Cluster power supply board for M1 and 3.5" HDD

Post Reply
User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

I'm building a storage cluster based on the M1.

I went for 4 nodes, each with a 500GB nvme and a 16TB datacenter HDD (loud but very effective 8-), giving me about 66TB of raw storage.

The M1 seems to be perfect for this, except... no 12V on the SATA power connector. :?

Which was just the perfect excuse to design a power supply module for the M1, supporting a couple of 3.5" HDDs:
m1-hdd-power.png
m1-hdd-power.png (320.64 KiB) Viewed 575 times

Project status

I've mostly finished the PCB design, and am now procuring the components (tricky thing in current times).

By the end of the week I'll have the PCBs ordered, and hopefully a few days later the first prototypes will be ready for testing!

So what does it do?

This board acts as the power supply for each cluster node, powering the node from a common 12V bus:
  • provides 12V power to the M1 module via the DC barrel connector
  • provides both 12V and 5V to 1 or 2 3.5" hard drives via the two white JST XH connectors in the image. These are the same as used by Odroid, so the cable kits are compatible.
  • lets the M1 monitor a few voltages and currents (interfaced via I2C on the expansion connector)
  • connects to the 12V bus using a 3 pin Molex Microfit-3.0 connector (12V, GND, on/off)
  • a hot-swap controller allows insertion/removal from a live system and provides undervoltage/overvoltage/overcurrent protection
Moreover, the 5V is generated independently, so that it doesn't strain the M1 dc/dc converters.

Installation

The mezzanine board is mounted on the 2x20 expansion connector, and uses one of the M1 mounting holes via a suitable hex standoff.

There will be a 15cm DC barrel connector pigtail soldered onto the two top-left pads. This provides power to the M1, and must be connected and ideally secured using a zip-tie ;)

Power supply bus

The module is meant to be powered from a 12V bus, and a 12V hot swap controller ensures the some degree of protection to the system. For more serious overvoltages, a TVS limits the voltage to safe levels, while hopefuly triggering a system-wide protection. I've decided against adding a fuse, as the hot-swap controller should protect against shorts.

I went for a three pin input connector, where shorting the third pin to ground turns the system off. This allows some degree of remote power cycle control.

Supply monitoring

A monitoring IC measures a few voltages and currents, and reports them via SMBUS (I2C):
  • 12V input from the supply bus
  • 12V output to the M1 SBC
  • 12V output to the hard drives
  • 5V output to the hard drives
It acts as a coulomb counter, allowing fairly accurate power consumption measurements. At some point I'd like to record those via Prometheus :)
These users thanked the author rmsc for the post (total 3):
mctom (Tue Aug 16, 2022 4:43 am) • tobetter (Tue Aug 16, 2022 6:39 am) • odroid (Tue Aug 16, 2022 9:41 am)

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

(reserved)

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

(reserved)

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

(reserved)

User avatar
mctom
Posts: 1961
Joined: Wed Nov 11, 2020 4:44 am
languages_spoken: english, polish
ODROIDs: OGA, XU4, C2, M1
Location: Gdansk, Poland
Has thanked: 224 times
Been thanked: 279 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mctom »

How cool is that, so many hardware projects all of a sudden! :D

Incredibly, just yesterday I was thinking how cool would it be to design a board that provides 12V to SATA drives on M1, but sitting right on SATA connectors and boosting 5V up to 12.. Clearly not the smartest idea so I didn't bother. :)

Cool stuff. Why won't you show us schematics too. :)
Punk ain't no religious cult, punk means thinking for yourself!

Maintainer of PiStackMon

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

mctom wrote:
Tue Aug 16, 2022 4:52 am
Cool stuff. Why won't you show us schematics too. :)
I think I can do that :D let me get this version finished and I'll prepare something. I can already share the (provisional) list of ICs used so far:

* MAX34565 hot swap controller
* AP62200 buck controller
* PAC1954 SMBUS voltage/current monitor

In the meantime there's all that unused area to the left of the PCB that's begging to be used... Any ideas?

I'm working on adding some hold-up capacitance to the board. This would allow for a safe shutdown in case of power failure, I just need to figure out how much capacitance is required.

I've also considered adding an MCU and transforming this into a proper BCM, but for my usecase that's probably overkill.

Ideas are welcome!

User avatar
mctom
Posts: 1961
Joined: Wed Nov 11, 2020 4:44 am
languages_spoken: english, polish
ODROIDs: OGA, XU4, C2, M1
Location: Gdansk, Poland
Has thanked: 224 times
Been thanked: 279 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mctom »

If the board is too big, simply make it smaller. You don't need all that GPIO, do you? :)
rmsc wrote:
Fri Aug 19, 2022 2:59 am
I'm working on adding some hold-up capacitance to the board. This would allow for a safe shutdown in case of power failure, I just need to figure out how much capacitance is required.
I don't think you can achieve that with capacitors. Let's assume stuff.
You deliver 12V to M1 that draws, let's say, 6W, especially if you want it to perform a graceful shutdown at this time.
You want to keep charge for some 30 seconds, so your circuit detects power outage, some polling process catches that event and initializes shutdown and so on.
It is also a well established fact that M1 works fine at 6.4V (it's something that I test all the time, no single issue so far). Let's simplify and call it 6V instead.

So, you need to store 6*30 Ws = 180J of energy.
Also, the capacitors will support M1 only as long as it's being drained from 12V down to 6V. That means only 3/4 of cap energy will be actually used.
I'm not handy with LaTeX to spell it all out, but capacitor's energy is E = (1/2) Vf^2 * C, where Vf is a voltage of fully charged capacitor. Since we can only drain capacitor until it's half full, the real energy we can get from it is:
E = (1/2) (3/4) Vf^2 C = (3/8) (12^2) C = 54 C.
That gives us C = 180J / 54 V^2 = 3.3 F capacitor.

Mhm, looks about right.
I found one for about $70:
https://www.mouser.pl/ProductDetail/KEM ... 52Bg%3D%3D

And that is why people use Li-Ion batteries for stuff like that. Oh wait, you never mentioned a capacitor in your post, did you? :lol:
That's what I assumed, since capacitors have capacitance. Batteries have charge. :)

I am working (slowly..) on a 18650 battery BCM for Odroid M1. I have a feeling we'll get out from a component shortage situation before I finish, ugh..

But anyway!
rmsc wrote:
Fri Aug 19, 2022 2:59 am
Ideas are welcome!
Your board is going to be amazing the way it already is, if everything works as you expect. :) Save additional features for another spin.
These users thanked the author mctom for the post:
rmsc (Sat Aug 20, 2022 12:36 am)
Punk ain't no religious cult, punk means thinking for yourself!

Maintainer of PiStackMon

User avatar
mad_ady
Posts: 10818
Joined: Wed Jul 15, 2015 5:00 pm
languages_spoken: english
ODROIDs: XU4 (HC1, HC2), C1+, C2, C4 (HC4), N1, N2, H2, Go, Go Advance, M1
Location: Bucharest, Romania
Has thanked: 647 times
Been thanked: 951 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mad_ady »

Nice idea!
If it weren't for a cluster I would have just soldered a wire between 12V in and the sata power wire. I've been meaning to do it, but haven't got time to get some thermal paste...
These users thanked the author mad_ady for the post:
rmsc (Sat Aug 20, 2022 12:36 am)

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

mctom wrote:
Fri Aug 19, 2022 3:44 am
If the board is too big, simply make it smaller. You don't need all that GPIO, do you? :)
Duh :D why didn't I think of that??
mctom wrote:
Fri Aug 19, 2022 3:44 am
Oh wait, you never mentioned a capacitor in your post, did you? :lol:
I did imply capacitors, except I was hoping for a few tens of milliseconds!! :lol:

I need to figure out how much time it takes to remount everything as read-only, and go from there:

Code: Select all

$ echo u > /proc/sysrq-trigger
But perhaps for now I'll just guesstimate and adjust later. Or just ditch this in favor of a proper UPS.
mctom wrote:
Fri Aug 19, 2022 3:44 am
I am working (slowly..) on a 18650 battery BCM for Odroid M1. I have a feeling we'll get out from a component shortage situation before I finish, ugh..
Ahh sorry, not that kind of BCM :D , this: https://en.wikipedia.org/wiki/Intellige ... _Interface

User avatar
mctom
Posts: 1961
Joined: Wed Nov 11, 2020 4:44 am
languages_spoken: english, polish
ODROIDs: OGA, XU4, C2, M1
Location: Gdansk, Poland
Has thanked: 224 times
Been thanked: 279 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mctom »

rmsc wrote:
Sat Aug 20, 2022 1:08 am
mctom wrote: ↑
Thu Aug 18, 2022 8:44 pm
If the board is too big, simply make it smaller. You don't need all that GPIO, do you? :)

Duh :D why didn't I think of that??
Check out this design, where I used only a few GPIO on XU4, just to make more room for LEDs, lol
viewtopic.php?p=340709#p340709
This approach is fun, because it poses low risk of misalignment on a 40-pin header. (XU4 has a 30-pin header, but you get the point)

So perhaps you could do something similar - use 6 or 8 pins on each side and place your smart circuits in between. :)
rmsc wrote:
Sat Aug 20, 2022 1:08 am
I did imply capacitors, except I was hoping for a few tens of milliseconds!! :lol:

I need to figure out how much time it takes to remount everything as read-only, and go from there:
I'd guess it depends how many write operations are pending. AFAIK Linux likes to store data in RAM and take its time to write it physically.
The safest bet is to force shutdown, but I like your idea of remounting stuff as read only. Or just going berserk and unmounting everything. Still not as gracefull, as some applications may not have a chance to complete their write operations.
But hopefully my analysis will be of any use to you. Note it doesn't take ESR into consideration.
rmsc wrote:
Sat Aug 20, 2022 1:08 am
Ahh sorry, not that kind of BCM :D , this: https://en.wikipedia.org/wiki/Intellige ... _Interface
No worries, I don't know what that is. ;)
Punk ain't no religious cult, punk means thinking for yourself!

Maintainer of PiStackMon

User avatar
mad_ady
Posts: 10818
Joined: Wed Jul 15, 2015 5:00 pm
languages_spoken: english
ODROIDs: XU4 (HC1, HC2), C1+, C2, C4 (HC4), N1, N2, H2, Go, Go Advance, M1
Location: Bucharest, Romania
Has thanked: 647 times
Been thanked: 951 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mad_ady »

Unmounting may not be allowed if there are open files (umount will report the filesystem is busy).
Not sure if remounting read-only runs a sync before to flush unwritten buffers. Otherwise that's what you need to do.

In theory you can do sync periodically via the watchdog (e.g. every minute) if you run it with -s, if you expect frequent outages or if the data consistency is critical, but it's not without problems. I had issues on a c2 with a 100Mbps link where I was copying a large file over the network and the sync call would not finish in a minute triggering the watchdog to do a reboot.

Another idea is to limit the filesystem write buffers (most likely available via sysfs), but that would reduce io performance.

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

Hey guys, sorry for the long silence! This is what the board looks like now:
m1-hdd-power-update.png
m1-hdd-power-update.png (941.64 KiB) Viewed 364 times

@mad_ady thanks for the feedback!

Unfortunately it turns out that at this voltage even a few milliseconds of hold-up time are really expensive (both in terms of space and cost). So I think it's best to rely on redundant power at the backplane level :)

Those big caps are there only for ripple reduction, and are huge compared to the rest. On the other hand, the PCB is tiny!

The BOM for this iteration is finished, but I'm still not quite happy with the unit cost (I'm making quite a few for the cluster). I may end up slashing the detailed voltage/current measurement circuitry, and instead just measure the power input.

I'm also toying with the idea of having redundant supplies, but I'm not sure if the dual connector belongs here or at the backplane.

User avatar
mctom
Posts: 1961
Joined: Wed Nov 11, 2020 4:44 am
languages_spoken: english, polish
ODROIDs: OGA, XU4, C2, M1
Location: Gdansk, Poland
Has thanked: 224 times
Been thanked: 279 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mctom »

Big caps? Where? :D

I love the Buck converter layout, clearly you knew what you were doing.
But you will regret those 0402 passives, unless you want to outsource the SMD assembly.

One way of reducing unit cost is making more units. :) You can always build some boards with not all chips soldered on, that's how I customized PiStackMon to make it more affordable. You may also find someone interested in your project and sell a few units.
Last contract I just fulfilled yesterday, I managed to reduce BOM cost by some 40% compared to the initial BOM, by finding alternative parts with local suppliers - in my case, tme.eu.

Redundant supplies belong to the backplane, IMO, unless some cluster units are more important than the others.

I used to work at the company that built systems which switched to battery power in case of emergency. Loudspeakers in the whole building shouted at people in German to get lost because the whole block is on fire. According to German safety standards, it had to withstand yelling for 8 hours after the fire started. What for? Nobody knows. But the batteries were some 80% of volume and 99% of weight.

Anyway, the system's input voltage had been deliberately selected just 1-2V above the battery standby voltage (49.6V?). Two reasons: The battery charging circuits were easier to build, and the redundancy boiled down to a single "ideal diode" circuit.
Punk ain't no religious cult, punk means thinking for yourself!

Maintainer of PiStackMon

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

Thanks!
mctom wrote:
Sun Sep 04, 2022 6:36 am
But you will regret those 0402 passives, unless you want to outsource the SMD assembly.
I've done plenty of 0603, but never had the courage to try 0402 :mrgreen: still not very confident, but at least it's a small board (and no deadlines this time) :D
mctom wrote:
Sun Sep 04, 2022 6:36 am
One way of reducing unit cost is making more units. :) You can always build some boards with not all chips soldered on, that's how I customized PiStackMon to make it more affordable. You may also find someone interested in your project and sell a few units.
Hmm, I could leave out the current monitoring IC, but I'd probably need to keep the current sense resistors. They are 0508's (reversed 0805s), very compact but not very standard. In hindsight using those was probably a bad decision :lol:

:arrow: I wouldn't really mind selling a few units to at least offset the cost of the leftover parts. If anyone's interested please let me know!

I've been relying on TME for quite a few years now. They're not as fast, but are cheaper and sell in smaller quantities compared to others, which is perfect for prototyping. Unfortunately they don't seem to be stocking ICs anymore :lol:

User avatar
mctom
Posts: 1961
Joined: Wed Nov 11, 2020 4:44 am
languages_spoken: english, polish
ODROIDs: OGA, XU4, C2, M1
Location: Gdansk, Poland
Has thanked: 224 times
Been thanked: 279 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mctom »

Ever since I tried soldering with a stencil I'm never looking back - perhaps even 0402 components would be possible, but I already have an extensive collection of 0603 passives so I prefer them. I even have preferred values, like omnipotent 10k. ;)
rmsc wrote:
Sun Sep 04, 2022 11:32 pm
They are 0508's (reversed 0805s), very compact but not very standard. In hindsight using those was probably a bad decision :lol:
You could have added a parallel 0603 footprint, or jumper pads. :)
I'm not fond of non-standard packages - makes the reuse of leftover parts so hard I might as well throw them away.
rmsc wrote:
Sun Sep 04, 2022 11:32 pm
I've been relying on TME for quite a few years now. They're not as fast, but are cheaper and sell in smaller quantities compared to others, which is perfect for prototyping. Unfortunately they don't seem to be stocking ICs anymore :lol:
They are faster than Mouser, they have a warehouse (and headquarters) some 300km away from me. I hear they do same-day deliveries in Łódź. ;)
But indeed the component selection sucks in there. You may count on the most basic parts. But they also offer cheaper Asian components that Mouser doesn't distribute. Hence the substantial savings I made with the last project, especially 50x P-MOSFET in SO-8 package.
In most cases I order Mouser anyway because there's always at least one part I'll need that is not available in TME.
Punk ain't no religious cult, punk means thinking for yourself!

Maintainer of PiStackMon

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

mctom wrote:
Mon Sep 05, 2022 12:57 am
In most cases I order Mouser anyway because there's always at least one part I'll need that is not available in TME.
Exactly! Most passives and discrete semiconductors from TME, more complex ICs from Mouser. They're reasonably fast here, can't complain. I still sometimes use RS, but now mostly for emergencies.

I think I'll be focusing on the power "backplane" now, as that will probably inform the decisions regarding the individual nodes.

I think I've settled on 2 redundant (ideal diode or'ed) power inputs and 4 node power outputs per backplane board. That allows plenty of power redundancy configurations, and yet keeps the individual node boards simple. There should be some auxiliary outputs as well for fans and other 12V devices (network switch 8-) ).

User avatar
mctom
Posts: 1961
Joined: Wed Nov 11, 2020 4:44 am
languages_spoken: english, polish
ODROIDs: OGA, XU4, C2, M1
Location: Gdansk, Poland
Has thanked: 224 times
Been thanked: 279 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mctom »

I do have a network switch lying around that I converted from 230V supply into 5-30V DC input, with "That Buck converter module everybody uses". It's 100Mb switch so I'll be lucky to trade it for 1 unit of currency.

So, are you going to use an integrated ideal diode, or one of those Linear / Analog drivers? Those were quite expensive a few years ago, I was wondering if anything changed on that matter.
Punk ain't no religious cult, punk means thinking for yourself!

Maintainer of PiStackMon

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

mctom wrote:
Thu Sep 15, 2022 5:04 am
So, are you going to use an integrated ideal diode, or one of those Linear / Analog drivers? Those were quite expensive a few years ago, I was wondering if anything changed on that matter.
Given the current involved, it must be a driver plus an external FET. Those from Linear/Analog are not only expensive, they are next to impossible to find :lol: the same for Ti and others.

Right now I'm considering the NCV68261 from Onsemi, as it's fairly inexpensive and even provides reverse polarity protection up to about -40V. The only downside is that it doesn't do anything else, so no UV/OC protection for downstream. And it's new, so not many distributors have stock yet.

I'm looking into other solutions, but they are probably more expensive. We'll see.

User avatar
mctom
Posts: 1961
Joined: Wed Nov 11, 2020 4:44 am
languages_spoken: english, polish
ODROIDs: OGA, XU4, C2, M1
Location: Gdansk, Poland
Has thanked: 224 times
Been thanked: 279 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mctom »

And it has an impossible case to solder :(
I'm not entirely sure how do you expect an ideal diode circuit to protect against downstream undervoltage, that's pretty much an initial condition.

It is possible to design something like that from discrete parts. That'd be quite complex and big and clunky, but at least completely independent from the winds of supply chain. The biggest problem is there are no comparators supporting inputs at Vcc rail, and dividing those voltages down to reasonable levels is another error source.

So it's probably more reasonable to get a chip instead.. Depends how repeatable this design is supposed to be.
Punk ain't no religious cult, punk means thinking for yourself!

Maintainer of PiStackMon

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

mctom wrote:
Thu Sep 15, 2022 7:05 pm
And it has an impossible case to solder :(
I'm not entirely sure how do you expect an ideal diode circuit to protect against downstream undervoltage, that's pretty much an initial condition.
Those controllers also support two mosfets back-to-back, so that the output can be turned off in case of an upstream undervoltage.
mctom wrote:
Thu Sep 15, 2022 7:05 pm
It is possible to design something like that from discrete parts. That'd be quite complex and big and clunky, but at least completely independent from the winds of supply chain. The biggest problem is there are no comparators supporting inputs at Vcc rail, and dividing those voltages down to reasonable levels is another error source.

So it's probably more reasonable to get a chip instead.. Depends how repeatable this design is supposed to be.
Not so long ago I designed something similar using a pretty vanilla low voltage rail-to-rail discrete comparator. The trick is to float the comparator Gnd to say 3.3V below Vcc (a simple zener+resistor shunt regulator should sufice). A PNP transistor at the comparator's output would then "translate" it to normal Gnd. Add the proper protections to keep voltages within safe limits, and you're ready to go.

That said, too many moving parts, and a lot to mess up. So you're absolutely right, unless there was no alternative, I would just get a chip 8-)

User avatar
rmsc
Posts: 17
Joined: Sun Jul 03, 2022 3:09 am
languages_spoken: english
ODROIDs: 4x M1
Location: Portugal
Has thanked: 2 times
Been thanked: 4 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by rmsc »

My bable can be a bit opaque, so here's a what I mean:
floating-comparator.png
floating-comparator.png (104.14 KiB) Viewed 127 times

User avatar
mctom
Posts: 1961
Joined: Wed Nov 11, 2020 4:44 am
languages_spoken: english, polish
ODROIDs: OGA, XU4, C2, M1
Location: Gdansk, Poland
Has thanked: 224 times
Been thanked: 279 times
Contact:

Re: Cluster power supply board for M1 and 3.5" HDD

Post by mctom »

Hmm.. so the Zener business is to just keep the comparator power supply voltage at about 3V3, but close to Vout? Yeah that should work.

I wasn't aware there are actual rail-to-rail input comparators already. Just a few years ago "rail-to-rail" actually meant "50mV away from each rail". Now I have easily found one that would be fit for this purpose. Or perhaps I wasn't looking hard enough back then.
Oh yes, and that comparator would require lower input voltage offset than a MOSFET voltage drop when on.

Otherwise, the idea looks sound to me. Too big to build a single ideal diode, but for many diodes could actually make sense (a shared Zener and 4x comparators).
Punk ain't no religious cult, punk means thinking for yourself!

Maintainer of PiStackMon

Post Reply

Return to “Projects”

Who is online

Users browsing this forum: No registered users and 1 guest