POC Computer Version One

[GARTHWILSON]

Quote:

Of course, no dense layout can be done without via,

In digital SMT work, that would be true, but I've done some thru-hole analog stuff with densities sometimes exceeding 40 parts per square inch, putting up to 11 resistors and diodes under each 14-pin IC, staggered in five rows. Those boards had no vias (there was no room for vias!); but in thru-hole, every pin goes to all the layers anyway.

Quote:

so the other trick I used was extra small via, which introduce less inductance into the circuit.

I generally consider .015" diameter holes with .035" pads to be the minimum reliable size for .062"-thick board (at least the minimum without paying a heavy premium), and .008" holes and .028" pads the minimum for .031"-thick boards. Any smaller and it becomes difficult to make the thru-plating reliable. [Edit, 2018: Board manufacturers are telling me now that they can reliably do much smaller holes, and that .008" vias in a .062"-thick board are fine; so I'm designing a board for work with that size of vias now.] I do tell the board manufacturer however that although other hole sizes specified are finished hole sizes (ie, after plating), it's ok if vias are filled with solder.

[OwenS]

Sorry; I didn't realise that the board was 4 layer; I must say, for PTH chips (And those in PTH sockets) it does seem a little overkill though

(As a final note, there is one other reason to ground pour your board: Thermal conductivity! The one I currently have out to manufacture will be stuck to a heat sink, with the ground pour effectively acting like a heat spreader for the chip)

[BigEd]

OwenS wrote:

a little overkill though

yeah - it's an interesting discussion, and these are crucial issues for the consulting engineer, but I wouldn't want to discourage any beginners: up to a few MHz, with 5v supplies, for hobby-grade reliability, anything goes. If you want to sell something, or if you've gained a bit of experience, there's always more knowledge you could apply.

Imagine if you never dared to hook up an LED flasher on a breadboard because you had more to learn about ground planes.

[BigDumbDinosaur]

OwenS wrote:

Sorry; I didn't realise that the board was 4 layer; I must say, for PTH chips (And those in PTH sockets) it does seem a little overkill though

Overkill? The inner layers made it so much easier to route the board, since I didn't have to concern myself with ground and power trace placement. It matters not that most of the parts are PTH. The benefits of the 4-layer design are still valid.

Quote:

(As a final note, there is one other reason to ground pour your board: Thermal conductivity! The one I currently have out to manufacture will be stuck to a heat sink, with the ground pour effectively acting like a heat spreader for the chip)

There's nothing on this board that runs remotely hot enough to be a concern. The disadvantages of filled planes substantially exceed the essentially non-existent benefits.

[BigDumbDinosaur]

In a marathon slobbering session last night, I assembled my POC 1.0 computer. Amazingly enough, I was able to slobber the SOJ32 SRAM package on the first try and not get any faults, bridges, etc. The trick, as described in a U-Toobe video I watched to see how it's done, is lots of flux (I used a flux pen, which works very well), a very small and clean soldering iron tip, and a modicum of patience. Once the SRAM was done the rest was easy, being all through-hole. I made one assembly error in reverse-installing an electrolytic but fixed that. I will post pictures of the completed assembly as soon as I have them.

All the critical circuits have been metered out and it appears there are no faults. It should be possible to put the juice to this thing and not fill up the shop with acrid smoke or crowbar the power supply. I won't have time today to power it and see what happens—perhaps on the day after Christmas. I'm afraid my wife may try to stick the soldering iron into where the sun don't shine if I work on this thing on Christmas Day.

Speaking of working with small SMD packages, has anyone here devised an alternative to hand soldering to mount those things? Anyone figure out how to make a reasonably good equivalent to a reflow oven?

[kc5tja]

I understand a lot of people use a relatively inexpensive toaster oven for this work, but it obviously limits the size of the printed circuit board you can use.

[OwenS]

Skillets are also reputedly popular

[BigEd]

Hot air paintstripper? Or hot air attachment to a butane soldering iron?

[GARTHWILSON]

Judging from what I've seen happen when I tried to use a stripper to salvage big, expensive parts from bad boards, I wouldn't recommend it for construction. Hot air is is ok if the airflow is very low and you can control it precisely. Someone I talked to at the break at a seminar however was using a toaster oven for low-volume assembly in their home business.

[BigDumbDinosaur]

GARTHWILSON wrote:

Someone I talked to at the break at a seminar however was using a toaster oven for low-volume assembly in their home business.

How did they determine the correct temperature to use and the amount of time to "toast" the PCB? It sounds kind of dicey to me.

BTW, contrary to my earlier post, I decided to put power to the POC today and at least determine that the reset circuit was working. It was, so I figured as long as I had the thing attached to the power supply, I might as well put the Ø2 oscillator in the board and see if there was any sign of life. My ancient Beckman 'scope (mid-1970s vintage) displayed a lovely square wave on the Ø2 circuit, so that's out of the way.

I should have stopped there, but since I had a clean Ø2 (as clean as the old 'scope can display it), I decided to insert the MAX238 transceiver and verify that the required EIA-232 logic voltages were present. They were.

Next step, which won't happen today (my wife told me to quit fiddling with my toys and get ready for our company that's coming here for Christmas Eve), is to plug in the MPU, EPROM and Dallas timekeeper and see if she goes or blows.

Ain't electronics fun?

[GARTHWILSON]

Quote:

How did they determine the correct temperature to use and the amount of time to "toast" the PCB? It sounds kind of dicey to me.

Experimentation, I suppose. Putting a temperature sensor on the board would not be very difficult.

[BigDumbDinosaur]

GARTHWILSON wrote:

Quote:

How did they determine the correct temperature to use and the amount of time to "toast" the PCB? It sounds kind of dicey to me.

Experimentation, I suppose. Putting a temperature sensor on the board would not be very difficult.

Obviously, the temp would have to be sufficient to melt the reflow on the PCB, with a little extra to spare. The metal parts would heat much faster than anything else, so the oven could be very hot to get the process over with as quickly as possible and reduce heat soak into the parts themselves. I'd be reluctant to try this without more info.

[GARTHWILSON]

Quote:

Obviously, the temp would have to be sufficient to melt the reflow on the PCB, with a little extra to spare. The metal parts would heat much faster than anything else, so the oven could be very hot to get the process over with as quickly as possible and reduce heat soak into the parts themselves. I'd be reluctant to try this without more info.

The process could have more than one step, like bringing it all up to a lower temperature first and keeping it there for awhile to reduce thermal gradients later when you actually melt the solder. The boards I saw pictures of that they made were much smaller, and being production, they had more, so they could experiment some, and any losses would just be part of the investment.

Semiconductors themselves can handle much higher temperatures than you might think though. Some other parts like capacitors and connectors can't handle nearly as much. When I worked in applications engineering at the VHF and UHF power transistor manufacturer, I saw, while thermal scanning with the infrared microscope, transistors actually operating (not just in storage or being soldered) at hundreds of degrees C. I don't remember exactly anymore, but I know it was above 350°C. It was not a normal situation, being that there was a faulty die-attach to the package and the amplifier was putting out hundreds of watts into a deliberate 3:1 SWR worst-phase mismatch for testing instead of a perfect 50Ω load. The engineering manager came by and was not amused that the temperature got so high. The transistors' life would be very short at that temperature, but it was not destroying them in the short time that would required for soldering. Some semiconductors are unable to handle such high temperatures because of the chemical reactions with the packaging plastics against the die; but I was scanning bare dice. I mostly worked with MOSFETs but some bipolars too. I don't remember which one this was. The MOSFETs could handle the high SWRs far better than the bipolars which would go into thermal runaway and blow up.

I never did find out what temperature they use for the die attach; but the die was literally brazed down to a metalized ceramic substrate which was an excellent conductor of heat and would let the heat out into the heat sink underneath as long as there were no voids in the brazing. No soldering iron could melt the brazing material. We frequently made fingernail-sized transistors actually dissipate hundreds of watts, and the heat sinks had channels of water running through them with a pump under the workbench. I had a 4,000-watt power supply at my workbench and an RF dummy load that was bigger than a PC tower case.

[BigDumbDinosaur]

GARTHWILSON wrote:

The process could have more than one step, like bringing it all up to a lower temperature first and keeping it there for awhile to reduce thermal gradients later when you actually melt the solder.

Here's some info about using a toaster oven as a reflow oven. It would seem the process shouldn't be too tough with 50 mil pitch components. The author gives some info on temperatures and timing but is careful to mention that it's "art." Just don't crank up the temperature too high or your new creation becomes "black art." As I have three unpopulated boards left, as well as some extra SRAMs, I may give it a try. The worst that could happen is my board and SRAM (literally) become toast.

[Nightmaretony]

Used to use a hot air gun to remove SMT parts at a previous job. Was able to salvage and reuse chips nicely. The trick was to keep the heat over a large area so it all heated up evenly, or you would warp the board.

Soldering was by hand, though...