[TriEmbed] T-962 reflow oven notes

Pete Soper pete at soper.us
Fri Sep 26 11:54:21 CDT 2014


Sorry, I wasn't clear. We can get another POS T-962 oven for $265. :-)  
Here is a prototyping reflow oven 
<http://www.ddmnovastar.com/reflow-ovens/low-volume/gf-b-ht-reflow-ovens> that's 
on the other side of the virtual and actual planet from the T-962. The 
price is from some other planet (20X the T-962 price).

But compared to an electric skillet or many homebrew broiler oven 
setups, the T-962 is the cat's meow. As I said, there is obviously a 
huge lattitude within which people find some level of success. There are 
no doubt makers who think I'm completely silly: Hey, just crank up your 
kitchen oven, stick the stuff in there, watch it carefully and use skill 
and experience and you're done. I'm reminded of a recent PBS documentary 
showing a master black smith forging two pieces of metal together with 
one big hammer such that it was impossible to to tell they had been two 
pieces. You and I could do that too, with enough hours of practice. :-)

(We should press on with the project to make PCBs specifically designed 
to test our soldering processes. Another thing I'll be pressing real 
soon now is pointing one of my microscopes through the oven window to 
hopefully watch the melting/reflow process. It's embarrassing, but I 
simply haven't found the time coincident with reminder of how valuable 
this might be to lash it together.

On a related note: Beware of sending off orders to China right about 
now. They start a 10 day holiday on Wednesday and not much of anything 
is going to happen to orders to places such as DirtyPCBs 
<http://dirtypcbs.com/> until that's over.)

The direction I'm obviously headed is to replace the entire controller 
in the oven if I can't make it more accurate by connecting alternative 
probes into it and arranging for the probes to be stuck to the PCBs. 
With a pair of Adafruit type K probes and a pair of breakouts that 
translate the probe signals to SPI or I2C digital data, then with a 
couple of solid state relays and/or power FETs that are already in my 
parts drawers, "just add software", right? There are no doubt thousands 
of freely available PID controller "sketches" out there, and via use of 
something like a Raspberry Pi the very seriously good PID software could 
no doubt be lined up together with thermocouple calibration, training 
modes, yada yada.

But we can't do that right now.  I've got to be very careful not to 
screw the oven up while we're completely dependent on it. I'm nervous 
enough as it is with it next door with it's guts exposed while I put the 
kapton tape on it.

But knowing what's really going on with the PCB temps, especially in 
multiple positions, would be a very good thing.

Here's my plan:
    1) Get pairs of type K probes and 31855 breakouts exactly like the 
ones Jon is using. I want to order at least two immediately after 
hitting "send" on this msg, but if you'd like me to expand that to a 
larger number that would be cool (I mean great!)
    2) Slap this together with a Raspberry Pi on my house LAN.
    3) When it's possible, compare the performance to Gene's Fluke 289, 
add calibration tables, etc.
    4) When the risk can be controlled, experiment with this lashup 
controlling the heating and cooling of the oven. In other words, 
bypassing all of the oven guts except the heating elements and fans.

I think it's possible to do steps 1 and 2 in one week.

The other thing that's worth exploring immediately is pointing the 
dental microscope through the oven drawer's window and putting the image 
onto a computer screen while the oven is running. If we can get decent 
pictures in combination with decent "real" temperatures we can make this 
thing very good before.

-Pete


On 09/25/2014 10:08 PM, Charles McClelland wrote:
> Pete,
>
> Wow, progress or at least some significant insights.
>
> All this makes me wonder - “what is our time worth?”.   It sounds to me that, without a significant amount of time invested in one of the solutions you proposed below (thermocouples on the boards, custom profiles, or fan modifications), we will not get a consistent heat profile across the tray.  It may be that there is enough tolerance in the reflow process to make this ok but, I am concerned about the rework we may run into with a large batch.  I guess we could also adopt the “small batch” process of building and baking each board individually.
>
> If you think we can get a better oven for $265, I would be willing to split the cost with you.  I know you are very busy and I know you have put a lot of time into this process already (can’t wait to see the new setup).
>
> I appreciate all you have done to help me with the baking of these boards.  I look forward to our next session.
>
> Thanks,
>
> Chip
>
>
> On Sep 23, 2014, at 9:23 PM, Pete Soper <pete at soper.us> wrote:
>
>> Chip,
>>   I've been very busy lately and won't be able to publish some oven measurements for a while longer. But in a nutshell, the thermocouple probes are sitting in mid-air and do not properly track the temperature of the PCBs in the oven: they lead the heating and cooling such that the oven's notion of temperatures is wildy optimistic at the start and end of a cycle. For a goodly part of the middle period the PCB temperature is *exactly right* for a small amount of PCB in the oven. That is, for the point at which the PCB has heated up to the same temperature as the air surrounding the oven probes the accuracy seems nearly perfect (and then it goes to hell again when the cooling fan comes on). But for large amounts of PCB (two 4x4" boards) the situation is just hopeless: there is error all the time,  it's grossly off at the beginning and end, and *the boards do not get hot enough to reflow*. Other than that it's great. :-) So currently with a few small boards the profile is WAY longer in time than Kester calls for with the solder paste we've been using. The heating is too slow and the cooling is way to fast, with the board holding a high heat for very much longer than called for.
>>   But for a few small boards like we've been doing, when the cooling "overshoots" and dips badly that is not reality: the temperature is not dropping that fast, the temperature of the oven's thermocouples are dropping that fast. On the other hand, we did see uneven heating when we had a handful of little boards spread across the drawer. I think we can avoid that by making the boards physical relationship to the radiant heaters more uniform instead of having boards directly under them and other boards getting "late winter" illumination at a shallow angle. What I'm trying to say is we should keep the relative distance between heaters and board surface as uniform as possible. The drawer could perhaps be marked to show where the tubes sit.
>>   What I think is called for is arranging for the oven thermocouples to sit on the surface of the boards to be soldered. Gene did this at HP or someplace like that and they very accurately measured the board temps. (and this was the scheme for the measurements above: a thermocouple pressing right against a pad on a board about 2x2". I haven't pulled the oven apart far enough to establish whether there's any excess probe wiring that would support sticking the probes on the board surfaces or whether the wires would have to be extended. And with this scheme a little air movement might tend to arrange the convective cooling that seems so sensible at first blush. But it isn't clear the oven is capable of doing the right thing with a PCB that covers most of the drawer space.  What all this tells me is that there's a very wide lattitude with this kind of soldering: it's hugely more forgiving than one might think. (I agree with the notion that you could just watch a board in a toaster oven and solder it by the seat of your pants, while still believing the accuracy of this and frying pan techniques is probably off by a mile in relation to what the SMD component manufacturers want you to do.)
>>   A second best approach might be to heuristically determining a custom heating profile that intentionally tricks the oven into heating up and cooling off faster.  There would have to be multiple fake profiles depending on the total area of PCB inside. This is a total kludge, but doesn't involve messing with the guts of the oven. I'm not keen to screw the oven up and leave us wondering how to continue the sheer adventure of soldering the chips that cannot be hand soldered. :-)
>>   The mod that guy published that turns the main cooling fan on and off in concert with the heating tubes being on is, in my opinion, not worthwhile. In my opinion goosing the fan, waiting two seconds, then cycling it on and off while the heat is on will simply stress the PID control algorithm at best. I could see arranging for a small amount of *constant* air movement, but even that is very likely to exacerbate the spread between what the oven thinks the temperature is vs the real PCB temperature.
>>   Finally, the notion that the masking tape is creating the smell from this oven seems to be a red herring. The masking tape doesn't get much over 90F. When I opened the top of the oven and looked inside I became extra suspicious about this masking tape business, because 1) the tape is a couple inches from the control electronics (would they really want that to get chronically baked?) and 2) the tape looks and feels nearly new. It has nothing of the "is this tape or some kind of glue layer?" look of old, abused masking tape. There's no doubt at all that the vaporizing flux and stuff in the solder paste stinks to high heaven, though! (I got the killer ductless fume hood set up, by the way. Pictures coming when my schedule lightens up.)
>>   Also, I found this oven for $265 shipped from China, which is kind of amazing (that they could ship it for free). The larger T-962A that is much more popular is still commanding a much higher price.
>>   Gene Kahn helped me with some of these measurements and we used his amazing Fluke meter with a type K thermocouple. I cheated by video recording the oven with the meter next to it and then fast-forwarding the recording to quickly jot down elapsed times and temps at intervals. I recommend this to folks who would go stir-crazy watching the oven for a dozen eight minute runs. :-)
>>
>> -Pete
>>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.triembed.org/pipermail/triembed_triembed.org/attachments/20140926/761ef408/attachment.htm>


More information about the TriEmbed mailing list