UPS design (APC Back-UPS 650)

I have an old UPS unit that I found in a trash pile at a Hamfest = probably 5=20 or more years ago, and I figured I might be able to use it as a 12 VDC = to=20

120 VAC inverter. So I opened it up, and removed the two 6V SLA = batteries=20 (which had cracked, probably from freezing). I had thought the weight of = the=20 unit (about 22 lb.) was from the batteries, but there is also a rather = large=20 (3.5" x 4" x 4") transformer which probably weighs at least 10 lb. I had =

expected a much smaller ferrite transformer such as found in most = inverters.

So I thought maybe it was used in such a way as to create a good = waveform,=20 but when I powered the unit with a 12V supply (1.08A), I found that it = was a=20 "modified sine wave" which was just a series of positive and negative = pulses=20

175 V peak and 2.6 mSec wide with a period of 17.2 mSec (57 Hz), and it = has=20 an overshoot at turn-off. This indicates that it is not being driven by = a=20 true bridge but just a very simple PWM push-pull drive. I found two heat =

sinks, each of which have four TO-220 devices, and I found a rectangular =

pulse train 11.5 volts and 3.6 mSec wide, so they must be MOSFETs.

It has quite a large and complex PCB, with a 4" x 6" section having = about 10=20 DIP ICs and associated components. No SMDs. The unit has a DB-9 serial = port=20 for computer connection. The date on the PCB is 1997. There does not = appear=20 to be any sort of microprocessor or "smarts", and the only IC I did not=20 recognize was an MF4CN-50 which is a low pass filter=20

formatting link
PDF.=20 So, what does the serial port connection do, and how does it do it? Here = is=20 the user manual: =
formatting link

I don't think there's much value to this thing, except maybe the big = tranny=20 which might make a good high current low voltage supply. Probably about =

15=20 VCT at 30 amps or so.

Are more modern UPS units made this way? Seems strange not to use high=20 frequency magnetics to generate a 175 VDC rail and then use a bridge to=20 generate the output waveform. And now that it's easy to make a PWM = inverter=20 with a high frequency carrier like a motor control, a fairly good sine = wave=20 could be made. And a crystal could be used to get a very accurate 50 or =

60=20 Hz waveform, although it's usually not much needed except for an alarm=20 clock. Actually, most computer equipment uses switching supplies and=20 probably would operate just as well, if not better, with the straight DC =

rail (although perhaps at less than the 175 volts).

Any thoughts? Thanks.

Paul=20

Reply to
P E Schoen
Loading thread data ...

Yes. Its more or less APC's standard recipy for up to 2500VA. This design is very reliable. I've never seen one fail.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
 Click to see the full signature
Reply to
Nico Coesel

Yes. It's not even a "stepped sine wave" -- think of the primary as a center tapped coil with FETs tugging alternately on one side, then the other.

The "difference" between various VA models is the number of FETs paralleled for each "side" of the coil.

I don't think the serial port does anything besides passing *levels* to the monitoring PC. I.e., the "data" isn't the sort of "ASCII text" that you might otherwise believe.

These are typically "consumer kit" -- even the models marketed for SOHO use. I inherit probably two a month from folks (usually

1000-1500 VA models) suffering from bad batteries ("Gee, it costs as much to replace the batteries as I paid for the darn thing to begin with!" or "Yikes! I just replaced the batteries a year or two ago -- and we haven't had any power outages in that time...") or blown FETs (I have three XS/RS 1200's with that problem, now).

For most folks, UPS's *sound* like a great idea! Likewise, various "(data) backup devices". In practice, the novelty usually wears off pretty quick. :> I have 6 deployed here, currently, and they tend to be more of "outlet/receptacle multipliers" than real UPS's. I suspect they have alarmed more times from battery failure than from actual power outages. But, our power tends to be pretty reliable (below grade service seems to make a difference -- esp in the warm/dry environment). So, the most annoying thing is brief "flickers" when something, somewhere might be switching to cover a changing distribution load (most of my machines will survive a 1 second "outage" without even requiring a UPS).

I previously had a 5KW unit that proved impractical to use (the size of a dishwasher). And, dangerous to work on (the

*120* VDC battery). Routing the conditioned power to various loads meant a long "extension cord" (or, a dedicated circuit!).

Some of the more "professional" units offer a better mix of features, output waveform quality, etc. (I was offered a large unit -- the battery bank being effectively a 40U rack -- but was smart enough to realize I could never afford to keep replacing the slew of batteries it needed!)

I've also seen smaller units used as DOOR STOPS to take advantage of their size and weight (the power cord serves as a convenient "handle" to drag it to where you want/need it!)

Reply to
Don Y

I suppose a normal load would clean up the waveform to some extent. The=20 "ringing" was at a very low frequency and probably a result of surge=20 suppression components on the 120 VAC side.

I figured that it simply uses one of the RTS, CTS, DTR, or DSR lines as = a=20 power fail indicator, and the PC software does all the rest. It probably =

doesn't have the ability to read the battery level or load or anything = like=20 that. Just a signal to save open files and perform an orderly shut-down = of=20 the computer.

That's probably all it's good for, IMHO. The surge protection components =

might be useful, but more conveniently located in a power strip. Seems a =

shame to throw away a functioning unit, but I use a laptop for all my=20 computing needs, and it has its own UPS in a fully charged battery that=20 allows about an hour of continued use after a power failure (and the = screen=20 provides a nice emergency light so I can find my flashlight).

I went to the Timonium Hamfest today, hoping to find some inexpensive = SLA=20 batteries, but didn't find any. I have a 17 A-H battery (date code =

1999?) I=20 bought there at least 5 years ago. I had it in an old garden tractor for =

several years, and I was amazed that it still had juice after being = outside=20 and only recharged once a year or so. It still seems to be going strong.

My project is a small electric tractor that has a three phase motor, and = I=20 plan to use a couple of 12V batteries and inverters to generate about =

300=20 VDC which I'll use with a 2 HP VF motor control. I know the 17 A-H SLA = is=20 really too small, but a pair of them should allow maybe 10 minutes run = time=20 at 1 HP average. The best price I found on-line was $33 for 18 A-H SLA,=20 while I can get a regular lead-acid garden tractor starting battery for=20 about $30 with 200 CCA (probably 20-30 A-H), and a 100 A-H deep cycle=20 battery for $85.

Thanks,

Paul=20

Reply to
P E Schoen

I have a couple of APC BackUPS 750T's that I got from eBay. One of them even had good batteries. One of them runs my NAS boxes at home, and the other runs my office server, a 16-core AMD Magny Cours box running

64-bit CentOS 6.2. Clicking on the power icon brings up a window that has all sorts of info about the UPS, including load levels, AC line voltage and frequency, the battery's state of charge, and so on. So there's a lot more info there than you suppose.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
 Click to see the full signature
Reply to
Phil Hobbs

Different models have different capabilities. Some provide more detailed reports on the status of batteries (including external "add-on" packs), actual load seen by the UPS (depends on whether you have an online or offline UPS), uptime estimates, programmable alarm thresholds, support for multiple UPS's, integration within the "management console", support for load shedding, event logging, etc.

Of course, the features you get vary with the cost/size of the UPS! :>

I have small units on each of my switches, name server, etc. I.e., things that need to work for anything to be able to talk to anything else. Each workstation has a ~1000-1500 unit to keep the machine and displays "up" -- I don't worry about printers (most of my printers have big power requirements) or other little peripherals. A power interruption would just be a "small inconvenience" for those ("Crap! I'll have to rescan that photo/sheet" "Grrrr... that *plot* is now toast" etc.).

OTOH, the operations performed by the workstations tend to be a bit more problematic to resume ("Hmmm... have to rerender that

3D animation").

But, the biggest advantage is having a single power switch for each workstation+displays+core peripherals (motion controller, tablet, etc.). Otherwise, I have to shut down each device individually and/or turn off entire *benches*.

Do you really intend that to be a *tractor*? (workhorse) You can consider looking into the power plants in electric "fork lifts". You'd have to change the gearing as they tend to be designed for torque at the expense of speed...

If you just want a "toy" to cart you around, motors from electric wheelchairs are a viable option (300-800W 24V). These are often "discarded" when their owner passes away or "insurance" opts to cover a replacement (possibly because the user's needs have changed -- most chairs are "fitted" to the needs of the user... eye tracking, head tracking, joystick, etc.)

Good luck!

Reply to
Don Y

Modified sine is so bad I would rather lose power.

I use a double conversion UPS. They convert the mains to DC where it is easy to filter and cap load, then produce a true sine (I think 3% THD). The advantage is there is no required time for the UPS to kick in. It is always on. The disadvantage is it is always on. Thus you have cooling fans and a bit of efficiency loss.

APC makes them, but you won't find them in consumer stores. I'm running an Opti-UPS. Same basic idea.

The log catches a few glitches once in a while. I caught a few live (heard the relay flip to kick in the battery) during a recent wind storm.

If you go double conversion, you need to inspect your load carefully. This is a case where the UPS is actually working rather than just sitting there.

Reply to
miso

I used to have a UPS like that. The degauss coil in my monitor killed it.

--
Failure does not prove something is impossible, failure simply
indicates you are not using the right tools...
 Click to see the full signature
Reply to
Nico Coesel

These are often called "on-line" converters.

There are other advantages: the UPS never sees the "sudden" load thrown on its output; the output operates at a constant line frequency; etc.

If the load can't handle the switch-over time (typically O(10ms)), then an offline UPS can prove to be useless -- the power to the load disappears for that period of time even if the batteries, UPS, etc. are "good".

Note that fans are *always* on (i.e., keep big UPS's in some non-living space -- because the converter is always carrying the load!

You also have a funky power factor that you present to the utility since the UPS looks like a bridge feeding a large cap (big enough to carry the load between AC cycles). For users (businesses) looking for true uninterrupted power (e.g., with a diesel genset coming online to restore power *to* the UPS) this can require the genset to be needlessly (?) oversized.

Efficiency loss can hurt considerably since you have two nonideal converters operating in cascade. The heat thrown off testifies to the losses encountered. Normally, the input rectifier only processes power to charge the batteries -- thereafter, sees no losses until the batteries once again need charging (at which point, power is presumably available and the inverter isn't running).

Of course, you can *buy* your way around these problems... :>

Reply to
Don Y

The other advantage of on-line is you get brown out protection. I still prefer using the term double conversion because online has so many other meanings. But you are correct in that most units are sold as online.

Well it claims PFC of 0.97 and efficiency of 92%. Of course at full load, which isn't reality. All the computers it feeds are PFC corrected and high efficiency.

The only thing I found kind of lame about the unit is the front panel is plastic, which in turn snaps into the metal box. The fit is kind of loose.

Expensive, but it drives a few computers and a monitor. Not that computers are all that expensive these days, but the hassle of getting everything working again makes the on-line UPS worth the effort.

It also has a bypass mode, which to my understanding means it acts like a conventional UPS, i.e. only turns on when needed, but supplies clean power. I don't use it that way.

The APC and Opti units are made in China unfortunately. The reason I got the Opti is it has an intelligent fan. You can hear it increase speed when I turn on a PC due to the surge, then it drops in noise again. The APC unit didn't have that feature. Note 40dB is mighty loud.

Similar Tripplite item. You can even get "hospital grade" versions.

Getting reviews on these items it kind of tough since they are soho grade. Data centers use bigger UPSs and consumers buy the junk modified sine. Data centers sometimes use computers that take 48V input rather than AC.

Reply to
miso

"Line interactive" designs offer some correction. And, FERRUPS.

I always hated the online/offline distinction as it was a "relative" one that depended on your interpretation of "on" vs "off". I have to *think* before refering to one or the other (it would be nice if it was more intuitive)

Ever since bakelite, plastic has been "here to stay" (especially for formed pieces)

Understood. As I said, I think it depends on the reliability of your mains power. Here, the outages we've had number ~0.5 per year! Unfortunately, some have been very lengthy (like when the neighborhood distribution transformer burned up taking out power to ~100 homes. They don't replace those things very quickly -- though they managed to run a feed in from a neighboring circuit to bring some power back).

OTOH, when I lived in Illinois, our neighborhood suffered 1-2 outages per *month*.

EVERYTHING is made in China! :>

You can also purchase aftermarket fans that monitor air temperature and self-adjust. Some will even alarm when the fan quits.

The problem with most larger UPS's is they are designed for folks with a "UPS budget". Who plan on servicing them regularly, etc. My attitude is that an outage is an opportunity to "do something different". If I *really* need to keep working, I can move to a laptop and power the modem from a UPS (almost "forever" with the amount of power it requires).

A more *practical* use for the UPS's is to use them to power CFL/LED lighting. The neighbors were all wondering why *we* had power ("lights") while they didn't during the last outage.

Reply to
Don Y

Maybe the higher output units (or more recent models) have more=20 monitor/control features. I can't see how all that information can be = passed=20 to the serial port without some sort of "smarts", like a microprocessor. =

There seem to be maybe 4 transistors and a dozen resistors, and an=20 optoisolator, connected to the serial port connector. It seems there are =

4=20 input pins other than RXD that might be used for some information, but = it=20 would be essentially go/nogo. They are CD, DSR, CTS, and RI.

I was able to disconnect the gate drives for the two banks of MOSFETs, = and I=20 hooked them to a PIC16F684 which has PWM output. I set it for 500 Hz, =

100%=20 duty cycle, with about 2 uSec deadband, and I used an LM324 as a simple = gate=20 driver to get 10-12V gate drive. The MOSFETs are IRFZ46 which are 50V, = 50A,=20 24 mOhm. When I first tried it, I got 360V P-P with 11.8V in at 3.9A, = but=20 something was heating up. I think it was the overvoltage / surge = protection=20 components, and when I disconnected the secondary from the board, I got = 360V=20 P-P at 1.47 amps, and no apparent heating. So I did some load testing:

1800 ohms 172V 4.10W 6.0V 1.22A 7.32W

1800 ohms 238V 7.86W 8.0V 1.58A 12.6W 1800 ohms 344V 16.4W 12.0V 2.82A 33.8W 300 ohms 171V 23.8W 6.0V 4.38A 26.3W 300 ohms 200V 33.3W 7.0V 5.00A 35.0W

For some reason the efficiency with a 300 ohm load seems a lot better = than=20 with 1800 ohms. It may have something to do with the power supply going = into=20 overcurrent limit, or maybe there is something about running it at 8 = volts=20 or more. My next step will probably be to use a heavy duty supply or=20 directly from a battery, and I should be able to get 108 watts into the =

300=20 ohm load. I have several of these resistors (91 watts each), so I can = test=20 up to about 500 watts. I also have some finstrip heaters which are 26 = ohms,=20 and that will give me 1246 watts if the voltage holds. The battery draw = will=20 be about 100 amps at this point. So maybe I'll use two heaters in series = for=20 600 watts and 50 amps from the battery.

Final test will be adding the FWB and capacitors to get 360 VDC which = should=20 supply the VF drive for the motor. But I'm a bit worried about the = start-up=20 current surges into the capacitors, especially with a square wave. I'm=20 considering making a current limiter, but at such high currents there = may be=20 unacceptable losses. Best thing might be to use a 100A 100mV shunt and = have=20 that connected to the PIC. The A/D converter might be fast enough, but=20 otherwise the comparator may be the better choice, and it could shut = down=20 the PWM drive. The Back-UPS has two 40 amp fuses in parallel, but I = don't=20 want to rely on them except for ultimate fail-safe protection. The=20 simulation shows 566 amps RMS during start-up and 84 amps RMS once the=20 output has stabilized at 20V into 104 ohms and 757 watts.

Maybe I'll have more results in a day or two.

Paul=20

Reply to
P E Schoen

I just tried reading input current at various supply voltages. Seems = like it=20 is starting to saturate, but unexpected at 500 Hz. Maybe it's because of =

core losses, since this is an ordinary E-I transformer and not a toroid = as I=20 previously used.

6V 0.57A 7V 0.63A 8V 0.69A 9V 0.77A 10V 0.92A 11V 1.09A 12V 1.44A 360V p-p 13V 1.93A

For the toroid, I measured:

6V 0.93A 151V p-p 7V 1.03A 172V p-p 8V 1.15A 192V p-p 9V 1.30A 220V p-p 10V 1.43A 241V p-p 11V 1.59A 263V p-p 12V 1.80A 289V p-p 13V 2.03A 309V p-p

The toroid current is just about linear with voltage, but for the E-I=20 tranny, the current at 13V is more than three times that at 6.5V. Any = ideas=20 about why this is?

Paul=20

Reply to
P E Schoen

I'm not concerned about down time as much as data loss. The UPS is set up to shut down the computers.

Reply to
miso

"P E Schoen" wrote in news:jljjmp$gnb$ snipped-for-privacy@dont-email.me:

Core saturation?

Reply to
Sjouke Burry

It has the characteristics of core saturation, but this is at the rated=20 voltage for the transformer (12 V peak rectangular wave), and at 500 Hz, =

which is more than 8 times its rated frequency of 60 Hz. The output = voltage=20 is a pretty good square wave, with no visible droop, which might be = expected=20 at saturation (although the regulated supply might hold it steady even = if=20 current is sharply increasing).

But there is a slight difference in the output waveforms. The E-I=20 transformer shows a more or less smooth transition from the positive to = the=20 negative peak, while the toroid shows a faster drop to zero, then a = brief=20 period of off time, and then another rapid rise to the opposite peak. I=20 think my driver has a 2 uSec deadband, and if the transition is linear=20 during this time, rather than stepped as seen in the toroid, there is=20 probably considerable power dissipated during that time. I should really =

measure the current waveform, but if there is core loss due to = switching, it=20 would show up as additional power dissipation.

It is still within acceptable limits for an inverter rated at 600 VA,=20 however. I could also change the frequency to 100 Hz or 60 Hz to see if = the=20 power is reduced. I like the idea of a higher frequency so I can get = less=20 ripple for the same size capacitors.

Paul=20

Reply to
P E Schoen

LOL!....

Reading all the posts in this thread is like listening to a buncha old druids arguing about the merits of some twig/animal/leaf/star/tree and how they should worship it.

Don't get me wrong, as an old geezer who's played an electrified stringed instrument for over 50 yrs and am only now, as a hobbyist, learning about electronics and tubes and such, I'm sure I'll go to my grave jes as whacked out as the rest of you "experts", and laughing my ass off about it all. ;)

nb

Reply to
notbob

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.