging and plugging from mains
very fast and many times and it was rock steady.  So, there must be something
wrong with the EATON UPS.  Maybe I will replace all the capacitors and the
comparator & opamp.

Below is the transition of the Cyberpower CP900EPFCLCD I've just bought
but with a 200Watt load connected.  It's output is better than mains...
Highly recommended!   

[Jeroen3]

Maybe you should buy an atx power supply with a higher nominal load so there are larger smoothing caps.

[mij59]

Below are some of the EATON UPS power transitions without any load connected:

Interesting data, seems the ups switches to battery power in less than 20 msec, wondering if several  bounces, like  in the second output pulse after the triggering could affect the power supply of the computer

[hgg]

Maybe you should buy an atx power supply with a higher nominal load so there are larger smoothing caps.

My current PSU is a 500W ATX. 
The Cyberpower waveform is with a 200W load (PC & LCD) connected on the 500W PSU.

There is nothing wrong with my PSU.
The problem is in the crappy EATON UPS.

Interesting data, seems the ups switches to battery power in less than 20 msec, wondering if several  bounces, like  in the second output pulse after the triggering could affect the power supply of the computer

And this is from what I managed to do manually by playing with the mains plug to the UPS.
From what I've observed during brownouts, it gets much worse than that with faster and continuous
transitions.  Our mains power is not the best...

[wraper]

My current PSU is a 500W ATX. 
The Cyberpower waveform is with a 200W load (PC & LCD) connected on the 500W PSU.

There is nothing wrong with my PSU.
The problem is in the crappy EATON UPS.

There is nothing wrong with EATON waveforms, especially something that could cause computer PSU to shut down. Also why are you capturing waveforms without any load? Unless you show that something wrong with waveforms under load, I say your ATX PSU is faulty crap.

[hgg]

I've just connected the EATON UPS to a different PC (250watt load) with a different PSU and tried again to
recreate the problem and captured some transitions.

First a very well behaved transition : (blackout)



...and below the real problem of the EATON UPS captured:  (brownout)



Power was lost for 240ms !!!
The UPS is definitely faulty.
(Looking at the graph can you think what can create this type of behavior? )

But the initial problem of my PC rebooting had two causes.  One was the actual hardware
fault above and the second problem was with my PSU.  But not because it was of low quality. 
On the contrary.  It is the Corsair CX500.  (I have been using it for just over a year.)

http://www.corsair.com/en-us/builder-series-cx500

The "problem" is, that is has active PFC for increased efficiency. 
This type of power supply does not like square wave approximations
and the output from the EATON is not even close to an approximation...

[bktemp]

It looks like the battery in the UPS is dead or the output is heavily overloaded. The voltage is regulated by chaging the duty cylce of the generated square wave (actually the phase shift between both half bridge drivers). The peak amplitude is determined by the battery voltage and the tranformer ratio.
Initially the UPS starts with a typical duty cycle, but the voltage is way too low. Therefore duty cycle increases, but the battery can not supply enough current and the voltage falls further. At the end the PC shuts down and the the voltage recovers without the load.

[hgg]

It looks like that but its not.
The battery is in mint condition and fully charged.
The PC does not have an APFC PSU and did not shutdown or reboot.

[FlyingHacker]

Maybe measure the current flowing from the battery (keeping in mind it may be a lot more than your meter can handle) and see if that is in line?