The Propeller 1-2-3 FPGA PCB is now on the Pick-n-Place SMT line. This four-layer board uses 75+ feeder locations and places components on both sides of the PCB.
Will keep everybody posted on the initial results.
No, we've placed these before on prior FPGA boards. The solder stencil process and reflow profile are the most complex aspects of dealing with BGAs. They're potentially troublesome like QFNs, but we'll succeed.
Actually, that board is built by Terasic in Taiwan. I'm having trouble inserting images into this message (we're meeting at lunch about our forum repair/upgrade), but do a Google for "Parallax Cyclone FPGA" or "Parallax FPGA Altera QuickStart" and you'll find our prior FPGA assemblies.
You see, enough time goes by and we forget how complicated it is to build these boards, and we find ourselves right back at the P&P doing it again
Actually, that board is built by Terasic in Taiwan. I'm having trouble inserting images into this message (we're meeting at lunch about our forum repair/upgrade), but do a Google for "Parallax Cyclone FPGA" or "Parallax FPGA Altera QuickStart" and you'll find our prior FPGA assemblies.
You see, enough time goes by and we forget how complicated it is to build these boards, and we find ourselves right back at the P&P doing it again
Ken Gracey
So I'm guessin' it was the one I found on the 2009 CD:
The Propeller 1-2-3 FPGA PCB is now on the Pick-n-Place SMT line. This four-layer board uses 75+ feeder locations and places components on both sides of the PCB.
Will keep everybody posted on the initial results.
Ken Gracey
Thanks Ken
(just inserted photos inline for easier view)
What is the setup time for a board on that machine? When you say troublesome, do you mean the QFN's require rework?
Hey T Chap!
The setup time on the P&P for this particular board is about two full days of loading parts on feeders. This includes time required to cycle count the parts and check inventory prior. The programming of the P&P for this board is at least three days, so we're into this five days before we actually press the green button.
QFNs require precise placement on the PCB and require more attention to reflow oven temperature profiles.
The programming of the P&P for this board is at least three days, so we're into this five days before we actually press the green button.
That sounds a lot of time ?
Better PCB design packages should be able to create an almost machine ready placement file, referenced to any point you choose.
The only unknown from the PCB design CAD info, is the relative Decal to Tape rotation fix-up, usually one of 4 90' choices.
That can be managed with a Part attribute, so as you learn the Decal to Tape rotation for each SMD part, it can not need manual entry next time.
(likewise for Z info, if that is needed too)
Many pcb houses don't use flat pcb stock so, unless you preflatten your pcbs, ,,, .... qfn is teh suxor.
You can see this if you take a stack of pcbs and restack them, but flip every other pcb.
I found qfns were problems with too little solder paste, those legs suck up lots of solder.
Rarely did I have bridges, but often I had a floating leg due to a lack of paste combined with pcb flex.
I can spot a solder bridge with my bare eyes on this pitch.
But if it had a loose leg there was virtually no way to notice, save for the missing ski slope solder.
Even ones that had no slope were welded, so it wasn't a good indicator, but the entire pcb would fail a JEDEC/IPC inspection due to a lack of paste.
Coarse teh bozz didn't see it that way. (apparently solder paste is expensive)
Parallax, you've come full circle.
This pcb is dated 2003.
@Frank: Start mooching little bits of money now. Then the "big ask" is much smaller. Just tuck it away...
I've already started skipping coffees. Got $50 already. By the time Parallax gets this done, I'll be a big chunk of the way there.
Since I got the DE2 on academic, I'm stuck with it. What I'll probably do is find a student somewhere, who I can trust not to sell it and keep the agreement with the nice vendor, who let me in cheap.
Comments
Is this the first BGA that you have done?
No, we've placed these before on prior FPGA boards. The solder stencil process and reflow profile are the most complex aspects of dealing with BGAs. They're potentially troublesome like QFNs, but we'll succeed.
Ken Gracey
http://www.parallax.com/product/60050
Much success!
Actually, that board is built by Terasic in Taiwan. I'm having trouble inserting images into this message (we're meeting at lunch about our forum repair/upgrade), but do a Google for "Parallax Cyclone FPGA" or "Parallax FPGA Altera QuickStart" and you'll find our prior FPGA assemblies.
You see, enough time goes by and we forget how complicated it is to build these boards, and we find ourselves right back at the P&P doing it again
Ken Gracey
So I'm guessin' it was the one I found on the 2009 CD:
I like the name, hopefully P3 is an easier process as a result of having this
First choice I see is the A7, but does the PCB design also support fitting of smaller FPGAs ?
Looking forward to their availability.
Thanks Ken
(just inserted photos inline for easier view)
With our newly-broken forums I'm unable to figure out the steps to do this. Can you tell me how to do this?
Thanks,
Ken Gracey
Testing:
-Phil
Hey T Chap!
The setup time on the P&P for this particular board is about two full days of loading parts on feeders. This includes time required to cycle count the parts and check inventory prior. The programming of the P&P for this board is at least three days, so we're into this five days before we actually press the green button.
QFNs require precise placement on the PCB and require more attention to reflow oven temperature profiles.
Ken Gracey
That sounds a lot of time ?
Better PCB design packages should be able to create an almost machine ready placement file, referenced to any point you choose.
The only unknown from the PCB design CAD info, is the relative Decal to Tape rotation fix-up, usually one of 4 90' choices.
That can be managed with a Part attribute, so as you learn the Decal to Tape rotation for each SMD part, it can not need manual entry next time.
(likewise for Z info, if that is needed too)
You can see this if you take a stack of pcbs and restack them, but flip every other pcb.
I found qfns were problems with too little solder paste, those legs suck up lots of solder.
Rarely did I have bridges, but often I had a floating leg due to a lack of paste combined with pcb flex.
I can spot a solder bridge with my bare eyes on this pitch.
But if it had a loose leg there was virtually no way to notice, save for the missing ski slope solder.
Even ones that had no slope were welded, so it wasn't a good indicator, but the entire pcb would fail a JEDEC/IPC inspection due to a lack of paste.
Coarse teh bozz didn't see it that way. (apparently solder paste is expensive)
Parallax, you've come full circle.
This pcb is dated 2003.
I've already started skipping coffees. Got $50 already. By the time Parallax gets this done, I'll be a big chunk of the way there.
Since I got the DE2 on academic, I'm stuck with it. What I'll probably do is find a student somewhere, who I can trust not to sell it and keep the agreement with the nice vendor, who let me in cheap.
Maybe he got a one-off from the January run? They looked like they where going to run a few through Pick-Place and reflow.
Hehe, maybe. I don't so, though. He just did a sizing exercise, as David suggests there's no requirement to even load a chip up for that.