Prototype KMPKT Dynamo DC-12V

[aquelito]

You've got yourself a customer !

Great solution to power a thin Mini-ITX / mSTX w/ discrete GPU build using a single 19V AC adapter.

Two comments though :

- how do you expect to sync your DC board with the motherboard in a "thin Mini-ITX / mSTX w/ GPU using a single 19V AC adapter" scenario ?
- while it fits vertically in small cases (for instance in my next ISK-110 build), it won't fit vertically into a 1U unit (44mm, including panels).
Possible to decrease the width to 38/39 mm ?

 

[cleveland]

You've got yourself a customer !

- while it fits vertically in small cases (for instance in my next ISK-110 build), it won't fit vertically into a 1U unit (44mm, including panels).
Possible to decrease the width to 38/39 mm ?

+1 on that, sir.

ISK-110 is the front door of the SFF world and, as such, deserves some love.

 

[iFreilicht]

Possible to decrease the width to 38/39 mm ?

Looking at the connectors, there's no way he can slim it down that much.

ISK-110 is the front door of the SFF world and, as such, deserves some love.

It does fit in a vertical orientation now

- how do you expect to sync your DC board with the motherboard in a "thin Mini-ITX / mSTX w/ GPU using a single 19V AC adapter" scenario ?

What do you mean by "sync"?

 

[Kmpkt]

- how do you expect to sync your DC board with the motherboard in a "thin Mini-ITX / mSTX w/ GPU using a single 19V AC adapter" scenario ?

The V-out is simply going to pass through whatever the AC-DC brick provides. So if you use a 19V brick and put this unit between your 19V powered motherboard and the brick, it will simply sip off what the GPU and SATA needs while allowing full power flow through to your motherboard. More simply put, you can think of it as a power splitter (in a sense). As long as the components do not exceed the rated wattage of the brick, there should be no issues. If you have a thin-ITX with a two pin header, you could just make a 2 pin to 2 pin connector and plug it into two of the six pins on the 6 pin V-out (on 12V and on GND). The same would be true of a 4 pin solution. A 6 pin simply allows for a more powerful 19V device to be powered downstream.

- while it fits vertically in small cases (for instance in my next ISK-110 build), it won't fit vertically into a 1U unit (44mm, including panels).
Possible to decrease the width to 38/39 mm ?

It has a 43.5mm width specifically to allow it to accommodate 1U builds and this was a conscious design decision. While I'd love to accommodate everything people are doing here, like any product some sacrifices had to be made. To get it narrower, I'd need to sacrifice a PCIe slot and maybe even SATA which are pretty significant changes. Also I'd really like to see someone daisy chain two of these and a 160W DC DC with the new Eurocom 780W (300W + 300W + 160W = 760W!) for some bifurcated dual Titan X goodness on an AsRock X99 ITX board and that requires 2 PCIe per card : )

 

[Kmpkt]

Updated OP with updated dimensions (138mm x 43.5mm x 26mm) and new renders of the board. Again any suggestions for improvement are welcomed and PM me if you're interested in being added to the email list for launch. Cheers!

 

[iFreilicht]

Also I'd really like to see someone daisy chain two of these

Hm, what's the maximum current rating on the pass-through? It might be a bit of a stretch to put another one of these and a direct plug DC/DC on just the 6pin. Probably better to make a Y-splitter and connect them in parallel.

 

[EdZ]

What do you mean by "sync"?

I think he means: How do you prevent the Dynano from providing 12V power to peripherals until the motherboard has requested power from the PSU. i.e. to avoid the state where your GPU is receiving 12V power from the PEG connectors, but the motherboard is powered off.

 

[Kmpkt]

Hm, what's the maximum current rating on the pass-through? It might be a bit of a stretch to put another one of these and a direct plug DC/DC on just the 6pin. Probably better to make a Y-splitter and connect them in parallel.

Point taken. Larry seemed to think it would be okay, but I'll double check the ratings and get back to you. That being said, splitting to two units should be quite simple in theory.

I think he means: How do you prevent the Dynano from providing 12V power to peripherals until the motherboard has requested power from the PSU. i.e. to avoid the state where your GPU is receiving 12V power from the PEG connectors, but the motherboard is powered off.

I actually thought that in this scenario the PCIe bus passes along a signal to the GPU to turn on with the correct timing. Am I missing the point of the question?

 

[aquelito]

I think he means: How do you prevent the Dynano from providing 12V power to peripherals until the motherboard has requested power from the PSU. i.e. to avoid the state where your GPU is receiving 12V power from the PEG connectors, but the motherboard is powered off.

That's exactly the question have in mind !

@Kmpkt, as I have asking you in PM, in the case of a Thin Mini-ITX build with GPU / eGPU :

I actually thought that in this scenario the PCIe bus passes along a signal to the GPU to turn on with the correct timing. Am I missing the point of the question?

In my scenario, I am using a powered riser + 6-Pin PEG to feed my GPU (GTX 1060), using an HDPLEX 250.
If the HDPLEX is not jumped, the GPU won't boot as the HDPLEX is powered off.

@QinX small MOSFET circuit allows to sync the HDPLEX with the mobo, by "jamming" its PS_ON and GND pins every time the mobo wakes up.

How do you do if your board does not have a PS_ON function ?

 

[EdZ]

I actually thought that in this scenario the PCIe bus passes along a signal to the GPU to turn on with the correct timing. Am I missing the point of the question?

In theory the GPU (or any other high power PCIe card) should be able to power up/down normally while 12V is supplied to the PEG connectors. In practice, no compliant ATX or SFX power supply should be continuously sending 12V to the PEG connectors, so this may be an edge case that some cards are not rigorously tested in. There may be other 12V devices that you may wish to power that would also be continually 'on'.
There is also the case of wanting to power a GPU through a powered riser (in the case the motherboard cannot provide the 5.5A needed on its own). In this case, a switched power supply is definitely needed.

 

[Kmpkt]

Okay I've got it now. So there is (or at least was) a jumper on the board as well as a two pin switch header to allow for turning the whole board from always on to switchable for eGPU applications. In this case I think you could use a dual pole bulgin to solve this problem if I'm not mistaken. If there is another more elegant solution for this problem or I am mistaken, please let me know as I'd prefer to get it built in for the prototype. I've contacted Larry to confirm that this functionality is still in place. Cheers.

 

[aquelito]

@QinX already tried this MOSFET-based solution in his various thin Mini-ITX builds.
This is a very elegant solution

One end of the MOSFET circuit is plugged into any header of the motherboard providing 5v or 12v (fan header, SATA header, USB header...).
The two other cables are plugged into the PS_ON and GND pin of the DC board 24-Pin connector.

When the motherboard turns on, a 5v or 12v is applied to the MOSFET, which "shorts" both PS_ON and GND pins.
The HDPLEX starts.

Same as the paperclip trick, but controlled by the motherboard through a MOSFET.

 

[Kmpkt]

So this was Larry's reply to my the above two questions:

No limit on the pass through. They are welded together. Think them as a one DC input.

Since there is PWR on two PIN on the 19V-12V sames as the PS_ON and Ground on the 24PIN port, maybe a Y shape PWR wire connect to 19V-12V and PWR button will do?

So as far as the passthrough goes, the only real limitations would be on the wires used to connect the units as well as the crimps on either end. The rated amperage (from my brief google search) for 18AWG is 10A and the Minfit connectors are rated for 9A (I'm probably wrong on one or both of these counts, I'm sure someone could correct me or confirm these values). Given the input range of this board is 16-24V, at 27A you're talking safe transmission limits of 430 to 650W and 510W @ 19V.

As far as dual pole solution, I believe this is exactly what Larry is suggesting and should correct the aforementioned problem. Again please correct me if I am mistaken, unfortunately I am a physiotherapist and not an electrical engineer.

 

[Kmpkt]

Okay this thread has kind of confused me so I need a couple of questions answered since I am getting conflicting information from a couple of different sources. What I understand from what Aquelito is saying is the HDPlex 250's default is an off state with no power to the GPU. Once the power button is pushed and the board is jumped, then the board turns on and sends power to the GPU. Is this correct, or am I confusing the matter?

What I know is that the Dynamo has a jumper to set the state between the 12V being always on and switchable off. If the board is in an always on state, then this shouldn't be an issue should it? My understanding is that if the jumper is set to always on, the GPU will essentially be powered but not active. Once the PCIe signal following motherboard power on is received then it will activate, no? There will also be the option if the jumper is in the off position to then use a fan header to act as the syncing switch which I believe is exactly what Qinx did.

I hate to seem like an idiot, but can someone please clarify exactly how motherboard and GPU power up works so I am clear on it? I thought I was and now I'm not so sure.

 

[aquelito]

Such as you @Kmpkt , I am no electronic engineer, just an interior designer !

From what I have experimented, the HDPLEX used with a thin-Mini ITX board doesn't start unless its 24-Pin connector has been jumped :

- either with a paper clip jammed into the PS_ON and GND pins : then the HDPLEX is always ON, as long as the AC adapter is ON.0

- or with Qinx MOSFET circuit, which allows to start the HDPLEX, as would do a "standard" a mini-itx board.
He was kind enough to build and send me one to test ; however I either f***ed up the pins or it was damaged during transport and I cannot fully test it right now (I'm just using it as a "paper clip").

So, right now, my HDPLEX is always ON and I have to switch off the master command of my multiple plug to shut it down.
No issues so far but I do not like the idea to have my powered riser constantly fed with power.

You might want to get closer to @QinX regarding those questions as I believe he has the most extensive experience regarding those issues.

 

[EdZ]

I'll repost my PM here for discussion/correction:

In theory, a PCIe card should ignore any power applied through the PEG connectors until the normal power-up process is initiated through the motherboard.
In practice, a situation where a motherboard is turned off but the PSU is continuously providing 12V should never occur with an ATX12V compliant PSU, so whether cards behave as expected cannot be taken for granted (could end up with anything from it all working OK, to the inconvenience of a card not powering back on properly, to VRMs being powered in the absence of active cooling).
A secondary issue is powering any other peripherals (e.g. fans, pumps, LEDs, etc) from the Dynamo would not work if the power is always on, as those peripherals would be continuously powered even when the PC is 'off'. This is more for convenience of wiring peripherals to a separate board, rather than wiring things all the way back to the plug-in module.

If the board already has a power-on trigger, then that's all that is needed.

The ideal way to trigger this would be to watch the PS_ON line on the 24 pin ATX connector. According to the ATX 2.4 specs (page 19 para 3.3.3), the PS_ON line will drop below 0.8V when the motherboard wants the PSU to power up. If the Dynamo board includes a high-impedance trigger pin (so it doesn't pull the PS_ON line low itself) then that can be hooked to the PS_ON line (pin 16, green wire) to trigger both PSUs at the same time. If the Dynamo board itself cannot accept this trigger, a small external board could be used to monitor the PS_ON line and provide the correct trigger.

A future plug-in DC revision could even provide a dedicated 'outboard PSU' trigger connector, or at least an easy to access header to reach the original PS_ON line. For current plug-in DC boards, the easiest ways to access the PS_ON line would be to solder a wire to the exposed pad on the side of the plug-in board (where the pins protrude form the top of the Mini-fit Jr connector), or to slip a fine wire into the Mini-fit Jr socket on the plug-in module, then push it onto the motherboard, trapping the wire in contact with the pin without needing to solder anything.

A simpler, but less reliable, way to trigger would be the PWR_OK line, which goes to +5V once the PSU is on. The problem with this is that the motherboard uses this as the 'power supply is all OK now' trigger and starts turning on peripherals, so if the Dynamo is just starting to power on after this line goes high, it may not be ready to supply power if peripherals demand it immediately. This would be one of those "probably OK most of the time, but could cause difficulty to track down edge-case problems" situations.

If Larry at HD-Plex can confirm whether any of the exposed 12V or 5V pins on the HD-Plex 160W go high long enough before the PWR_OK line to allow for the Dynamo to power up, that would be the easiest wiring solution for an end user. The problem is that would only apply for that particular plug-in solution, rather than being a more general solution.

 

[aquelito]

Some interesting reading regarding a similar eGPU application :

https://www.techinferno.com/index.p...ter-discussion/&do=findComment&comment=113469

 

[Kmpkt]

We have built in a variant of the @QinX fan header mod so any DC-DC unit can be used with the Dynamo without worrying about the board being out of sync in an always on state. All you will need to do is us a fan header to send a 5V signal back to a header on the Dynamo on startup. This board will be set to one state or the other (always on vs switchable) via a jumper. From my research, this likely will not be necessary in a lot of cases anyhow as GPUs and hard drives/SSDs SHOULDN'T turn on even with the Dynamo in an always on state until they receive a signal to do so from the motherboard. That being said, I don't want this being something that ruins peoples' custom builds as I really want this supply to make creative builds as easy as possible for the community. We are also trying to build in PS_ON monitoring so that if the end user wants to run a thin PS_ON cable back along the 4 or 6 pin connector to enable normal PSU functionality, they will be able to do so. I hope this is the fix you were looking for @aquelito - thanks for the awesome feedback. Also thanks to @EdZ for helping me figure all of this out. Nice to have such a knowledgable community here.


Also have updated dimensions as they have changed slightly since the last update. Dimensions are now looking like 130mm x 45mm x 26mm.

 

[aquelito]

Thanks a lot for your follow-up @Kmpkt, that's great product design.
Again, these are empirical observations from my limited experience ; double check everything with the more knowledgeable people out there.

Eager to see the first prototype

 

[Kmpkt]

If there are any other features people would like to see, please let me know ASAP as we'll be getting first prototypes made early next week.