ACA630 thoughts

About the ACA630's method to fasten the piggyback socket onto the CPU.

These kinds of sockets are hard to make work in any case. What special here is that so many have the same problem with the same product. For example, I don't remember having this with my A608 boards. (They are lighter and less hot.) Others have had this with their 030 accelerators (But they didn't have this fastening solution.)

Usually it's (kinda) just a matter of pressing it down and then it sits there.

On the others, when you had trouble, making sure it was flush and hotgluing it was usually enough. ACA630 is different. The socket is in the corner, and there's a huge slab of Aluminium covering the whole board except the socket, absorbing heat.

What stops the socket from popping off by the weight of it and the socket being in the corner is two legs and two screws. Now, normal plastic expands much more than steel metals. This would perhaps make the screws bend down the board, which is bad but you usually have to bend a PCB very much to make it malfunction (and then permantently, not intermittently as in our case). However, having graphics card type springs under the screws might at least make you sure that's not an issue anymore, that you're treating your card right ;)

But there's really nothing holding down the socket at the rear end of the A600. I actually hotglued my stands, but I think that's not the issue so much as making sure the socket is like a magnet to the 68000 and stays that way. I can only see the stands affecting things if the socket was soldered slanted or somehow too near the surface of the ACA PCB so that the stands lift the rear end of the PCB and the screws pull it down, ie. the socket is just slightly slanted over the 68000.

With the heat generated in the A600 (feels like 60 C but might be just 45), *just* doing it with hotglue and maybe some pillow or such (under the rear end to give some weight support but not risk pushing it up with something hard) might not give optimal results.

Still I would like the force that holds the ACA to the 68000 to be only around the socket, so that no matter what else expands in the A600, the socket stays like glued to the 68000, with only the normal heat expansions that other accelerators have.

Anything that is used to fasten it should either be soft and just keep it from failing its 'designed-for' just-pop-it-on-and-it-stays function, so that heat expansion at least doesn't move the socket, or the same material as the socket to have the same expansion.

Here's where you need imagination. :) Some block of not-too-soft styrofoam-like material (but not styrofoam, it's not springy enough) that makes the A600 case top press on exactly the right place when the case is screwed shut, maybe. It would have to be glued to the ACA, since the case (well, keyboard) is slanted.

Glued (and isolated) plastic nut on the mobo along the left edge of the ACA PCB, and a plastic screw with a big underside/washer, half of which is over the edge and with which you can secure 3/4 of the corners of the socket, at least. And removing the current screws and standoffs, of course.

I keep returning to the big slab of Aluminium that gets so hot. Replacing it with a more conventional cooling solution would take a bit of the weight off, so that maybe hotgluing 3 corners of the socket might help.

-~-

And through all this, I couldn't deomnstrate that heat (and thereby heat expansion) was the villain, and yet it seems the only cause for any change in 'socket-in-placedness'.

Perhaps the socket height is just slightly too high or low, so that the pins are on the verge of being pushed to a not-conductive-enough-for-power-signal position by height expansion of the socket alone - even if it's ever so flush with the mobo. The hotglue would give a little and that would be enough.

And again, if heat expansion affected it directly, how can it work perfectly at the end of an 8-hour session, day in and day out, and then suddenly starting to consistently fail to boot every time, regardless of being booted while it's warm or room temp?

Something happened to it. Reseating fixes it (for almost all). Maybe you have to reseat it when it and the A600 are both warm or both room temp, and it will always work to reseat it.

If it's mechanical rather than heat, a 68000 resolder would make the problem never ever appear again. This needs a poll I think :) Or a test. If it ever happens again, then that is NOT the reason.

There are many special features of this accelerator, it's wonderful to use and has many fabulous options, but also has a very unconventional design. For certain reasons, of course.


-~-

If it can be proven that a resolder makes the problem disappear, never to return, I'll concede.

But I think it's one of those times when you should use your accumulated experience, and modify the design to make it work like others. And also to just generally take care of your hardware. I wouldn't subject other PCBs to the kind of heat that's inside my A600, and even if it turns out it's not the culprit for the ACA problem it could reveal weaknesses in designs of other expansions you put in.

So I come back to the original idea: remove the Alu flange and thick gob of heat-conductive. Add proper, normal cooler with razor-thin PC cooling paste to keep the CPU operating. Memory cooling flanges if necessary. If the heat in the case is too much, blow it out the case with a nice little fan.

Fasten the socket like you know it works for other accels.

Simply, taking care of your computer.

Then we'll see.

-~-

We all want to expand and speed up our Amigas, but since we love to have the computer under our fingers, we'd like it to be completely quiet.

At some point, there will be an amount of expansions inside that makes this impossible, and it's best to give in a little - either swap cases, modify expansions, or - God forbid - add a fan.


/Photon 'too much time on his hands' Erlandsson signing off...

 

Comment this post

  • Required fields are marked with *.
  • Comments may be moderated.

If you have trouble reading the code, click on the code itself to generate a new random code.
Security Code: