Production Project CPR - Customizable PCI-E Riser (up to X16)

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
Hello, friends!
I want to share with you my ideas and developments, to get feedback or even help!
The essence of the project is to create a universal platform for PCI-E riser, which allows to use it with various devices - from nettops, laptops to stationary PCs.
Local page devoted to development: https://vk.com/x4p_riser
Proceeding from this, the following requirements were formulated:
- Compactness
- External power 12v (with dc-dc 3.3v)
• Ability to use both ATX power supplies with autostart when you turn on the main device (PC), and 12v AC-DC
- Detachable connectors for connection to PCI-E x1, x4, x8, x16, miniPCI-E, M2 ...

More about the implementation approach
-Using in the riser only data lines, clock signals, PERST, WAKE and + 3.3v as external power management. The remaining lines are not critical.
-Interface cable - micro coaxial or twin-axial cable assembly with connectors I-PEX Cabline VS 30pin 0.5mm pitch
-Multifunctionality of ATX power connector (8pin pci-e)
1. Input connection using pci-e 6pin connector (autostart not supported)
2. Input connection using the 8pin pci-e adapter on the 4pin + ATX 20pin CPU (the touch-sensitive 4th pin through the transistor is closed in. The transistor control will allow to realize autostart on a separate PSU)
3. Output (up to 75W) 6pin pcie connection when using 12v AC-DC via 5.5 * 2.5 connector - will allow you to connect video cards with external power supply with a total power consumption up to 150w (maximum power limit due to the connector and mosfet as a key)

At the moment, the components for the main module are selected and placed, and the complement is acquired:





It is necessary:
- Make the final layout of the main module and the separable connectors for different interfaces
- To choose a manufacturer for making cables according to my schemes (negotiations are already under way, I hope for a positive result and a reasonable price)
- Collect money to buy missing components, order cables, order printed boards)))
- Build and test!
I await your comments!
And sorry for my English)
 
Last edited:

AleksandarK

/dev/null
May 14, 2017
703
774
Wow man, that is really awesome work you are doing. What should be the final price?

Also, could you translate your page to english?
That way more people could read it. I am fine with it Russian and i know to read and speak it, but others may not.
 

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
Wow man, that is really awesome work you are doing. What should be the final price?

Also, could you translate your page to english?
That way more people could read it. I am fine with it Russian and i know to read and speak it, but others may not.

It's nice to read such reviews) I'll try to answer the questions.

Translation, who is interested in, I can offer to look through the translator Google - the fact is that I myself through it communicate. Just spread the machine translation does not see the point. In addition, I in my report summarized the current state of affairs, so that there are no special differences. And if it's interesting, I'll post news in vk group and there.

For the price - if we talk about a small batch - then around $ 100 is expected cost. It is difficult now to calculate more accurately. But a couple of samples will cost about $ 250 per piece.
 
  • Like
Reactions: AleksandarK

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
Added for greater visibility 3D model of the video card. This allowed to remove some errors in the placement of components:



Also, there was an idea to transfer the DC power connector on the back side of the PCB and place it between the pci-e bracket teeth:



Which solution is more like?
 

EdZ

Virtual Realist
May 11, 2015
1,578
2,107
As elegant as that centred DC jack placement is for PCB size, it also makes the jack incompatible with the existing PCIe slots on almost any case: not only would it be blocked by the case rear panel, but it projects below the plane of the PCB the card would otherwise plug into, so the motherboard tray would prevent it from fitting. This is a problem if (for example) you want to use the adaptor to add a second card to an ITX board with an m.2 slot, and mount that in case like the NCase M1.
 
  • Like
Reactions: AleksandarK

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
About the lock completely agree, but in such cases the computer power supply still sticks and use of an external DC looks inappropriate. There will be a 6-pin PCI-E connector. So I do not see this as a great tragedy.
But this solution allows you to reduce the size on the back of the video card, which will be useful, despite the increase in width from the front side - still the printed circuit board of the riser will be narrower than the width of the two-slot cooling system. .
 
  • Like
Reactions: owliwar and Elerek

EdZ

Virtual Realist
May 11, 2015
1,578
2,107
It's not the access to the Dc jack, but that the DC jack extends below the PCB. The PCB (being attached to the PCIe slot underside) is at the same Z-height a motherboard PCB would be at. That means the space behind the PCB before you hit the solid motherboard backplate is the same as the standoff height, or 6.4mm. If the DC jack is taller than that, it will prevent the card fitting as it would be pushed higher than the slot opening. The modelled DC jack appears to be the same height as the Mini-Fit Jr. connector, which is 10mm high.
 
  • Like
Reactions: AleksandarK

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
It's not the access to the Dc jack, but that the DC jack extends below the PCB. The PCB (being attached to the PCIe slot underside) is at the same Z-height a motherboard PCB would be at. That means the space behind the PCB before you hit the solid motherboard backplate is the same as the standoff height, or 6.4mm. If the DC jack is taller than that, it will prevent the card fitting as it would be pushed higher than the slot opening. The modelled DC jack appears to be the same height as the Mini-Fit Jr. connector, which is 10mm high.
I understand you and you are absolutely right. If we consider this riser for cases with the provided placement of the video card. But I see my idea more applicable for such solutions called egpu. And in the long run, nothing flushes this connector, if there is such a need)
 

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
It's not the access to the Dc jack, but that the DC jack extends below the PCB. The PCB (being attached to the PCIe slot underside) is at the same Z-height a motherboard PCB would be at. That means the space behind the PCB before you hit the solid motherboard backplate is the same as the standoff height, or 6.4mm. If the DC jack is taller than that, it will prevent the card fitting as it would be pushed higher than the slot opening. The modelled DC jack appears to be the same height as the Mini-Fit Jr. connector, which is 10mm high.
I still thought about your remarks, and now I'm inclined to completely get rid of the DC 5.5 * 2.5mm connector, but leave all the electronics to work with it (mosfet, protective diode, fuse). This way it will be possible to use this connector for DC power supplies with the help of such adapters:
 

EdZ

Virtual Realist
May 11, 2015
1,578
2,107
That would work well, and it'd also allow those who want to use the DC in jack to position it on their case where it can be accommodated rather than having to drill an extra hole in the PCIe bracket area.
 

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
At the Christmas weekend worked out the final version.
- DC connector removed
- Added control function for input voltage with indication of over-tolerance (12v ± 5%)
- Added indicator of the activity of the riser
- The components of circuit protection from voltage surges and short circuits are transferred to the reverse side
Schematic description:



 

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
I have not published news for a long time!
Here at last hands have come to collect and test the first samples!
There were difficulties and problems - the DC-DC converter was unstable, I had to use an external module instead.
The rest of the functions work correctly:
- Initiation of the permissible voltage value 12v
- autostart external ATX PSU
- autostart when using simple 12v power supply
When assembling the first sample, the connectors of the micro-coaxial cables were damaged - it could only be launched in the PCI-E x1 mode

For the second sample, the board was unplugged without additional functions - only power (powered by the same PSU as the PC) and interface connectors.
And it worked! I managed to run 8 lines! Because of the old equipment, I could only test Gen1.

 

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
Just managed to quickly test on a more modern PC under load.
In the test settings, the maximum anti-aliasing was set - this led to a low FPS value. The goal was simply to create a load that would translate the work of the bus from Gen1 to a faster mode.
As a result - Gen2 x8 works stably. But to switch to Gen3, the load from the test is not enough - even without a riser, Gen2.
 

kotproger

SFF Lingo Aficionado
Original poster
Jul 9, 2016
136
131
vk.com
Mind blowing design - PCIe is dark magic to even myself!
But there are freely available various recommendations for designing and wiring boards. This did not make me a "mage of level 80" in part of PCI-E, but it was quite possible to get the working version.

In the same way, a trial was conducted on the shortcomings and mistakes of the first samples. So the completely finished board design is ready.