This is an "extended run" APC UPS as designated by the "XL" suffix on the part number. 2 things make it extended run: 1) the secondary battery input and 2) additional cooling fans. The extended run models are "highly desirable" for use as straight inverters because on the standards models, the transformers will overheat within about 30 minutes of heavy continuous use. Of course, now that I told you that the transformer is the "weak link" you can add your own additional cooling including thermal compound between the bottom of the transformer and the case !
Nice explanation Dave, I really like your videos on DFM. A quick comment on how the charging work: Voltage reflecting on the battery side of the transformer (primary side) will never be high enough to charge the battery through the four mosfet body diodes. (battery shouldn't be charged this way because you cannot control the charge current and stop charging appropriately) In order to charge the battery, the mosfet H-bridge works like a boost converter, boosting the voltage to high enough to charge the battery, and at the same time control the charge current and do PFC (Power factor correction). In inverter mode, the H-bridge works like a buck converter having output voltage swing at line frequency. Half-bridges or full bridges are bidirectional converters. You can easily control the power flow both ways by applying correct PWM. Usually, an algorithm called dq control (similar to field oriented control) is used to control the power flow. For example, when vector q is positive, power flow forward, and when q negative, power flow backward. This idea is not new, but applying the idea to UPS to simplify the design is really smart.
why don't they teach it in college, the topology practical design. When I opened up my 800VA APC UPS, I wonder how there's only one power transformer who's responsible for AVR, charging, and inverter, with only 3 cable for high voltage and 2 cable for low voltage. But after all these years why don't they switch to high frequency transformers, instead staying with the bulky heavy transformer. I noticed these designs get warm even without load consuming around 20 watt, that amount does affect the energy bill for 400W UPS running 24/7, and all UPS I can find does that except for the DC UPS's.
This video is very timely, in my case, I watched this video as a study break from studying for my Power Electronics final exam. It was entertaining AND informative. Thanks Dave! Keep up the great work!
To answer your question @28:43. The MOSFETs in the inverter are also used as the rectifier. It called a "Dual Headed Synchronous Rectifier"(DHSR). In a DHSR configuration. The controller synchronized to the mains/line and then switched the MOSFETs on and off in sync of the AC wave form. It also uses PWM to control the battery current as well as perform power factor correction. Hook your oscilloscope up to it and look at it.
Actually I do believe that they are turning on the mosfets at the same time that the body diode would be on (as stated in the comment you commented on). It would have been a simple enough optimization to figure out, and they would have the control circuitry that would otherwise be dormant. It could also allow them to do some power factor correction to make the charge circuit seem more like a purely resistive load.
There is another type of UPS - transformerless double conversion. Just had a 40kva 3 phase unit installed. Batteries are in series and the DC side runs at 480v. This is common for large, fixed installations. It can be bypassed externally without dropping the load e.g. for servicing. When going out of external bypass it will automatically sync the inverter with the mains supply after which you can flick the bypass off. No interruption to critical load at all.
How do you not know what Best Power is? I mean the company has been split into pieces and sold for quite a while now but that place was such a legacy. It's building is still standing in Necedah, WI being rented out by the original owner. That place is such a damn throwback, my parents met while working there and my great grandmother was one of the first to get an award for exemplary work. I was practically raised there. On the upside, since closing down, my dad has started up his own business based around UPS.
Great video, helps me a lot to figure out my scrappy $20 APC Smart-UPS 2200VA. i GOT THE SAME ! Difference is that i have one without the slide out battery bay, Only 2 big capacitors placed far behind the path of cooling air through the H bridge (Seems logical). Basically you have 8 small NP7-12 Lead Acid batteries in a weird configuration. APC made 2 separate 24v sections (with 2 Anderson connectors) in series (48v 14A total) with a 100A fuse in the middle. Both sections consist of again 2 individual banks of again 2 NP7-12 batteries in parallel (= 24v 7A) again in series = 24v 14A. The leads get bigger at every level. Basically the charging current is only 1A max (Maybe even a lot lower) for these 8 tiny 12v batteries switched that way. The total charge current at 48v is about 4A (only 200 Watt). Discharging each quickly ramps up (ads up voltage and amperage in the chain) the capacity of this 48v battery without damaging them individually. I only wonder about the charging profile. Does this Interactive system stop charging at all at some point if batteries reach the full capacity and if not does this explain why all the old AGM batteries in these units seem exploded or expanded with bubbles on all sides? Does this pyramid of 2x2x2 batteries equalize automatically if you leave the power disconnected? I ask because it would be possible to replace every single NP7-12 battery with a 10 x AA (1.2v) Ni-MH battery pack, Haha. Picture link: www.dropbox.com/s/leg4t0lyqml42mh/2020-03-14%2018.40.03.jpg?dl=0
Pretty sure it's OK to use the body diode for 50Hz rectification, most mains frequency diodes are pretty crummy in terms of switching/reverse recovery performance. Not sure about the body diode, but (some?most?) MOSFETs have a positive temperature/resistance coefficient, thus they share current pretty well. Series diodes are much more of a pain since you have to get voltage sharing. As for the large number of caps, that's not unreasonable for single phase -> 100Hz power flow and all...
I seen this and it took me back. I worked for a large UPS maker for 10 years. I went to school for APC Anton Piller Belkin Kw controls and a few others . The company also did motor gen sets up to 500kva I could tell you every connection and what to look for to repair it. They had on line UPS and non online UPS systems . I even set up the systems for several companies including Blue Cross . I could tell you so much about UPS systems it would make your head swim.
Nice feature of the APC SmartUPS, if you have a power failure, you can turn them on and they will run off the internal or external batteries w/o any mains connection. Many of us use these older models that nobody wants for our older amateur radios, aka Ham Radios. These older circa 1975 radios have internal transformer they do not like any square or modified sine waves. The APC SmartUPS put out full sine wave and we can run the radio off batteries for emergency use.
I disassembled a large online UPS, and it didn't follow that schematic. It was near identical to your "offline" schematic - the main difference was that the step-up transformer was an SMPS and went before the inverter. The battery voltage was stepped up from 400 volts to 700 V (+/- 350 volts), allowing a simple push-pull topology inverter. The difference is that in an offline UPS, the load is directly supplied by the mains. If the mains fails, then the transfer relay switches the load to the inverter. In the online UPS, the load is supplied by the inverter, unless the inverter fails, in which case it transfers back to mains. The common reason for inverter "failure" is actually the output being severely overloaded or shorted (e.g. due to a fault on a single load where a UPS supplies multiple individual loads). In this case, the inverter voltage drops under the overcurrent stress, and the load is transferred to mains with its high current supply capability. This can also happen if regenerative loads are connected to a UPS (e.g. motor drives) and the reverse power exceeds the handling capability of the inverter causing it to trip out due to DC bus overvoltage. There is usually a difference in the type of switching. Offline UPS systems are designed to be cheap, so tend to use a relay. Online UPS systems are designed to be high end, so the delay of relay switching is unacceptable, so online UPS systems usually use IGBTs to switch between inverter and bypass mode.
Invensys sold off the UPS business to Eaton Corp, they already made a lot of the panel gear that the UPS uses so it was a great fit for them. They build the big systems here in Raleigh NC. Liebert is big on UPS and Data Center Cooling, they are part of Emerson. APC was bought and is now part of Group Schneider who owns Square D, a big electrical panel and circuit protection corporation. They made a lot of stuff here in Zebulon, NC but moved all of it to Mexico.
Sorry for two replies - now I think about it they have some form of active rectification control since it has to be a boost device (acting as an inverter it is a buck device). Hence the short circuit "charge state". But it may not be that high frequency to bother the body diodes.
Great video. I've learned a lot from watching your videos. I liked the bit of theory at the beginning. It helped me understand better what was going on and being said during the teardown. Keep up the good work
I've seen almost identical transformers in 'smaller' rack mount supplies, but they had just one and had 24V battery banks. Pairing them allows for fitting the whole thing in a 3U case without having to design a custom squat transformer. Also, small ones (
Just googled and found this on the wikipedia page: "Body diodes may be utilized as freewheeling diodes for inductive loads in configurations such as H-bridge or half bridge. While these diodes usually have rather high forward voltage drop, they can handle large currents and are sufficient in many applications, reducing part count, and thus, device cost and board space." Exactly what you said dave!
Free wheeling diodes are needed on inverter designs, even if they are not used in this particular charger implementation. In high end designs, discrete high speed diodes are used, to reduce power disipation on the substrate diodes
Amazing video! Thanks for explaining how each type actually works. One tiny correction, though: at 6:22, the fact that for a given power, a lower voltage implies a higher current, is not Ohm's law, it's the relation between those three quantities (what we may call the power equation): P = V * I. Ohm's law has nothing to do with that.
THANK. YOU! SO many videos on here that just feel like surface level, no real understanding, reviews / analyses. Takes an Aussie to actually do a thorough job... You're Australasia's Gerald Undone mate! :D Thanks again
Hey Dave, the same big transformers that step up the voltage is also used to charge the batteries. That is why they put a lot of relays and series of output wires in the secodary side of the transformer. During mains off, the side of transformer connected to the heat sink serves as primary and when mains on ,it turns to seconďary to charge the battery. If you look around the heat sink, you will see some rectifiers.
Thought I'd make a correction on you DaveCAD @20:00.The H-bridge in the UPS is all N-channel. The N-channels are connected backwards in place of your P-channels. The reason for using N-channel over P-channel is that N-channels have a lower RDS then there P-channel counterpart.
I tend to agree Binray - it won't be switching particulatly fast. There is a fair bit of series inductance in those cores (needed for the boost/active rectifier) and 13uF of output capacitance (3x4.7uF) to filter the sine wave.
I got one hooked up to my 3d printer. No more shall power outages waste my material! Also, a bit of a question: If I have a triac-based power supply for a high power (~400W) device, would those pulses of high current damage the UPS? P.S. The device would be a heater is for the heated build plate. It really only needs to be UPS-protected for long periods of time because of the thermal mass.
IBM used to sell their mainframes with a system which produced fully synthetic AC power. It was a huge machine with six independent three-phase inverters.
APC did have some interesting design for UPS's. And they're HQ'd right here in Rhode Island. It's funny - two companies I know of in RI are Alesis of the synthesizer fame, and APC.
Funny you would say that. I read an article in a power utility magazine about switching the power grids of the world over to dc in the next 20 years. Advances in solid state technology is making it possible to build sub-station and pole-pig transformers into switching power supplies DC2DC as well as AC. Both AC and DC system can and will both co-exist until the system migrates over to full DC.
Good video Dave, and please, don't worry yourself over what the over-entitled fringes of the comments demand you do. keep up the great work, you're an inspiration
Totally agree. Google operate on such a huge scale even a small energy saving per server makes it very cost effective. The scale also allows them to make it worthwhile designing their own hardware. Years ago i had some servers with a hybrid PSU made by a company called Magnum Power Solutions in Scotland. Like a traditional PC PSU but with UPS functionality. Very neat design, it would run off the internal battery if mains failed. However it could not cold-start from DC.
Absolutely, I was thinking about getting a UPS for my networking gear at home (modem, router, switch) and then thought "Hey, all this runs on 12V DC, why use a 120V AC inverter to waste battery" and decided that I should just build one with maybe a 24V battery and some efficient switching regulators, plus a float charger. It'd probably last longer too.
Got a couple in the garage that I need to do a teardown on sometime. Two SmartUPS 2200RMXL (one 5U, one 3U), a Galatrek Microbak ferroresonant 1.5kVA UPS and a transformerless NuWave Powervalue 12kVA.
This topology likely uses an "active" rectification technique, think full bridge rectifier but substitute the diodes with mosfets. By turning the right mosfets on and off at just the right time you essentially have an extremely efficient full bridge rectifier. The mosfets drop far less voltage than diodes.
The remains of Best were ultimately gobbled up by either APCC or their parent, Schneider. I can't remember who bought MGE (one of the last intermediaries that had Best) ... I had a Best FerrUPS that used that topology, big bugger. Carving circuit big enough that when I hacked it and added several more external SLAs it didn't decrease the charging time by much.
that option bay could be the information centre, where it gives you all the readings and status. if one doesn't want to hook it up to laptop via USB and software.
As to why, all I can say is some equipment doesn't play nicely with square waves, possibly due to the design of the power supply. Another reason might be it uses phase angle control (such as in a light dimmer or hand held non-cordless power tool). As for the cost, I'd say it's mostly due to the parts. To make a sine wave you need to do Sinusoidal PWM, which requires fast switching and either extra electronics or a reasonably fast processor.
I grabbed two HP R3000 XR UPS's a while ago, unfortunately UPS(the shipper) managed to throw it off a cliff. Arrived in pieces but I managed to fix one of the two units. I'll post some pictures on the EEVBlog forum if I can find them.
@Giza ,Like you could do any better working in the patent system. What you are suggesting is obvious to everyone. Just because technology switches rapidly doesn't mean the paths that were taken to get to this point is trivial.
Love this "Fundamentals & Teardown" mix. Even if you know the Fundamentals, it is a lot of fun to actually look into the device. Will open up my UPS any time soon :D Mine might even need a bit of repair as well.
I have an APC Smart-UPS XL Modular 1500VA. acquired from a local server management building after upgrades. I also have a few of those small 400W personal computer ones, only one with a battery still, the rest dried out.
I have used many, many huge (100kva and up) for data centers. The "U" in UPS means UNINTERRUPTABLE. This means ZERO power transfer time delay and NO PHASE shift. The output power is completely obtained by the inverter. This is what a true UPS does anything else is a STANDBY lower supply. I had one in a large data center that supplied global credit card processing systems (all mainframe stuff) the UPS room was 625 sq feet. I contained racks of batteries and a unit on the wall that just buzzed. It supplied 3 phase power to the data center. With all thay power we only had 45 min if battery power. We didnt need much because we had a generator that would activate if we lost power so at most the batteries were needed for 2 or 3 min until generator stabilized
Felicitaciones! Me parece un video muy completo y has ganado un seguidor, reunir todos estos datos, la presentación y la explicación han sido espectacular!, gracias por tu dedicación y tiempo,have a nice day!
I was having blackouts and was worried that the sudden power drops may damage my TV. I bought a 115 amp 12 volt solar battery. I connected a really good charger to keep it charged up. I then bought and installed a 400 watt sine wave inverter. I put fuses in for safety. This is all in my garage. My TV and DVD player only draw 105 watts of electricity. I only use the TV for 3 hours per night. The TV power cord is full time connected to the inverter. The inverter is switched on and off via a remote control. This all cost 480 dollars and I have had no worries so far for over 2 years. The power goes down but my TV is safe and just keeps right on going. The thing that you pulled apart is just an expensive wank on as far as I can see. More for rich geeks than us "ordinary folks".
+7316bobe UPSes are very common and useful devices, and although he gave your home PC as an example, they are not often used by consumers. These are mostly used for commercial industrial applications or to provide clean power and graceful shutdown for important equipment. These things can be small bricks for one device or massive units for many. For an individual device you won't see these often in someone's home, but you will see it connected to point of sale systems in stores and banks. Your solution is interesting but not very practical for other implementations, especially since it's a solar battery. It's silly to build something for extra reliability that draws its power from an unreliable source.
Not in this design, they are used to filter the h-bridge rectified back-fed mains into the battery, which isn't going to like AC rippled DC charging current.
also looking forward to a followup video with oscilloscope poking. Too bad it's a rather "uninteresting" design internally(it's just "beefy" for the power), i've seen a SURT5000 (realtime, online) and it's an absolujte joy to see inside, they don't have a single big transformer, it's all toroidal core(and several) high freq transformers inside, also it had ~ 8 BIGGER caps as well and the pack operates at 96V
So, the purpose of the patent might be to make product manufacturing cheaper - no separate charger required. But for unknown reasons I don't notice that APC UPSes would be cheaper than others. On the other hand, now almost everybody seems to be using similar design. I have seen some Powercom UPSes having separate chargers in their midrange UPSes but that was long time ago, so maybe now they also use the same charger-less design.
You're a brave man poking around those circuits with a metal screwdriver. I'm sure the circuitry still has quite a bit of capacitance left over that could stop your heart.
The charge current for the batteries is much lower than the discharge current, so they can get away with it. I would expect the charge current to be a few amperes at most as SLA batteries do not like high charge currents.
22:58 I notice a good air path to the main cooling fan around those capacitors, so they should stay comfortable enough for the time it takes your server to shut itself down gracefully. I doubt a company like APC would cheap out on electrolytics in something that is ostensibly mission-critical infrastructure.
I prefer having AC on the power grid because a lot of high power devices use it and it's easy to convert voltages with transformers and capacitors. Building oscillator circuits (to convert DC to AC) is a nightmare, but using diode bridge for the opposite is cheap and simple.
have had an UPS on my computer system since 2006. Saved my arse plenty of times.. Maybe time to change the batteries in it. But the diagnostics say all is 'Normal' Vis-à-vis: UPS Status: High Efficiency Auxiliary Mains Star Voltage (V): 240 Output Star Voltage (V):: 237 Output Load (%): 30 Battery Capacity Remaining (%): N/A Battery Voltage (V):: 27.6 Internal UPS Temperature (°C): 25
I want your 555 timer circuit shirt! Lol. I've read up on this before, but i have to say, i'd much rather watch you explain sometimes.. I'm often told i can bring some things to life which are otherwise boring but you do it so damn well.. You should present how it's made/how it works/other similar programmes because you make everyone else look shit!
31:26 you are right but sightly wrong they used Not h bridge MOSFETs as rectifier instead they used the Body diodes of the MOSFET used in H bridge , as rectifier .
Always walk away a little smarter after watching any of your videos. The big UPS players since the 1980's are Liebert, Best Power Systems, Exide and a few smaller ones. APC got their start making inverters for solar panel systems but that business was not profitable once Congress in the US took away the IRS tax credits for solar power. APC then turned their expertise into UPS systems. Exide was spun off to Invensys, then that got sold off too.
This should not be a problem at all. First it's just 50Hz, so no worries about the speed of the diodes. Then according to the datasheet the diode can take 75 amps (300A pulsed). Most LA batteries don't like charge currents over 0,3C. So it would take approx. at least a 300Ah battery to overload a single diode. Sure doesn't look like one. Btw. since the diodes are thermally coupled through the heatsink, it's very improbable that only one of the parallel diode will conduct anyway.
Love the way the bridge is used as a rectifier. 👍🏻
Genius !!!
This is an "extended run" APC UPS as designated by the "XL" suffix on the part number. 2 things make it extended run: 1) the secondary battery input and 2) additional cooling fans.
The extended run models are "highly desirable" for use as straight inverters because on the standards models, the transformers will overheat within about 30 minutes of heavy continuous use.
Of course, now that I told you that the transformer is the "weak link" you can add your own additional cooling including thermal compound between the bottom of the transformer and the case !
Nice explanation Dave, I really like your videos on DFM.
A quick comment on how the charging work:
Voltage reflecting on the battery side of the transformer (primary side) will never be high enough to charge the battery through the four mosfet body diodes. (battery shouldn't be charged this way because you cannot control the charge current and stop charging appropriately)
In order to charge the battery, the mosfet H-bridge works like a boost converter, boosting the voltage to high enough to charge the battery, and at the same time control the charge current and do PFC (Power factor correction).
In inverter mode, the H-bridge works like a buck converter having output voltage swing at line frequency.
Half-bridges or full bridges are bidirectional converters.
You can easily control the power flow both ways by applying correct PWM. Usually, an algorithm called dq control (similar to field oriented control) is used to control the power flow. For example, when vector q is positive, power flow forward, and when q negative, power flow backward.
This idea is not new, but applying the idea to UPS to simplify the design is really smart.
How the H-bridg work as rectifier
why don't they teach it in college, the topology practical design. When I opened up my 800VA APC UPS, I wonder how there's only one power transformer who's responsible for AVR, charging, and inverter, with only 3 cable for high voltage and 2 cable for low voltage.
But after all these years why don't they switch to high frequency transformers, instead staying with the bulky heavy transformer. I noticed these designs get warm even without load consuming around 20 watt, that amount does affect the energy bill for 400W UPS running 24/7, and all UPS I can find does that except for the DC UPS's.
Well done Dave, a hefty Tuesday video bringing us a complete explanation of these various UPS technologies. Great tutorial!!
This video is very timely, in my case, I watched this video as a study break from studying for my Power Electronics final exam. It was entertaining AND informative. Thanks Dave! Keep up the great work!
Hi
To answer your question @28:43. The MOSFETs in the inverter are also used as the rectifier. It called a "Dual Headed Synchronous Rectifier"(DHSR). In a DHSR configuration. The controller synchronized to the mains/line and then switched the MOSFETs on and off in sync of the AC wave form. It also uses PWM to control the battery current as well as perform power factor correction. Hook your oscilloscope up to it and look at it.
Usually that's for high voltage isolation. Helps reduce surface leakage and capacitance a bit.
The option slot is for installing additional interfaces i.e. giving the UPS a web UI so you can manage it remotely.
Actually I do believe that they are turning on the mosfets at the same time that the body diode would be on (as stated in the comment you commented on). It would have been a simple enough optimization to figure out, and they would have the control circuitry that would otherwise be dormant. It could also allow them to do some power factor correction to make the charge circuit seem more like a purely resistive load.
I am rewatching this video and I think it is Dave's best video ever!
Awesome find!
There is another type of UPS - transformerless double conversion. Just had a 40kva 3 phase unit installed. Batteries are in series and the DC side runs at 480v. This is common for large, fixed installations. It can be bypassed externally without dropping the load e.g. for servicing. When going out of external bypass it will automatically sync the inverter with the mains supply after which you can flick the bypass off. No interruption to critical load at all.
How do you not know what Best Power is? I mean the company has been split into pieces and sold for quite a while now but that place was such a legacy. It's building is still standing in Necedah, WI being rented out by the original owner. That place is such a damn throwback, my parents met while working there and my great grandmother was one of the first to get an award for exemplary work. I was practically raised there. On the upside, since closing down, my dad has started up his own business based around UPS.
Great video, helps me a lot to figure out my scrappy $20 APC Smart-UPS 2200VA. i GOT THE SAME !
Difference is that i have one without the slide out battery bay, Only 2 big capacitors placed far behind the path of cooling air through the H bridge (Seems logical). Basically you have 8 small NP7-12 Lead Acid batteries in a weird configuration. APC made 2 separate 24v sections (with 2 Anderson connectors) in series (48v 14A total) with a 100A fuse in the middle. Both sections consist of again 2 individual banks of again 2 NP7-12 batteries in parallel (= 24v 7A) again in series = 24v 14A. The leads get bigger at every level. Basically the charging current is only 1A max (Maybe even a lot lower) for these 8 tiny 12v batteries switched that way. The total charge current at 48v is about 4A (only 200 Watt). Discharging each quickly ramps up (ads up voltage and amperage in the chain) the capacity of this 48v battery without damaging them individually. I only wonder about the charging profile. Does this Interactive system stop charging at all at some point if batteries reach the full capacity and if not does this explain why all the old AGM batteries in these units seem exploded or expanded with bubbles on all sides? Does this pyramid of 2x2x2 batteries equalize automatically if you leave the power disconnected? I ask because it would be possible to replace every single NP7-12 battery with a 10 x AA (1.2v) Ni-MH battery pack, Haha. Picture link: www.dropbox.com/s/leg4t0lyqml42mh/2020-03-14%2018.40.03.jpg?dl=0
Pretty sure it's OK to use the body diode for 50Hz rectification, most mains frequency diodes are pretty crummy in terms of switching/reverse recovery performance. Not sure about the body diode, but (some?most?) MOSFETs have a positive temperature/resistance coefficient, thus they share current pretty well. Series diodes are much more of a pain since you have to get voltage sharing.
As for the large number of caps, that's not unreasonable for single phase -> 100Hz power flow and all...
This guy is really good with explanations and illustrations.
I seen this and it took me back. I worked for a large UPS maker for 10 years. I went to school for APC Anton Piller Belkin Kw controls and a few others . The company also did motor gen sets up to 500kva I could tell you every connection and what to look for to repair it. They had on line UPS and non online UPS systems . I even set up the systems for several companies including Blue Cross . I could tell you so much about UPS systems it would make your head swim.
Nice feature of the APC SmartUPS, if you have a power failure, you can turn them on and they will run off the internal or external batteries w/o any mains connection.
Many of us use these older models that nobody wants for our older amateur radios, aka Ham Radios. These older circa 1975 radios have internal transformer they do not like any square or modified sine waves.
The APC SmartUPS put out full sine wave and we can run the radio off batteries for emergency use.
I disassembled a large online UPS, and it didn't follow that schematic. It was near identical to your "offline" schematic - the main difference was that the step-up transformer was an SMPS and went before the inverter. The battery voltage was stepped up from 400 volts to 700 V (+/- 350 volts), allowing a simple push-pull topology inverter.
The difference is that in an offline UPS, the load is directly supplied by the mains. If the mains fails, then the transfer relay switches the load to the inverter. In the online UPS, the load is supplied by the inverter, unless the inverter fails, in which case it transfers back to mains.
The common reason for inverter "failure" is actually the output being severely overloaded or shorted (e.g. due to a fault on a single load where a UPS supplies multiple individual loads). In this case, the inverter voltage drops under the overcurrent stress, and the load is transferred to mains with its high current supply capability. This can also happen if regenerative loads are connected to a UPS (e.g. motor drives) and the reverse power exceeds the handling capability of the inverter causing it to trip out due to DC bus overvoltage.
There is usually a difference in the type of switching. Offline UPS systems are designed to be cheap, so tend to use a relay. Online UPS systems are designed to be high end, so the delay of relay switching is unacceptable, so online UPS systems usually use IGBTs to switch between inverter and bypass mode.
Invensys sold off the UPS business to Eaton Corp, they already made a lot of the panel gear that the UPS uses so it was a great fit for them.
They build the big systems here in Raleigh NC.
Liebert is big on UPS and Data Center Cooling, they are part of Emerson.
APC was bought and is now part of Group Schneider who owns Square D,
a big electrical panel and circuit protection corporation. They made a lot of stuff here in Zebulon, NC but moved all of it to Mexico.
Sorry for two replies - now I think about it they have some form of active rectification control since it has to be a boost device (acting as an inverter it is a buck device). Hence the short circuit "charge state". But it may not be that high frequency to bother the body diodes.
Great video. I've learned a lot from watching your videos. I liked the bit of theory at the beginning. It helped me understand better what was going on and being said during the teardown. Keep up the good work
I've seen almost identical transformers in 'smaller' rack mount supplies, but they had just one and had 24V battery banks. Pairing them allows for fitting the whole thing in a 3U case without having to design a custom squat transformer.
Also, small ones (
Just googled and found this on the wikipedia page: "Body diodes may be utilized as freewheeling diodes for inductive loads in configurations such as H-bridge or half bridge. While these diodes usually have rather high forward voltage drop, they can handle large currents and are sufficient in many applications, reducing part count, and thus, device cost and board space." Exactly what you said dave!
Free wheeling diodes are needed on inverter designs, even if they are not used in this particular charger implementation. In high end designs, discrete high speed diodes are used, to reduce power disipation on the substrate diodes
The patent does not expire because the company goes bust. Of course if there is no one left to sue you that does help a lot!
Amazing video! Thanks for explaining how each type actually works.
One tiny correction, though: at 6:22, the fact that for a given power, a lower voltage implies a higher current, is not Ohm's law, it's the relation between those three quantities (what we may call the power equation): P = V * I. Ohm's law has nothing to do with that.
I must catch up on your videos Dave ! Did not know you had this video on UPS! GREAT video !
THANK. YOU!
SO many videos on here that just feel like surface level, no real understanding, reviews / analyses. Takes an Aussie to actually do a thorough job... You're Australasia's Gerald Undone mate! :D Thanks again
A teardown with whiteboard fundamentals is awesome!
Hey Dave, the same big transformers that step up the voltage is also used to charge the batteries.
That is why they put a lot of relays and series of output wires in the secodary side of the transformer.
During mains off, the side of transformer connected to the heat sink serves as primary and
when mains on ,it turns to seconďary to charge the battery. If you look around the heat sink, you will see some rectifiers.
Thanks Dave, very well explained. I've been using these units for quite some time and finally could see how they work!
Yes, I have several Online and Line interactive UPS's and you can hot swap batterys
Thought I'd make a correction on you DaveCAD @20:00.The H-bridge in the UPS is all N-channel. The N-channels are connected backwards in place of your P-channels. The reason for using N-channel over P-channel is that N-channels have a lower RDS then there P-channel counterpart.
My brother's UPS is a Back-UPS XS 1200...the XS series is a line-interactive UPS with a built-in power conditioner, similar to what OneAC makes.
I tend to agree Binray - it won't be switching particulatly fast. There is a fair bit of series inductance in those cores (needed for the boost/active rectifier) and 13uF of output capacitance (3x4.7uF) to filter the sine wave.
I got one hooked up to my 3d printer. No more shall power outages waste my material!
Also, a bit of a question: If I have a triac-based power supply for a high power (~400W) device, would those pulses of high current damage the UPS?
P.S. The device would be a heater is for the heated build plate. It really only needs to be UPS-protected for long periods of time because of the thermal mass.
Good one! I liked the way this combined theory and tear down told a story.
i loved the charging system the way it fed it back , very clever
IBM used to sell their mainframes with a system which produced fully synthetic AC power. It was a huge machine with six independent three-phase inverters.
Thanks Dave, For excellent explanation about UPS system.
APC did have some interesting design for UPS's. And they're HQ'd right here in Rhode Island.
It's funny - two companies I know of in RI are Alesis of the synthesizer fame, and APC.
Great video Dave! I enjoyed the theory and teardown coupled together, was very informative.
Using the by pass diodes as bridge retifier is nice!
Funny you would say that. I read an article in a power utility magazine about switching the power grids of the world over to dc in the next 20 years. Advances in solid state technology is making it possible to build sub-station and pole-pig transformers into switching power supplies DC2DC as well as AC. Both AC and DC system can and will both co-exist until the system migrates over to full DC.
That is a real piece of good engineering...and one proove why APC is one of the most reliable.
Good video Dave, and please, don't worry yourself over what the over-entitled fringes of the comments demand you do. keep up the great work, you're an inspiration
So good, love how you explain everything, keep it up! Cheers from Sweden!
Totally agree. Google operate on such a huge scale even a small energy saving per server makes it very cost effective. The scale also allows them to make it worthwhile designing their own hardware. Years ago i had some servers with a hybrid PSU made by a company called Magnum Power Solutions in Scotland. Like a traditional PC PSU but with UPS functionality. Very neat design, it would run off the internal battery if mains failed. However it could not cold-start from DC.
Nice video Dave,
i use the 750VA Smart UPS rack version of this series for my pc
and i am very satisfied with it
I like the hybrid lecture teardown idea with a twist!
Absolutely, I was thinking about getting a UPS for my networking gear at home (modem, router, switch) and then thought "Hey, all this runs on 12V DC, why use a 120V AC inverter to waste battery" and decided that I should just build one with maybe a 24V battery and some efficient switching regulators, plus a float charger. It'd probably last longer too.
Got a couple in the garage that I need to do a teardown on sometime. Two SmartUPS 2200RMXL (one 5U, one 3U), a Galatrek Microbak ferroresonant 1.5kVA UPS and a transformerless NuWave Powervalue 12kVA.
This topology likely uses an "active" rectification technique, think full bridge rectifier but substitute the diodes with mosfets. By turning the right mosfets on and off at just the right time you essentially have an extremely efficient full bridge rectifier. The mosfets drop far less voltage than diodes.
The remains of Best were ultimately gobbled up by either APCC or their parent, Schneider. I can't remember who bought MGE (one of the last intermediaries that had Best) ... I had a Best FerrUPS that used that topology, big bugger. Carving circuit big enough that when I hacked it and added several more external SLAs it didn't decrease the charging time by much.
that option bay could be the information centre, where it gives you all the readings and status. if one doesn't want to hook it up to laptop via USB and software.
This is terrific, thank you Dave!
Very cool review, thank you. If you would have done the review of UPS online class would be just super.
Yes, this one uses only N channel MOSFET's according to the schematic.
The way comments were arranged, it shows that the video was created before the "reply" feature is added in TH-cam
Another insightful video, can't wait to see your videos as soon as they come out!
As to why, all I can say is some equipment doesn't play nicely with square waves, possibly due to the design of the power supply. Another reason might be it uses phase angle control (such as in a light dimmer or hand held non-cordless power tool).
As for the cost, I'd say it's mostly due to the parts. To make a sine wave you need to do Sinusoidal PWM, which requires fast switching and either extra electronics or a reasonably fast processor.
At 14:24 + 15:06 - They are actually 16 amp connectors but could only be supplied with 15 amps anyway from an Australian type plug
To get a sine wave output it must run PWM at a higher frequency.
this technique is used in very similar way to recuperate energy from BLDC motors back into batteries.
i use my back-ups 450 with a 12v 75ah sla batterys in parallel and i have a lot of time when the main goes out ! nice vid !
I grabbed two HP R3000 XR UPS's a while ago, unfortunately UPS(the shipper) managed to throw it off a cliff. Arrived in pieces but I managed to fix one of the two units.
I'll post some pictures on the EEVBlog forum if I can find them.
@Giza ,Like you could do any better working in the patent system. What you are suggesting is obvious to everyone. Just because technology switches rapidly doesn't mean the paths that were taken to get to this point is trivial.
Love this "Fundamentals & Teardown" mix. Even if you know the Fundamentals, it is a lot of fun to actually look into the device. Will open up my UPS any time soon :D Mine might even need a bit of repair as well.
i built a 200W ups for learning with same charging technique few years back. didn't know it has Patent.
I have an APC Smart-UPS XL Modular 1500VA. acquired from a local server management building after upgrades. I also have a few of those small 400W personal computer ones, only one with a battery still, the rest dried out.
Fun! I just recycled one of these. Interesting to see the inside of it.
I have used many, many huge (100kva and up) for data centers. The "U" in UPS means UNINTERRUPTABLE. This means ZERO power transfer time delay and NO PHASE shift. The output power is completely obtained by the inverter. This is what a true UPS does anything else is a STANDBY lower supply.
I had one in a large data center that supplied global credit card processing systems (all mainframe stuff) the UPS room was 625 sq feet. I contained racks of batteries and a unit on the wall that just buzzed. It supplied 3 phase power to the data center. With all thay power we only had 45 min if battery power. We didnt need much because we had a generator that would activate if we lost power so at most the batteries were needed for 2 or 3 min until generator stabilized
You were reading my mind for a future video Dave. Thanks
Felicitaciones! Me parece un video muy completo y has ganado un seguidor, reunir todos estos datos, la presentación y la explicación han sido espectacular!, gracias por tu dedicación y tiempo,have a nice day!
35:43 Made in India..... Wow....😲
I was having blackouts and was worried that the sudden power drops may damage my TV. I bought a 115 amp 12 volt solar battery. I connected a really good charger to keep it charged up. I then bought and installed a 400 watt sine wave inverter. I put fuses in for safety. This is all in my garage. My TV and DVD player only draw 105 watts of electricity. I only use the TV for 3 hours per night. The TV power cord is full time connected to the inverter. The inverter is switched on and off via a remote control. This all cost 480 dollars and I have had no worries so far for over 2 years. The power goes down but my TV is safe and just keeps right on going. The thing that you pulled apart is just an expensive wank on as far as I can see. More for rich geeks than us "ordinary folks".
+7316bobe UPSes are very common and useful devices, and although he gave your home PC as an example, they are not often used by consumers. These are mostly used for commercial industrial applications or to provide clean power and graceful shutdown for important equipment. These things can be small bricks for one device or massive units for many. For an individual device you won't see these often in someone's home, but you will see it connected to point of sale systems in stores and banks. Your solution is interesting but not very practical for other implementations, especially since it's a solar battery. It's silly to build something for extra reliability that draws its power from an unreliable source.
Not in this design, they are used to filter the h-bridge rectified back-fed mains into the battery, which isn't going to like AC rippled DC charging current.
Yep, they PWM at high frequency, but the duty cycle is modulated at 50/60Hz.
Thank you I really good to find service manual when I don't need them, when I need them I cannot find nothing free!
also looking forward to a followup video with oscilloscope poking.
Too bad it's a rather "uninteresting" design internally(it's just "beefy" for the power), i've seen a SURT5000 (realtime, online) and it's an absolujte joy to see inside, they don't have a single big transformer, it's all toroidal core(and several) high freq transformers inside, also it had ~ 8 BIGGER caps as well and the pack operates at 96V
Great video.. I'd love to see some some scope captures of the current flowing into the batteries during charge. Have a current probe?
So, the purpose of the patent might be to make product manufacturing cheaper - no separate charger required. But for unknown reasons I don't notice that APC UPSes would be cheaper than others. On the other hand, now almost everybody seems to be using similar design. I have seen some Powercom UPSes having separate chargers in their midrange UPSes but that was long time ago, so maybe now they also use the same charger-less design.
Beautiful. Appreciate you work, sir
Very professional and very informative. Thanks for making and sharing this video.
You're a brave man poking around those circuits with a metal screwdriver. I'm sure the circuitry still has quite a bit of capacitance left over that could stop your heart.
The charge current for the batteries is much lower than the discharge current, so they can get away with it. I would expect the charge current to be a few amperes at most as SLA batteries do not like high charge currents.
The online kind of UPS can be designed without the switching mechanism using a bit of diode magic.
Also known as 'buck' topology.
Please Explain i need to know
Percussive maintenance. LOL! Excellent shop term! Never heard that one.
22:58 I notice a good air path to the main cooling fan around those capacitors, so they should stay comfortable enough for the time it takes your server to shut itself down gracefully. I doubt a company like APC would cheap out on electrolytics in something that is ostensibly mission-critical infrastructure.
I prefer having AC on the power grid because a lot of high power devices use it and it's easy to convert voltages with transformers and capacitors. Building oscillator circuits (to convert DC to AC) is a nightmare, but using diode bridge for the opposite is cheap and simple.
have had an UPS on my computer system since 2006. Saved my arse plenty of times.. Maybe time to change the batteries in it. But the diagnostics say all is 'Normal'
Vis-à-vis:
UPS Status: High Efficiency
Auxiliary Mains Star Voltage (V): 240
Output Star Voltage (V):: 237
Output Load (%): 30
Battery Capacity Remaining (%): N/A Battery Voltage (V):: 27.6
Internal UPS Temperature (°C): 25
I think I learned something new today, thanks Dave!
Thank you for this video. I subscribed and I stand to learn a LOT from you.
The specs tell you it's a sine wave inverter.
I want your 555 timer circuit shirt! Lol. I've read up on this before, but i have to say, i'd much rather watch you explain sometimes.. I'm often told i can bring some things to life which are otherwise boring but you do it so damn well.. You should present how it's made/how it works/other similar programmes because you make everyone else look shit!
31:26 you are right but sightly wrong they used Not h bridge MOSFETs as rectifier instead they used the Body diodes of the MOSFET used in H bridge , as rectifier .
Oh you already mentioned it later here 31:56
Yes, there are many different flavors.
How would you disable charging of the battery only?(in case of multiple ups on same battery). Without using diodes in series with the battery.
Always walk away a little smarter after watching any of your videos.
The big UPS players since the 1980's are Liebert, Best Power Systems, Exide and a few smaller ones.
APC got their start making inverters for solar panel systems but that business was not profitable once Congress in the US took away the IRS tax credits for solar power.
APC then turned their expertise into UPS systems.
Exide was spun off to Invensys, then that got sold off too.
This should not be a problem at all. First it's just 50Hz, so no worries about the speed of the diodes.
Then according to the datasheet the diode can take 75 amps (300A pulsed). Most LA batteries don't like charge currents over 0,3C. So it would take approx. at least a 300Ah battery to overload a single diode. Sure doesn't look like one.
Btw. since the diodes are thermally coupled through the heatsink, it's very improbable that only one of the parallel diode will conduct anyway.