Cheap, powerfull DIY UPS question....

[Brane2]

I have a small company with a small workshop, consisting some soldering
tools, a few computers etc.

Some time ago, we have updated 3 of our rather old machines to dual
Opterons etc and hooked two of them onto 1200VA UPS (they are constantly
powered on).

Some time after that, I have started noticing heftier electricity bills
than usual, but it took me some extra time to isolate the cause.

Each of the Opterons is drawing constantly 300W per se (without
monitor), moreover it draws some 40ish W while being in standy by
(powered-off by switch) !

For comparison, Athlon on 1300 MHz churns 60W on full load and less than
5W when in standby...

This is measured directly on the computer input, without factoring_in
the losses in UPS itself...

As it turns out, after certain power level, all quality PSU computer
units have extra filtering, correction factor circuitry etc and all this
apparently comes with extra losses. At least with my Sparkle 460W units.

Extra UPS protection only makes matters worse. Quality models do not
"hop_in" when AC goes out, but they work constantly.

So, with UPS, power goes through 4 (!) converters:

-step_down converter in UPS
-step_up & AC converter in UPS
-various cos fi etc gadgets
- step down conversion to 5V/3.3V/12V etc....

Fed up with constant ventilator noises, high losess,short battery life
etc, I have an idea to try.

I plan to use series chain of automobile accus (20x12V/66Ah), which
would be filled more or less directly from mains (O.K., not directly,
but through simple step_down conversion).

Computers would be powered directly from accus (240-280V DC), possibly
bypassing the cos fi corr. & filtering crap in Sparkles).

I plan to power that way all the monitors, too, but will have to make
modifications for degaussing coils...

Any thoughts, advices ?


Regards,


Branko

 

[Neil]

And buying a ton of batteries, enough to make up over 200 volts, and having
to store them somewhere is cheaper, and easier.....how??!!!
Kim

 

[Brane2]

And buying a ton of batteries, enough to make up over 200 volts, and having
to store them somewhere is cheaper, and easier.....how??!!!
Kim

A "ton of batteries" would cost me cca 5x monthly electricity bill and
it would mean only marginally bigger expense than what I have paid for
1200 VA UPS unit already.

Storage space for the batteries is also not a problem.

Branko

 

[john KB5AG]

seems like your problem is with the computers being hogs, not the UPS's.
How are you going to come up with +/-12v and +/- 5v at the required amperage
(especially the 5v which will be a large load)? And how are you going to
sequence your power up/down voltages so that the computers like it?

 

[john KB5AG]

after re-reading, I guess you're going to power the switchers in the 'puter
from DC not AC? Still don't see what that will get you. Do your UPS's make
noise all the time? Mine don't. The 'puter will still need all its fans...

 

[Brane2]

after re-reading, I guess you're going to power the switchers in the 'puter
from DC not AC? Still don't see what that will get you. Do your UPS's make
noise all the time? Mine don't. The 'puter will still need all its fans...

Yeah, that was the idea.

I plan ot get from it:

-much longer autonomy. Currenlty my UPS batteries last for 23 mins with
one machine on and 7-8 mins with both machines running.
New scheme should get me at least 8 hours, if not a whole lot more in
case of need...

- much higher efficiency, due to simpler conversion path, lesser mass
and volume constraints etc.

Simpler conversion path means that Accus would be filled more or less
directly from AC (there will be a simple switching regulation and an
gigantic choke, though). But due to very low switching frequency at
convenient points, low saturation of the choke core and low currents
overall I expect low losses.

I plan to regulate the thing by switching it on/off for integral number
of periods, so switching would be done at AC zeropoints. Assuming that I
get cos fi correction about right, losses at that moments will be minimal.

After that, all should be easy. Computer units run on 300V DC internally
ass it is. First thing inside (after a simple overvoltage spikes
portection) is greatz and elko- which feeds ~ peak voltage DC to the
rest of the circuitry.

So, PSU shouldn't have a problem with this arrangement, moreover it
should run even better. Due to powerline polution standards and other
constraints, in built elko can't be too big, meaning that DC voltage
will have considerable AC component.

With this arrangement, AC noise should be much lower.

Using DC means alko that I don't need anymore cos fi correction, extra
filtering etc inside high end powerful PSU units (400W +), which really
waste energy.

That is the general idea...

 

[Ken Taylor]

Yeah, that was the idea.

I plan ot get from it:

-much longer autonomy. Currenlty my UPS batteries last for 23 mins with
one machine on and 7-8 mins with both machines running.
New scheme should get me at least 8 hours, if not a whole lot more in
case of need...

- much higher efficiency, due to simpler conversion path, lesser mass
and volume constraints etc.

Simpler conversion path means that Accus would be filled more or less
directly from AC (there will be a simple switching regulation and an
gigantic choke, though). But due to very low switching frequency at
convenient points, low saturation of the choke core and low currents
overall I expect low losses.

I plan to regulate the thing by switching it on/off for integral number
of periods, so switching would be done at AC zeropoints. Assuming that I
get cos fi correction about right, losses at that moments will be minimal.

After that, all should be easy. Computer units run on 300V DC internally
ass it is. First thing inside (after a simple overvoltage spikes
portection) is greatz and elko- which feeds ~ peak voltage DC to the
rest of the circuitry.

So, PSU shouldn't have a problem with this arrangement, moreover it
should run even better. Due to powerline polution standards and other
constraints, in built elko can't be too big, meaning that DC voltage
will have considerable AC component.

With this arrangement, AC noise should be much lower.

Using DC means alko that I don't need anymore cos fi correction, extra
filtering etc inside high end powerful PSU units (400W +), which really
waste energy.

That is the general idea...

 

[Ken Taylor]

Yeah, that was the idea.

I plan ot get from it:

-much longer autonomy. Currenlty my UPS batteries last for 23 mins with
one machine on and 7-8 mins with both machines running.
New scheme should get me at least 8 hours, if not a whole lot more in
case of need...

- much higher efficiency, due to simpler conversion path, lesser mass
and volume constraints etc.

Simpler conversion path means that Accus would be filled more or less
directly from AC (there will be a simple switching regulation and an
gigantic choke, though). But due to very low switching frequency at
convenient points, low saturation of the choke core and low currents
overall I expect low losses.

I plan to regulate the thing by switching it on/off for integral number
of periods, so switching would be done at AC zeropoints. Assuming that I
get cos fi correction about right, losses at that moments will be minimal.

After that, all should be easy. Computer units run on 300V DC internally
ass it is. First thing inside (after a simple overvoltage spikes
portection) is greatz and elko- which feeds ~ peak voltage DC to the
rest of the circuitry.

So, PSU shouldn't have a problem with this arrangement, moreover it
should run even better. Due to powerline polution standards and other
constraints, in built elko can't be too big, meaning that DC voltage
will have considerable AC component.

With this arrangement, AC noise should be much lower.

Using DC means alko that I don't need anymore cos fi correction, extra
filtering etc inside high end powerful PSU units (400W +), which really
waste energy.

That is the general idea...

Whoops, shot off a blank, so to speak....

If you're capable of building a UPS like that, and modifying the PC
switch-modes, good luck to you. However it'd be a whole lot easier to get
48VDC input power supplies for the PC's and a new UPS with 48VDC output.

Ken

 

[Brane2]

If you're capable of building a UPS like that, and modifying the PC
switch-modes, good luck to you. However it'd be a whole lot easier to get
48VDC input power supplies for the PC's and a new UPS with 48VDC output.

Ken

What would I gain from it ? I would have to get hold of 48V models,
which are, I presume, much more expensive than off_the _shelf 230V models.

Besides required modification seems to be a minor one, if any.

I think elcheapo PSU unit should be able to run off 300V DC without any
modifications...