Log Project 🐦RoadRunner (Battery Powered Portable PC, in S4M-C #524)

[Choidebu]

I should have realised it - the dc-ups would flicker when it reached charging voltage and current dips to 20mA. It didn't last night. Let's hope these light green cells is not damaged permanently. It's now going up to 3.5V, fingers crossed.

My problem was these XT 60 connectors doesn't lend itself to easy current measuring, it was my fault for desoldering the screw terminal, that'd've been way easier to do with.

All in all, the light green cells are proven to be the bottleneck. And I'm not sure how a bms could help us in that regard.

Movie's finished, will shut this down and set a proper charging current, then I'll leave it plugged until the board deemed it to be full then redo the test.

All in all, disregarding the small recharges, we got 1.5 hrs youtube/browsing + 1 hr idle (can't be sure, might be 2 hrs) + 45 mins movie playback.

If we are to sum all capacities together, the whole pack is 24.300 mAh with 3.7V nominal, which is ~90Whr. Given that we run them at 85% capacity (max of 4.03V instead of 4.2V, figure taken from a study about SoC-OCV approximation method), that means only 76.5 Whr.

So either my PC's pretty efficient or the load just not taxing enough.

 

[Choidebu]

One thing I just noticed:

PSU fan just straight up run 24/7. And I've already shortened my cables... Strangely enough the fan didn't spin when I test them on days ago, but now they do.. thought they're only spin on certain load...

Gotta extend them to motherboard now..

 

[Choidebu]

(Deleted, double post)

 

[Choidebu]

BAD NEWS....

I burned my dc-ups board.

It's shorting out at the battery screw terminal now, the shunt resistor near it is also blown.

I dissasemble the rig again, it's only been charging slowly while the pc was off. Like I mentioned earlier I was going to adjust the charging current.

Before that spent some time desoldering the screw terminals at input/output, then solder the wires directly onto that. Then do the reverse on the battery end. All that effort gone down the drain.

It was working fine, I'm adjusting the current but it doesn't seem to adjust so I thought I'd check battery voltage - dumb me I check while it is charging. Again. The manual explicitly said not to do that. In the process also sparked the terminal couple times. Maybe more. Last nail in the coffin I finally realised that the battery is already at 24V open circuit and it doesn't wanna go up from there. Then I had the brilliant idea of adjusting the current potentiometer while the battery is not connected, then I plugged it in and crackle and smoke later, my shunt is sizzling and burnt smelling, me unplugging the battery uselessly. I had the audacity to try again, this time shorted out the battery and sparked it again.

Yeah..

Bummer.

Will need to spend 37$ and a month to get a new one here....

 

[Valantar]

BAD NEWS....

I burned my dc-ups board.

It's shorting out at the battery screw terminal now, the shunt resistor near it is also blown.

I dissasemble the rig again, it's only been charging slowly while the pc was off. Like I mentioned earlier I was going to adjust the charging current.

Before that spent some time desoldering the screw terminals at input/output, then solder the wires directly onto that. Then do the reverse on the battery end. All that effort gone down the drain.

It was working fine, I'm adjusting the current but it doesn't seem to adjust so I thought I'd check battery voltage - dumb me I check while it is charging. Again. The manual explicitly said not to do that. In the process also sparked the terminal couple times. Maybe more. Last nail in the coffin I finally realised that the battery is already at 24V open circuit and it doesn't wanna go up from there. Then I had the brilliant idea of adjusting the current potentiometer while the battery is not connected, then I plugged it in and crackle and smoke later, my shunt is sizzling and burnt smelling, me unplugging the battery uselessly. I had the audacity to try again, this time shorted out the battery and sparked it again.

Yeah..

Bummer.

Will need to spend 37$ and a month to get a new one here....

Bummer! But if you already have to spend that much, why not go all the way and splurge on an openUPS? From what I can tell that should also take care of your battery balancing issues and so on, and it supports up to 6S. The 10A output rating might be a bit low, though?

 

[Choidebu]

Well price wise:

Dc-ups - 28 aud (+ 32 for a new one)
Bms - 7 (+ 8 for a new 6s one)
Arduino micro - 8
INA219 breakout - 5
XT60 and other connectors, cables - 10

Total 58 aud, plus another 40 for my mistake.

Openups 2 is 159 aud from minibox.com, plus shipping from the us.
Openups is 174.

Spec wise:
This dc-ups can do 15A (to be safe I only do 8-9A) non regulated output.
Openups2 is max 5A on 12V
Openups can do 6A (10A peak, 30 secs) on 24V

With regulated output I'd prefer a 12V output coupled with a 12V plugin unit though..

Size wise:
Openups2 is too big.
Openups is 4"x2" like the epp
Dc-ups.. well.. smaller

About balancing, no bms can help with underperforming cells. It helps with diverting power when you're charging them, but on discharge all it can do is protecting it and shutting the whole pack down when that one underperforming cell dips.

I don't have a lot of money to spend on this, if a $100+ part broke I'd have no means to replace it right away.

 

[Choidebu]

11.11 sale in 6 hrs. Trying to decide if I need the $7 daughterboard to provide undervoltage and low start protection. I'd def shell out extra $1 for heatsink version though.

I might still be able to do one testing - the pack is full after all. After that reckon I can just CV them with the epp at 24V

 

[Choidebu]

Ordered a replacement board on free shipping so god knows when that is. 28$ though *shrug*

Also bought locally an arduino pro micro, an INA226 (just so happen the store got this one instead of INA219) and a 6S bms (no balance, won't need it I figure) for around 15 aud!

Also put on order for 6 refurbished 18650 cells to replace thess light greens, 1$ each. Annd... 10mOhm shunts to replace the one on the INA current sensor, for another 1$.

These would be here sometime this week.

That'll keep me busy for a while.

 

[Choidebu]

Batteries arrived yesterday,

And by today the connectors, and tidbits, and my v3 enclosure also.

After a bit more fiddling..

Looks pretty good IMO. But this had its clearance hand-adjusted, and I also found the cross sections to be fragile and not much point at all.

They say in design we iterate and remove as much as we can, to end up with the essentials and nothing more. How true it is...

I'm getting rid of the 4 cross sections altogether, pack all cells vertically with no space (just like current horizontal row).

By doing this I frees up 3mm or so at the bottom that I can fit there, one additional panel that extends from the front to the back, raised so that the front and back panel now becomes the "feet".

This removes the need for current two "cable run" slots at the bottom, because before I had to weave around the cells to create space for them.

It would also, possibly, allows the battery pack to be put in or taken out without taking the enlosure out of the case.

Not too mention should be cheaper to produce ^__^ v1 -> 5.5$, v2/v3 -> 8.5$, v4??

So stay tuned for it!



Funny story, I was still waiting for one more package containing the would-be monitoring module. Checked today and realised that I didn't put an order in -_-!

So while redoing so, I came across Adafruit 8-bit Neopixel board up there (image is linked), which got me thinking..

The dimensions fit skyslots perfectly! I measured 18x53 mm-ish (NOPE had a brain fart), this is 10.22x51.1 mm. I'd just need to cut more acryllic in skyslots shape, secure it in the slot with silicon sealant, hotglue the board on, solder a 4 pin connector and done!

Don't know what I'll do with it yet, but, you know, it's an 8 rgb led, we'll figure it out

Edit: idk why I thought it was 18x53. It's 7.5x32.5 cm. Fml

 

[Choidebu]

Oof forgot to mention I did test the cells, and got 4 good cells; 2x 1800-1900mah, 2x 1500-1600mAh, one 1200mAh and one 700mAh.

My best light green cell is 900mAh.

So I'm ordering more of these burgundy cells, reckon 4 more should do it. Fingers crossed!

 

[Choidebu]

Finished v4 design of the enclosure, 215x227, a bit smaller than my first 243x243, but added some engraving so the cost might be higher still.

It's now just 7 parts: top, bottom, front and back panel, 2 side panels one mid section.

Shifted the 50mm fans to the inside, gonna add some cable (balance, output) routes, maybe some holes for zip ties.

All but front panel will be glued, and fans would need to be mounted before gluing. No way around this unfortunately - there is no more clearance available for additional brackets. If only they can be screwed from outside, through the fan on to the acryllic.

 

[Choidebu]

Made a bracket to hold the EPP in place.
Cut two parts out of a laptop LCD hinge and epoxied them together (would redo this with JB Weld), then drill some holes.

It could've been a simple case of drilling 4 holes in the frame, but my MINI came with a cutout in the front-left frame (for oversized gpu w/ custom bezel) that are just large enough to interfere. Plus it'd mean I'll have to remove the bezel every time I need to remove the PSU.

The bracket is mounted to the EPP. That standoff is only there because I didn't have any M3 nuts around.

From the top this might be clearer.. Width now becomes exactly 57mm.

Put in place, now you can see how both screw holes lines up in where originally Drive Bracket would mount to.

Again, No M3 nuts... And below is a further away shot - You can see the GPU cutout. I know this only mounts one side, but given everything is so snug it's actually pretty secure. I've got just enough clearance to slip a thermal pad to the other side of the EPP, where the heatsinks are, to spread the heat to the case panel.

My new cells arrived. I bought another 4, seller gave me 5 because I got 1 bad cell before, awesome. So I have now, 1 group of three ~1900mAh cells, 1 group of three ~1500mAh, 3 bad cells (yes... got me 2 more ~600mAh), and 2 decent ones (~1100mAh). Well it's just the risk of refurbished cells - 11 total, 3 bad - 27% failure rate.

While that was charging I took a shot at laying out components for everything else - the dc-ups, arduino, and the bms.

The width is ~55mm once cut, length is 8 cm or so. The bms doesn't have holes, reckon I'll just hotglue it to the underside and solder the leads to cables, to that balance socket.
Gonna put a switch in the free space above that XT60 socket, to kill the battery.

Next up, solder the new cells in the pack and test them out (without the dc-ups, I know.. - but I've got a battery low voltage alarm now!). New enclosures should be here within 2 days too.

 

[Choidebu]

Managed to solder my pack before my soldering iron broke on me.. didn't get to do an xt60 to 4 pin molex I needed to test the pack.

Probably a good thing, otherwise I'd stay up much later.

And today I found my old, college time soldering iron gun! Hope it works, never enjoyed the gun style, but beggars cant be choosers.

Edit: tried to do some cost breakdown, I did but closed the browser accidentally and lost it all.. anyway ended up being 307 aud incl. 32 in shipping. My gosh.

Deducting tools like the nitecore and tester, broken/unusable parts, and prototype enclosures, it becomes 180.

Deducting the psu (costs me 52), 128.

Deducting refurbished cells, 115.

Yeah 100-120 sounds about right

 

[Choidebu]

Testing complete!

Started off with 24V pack voltage. The new cell groups were 4.1 something and were each put at the positive and negative end.

Plugged the pc to the pack, then lipo alarm to the balance lead.

First boot, dropped to 23.2V

Then movie playback, while monitoring the alarm display.

1 hour and 45 mins in I was feeling pretty good, but then the alarm went off. The fourth group have gone below 2.7V. I turned the alarm off, then at 2 hour mark the first group dipped to 1.6V. Pack voltage was still at 18V.

Shut the pc down, voltages went back to 2.5V for that group, the fourth settles at 2.8V. Others still going around 3.3V.

Yeah I'm not sure why - I've heard anecdotes that the cell group closest to cathode (positive lead) is always the first to go in a laptop battery pack. Let's just say I witnessed it firsthand.

2 hours for video playback.

Will solder bms tomorrow and trickle charge the pack with 24V, then play a gane. Fun times.

 

[Choidebu]

So I've been scratching my head of what's happened and why..

This #1 cell was he new burgundy ones with 1900mAh, they should've hold...

Did I heat them too much while soldering them?

I came across pack balancing techniques, then read about bottom balancing. The symptoms are close to mine; weaker cells plummets suddenly while the other just seemed to chug along just fine.

See parallel cells balance themselves right away while connected. Stronger cell will charge the weaker ones, and we end up with shared capacity closer to n times the weakest cell in the group. This is happening all the time, which is why we like to ensure similar capacity to be grouped together to avoid it happening too much.

But series cells won't exhibit this before they are connected to a load. When they all discharging at the same rate (more a product of internal resistance - how fast they discharge and how voltage changes throughout the process) it's fine and dandy but once a weaker cells starts to drag along, the stronger ones will keep trying to push current through them at the same rate, worsening the weaker ones' condition, a.k.a plummet them even faster. Think of soldiers running tied to each other - then imagine if one is limping along.

The key could be to start at the bottom. Drain each cell groups at around 2.7V and let them settle for a few hours to about 3V, repeat the process so that all of parallel groups have the same settling voltages. With this we established that the cells have same bottom threshold or effectively zero capacity for our usage.

And then we charge the pack as a whole, setting top threshold lower than what we usually do, to account for differences of each cells. And in this case, we already are doing so at 4.02V.

This technique is called bottom-balancing, as opposed to top-balancing which is used by balance charger and BMSes in general.

What this technique try to do, is match draining / discharging graph at the bottom right of the curve, which is where more things can go wrong at much shorter time.

I'll try this for now. While discharging them equally won't take long, charging them again would take a while.

Although I have another problem: now that my dc-ups is toast I don't have a CC (constant current) supply.

I could just charge them CV, but that would potentially unbalance the cells again. Why? CV charging is like an all-you-can-eat buffet - you take what you can - some take more some take less. CC charging effectively gives same amount of energy to each cells, mimicking the opposite which is real-world discharging process.

Worst case scenario I could monitor the current while adjusting my output voltage (with a help of a resistor at below <3.7V or <22.2V). But we're talking 12 hour+ I'm estimating here...

 

[Choidebu]

Although I have another problem: now that my dc-ups is toast I don't have a CC (constant current) supply.

You know what... I just went through my old toolbox from college, organize them and whatnot... and what did I find there... an LM317!! I can make a crude CC from this! Just need some resistors and capacitor, a potentio then I can hook it up to the EPP!

Edit: lol of course it wouldn't be so simple.

1. The thing is rated at 1.5A. But I'd be crazy to drive that much through a single TO-220 package. I do have a pretty big heatsink though
2. To get 1.5A I'd need 1 ohm 5 w resistor. I've got 2x 2.2 ohm 5 w, we can always paralel them.
3. Ooh I got a couple of 25v caps... these are perfect.

Edit2: Got the circuit working off the EPP. I calculated it to source 0.68A but on testing it shows a consistent 0.58A (tested with a couple 12V 3W led bulbs I had around, both parallel and series configuration). It's good enough though, might just be my cheap dmm.

I _can_ double the current, but this is my last LM317 - can't risk it.

 

[Choidebu]

It's half past midnight here and I'm still trying to drain these cells to a similar level.

I've got around 3.3V for most but two cells, which baffled me cos even though they are down to 2.5V in seconds with a 1A load, they just bounce back to 3.5V after a min or two.

Idk man...

In any case, they'll have to sit the night on drained state. Then I'll check their voltages tomorrow, maybe drain some more cells, then charge them as a pack.

[15 mins later] kay I'm discharging these two at 20 times less rate, 75 mA. We'll see...

[The following day] I'm relieved to see the pack rested and settled well. I'm looking at 3.3 to 3.5V. At this point I discharged the 3.5V cells, slowly to 2.75V cutoff. Took 2 tries to get it down to 3.37V. I also charged one group at 800mA for less than a minute to get it up to 3.45V, which settles at 3.35V.

I'm gonna leave it out a bit for an hour before charging the pack as a whole.

Edit2: the enclosure is here! Will update this post once assembled.

Edit3: here it is..

BEEP BEEP!

For this round I tested out different lasercut vendor, he's very attentive and went through the design to see possible problems. It took longer, but apparently once he's done cutting he went back to the design and tighten the interlocking slots, cut it again and sent me two sets. Awesome fella.

I was too excited, went and glued up one (the original) then I realised I forgot that the fan holes need to be countersunk and fans fitted first...

Good thing he sent me two!

 

[Choidebu]

This is my janky hacky constant current charging setup:

All seems good for now, I'm seeing low starting voltage (was 20.8V OCV) so it must be working. If no other fuckeries going on it should go relatively fast (~ 1 hour or so) to about 22.5 (cells are at 3.7-3.8V range) then it'll stuck there, prob climbed close to 23, for 8 hr or so.

Best case scenario it'd go on for 20 hrs, but my hope is diminishing and my confidence that I didn't screw them up during all that soldering and handling is low. If there's any of them crossed 4V within 8 hours that means they are f***ed. That means their internal resistance shot high and capacity went down significantly.
Edit: I screwed up on the calculations.. 22.5V in first hour was correct, and it did. But I calculated 0.6A to charge 30Ah to 70%:
10% in that first hour,
The next 60% is 18Ah, and I simply divided them by 0.6A (my cc output) which is 30 hrs.

This is wrong, because that is 30Ah over nominal 3.7V. My pack and the way it is charged right now is a simple 4500mAh (3P) over nominal 22.2V (6S).

So we can crudely estimate this relatively flat stage of charging to be 60% * 4500mAh ÷ 580mA = 4.65 hrs.
--------- correction done -----------


On to the good news...

Test fit. Had to file 1mm off the back panel, just the bit that fit inside the bracket there.

Oh you can see I managed to countersunk the holes and mount the fans.

Batteries in, ribbons to take em out. I might have to rethink if I do need that middle section at all. Currently it's too fiddly to keep the pack together and insert it, so I split them up, left and right.

What if I do like ebike ppl do - wire it up, shrinkwrap it, then just put it in, you know... like a sane person would. Who in their right mind would design a split compartment with individual cells anyway?

With some effort we can shut the side panel!

 

[Choidebu]

Made some miscalculation up there (edited) - this is hard ey.

So the pack did went up to 22.5V in one hour. Cell drift was 0.16V. Not good, but to be expected.

And now it's been another 3 hours, the drift became 0.14V (3.72V - 3.86V), pack voltage is 22.9V.

The CC supply's max voltage is 23.2V. At this rate they'll reach that in couple of hours.

So the estimate was good (I corrected above if you missed it) - 1 + 4.65 hrs to reach a pessimistic 60% of 4500 mAh (should be closer to 5000mAh+).

Once they reach 23.2 I'm gonna have to improvise... if drift is 0.13V by then, given average of 3.87V then min voltage is 3.8V and max is 3.93V.

Yep, probably safe to CV them then at 23.6V (3.93*6) for 30 mins, then 24.2V (pack upper threshold) for couple hours.

[1 AM update] 3.84 - 3.92, 0.08V drift. Pack voltage according to the lipo tester is 23.2V, but I doubted it since my DMM consistently registers ~0.6V below their measurement.
....
Sure enough, 22.7V on my DMM. Funny, even the ZB2L3 discharger shows similar voltage as this lipo tester. Thought my DMM was off, even tried another one, same result, and it sure did capture max voltage on my LM317 CC correctly, and the EPP's max tunable output. WTF is with these chinese stuff?

Well in any case this is good news. If it goes on to about 8 hrs that is 4,604mAh in them. Another couple hours or so in CV will add about one fifth of that, 5,568mAh. Given ~90% typical li-ion charging efficiency, that's 5000mAh usable.

Realistically speaking, I doubt I can run them reliably if I discharge them to 2.7 (16.2V). Things would just go awry again at that voltages. Now questioning myself again over the whole 5S - 6S situation...

Lesson learned again I guess.

With 5S you'd need top balancing, voltage is lower so more current is being pushed hence bigger voltage drop/sag, more burden on each cell. Dynamo is alright <16V though (didn't know this before), so we can possibly utilise more capacity out of the pack just running them, 3 - 4.2 / 15 - 21V. Of course, 15 cells < 18 cells to start with, so the jury is still out on that.

With 6S you'd need bottom balancing (and top balancing too eventually), less burden and possibly longer battery lifespan due to never charging them near their max, but drops fast and dangerously so we cannot run them to bottom even if dynamo supports it. I'm thinking 3 - 4 / 18 - 24V.

Just what did I get myself into??

[9AM update] I dozed off waiting for the pack to reach 23.2 on my DMM. Woke up 8AM, measured 23.4V - whatever, unplugged it. measured 4V on one, 3.8s and 3.9s on the others. I then tried plugging it to the epp, adjusted to 24V - sparks flying... pretty quickly but no damage thankfully. Got scared, searched my toolbox for a diode, find a bridge rectifier instead, soldered it to the cable, adjusted the output again to compensate for diode loss. It's plugged now, on CV 24.2V. Reckon 2 hours on this, which is tricky since I'm going to church in an hour.

 

[Choidebu]

Okay, here we go again...

Started 00:30, OCV 23.8V, on load 23V. Bear in mind this is on the lipo tester. Currently 0.4V lower than my dmm. Worse case, 0.6V lower.

Started off a 2-hour movie, see you in a bit. MiB international!

[2AM] 1.5 hr in, 21V. 3.27-3.65V.

[2:15AM] #4 (from -) cell group dipped to 3V, and in couple of minutes to 2.7V. Didn't wanna take that risk, shutting down. And it was just getting at the good parts.
This is more like it though, #4 is my old green cells which is the weakest cell in the pack at 3*1500 mAh. Looking back at before, seemed to mirror previous result.

All this tells me is that the pack does behave close to its weakest member. So no matter what I do this is still just a 4500mAh, 22.2V, 100Whr usable battery.

So this test usable capacity and battery life. I'll find some time to charge them again, and do another test, but this time to test throughput: stress out cpu and gpu and see how long the pack can withstand high current loads.