The stack exchange post says that over a few years, the replacement batteries become the dominant cost, so buying a high end UPS at say twice the costs, kind of almost doesn't matter after several years, the total costs approach each other - so buy a very good one.

Oh that's a very interesting viewpoint.
I'd have to figure out if that is the case for my stuff too. It makes sense though, because if we buy a 1000 watt unit sine and 1000 watt unit crap sine (ha ha) then when we replace the batteries it will be the same cost for both. So let's see...
Purchases both types: $200, $100 [totals 200 and 100] cost ratio: 2
Batteries after 2 years: $50, $50 [totals 250 and 150] cost ratio: 1.67
Batteries after 4 years: $60, $60, [totals 310 and 210] cost ratio: 1.48
Batteries after 6 years: $70, $70, [totals 380 and 280] cost ratio: 1.36
Batteries after 8 years: $80, $80, [totals 460 and 360] cost ratio: 1.28
Batteries after 10 years: $90, $90, [totals 550 and 450] cost ratio: 1.22

So we can see that the cost ratio does start to become the same after some years, but it takes a long time.
First set we see it drop by about 33 percent, then about 19 percent, then about 12 percent, then about 8 percent then about 6 percent.
So they do start to even out, but it takes a long time.
It takes 20 years to get down to 10 percent difference.
It takes about 30 years to get down to 5 percent difference.
After 40 years about 3.3 percent difference.
I guess we can say though that 200 and 100 are widely different, while 550 and 450 are not that much different in the long term.

This is based on a battery or battery set cost rising by $10 USD every 2 years.

This is the function I used:
f(x):=[%[1]+%[3],%[2]+%[3],%[3]+10,%[4]+2,float((%[1]+%[3])/(%[2]+%[3]))]
with initial values:
[200,100,50,0,0]
which is the two costs then the $50 then, then the number of years, then the ratio of the higher cost unit over the lower cost unit.
The first update becomes:
[250,150, 60,2,1.6667]
where we can see the cost increased from 50 to 60, the years increased from 0 to 2, and the ratio is 1.6667.
The 60 is applied after the next 2 years but it's updated at the present so it can be used later. The cost applied was actual 50.

BTW the notation "%[n]" just means the nth element of that vector of 5 numbers. So %[3] is 50 to start then 60 after 2 years.
the function f(x) is just applied over and over again, which updates everything every 2 years.

One thing to note here though is that not everyone will have to buy new batteries after 2 years. It depends on how much the power goes out and how long you have to run the UPS before you can shut down manually.

 

Hi,

You mean in comparison to the modified sine ?

Yes.
It seems that there are three requirements:
1) the correct peak value
2) the correct rms value
3) reasonably low harmonic content.

1) satisfies anything with a power supply with a rectifier
2) satisfies an ordinary resistive load
3) satisfies anything with a magnetic core which has core-loss.

A square-wave can pass 1) or 2) but not both.
The pulse waveform that likes to be know as "modified sinewave" can pass 1) and 2) but not 3).
The most cost effective option to pass all three would appear to be a pure-sinewave inverter.

 

A question I have is why don't these "noisy" UPS boxes, have a 1:1 transformer right before their outlet? surely that would filter out some of these harmonics and noise

Because a line-frequency transformer rated for the full output is large and expensive.
Emergency lighting systems DO have that requirement.

 

APFC's try to make the input current more sinusoidal,

I'm going to be pedantic here: APFCs try to make the input current waveform the same shape as the input voltage waveform.

 

OK this is a must read, this guy found that his pure sine UPS produced a cleaner shape than the incoming power line!

Take look, he has pics, spectrums etc.

https://superuser.com/questions/912679/when-do-i-need-a-pure-sine-wave-ups

A question I have is why don't these "noisy" UPS boxes, have a 1:1 transformer right before their outlet? surely that would filter out some of these harmonics and noise

Hi,

Some of the ones we made did have transformers on the output, but not 1:1 because the turns ratio worked into the design. The transformer would take the bridge output as input and then the voltage could be stepped up or down depending on the requirements.
Some of the transformers were very large though, and heavy. You might not be able to pick some of them up by hand. A 30kW three phase converter had three separate transformers each weighing in at over 50 pounds, I think pushing 100 pounds each. The converter was as big as a small closet.

The transformer(s) also provided good input to output galvanic isolation, which was a requirement for a lot of clients.

 

Hi,

Some of the ones we made did have transformers on the output, but not 1:1 because the turns ratio worked into the design. The transformer would take the bridge output as input and then the voltage could be stepped up or down depending on the requirements.
Some of the transformers were very large though, and heavy. You might not be able to pick some of them up by hand. A 30kW three phase converter had three separate transformers each weighing in at over 50 pounds, I think pushing 100 pounds each. The converter was as big as a small closet.

The transformer(s) also provided good input to output galvanic isolation, which was a requirement for a lot of clients.

What was the reason for those transformers then, cleanup the output sine or just galvanic isolation?

 

What was the reason for those transformers then, cleanup the output sine or just galvanic isolation?

You can wind them with a designed amount of leakage inductance, and connect a capacitor across the output so that if forms a 2nd order filter when combined with the leakage inductance.
You then drive them with PWM from a MOSFET bridge operating at 48V, and you have voltage step-up, PWM carrier removal filter, and galvanic isolation all in one component.

 

What was the reason for those transformers then, cleanup the output sine or just galvanic isolation?

Hi,

In most cases they had leakage inductance so it was part filtering too.
If galvanic isolation was not needed an inductor would be used instead, but that would be fairly large also so a transformer with leakage inductance was almost always used.
I can't remember a case where there was no leakage inductance.