Where Are the Cables???

[TheWhole___]

15 hours ago, Nystemy said:

Actually, putting stuff on both sides of the board is really nothing new. Well, in the PC space it partly is. But in the PCB manufacturing space it really isn't.

 

However, through hole components tends to require hand soldering one of the sides instead of wave soldering it. _{(But most PC motherboards likely don't see the tin wave to start with, since it doesn't really play well alongside SMD components, even when the SMDs are glued in. And yes, gluing in SMDs is a thing. And most motherboards I have seen don't show any signs of having been wave soldered, so either they do an amazing job at masking everything, or they simply use other soldering methods. Perhaps just SMDing the through hole ones as well, smearing in past into the through holes can lead to somewhat decent and repeatable results after all.)}

Nystemy is more or less correct. Some additions for those who are intrested:

There is a process for selective THT soldering. It's just a nozzle that "sprays" the tin where it needs to be.

If there are only simple SMT components (resistors, capacitors, ferrit beads) there is no problem with glueing them in place and put it over the wave with THT components. Without glue they would just wash away.

And "SMDing the through hole" is not a thing as far as I know, there wouldn't be enough tin for a good connection.

 

15 hours ago, Nystemy said:

For SMD components, then double sided boards are common place, and a lot of production houses skips gluing them down and simply does one side first, soldering it and everything before doing the other. As long there isn't any drift sensitive parts on it, then it is fine. _{(and with drift I mean a change in electrical specifications due to the fairly high heat of soldering.)}

 

For heavier components, usually connectors, then glue typically is used to hold stuff in place. But some connectors latch into a set of guide holes removing any need for glue.

 

Normally when designing a PCB you put all the heavy components on one side, so there is no problem with going through the reflow process twice. If thats no possible you have to look for ways to hold the components in place, like you said.

A components drift has nothing to do with that as long as you don't exceed the recommended soldering temperature.

 

But there can be temperature problems with some components. We had some at work some time ago with DC/DC converters, which themeselves were PCBs with a closed case. Our manufacturer soldered our boards to hot and the components inside the DC/DC converters came loose. That took some time until we knew what was going on...

 

In the industry you don't really go over 260°C, ideally about 240°C (250 for Pb-free) which is far cooler than hand soldering.

 

 

 

 

[Nystemy]

5 hours ago, TheWhole___ said:

Nystemy is more or less correct. Some additions for those who are intrested:

There is a process for selective THT soldering. It's just a nozzle that "sprays" the tin where it needs to be.

If there are only simple SMT components (resistors, capacitors, ferrit beads) there is no problem with glueing them in place and put it over the wave with THT components. Without glue they would just wash away.

And "SMDing the through hole" is not a thing as far as I know, there wouldn't be enough tin for a good connection.

 

Normally when designing a PCB you put all the heavy components on one side, so there is no problem with going through the reflow process twice. If thats no possible you have to look for ways to hold the components in place, like you said.

A components drift has nothing to do with that as long as you don't exceed the recommended soldering temperature.

 

But there can be temperature problems with some components. We had some at work some time ago with DC/DC converters, which themeselves were PCBs with a closed case. Our manufacturer soldered our boards to hot and the components inside the DC/DC converters came loose. That took some time until we knew what was going on...

 

In the industry you don't really go over 260°C, ideally about 240°C (250 for Pb-free) which is far cooler than hand soldering.

 

 

 

 

Yes, putting heavy components on both sides is when glue is needed, or a bit of hand rework.

Glue is also indeed useful for wave soldering over SMD components. However, fine pitched components tends to bridge in the wave, but clever placement and other sneaky tricks do help a bit. _{(larger pads on the ends of a package, and avoiding having pins on axis to the wave itself is two simple tricks. The solder resist also tends to helps a bit if put between pins.)}

 

In terms of "SMDing" through hole, it relies on paste being pushed into the through hole itself, this is what gives it enough paste to solder somewhat properly. However, it needs somewhat fluid solder paste that doesn't end up extruding out when the component legs is inserted. And to be fair, few people do this since hand soldering a connector isn't all that slow in practice. I have mainly used this technique in paces where there isn't room to hand solder, while being on the wrong side for wave solder, ie, it is niche. In terms of bond strength, it isn't actually that bad. Though, pressfit is starting to take over this niche.

 

In regard to component drift, I still mean a drift in behavior, it really isn't a major concern for digital logic, but all semiconductors do age faster at a higher temp, and some analog components even specifies their specifications after 1 reflow cycle. It doesn't matter much for most applications however. _{(If one buys a laser trimmed component one usually don't want to offset its value by thermally cycling it to 200+ C. And here hand soldering is usually the trick, since leaded components don't soak up as much heat if one only heats one pin at a time. Or one can just put it in a socket but here contact resistance is another consideration.)}

 

Soldering temps largely depends on the components and the soldering setup, not to mention the solder itself. But yes 220-250 C is typical, but going to 280 isn't uncommon, and a lot of IR ovens generally don't have "a temperature" but instead everything just generally heats up to "some" temperature above the solder's melting point, but a bit of airflow tends to keep the highest temps at bay and help the low ones up a bit. IR is fast and simple, but handling components with wildly differing IR absorption and thermal mass isn't its strong side. I personally use a vapor oven instead, but it has its own slew of downsides.

 

_{(I should probably point out that I do electronics manufacturing and designing for a living, and use these tools quite often. And in Gigabyte's case, I would go with hand soldering the back connectors, since they are more or less just power and a few pin headers. And I expect this is what they did, labor isn't too expensive and soldering isn't all that hard, especially if they are only going to make a few thousand boards of this "style first" design, something one can often request a higher price for in the end regardless.)}

[GhostRoadieBL]

How many other cases can work with a card like that? I know of a few which have grommets in the right area so it's likely to be quite a few.

 

There's probably a way to make custom flat profile cables for a 3090 to work with the system and a lot of motherboards are through soldered for power connections so adding a tail for other motherboards to work wouldn't be a ton of work.

[mariushm]

Eh, I see this as a half-assed job. 

 

Being a Gigabyte motherboard AND a Gigabyte video card, they could have taken it a step further and just EXTEND the pci-e x16 connector on the card and the slot on the motherboard with a segment just for power delivery. 

The current cards take in up to 65w on 12v through just 5 contacts in that small separate segment on the card, so it's possible to add a segment with more pins just for power. And, it still allows you to use regular video cards, with the downside of having wires visible. 

The downside would then be that the video card may not be usable in other motherboards, if other motherboards have some components right after the x16 slot, blocking the extended edge connector of the video card.

 

Other things that bother me... the fan headers popping up in random places on the back of the motherboard, same for usb headers, audio header. Same boring 0.1" spaced pins. 

They could have used a low profile right angle surface mount connector, maybe have 2 usb 2.0 headers (4 ports) in a single 12-14 pin 4 ports x 2 data wires (d- and d+)  plus power go to a small custom expansion board / pcb mounted that makes those standard size usb 2.0 headers

For the audio header, maybe they could have used some micro coax wires for the analogue output for cleaner sound (like the wireless antenna cables)  

 

EVGA already did some cool things with right angle connectors and cutouts on the edges of the motherboard so you could use right angle plugs and route cables directly to the back side, so just reorienting the traces on the motherboard to have the pins on the bottom is not clever enough. 

 

They could have also made the motherboard 12v0 and then have only 12v going into the motherboard - you can still use a small adapter cable to use regular atx power supplies.  You have the 10-12 pin connector instead of 24 pin, you have a 8 pin or 2 x 8 pin for CPU , they could add maybe 2 x 8 pin pci-e close to the pci-e slots if the power is delivered through motherboard.  They could group together basically the 10-12 pin (24 pin replacement) and the video card power connectors somewhere near the middle of the board, closer to pci-e slot, and the cpu power connectors at the top corner.

 

The board could have a strong 5v dc-dc converter for the RGB lightning and USB ports and to power 2-4 sata mechanical drives, and for everything else it could have 3.3v for M.2 drives in the front. 

 

[Ross Siggers]

You guys can all have your in-depth tech discussion, I'm just sitting here thinking....

"Isn't kerjigger a word from Futurama?"

Has Linus's extensive exposure to the show effecting his language? Am the Big Brain winning again?