On DIY Battery Powered DC UPS, for Portable SFF and Off Grid

[msystems]

I have been researching how to buy or build what is essentially a high current DC uninterruptible power supply and power pack for portable SFF / off-grid purposes.

You may be wondering what purpose this would serve over using a typical UPS/house battery, or just using a laptop. To be honest, I have no compelling reason for perfecting this system yet, but I foresee a scenario in the future when self-tracking VR/AR headsets improve in resolution to the point that they can replace external monitors for productivity applications, and also, if VR/AR OS' continue to improve to the point of simulating a desktop OS', as we are seeing with development of Oculus Dash, ect.

If this happens, the need for a portable "headless" VR/AR rig would be real. No external monitor would be necessary, you would just want what is essentially a suitcase with a beefy GPU and lots of battery life.

For off-grid use, there is already a reason to build this today, in that by running your rig on DC power will skip the inverter loss which happens twice, from the house battery to AC, and then from the psu's AC to DC, thus giving more battery life (assuming you can charge this system from DC).

After researching available solutions, it became obvious that modifying a traditional UPS with an inverter and AGM battery is way too bulky and inefficient to power such a system. The solution needs to use Lifepo4 or Li-ION chemistry. Consumer "Power Packs" (for phones and USB devices), while inexpensive, are not appropriate for this as they do not offer nearly enough discharge current.

A solution actually already exists in the form of OpenUPS, if one has the expertise to use it:

- Programable output 6-24V
- Default output is 12V
- Input/Output current 6A/10A peak
- Charge between 6-30V, 3A
- Voltage and current limits
- Li-Ion, Li-Po, LIFEPO4***, Lead Acid
- 1, 2S, 3S, 4S, 5S, 6S

And OpenUPS2, which is only a 2.5" drive footprint and offers programmable output at 12-24V/5A.

These are really cool! However 5A or even 10A is not enough for such a rig, and we need more current, like about 30A @ 16v which would provide a hefty 480 watts. After a few hours of searching I stumbled upon this video, which remarkably confirms this is indeed possible 100%:


All of the heavy lifting was done by the creator of this video. As you can see, he connected a 20v, 20,000mAh RC LiPo battery directly to the HDPlex barrel connector on the HDPlex 400w DC-ATX and it powers it perfectly. The fact that the HDPlex 400w accepts a variable 16v-24v is really useful here. As long as the battery pack discharges above 16v, it should power the rig without issues. This wide range should allow for a variety of battery wiring configurations to discharge within the voltage range.

The difficulty comes in picking the right battery for this application. I am not sure how viable it is to use a RC LiPo battery as those sometimes those do not last past 100 cycles and they are not the safest. Compare this to 18650 Li-Ion cells, which can be discharged thousands of times. Based on my limited research, a safer method would seem to use a Li-Ion battery pack, for example, in a 6S6P configuration.

Cell type 18650
Cell model Panasonic NCR18650B
Cells in total 36 x 18650 cells
SIZE Weight, max. 1.74 kg
Volume ~804.56 cm³
Length, max. 11.10 cm (approximate)
Width, max. 11.10 cm (approximate)
Height, max. 6.52 cm (approximate)
VOLTAGE Voltage, charge max. 25.20 V
Voltage, nominal 21.60 V
Voltage, discharge end 15.00 V
CAPACITY Capacity, max. 20.10 Ah
CURRENT Charge current, standard 9.75 A
Max. continuous discharge current 29.25 A
Peak discharge theoretic, 4 sec 54.00 A
C-rate (discharge, max.) 1.46 C
POWER Watts (discharge, max.) 631.8 W
ENERGY Energy, max. 434.16 Wh
Density gravimetric theoretic 249.00 Wh/kg
E-rate, discharge max. 1.46 E

This battery pack has a voltage range of 15v to 25.20v. By charging it to only 24v instead of 25.2v, and discharge at 16v instead of 15v, it should preserve the battery lifetime considerably. This is a highly configurable and longer lasting battery pack but is considerably more time consuming and expensive to create since it will need a programmable battery management system to control the cell balancing and match the voltage range to the 16v-24v input on the HDPlex. Many cheap chinese BMS seem to be available however it was spotty finding information on those. Here is an example of one. Higher quality ones (not china) seem to be considerably more expensive. It takes technical knowledge, time, tools and skill to wire the BMS to the battery pack.

Instead of doing that I am still looking for a battery pack which falls in the desired voltage range and is already built. Apparently "E-bike" batteries are similar to what is needed. Here is one for example which might work.

 

[owliwar]

uh, I really like this idea. an AR portable rig.
even if we do use a notebook, for the advanced power saving these chips have, A easy battery solution like this could really be useful for a lot of stuff.

how long a battery like these could last? I don't really know how to do the math haha

 

[msystems]

how long a battery like these could last? I don't really know how to do the math haha

For the 6s6p battery pack above, 3 hours @ 150w or 1.5 hours @ 300w.
For "2D" use, it should last in excess of 8 hours @ 50w. A SFF system should only draw 35-50 watts in desktop and idle really.

This does not come cheap though as high quality Li-Ion batteries alone will cost about $5 each, so for 36 of them that is already $180 USD and that does not include the balancing and charging system.

 

[darksidecookie]

I have looked into this for about the same reason; that being using the deskmini gtx in a backpack with battery for vr.
And i came to the same conclusion as you but thru a different route.

This was the website i came across: https://randomfoo.net/2016/05/30/diy-backpack-vr-system

And I also noticed this guy was on the forum under the name @lhl in this thread :https://smallformfactor.net/forum/threads/ono3-pelican-1430-to-s4-portable-vr-workstation.423/

 

[Kmpkt]

I've already developed this for my wearable PC. Runs off 19V Bosch 6.0A batteries and is scalable up to around 1300W. Compatible with HDPlex and Dynamo power systems. If you're interested hit me up.

 

[lhl]

Yeah @darksidecookie been there done that, let me know if you need any help but I recommend the OP just read through my writeups you linked, as I think they're reasonably thorough. As mentioned, a 5S or 6S battery makes life easiest as you can basically just plug it straight into an HDPLEX with barely any soldering (just need to an XT90 to barrel connector). You can pick up a Zippy 8Ah 5S 30C for $60. Note, the max size you can fly with reliably (100Wh limit) is a 5400mAh 5S - that gets you to 99.9Wh).

For a little more money you can put together your own 18650 or 20650/21700 cell battery, there are plenty of 10-20A options that should give you better energy density (Wh/kg) and handle the draw w/o problem (5Sx10A gives you 50A discharge rate, 30C is really overkill for a portable PC) but you'll need to do a more work (spot welder for connecting bus bars mostly). Either way, I've been pretty happy w/ the iSDT Q6 chargers.

I don't think a BMS is really useful/necessary for this sort of setup btw as they don't do a very good job balancing cells anyway - just use a balancing charger like mentioned above. If you're using a LiPo you may want to figure out adding a fuse or just slap on a low voltage alarm, but all standard cell batteries (18650 etc) sold in the US are required to have CID/PTC protection, and you can get additionally protective circuitry cells (which limit draw to 10A per cell, but that should be enough).

BTW, for those looking merely for untethered VR, for the past year I've been using the TPCAST on the Vive - it's not polished/perfect, but is a generally much better experience than a backpack PC.

 

[Reldey]

Not sure if this helpful for your goals, but I had a thread on my battery powered NUC with off the shelf parts here: https://smallformfactor.net/forum/threads/battery-powered-nuc-project.7143/

EDIT: My solution uses DC-DC power, as the NUC has a DC in barrel jack with the battery. Pretty efficient, lasts around 6 hours while powering the USB monitor.

 

[Choidebu]

So I've been learning and researching this bit also. An NUC is most viable atm with <90W wattage, because easily procureable, dc 19V output powerbanks exists that can provide 5A. The bosch 6Ah battery that @Kmpkt linked should also work. How long will it power it, is another story.

Here are my findings so far and my thought:

1. Battery chemistry: currently the king of energy density, that is, watt-hour per kilogram is Li-ion. Li-Poly or LiFePO4 is actually Li-ion too with minor differences. Lead Acid, NiCd or NiMH simply won't cut it for portable computing application. BUT the story doesn't end there....
2. 18650 cell is the most common li-ion that people have been diy-ing them into packs, powerwalls, even scavenged them from old laptop and tools batteries. Its name is mnemonic of its dimensions: 18mm in diameter and 65mm in length.
3. Capacity means f**k all. According to www.batteryuniversity.com 18650 Li-ion cells started off 2200mAh in 2013, all the way to 3400mAh in 2017. BUT there was oversupply of batteries and this means a lot of older ones are in circulation atm plus a lot of 'counterfeits' cells that are just scavenged and then repackaged cells. So your 2600mAh 2015 cell can only have 500mAh left in it if your seller is dubious. PLUS, the story still doesn't end there...
4. Even if you have 3400mAh brand spanking new cell from LG or panasonic GA type, you never want to drive these battery down to their rated capacity. Actually, a good BMS will never let you do that. Why? Longevity. A Li-ion cell rated for 2000 cycle will be down to about 40 cycle under extreme discharges. You wouldn't want a 200$ worth of battery pack that lasts you month, would you? So.... realistically speaking, 70-80% is the REAL capacity here, and 50% is the best if you really want long life for your pack, and we're talking 5+ years here.
5. Battery ages. And with it, capacity goes south. Even a well cared for battery will lose 50% its capacity in 2 years time under normal usage. Also nothing else kills batteries faster than heat. This is also one thing to consider when we start shoving them into PCs.
6. The real money is in BMSs, chargers, health monitors. These are the primary price drivers for commercial packs.
7. DIYing them is not cheap either. A recommended way to assemble these cells is spot welding, and a proper home grade equipment will set you back 200$.
8. Now the real setback for a dream of laptop-like portable pc: software. As close this is to hardware, there is no open, common way of a battery pack to talk to a computer. The interface is there, but everything else is proprietary. This is changing, though, one recently started project by Eric S. Raymond, https://gitlab.com/esr/upside is aimed to create an open source ups platform. A battery, after all, is just a ups from OS viewpoint.

So that's that...

So IF what I really wanted is a diy battery pack so that I can place the cells inside the case the way I want, not confined in a cuboid shape, with 3d printed enclosure, thermally insulated, capable of driving average of 100W pc usage (browsing/video) for 2.5 hours, and maybe 150W peak while gaming for 1 hour, which wattage also includes a monitor...

So what I need is a ~18V (assuming 3.6V/3.7V cells in series of 5) pack capable of 250Whr (2.5 hours * 100W) and can deliver up to 10A (150W ÷ 18V = 8.33A)....

250Whr @18V is 14000mAh, if that is ~70% then what I really want is 20000mAh. Conservatively speaking if one cell is 2000mAh that means I need 10 which is perfect because I only need 2 parallel series of 5 cells.

A 5S2P, 18650 battery pack.
Great. Where do I get one? Or how do I make one?

Am I doing the math right? Anything I overlooked?

Edit: a 10A peak means doubling down on nickel strip across assembled batteries: these are rated for 5A, more than that they start heating up and while they won't catch fire it wouldn't be good for the cells. But then again, maybe not... it's in a parallel of 2 anyway each would just carry 5A.

 

[Choidebu]

I've already developed this for my wearable PC. Runs off 19V Bosch 6.0A batteries and is scalable up to around 1300W. Compatible with HDPlex and Dynamo power systems. If you're interested hit me up.

See this is where I got confused. Looking at the pictures this exact battery is a 10 cell, 5S2P arrangement I described above. But WHY is it just rated at 6Ahr? Are they using longer lasting cells with less capacity?

The more I looked at it the more I think I understand nothing.

 

[Vlad502]

250Whr @18V is 14000mAh, if that is ~70% then what I really want is 20000mAh. Conservatively speaking if one cell is 2000mAh that means I need 10 which is perfect because I only need 2 parallel series of 5 cells.

A 5S2P, 18650 battery pack.

5s2p is 18.5V 4000mAh:
1batt 3.7V 2000mAh
1s10p 3.7V 20000mAh
5s1p 18.5V 2000mAh
5s10p 18.5V 20000mAh
needed cells are 50

But WHY is it just rated at 6Ahr?

Maybe 5s2p with good 3000mAh cells

 

[Choidebu]

5s2p is 18.5V 4000mAh:
1batt 3.7V 2000mAh
1s10p 3.7V 20000mAh
5s1p 18.5V 2000mAh
5s10p 18.5V 20000mAh
needed cells are 50

Maybe 5s2p with good 3000mAh cells

I knew my calculation is too good to be true, but can you tell me where did I go wrong and what's your version based on?

Edit: a 2016, 90W max (just by its brick, not actual wattage) laptop can go 5 hrs with a 9-cell battery. So I don't think my math are too far fetched? Maybe 5S3P is more realistic?

Edit2: Okay so I see now that mAh rating is the constant current that is needed to deplete a cell from 'full' State of Charge at 4.2V to 'zero' at 3V. But as I understand it, the actual capacity won't go down if you don't push these cell beyond 1C. Or does running 5A per series considered 'pushing' them above 3.4A (1C for newer cells)

 

[Vlad502]

Edit: a 2016, 90W max (just by its brick, not actual wattage) laptop can go 5 hrs with a 9-cell battery.

9-cell battery (3s3p 11.1v 9ah) 100Wh for 5h, power usage must be around 20W

Or does running 5A per series considered 'pushing' them above 3.4A (1C for newer cells)

For example: https://www.electric-skateboard.builders/t/max-discharge-current-from-18650-cells/2723/2

2C = 2 x Capacity so 2 x 3000mAh => 6A discharge
the 4P then multiplies that by 4 so: 6A x 4p => 24A max that the pack could handle.

 

[Choidebu]

I still doubt where you based your data on. Is it LiFePo4? Because lipos are slightly different: they are high current density as opposed to Li-ion's high power density, essentially targeted towards more current discharge capability in exchange for energy density. The current lipo cells I'm seeing is ~1500mAh, which explains this discrepancy.

I did some more reading, in http://www.ebikeschool.com/how-to-build-a-diy-electric-bicycle-lithium-battery-from-18650-cells/, Micah toll build a 30 cell, 10S3P battery which at first run time could deliver 8.54Ah (~300Whr at 36V and discharge of 0.5C => 1.467A * 3 parallel groups = 4.4A) with 2700mAh panasonic 18650pf cells. The math checks out.

BUT I did notice an error on my methods: I should be aiming at 1C or lower across each parallel group, e.g if I'm aiming for peak draw of 150W, that means 8.33A, so 2 parallel series of 3400mAh won't cut it, it has to be 3 at least. This is because running a cell over 1C will decrease its effective capacity and lifespan.

As an added bonus on going 5s3p is the possibility of using more common less bleeding edge, hence cheaper cells around 3000mAh or 2600mAh.

We can, go for 4s3p but 14.4V is annoying to work at..

 

[Kmpkt]

With respect to the Bosch batteries they are 5S2P arrangement and quite high quality cells. The newer ones (Core 6.3 Ah) even moreso.

 

[Phuncz]

@Choidebu mind you that laptop chargers are specced to support the maximum load of a laptop and the charging of the battery. Considering most laptops have their battery life rated for light browsing without peripherals and with a dimmed screen, together consuming a fraction of the power of the max, the rating isn't realistic to go by.

You see this with gaming laptops that are benchmarked for battery life during stress, there's rarely one that exceeds an hour on 90W loads.

 

[Choidebu]

I know.. I know... it's just that I'm fairly new to this so I'm trying to put my mind around it and get a ballpark going. If most 6 cell laptops can do that with ~2600mAh cells (yes I've been looking at used laptop batteries market) I think my use case is feasible with 15 cells and 30% more capacity in each cell (3400mAh).

Plus I do know those ratings are bull....

And about the brick must be capable of powering the device AND the charging? I also thought of that - essentially we're bottlenecked by the amount of parallel groups, so 3 * 1.7A (0.5C) = 5.1A, times 18.5 is ~95W. But we can get that down to 55W if we're willing to just do 1A.
Doing that, a 200W brick would be enough.

Well back from researching again, AliExpress is awesome! Roughly 40% cheaper cells and already pre-welded with nickel strips!

Now I'm getting my head dizzy with chargers... anyone knows any good 5s charger?

Also there's this tricky unsolved piece - power path manager. Essentially a UPS - board that can take dc input and battery, switches the load to dc input and charges the battery when dc input is there, and provide backup to battery when dc fails.

That upside project could not get here sooner!

Man I'm getting excited!

 

[Choidebu]

With respect to the Bosch batteries they are 5S2P arrangement and quite high quality cells. The newer ones (Core 6.3 Ah) even moreso.

Yes I've run the numbers the 6Ah ones came up at 3000mAh per cell, 6.3Ah is 3150mAh (just half it - the total pack capacity depends on how many parallel groups there are). They actually have 7Ah 'professional' model now, with a very professional price point.

I've thought about scavenging two of those, even using the embedded BMS and balance board, if exists... but price wise is not that different from getting newer cells with 10% more capacity. And that compared to used battery. Who knows what they have gone through.

 

[Choidebu]

250Whr @18V is 14000mAh, if that is ~70% then what I really want is 20000mAh. Conservatively speaking if one cell is 2000mAh that means I need 10 which is perfect because I only need 2 parallel series of 5 cells.

@Vlad502 , I have noticed my error. You are right. An 18.5V 20Ah with 3.7V 2Ah cells does not follow my calculations. Simply put, Ah =/= Whr, if I simply divide 20Ah by 2Ah, that will just calculate the amount of parallel group I need, not total cells.

So a xSyP battery, corresponds to mV nAh battery, where m is x times individual cell voltage, and n is y times individual cell Ah rating. Ah is not, strictly speaking, a power/energy capacity unit.

So... yeah... 5S10P battery. Totally NOT feasible. But... best case scenario with 20Ah and 3.4Ah cells is... 5S6P...

Argh... I'm gonna have to rethink my actual power usage. I guess portable custom computing over 100W (with dGPU) is still not very feasible? A system with Ryzen 2400g consumes ~40W on idle so there's that..

Sorry guys 3 post in a row for a brain dump....

 

[Choidebu]

You see this with gaming laptops that are benchmarked for battery life during stress, there's rarely one that exceeds an hour on 90W loads.

Missed this the first time @Phuncz , is that so... so I take it you're saying that the battery development is still not there yet?

 

[lhl]

@ChoidebuYou see this with gaming laptops that are benchmarked for battery life during stress, there's rarely one that exceeds an hour on 90W loads.

You can get pretty good power consumption numbers with PowerTOP and nvidia-smi - most gaming laptops will do 120-150W running full tilt even on battery-only downclocks. Most gaming laptops have 50-100Wh batteries (and shut you down at about 80% discharge), so you can do the math there.

@Choidebu I have some comments I'll write up since a lot of your research is... not quite right IMO. There is a lot of work on battery/offgrid solutions by people building DIY powerwalls (hbpowerwall, jehugarcia are good places to start) and the like (people building DIY solar/offgrid systems are good resources as well), and there's a lot of people doing portable power in the RC/ebike world - the ebikeschool guy you linked has a book DIY Lithium Batteries.