Completed DIY "laptop" / portable PC (7.39 litres, 17'' screen, 5700x, RTX2070)

[timginter]

Here is what I have for the battery configuration on a similar project.

AC-DC ----------------------->|------------------------DC-ATX
12.6V          |                             |
               |                             | (out: 12.3V)
             BOOST                         BUCK
             CONVERTER                     CONVERTER
               | (out: 25.3V 2-4A)           | (in: 18-35V)
               |                             |
               -------------------------------
               |
          MOSFET SWITCH
               |
              BMS
               |
           6S BATTERY

Huge thanks for both!
You have only 1 diode, though - do batteries feed current back into the boost converter, and the 12.6V PSU feed back to buck? Or are they specifically boost and buck only to prevent current flowing back?

AC-DC ----------------------->|------------------------DC-ATX
12.6V          |                             |
               |                     here    \/ ?
               |                             |
               |                             | (out: 12.3V)
             BOOST                         BUCK
             CONVERTER                     CONVERTER
               | (out: 25.3V 2-4A)           | (in: 18-35V)
               |                             |
    and here   /\?                           |
               |                             |
               -------------------------------
               |
          MOSFET SWITCH
               |
              BMS
               |
           6S BATTERY

 

[ruleh]

This specific boost converter doesn't doesn't feed any current back to the 12v line. I don't think boost converters do in general but not sure. The buck converter would feed current from 12v rail back to the battery rail with ~0.6V drop, probably due to a mosfet body diode but it doesn't because the battery voltage is always higher.

 

[Vlad502]

Here is what I have for the battery configuration on a similar project.

AC-DC ----------------------->|------------------------DC-ATX
12.6V          |                             |
               |                             | (out: 12.3V)
             BOOST                         BUCK
             CONVERTER                     CONVERTER
               | (out: 25.3V 2-4A)           | (in: 18-35V)
               |                             |
               -------------------------------
               |
          MOSFET SWITCH
               |
              BMS
               |
           6S BATTERY

It switches automatically between ac and battery and charges the battery while ac is connected.
I have only tested this up to 70-80W since that is the max the system draws right now.
I should be able to test higher loads once new parts arrive.
low quality pictures following shortly, hopefully...

Maybe this is the correct one:

AC-DC ----------------------->|------------------------DC-ATX
12.6V          |                             |
               |                             | (out: 12.3V)
             BOOST                         BUCK
             CONVERTER                     CONVERTER
               | (out: 25.3V 2-4A)           | (in: 18-35V)
               |                             |
          MOSFET SWITCH-----------------------
               |
              BMS
               |
           6S BATTERY

I`m interested in switch circuit diagram. Also IMO it can be replaced with automotive relay switching between both converters

Nice looking 6s bms with 18650 case

 

[timginter]

This specific boost converter doesn't doesn't feed any current back to the 12v line. I don't think boost converters do in general but not sure. The buck converter would feed current from 12v rail back to the battery rail with ~0.6V drop, probably due to a mosfet body diode but it doesn't because the battery voltage is always higher.

thanks for the explanation

Nice looking 6s bms with 18650 case

true, looks really neat

was there a reason you went with 6s for 70-80W setup? I was planning on 6s originally (I think I even ordered the same 6s case/holder as you), but I wanted to avoid stepping up/down to charge the batteries and to power HDPLEX ATX PSU (it has a 16-24V range - even if it's strict 16V cutoff 5S will be good until around 15% battery)

 

[ruleh]

Maybe this is the correct one:

AC-DC ----------------------->|------------------------DC-ATX
12.6V          |                             |
               |                             | (out: 12.3V)
             BOOST                         BUCK
             CONVERTER                     CONVERTER
               | (out: 25.3V 2-4A)           | (in: 18-35V)
               |                             |
          MOSFET SWITCH-----------------------
               |
              BMS
               |
           6S BATTERY

I`m interested in switch circuit diagram. Also IMO it can be replaced with automotive relay switching between both converters

The output of the boost converter is connected to the input of the buck converter.
The mosfet is just used as a (semi-)kill switch to prevent the discharge resistors of the dc-dc converters from draining the battery. A relay or simple manual switch with the correct rating would also work, though I don't know if the inrush current from the battery to the input/output capacitors could damage the switch contacts or not.

was there a reason you went with 6s for 70-80W setup? I was planning on 6s originally (I think I even ordered the same 6s case/holder as you), but I wanted to avoid stepping up/down to charge the batteries and to power HDPLEX ATX PSU (it has a 16-24V range - even if it's strict 16V cutoff 5S will be good until around 15% battery)

The main reason I went with 6s is this battery holder.
The battery voltage needs to be regulated anyways so the conversion losses can't be avoided.
Same is true for charging, especially so with current limiting. Though charging efficiency is not a priority right now.
The problem with the hdplex is that it also has conversion losses when running from an ac-dc supply.
With the planned total power consumption of 400W (yes, really ) having 20W loss can be a bit hard to deal with.
Also finding high power 12V supplies seems somewhat easier.

 

[timginter]

The main reason I went with 6s is this battery holder.
The battery voltage needs to be regulated anyways so the conversion losses can't be avoided.
Same is true for charging, especially so with current limiting. Though charging efficiency is not a priority right now.
The problem with the hdplex is that it also has conversion losses when running from an ac-dc supply.
With the planned total power consumption of 400W (yes, really ) having 20W loss can be a bit hard to deal with.
Also finding high power 12V supplies seems somewhat easier.

Nice! I can easily imagine 400W (my current rig is around 380 peak with the GPU) - makes sense with 12V and stepping up/down without HDPLEX.

I'm really curious how yours develop, let me know if you're posting the progress somewhere.

On tinkering note - 4pin PCIe power extension cable just came - couldn't help but cut into it straight away. Effect: neat 4pin PCIe to 5mm Jack to power the screen directly from HDPLEX's PCIe connector. Thanks for the tip @Choidebu

Spoiler: photos

 

[Choidebu]

AC-DC ----------------------->|------------------------DC-ATX
12.6V          |                             |
               |                             | (out: 12.3V)
             BOOST                         BUCK
             CONVERTER                     CONVERTER
               | (out: 25.3V 2-4A)           | (in: 18-35V)
               |                             |
               -------------------------------
               |
          MOSFET SWITCH
               |
              BMS
               |
           6S BATTERY

It switches automatically between ac and battery and charges the battery while ac is connected.

Color me interested as well.

Tell me, what do you have at -->| between boost input and buck output? How do you make sure the buck won't power the boost on ac fail?

Edit: oh you said they're directly connected.
So...what happen when battery voltage is low? Does the battery discharges just to charge itself again? Have you tested this thoroughly? I feel like it'll drop efficiency pretty fast under certain battery voltage, and you're basically powering both circuit incurring losses there.

I get that buck won't supply or feed back when ac input available since you got .3V difference there, but if you got a diode there, what power diode have only .3V drop?

 

[ruleh]

The ">|" is a diode. The actual part is a mosfet based (I think) ideal diode thing. At 6A it dropped ~0.03V. For comparison, the supply voltage itself drops ~0.05 at 6A and the buck converter drops ~0.08V at 6A.

The diode is supposed to prevent the buck converter from powering the ac and boost converter, which should prevent the battery from trying to charge itself.

When the battery voltage drops low, the buck converter draws more current from the battery until one of the cells reach their cutoff voltage and the bms disconnects the battery.

There are losses from the battery to the boost converter's discharge resistor and indicator led but the boost converter itself is off and doesn't consume any power while ac is off.

 

[Vlad502]

The output of the boost converter is connected to the input of the buck converter.
The mosfet is just used as a (semi-)kill switch to prevent the discharge resistors of the dc-dc converters from draining the battery

Then if i understand correct when charging while there is load on DC-ATX, you will have two parallel sources- ac-dc and boost-buck converter and boost converter will probably burn trying supply both buck converter and battery charge.

Mosfet part i still don`t get it

 

[ruleh]

It would be possible if the buck converter voltage is higher than the ac-dc voltage. However since the ac-dc supply pulls the 12V line to ~12.6V and the buck converter only up to ~12.3V, no current will should through the buck converter.
Also the boost converter is current limited so it shouldn't burn immediately.

Edit: The mosfet gate is connected to a switch which either pulls it low to let the mosfet conduct or pulls it high to block current flow from the battery to the dc-dc converter.
It isn't strictly necessary but it is nice to have.

 

[timginter]

@ruleh do you have links to the mosfet diode and mosfet switch you used? Huge thanks for explanations - your control of the voltage to prevent backflow is a really cool way to solve that problem.

I thought of a decoupling capacitor right before the connector to ATX PSU for a steadier voltage, and a schottky diode before the capacitor to prevent current flowing back, but even schottky's have 0.2-0.3V drop. After your posts, though, looking at the UPS board and its max 15A - that circuit + HDPLEX should be OK without the capacitor for between 225W and 315W (for 5s voltage).

I'm reading up on electrics while I'm waiting for parts to come, I'm trying to find info on connecting 2x 12V power bricks in parallel - if done correctly it would make a 2x Dell DA-2 £40 12V 440W PSU (although bulky and ugly) - perfect for a high-end eGPU if you tuck the bricks away behind the desk

 

[ruleh]

The ideal diode is this:

50A Ideal Diode Controller Solar Battery Anti-backflow Protection Module | eBay

Peak Start Current: 50A(10s). Continuous working current 30A (with no heatsink) within 60℃, If a large area heat sink is attached on the back and forced air cooling can reach 50A. Compared with Schottky diodes, it can provide a lower-loss path, and in high-power applications, it can provide a...

www.ebay.co.uk

For the mosfet switch I used two irf5210lpbf because they were the cheapest low resistance p-channel mosfets I could find.
They aren't the best choice though. Something like a bts555 would work much better for this (and would only need one).

I think the hdplex should have sufficient capacitors on board. At least that's what I would assume looking at it.

If you need high power 12v supplies, take a look at used hp server supplies and something like this:

Breakout Board with Button for HP PSU Server Power Supply GPU Mining | eBay

Find many great new & used options and get the best deals for Breakout Board with Button for HP PSU Server Power Supply GPU Mining at the best online prices at eBay! Free delivery for many products!

www.ebay.co.uk

If you don't want to use the breakout board, you need to bridge 2 pins to ground (possibly with a resistor) and solder some wires or connectors to the 12V output pins.
At least for the common slot supplies, the pinout is known and there are ways to play with the voltage.

Or something like this might work:

AC 110V-220V To DC 12V Switch Netzgerät Trafo Driver Treiber Netzteil LED Strip | eBay

Universal AC input / volle Palette. Intput Spannung (AC). 1 x Netzteil (nach Ihrer Wahl. ). Es ist gute Qualität & high-performance. Kühlung durch Konvektion der Luft. Luftfeuchtigkeit: 0 ~ 95 % nicht-Kondensation.

www.ebay.co.uk

The 480W version is what I use currently and it kinda works, except the noise is too high and it blew up once.

 

[timginter]

The ideal diode is this:

50A Ideal Diode Controller Solar Battery Anti-backflow Protection Module | eBay

Peak Start Current: 50A(10s). Continuous working current 30A (with no heatsink) within 60℃, If a large area heat sink is attached on the back and forced air cooling can reach 50A. Compared with Schottky diodes, it can provide a lower-loss path, and in high-power applications, it can provide a...

www.ebay.co.uk

For the mosfet switch I used two irf5210lpbf because they were the cheapest low resistance p-channel mosfets I could find.
They aren't the best choice though. Something like a bts555 would work much better for this (and would only need one).

I think the hdplex should have sufficient capacitors on board. At least that's what I would assume looking at it.

If you need high power 12v supplies, take a look at used hp server supplies and something like this:

Breakout Board with Button for HP PSU Server Power Supply GPU Mining | eBay

Find many great new & used options and get the best deals for Breakout Board with Button for HP PSU Server Power Supply GPU Mining at the best online prices at eBay! Free delivery for many products!

www.ebay.co.uk

If you don't want to use the breakout board, you need to bridge 2 pins to ground (possibly with a resistor) and solder some wires or connectors to the 12V output pins.
At least for the common slot supplies, the pinout is known and there are ways to play with the voltage.

Or something like this might work:

AC 110V-220V To DC 12V Switch Netzgerät Trafo Driver Treiber Netzteil LED Strip | eBay

Universal AC input / volle Palette. Intput Spannung (AC). 1 x Netzteil (nach Ihrer Wahl. ). Es ist gute Qualität & high-performance. Kühlung durch Konvektion der Luft. Luftfeuchtigkeit: 0 ~ 95 % nicht-Kondensation.

www.ebay.co.uk

The 480W version is what I use currently and it kinda works, except the noise is too high and it blew up once.

Huge thanks again!

Well... blowing up is not ideal Lack of noise is the appeal of 2x Dell DA-2, though - they are ridiculously cheap, silent, and come with a plug that can reasonably be modified for 8pin PCIe... HDPLEX 400W would have been perfect if it was 12V out of the box - it's silent and ridiculously small.

On tinkering note - I got the aluminium board ready for hardware. I had to take the GPU IO shield off, the holes make perfect mounting point, though. Had to drill 2 extra holes in the back of the backplate. Then marked holes on the metal plate. Drilled holes and put screws through. Self locking nuts hold the screws and act as washers/raisers for the PCBs. Motherboard was M4, the rest M3 so a bit of a faff.

Spoiler: photos

Waiting for the PCIe cable now, all is ready to mount the hardware.

Impossible to get the holes perfectly, though, I'd recommend drilling with a drill 1mm larger than the screw.
I drilled one and tightened one screw - that's the reference. I put the board on that screw, aligned and carefully drilled through the other holes on the PCB to mark the aluminium. Took the board off, put a drill bit 1mm larger than the screw and drilled all the way through. I didn't tighten the other screws all the way - they were a bit off. I put the boards back on all of the screws and tightened the remaining screws bit by bit until there was enough wiggle to put the boards on the bolts, but tight enough so the boards needed a bit of force to go on and didn't fall off on their own. A lot of trial and error

 

[Choidebu]

I get it now, thanks @ruleh .

Using voltage difference to ensure battery doesn't discharge on AC is okay I guess, but I can't help thinking longer transients will still activate the buck. There is ideal diode OR ICs out there (not sure if there's breakouts though) for that.

Please don't use universal psus - they have crazy swings and pcs, rarely constant load devices, is a very bad match. You might get away of using 480W for 80W max usage, but don't expect more than 200W out of it. Even then I wouldn't risk it.

reading up on electrics while I'm waiting for parts to come, I'm trying to find info on connecting 2x 12V power bricks in parallel - if done correctly it would make a 2x Dell DA-2 £40 12V 440W PSU (although bulky and ugly) - perfect for a high-end eGPU if you tuck the bricks away behind the desk

One thing at a time... People have tried to do this with no success so far. My guess is such load sharing circuit is too expensive for so little gain.

Server psus is the way to go once you go beyond 400W. Get a used one, hack it to submission.

 

[ruleh]

Drilled holes and put screws through. Self locking nuts hold the screws and act as washers/raisers for the PCBs. Motherboard was M4, the rest M3 so a bit of a faff

Does the gpu really need to be raised? If the gpu backplate was flat on the metal plate would it improve cooling?

I get it now, thanks @ruleh .

Using voltage difference to ensure battery doesn't discharge on AC is okay I guess, but I can't help thinking longer transients will still activate the buck. There is ideal diode OR ICs out there (not sure if there's breakouts though) for that.

Please don't use universal psus - they have crazy swings and pcs, rarely constant load devices, is a very bad match. You might get away of using 480W for 80W max usage, but don't expect more than 200W out of it. Even then I wouldn't risk it.

The buck is always active, needlessly consuming 0.04A without delivering any power while ac is on. It is fine if it supplies some power every now and then when the ac becomes unstable or overloaded.
One of the earlier ideas was to use a rather weak power supply with a decent voltage drop on heavy loads so that at some point both the psu and the battery would supply power.
It would require shutting down the boost converter though.

One of the reasons I didn't want to put a diode (or transistor or relay) on the buck output is that I wanted to use it's output capacitors for a bit of extra smoothing. However I don't think it makes much of a difference.

The led driver psu isn't the best for this, considering that it blew up once and has been really weird ever since then (and even before). Especially on power on and off it creates some nasty spikes on the dc output if the load is low. Having the boost converter and battery connected (and not full?) seems to reduce it for now. Another time the output voltage pulsed between 12.4V and ~5V for no good reason (other than that I removed the fan and discharge resistors ). Still I want to see how well it does over time with higher loads. If things get too dangerous I will switch to using a dps400ab5 server psu, once I find a fan for it.

 

[timginter]

Does the gpu really need to be raised? If the gpu backplate was flat on the metal plate would it improve cooling?

True. Not the two holes I drilled in the back - backplate sits on the aluminium plate. The holes from the IO shield are higher, though - I'm using M3 self-locking nuts as washers

 

[timginter]

One thing at a time... People have tried to do this with no success so far. My guess is such load sharing circuit is too expensive for so little gain.

No harm in reading. Unorthodox solutions are a bit of a theme for this forum

The problem seems to have both PSUs supply equal current, I've read about using series resistors, but I'm not a professional/experienced sparky and definitely wouldn't know what I'm doing without some heavy hand-holding. If one PSU taking the majority of the load is not a problem (and the other kicking in only when current exceeds PSU1), a rough DIY diagram I consistently found was:

Makes sense with my limited electrical knowledge. If I had a way to simulate 350W of load (other than buying an RTX3090 ) I'd actually try it just for the tinkering of it.

Server psus is the way to go once you go beyond 400W. Get a used one, hack it to submission.

I had a doubtful pleasure working with servers in my previous job - that noise is not something you'd want in your room

 

[REVOCCASES]

No harm in reading. Unorthodox solutions are a bit of a theme for this forum

The problem seems to have both PSUs supply equal current, I've read about using series resistors, but I'm not a professional/experienced sparky and definitely wouldn't know what I'm doing without some heavy hand-holding. If one PSU taking the majority of the load is not a problem (and the other kicking in only when current exceeds PSU1), a rough DIY diagram I consistently found was:

Makes sense with my limited electrical knowledge. If I had a way to simulate 350W of load (other than buying an RTX3090 ) I'd actually try it just for the tinkering of it.


I had a doubtful pleasure working with servers in my previous job - that noise is not something you'd want in your room

If you want to use two Dell bricks, the easiest way is to use one for the CPU / ATX PLUGIN and one for the PCIe Power 8PIN on your GPU. But I guess this doesn't work for your setup?

Using a Server PSU is not the worst idea IMHO. They are usually rated Platinum and actually do not need such a noisy / high airflow fan if you do not run them in some hot environment. If you go for a 500W ~ 600W Platinum, you can replace the fan with a Noctua or something and it will keep pretty silent up to 400W.

Just modded one myself into a nice little 500W 12V Brick:

 

[timginter]

If you want to use two Dell bricks, the easiest way is to use one for the CPU / ATX PLUGIN and one for the PCIe Power 8PIN on your GPU. But I guess this doesn't work for your setup?

That's what I'm doing now - one 24V 220W PSU for the mobo + CPU is enough, another 12V 220W for my GPU, but it won't if I decide to upgrade my GPU to anything more powerful.

Using a Server PSU is not the worst idea IMHO. They are usually rated Platinum and actually do not need such a noisy / high airflow fan if you do not run them in some hot environment. If you go for a 500W ~ 600W Platinum, you can replace the fan with a Noctua or something and it will keep pretty silent up to 400W.

Just modded one myself into a nice little 500W 12V Brick:

That's really neat! Good point with going for a more powerful one to keep current/fans/noise low relative to max. Which one did you mod and what are the dimensions of it, though?

 

[REVOCCASES]

That's what I'm doing now - one 24V 220W PSU for the mobo + CPU is enough, another 12V 220W for my GPU, but it won't if I decide to upgrade my GPU to anything more powerful.


That's really neat! Good point with going for a more powerful one to keep current/fans/noise low relative to max. Which one did you mod and what are the dimensions of it, though?

Right, I must have missed this part

It's a Liteon 500W FLEX size. Unfortunately the label peeled off so I can't give you the exact model. But there's a ton of options from Dell, HP, Delta, etc.

Final size with the case is smaller than a Dell brick. I made a size comparison here:

Completed - REVOCCASES RCC-BRICKMOD500 / build your own power brick with up to 500W output

While looking for some 12VDC alternative to the HDplex 400 AC-DC PSU I came up with the idea to develop a (cooling) case kit for the quite popular MEANWELL RPS-400, RPS-500, EPP-400 and EPP-500 power supplies. As you might know, they can deliver up to 400/500W if cooled sufficiently. With this...

smallformfactor.net