I promised Mitch a writeup on this, so here it is.
OSHPark - 2-4layer batch pooling service, charges by the square inch bounding box of gerbers.
Hackvana - 2-4layer direct ordering service, charges by setup cost + quantity, specs, and area.
Let me start by saying I have ordered many boards from Laen through his oshpark service, and have that service of his to thank for enabling my projects. However there are issues that I and others have experienced that might save somebody else some trouble.
4 layer, 5mil track/space, 10mil minimum drill.
Screenshot of the 0.8mm BGA which is why I need 10mil drill.
![](http://retroactive.be/lsu3_001_altium.png)
Drill export report
Tool Hole Size Hole Type Hole Count Plated Tool Travel
---------------------------------------------------------------------------
T1 10mil (0.254mm) Round 869 57.37 Inch (1457.13 mm)
T2 12.99mil (0.32995mm) Round 4 0.14 Inch (3.60 mm)
T3 13mil (0.3302mm) Round 16 0.63 Inch (16.00 mm)
T4 27.56mil (0.70002mm) Round 5 0.33 Inch (8.50 mm)
T5 35.43mil (0.89992mm) Round 10 0.90 Inch (22.86 mm)
T6 36.22mil (0.91999mm) Round 4 0.28 Inch (7.00 mm)
T7 76mil (1.9304mm) Round 3 0.34 Inch (8.60 mm)
T8 86.61mil (2.19989mm) Round 2 0.39 Inch (9.90 mm)
T9 90.55mil (2.29997mm) Round 2 0.47 Inch (12.04 mm)
T10 100.39mil (2.54991mm) Round 1 0.00 Inch (0.00 mm)
T11 125.2mil (3.18008mm) Round 4 7.47 Inch (189.81 mm)
---------------------------------------------------------------------------
Totals 920 68.32 Inch (1735.45 mm)
Sep 29 - Run 1 ordered. $75
Oct 11 - Run 1 shipped.
![](http://retroactive.be/lsu3_001_run1.jpg)
![](http://retroactive.be/lsu3_001_run1_zoom.jpg)
Minor plating failures, maybe salvageable if all pads had oxidation removed immediately before reflow. However all 10mil hits were upsized to 13/14mil by the fab, resulting in catastrophic breakout. Inner 16 power vias are 13/21mil and were not negatively affected.
Oct 15 - Run 2 ordered. Complementary re-run after talking to Laen (Thanks)
Oct 25 - Run 2 shipped.
![](http://retroactive.be/lsu3_001_run2.jpg)
![](http://retroactive.be/lsu3_001_run2_zoom.jpg)
Plating failures, severe overetch and again overdrill.
My project needed to progress and it looked unlikely that I would be able to get it done at all.
Enter hackvana. Mitch is an Australian fellow that travels to Shenzhen and works directly with a board shop there. He does not pool batch orders like Oshpark. This means the minimum order is 5 pcs and you pay all the setup costs. At this point I was more interested in getting a working PCB.
Quote from Mitch was $123 plus shipping. Turn time was about 9 days.
![](http://retroactive.be/lsu3_001_runh.jpg)
![](http://retroactive.be/lsu3_001_runh_zoom.jpg)
Great plating, proper 10mil drills, adequate silk and soldermask. The only beef I had is that the fab grew the copper layers by about 2mil to prevent drill breakout. This is a pretty common chinese fab trick. Mitch says this can be bypassed if needed (obv. for stuff like controlled impedance diff pairs). It does not create any shorts but does mess with track impedance. E-testing is claimed but I do not see any probe marks or evidence of testing.
In any case I assembled the PCB and it brought up OK.
![](http://retroactive.be/lsu3_001_assemble.jpg)
Back to OSHPark -----
Nov 7 - Run 3 Oshpark started (Again complementary after talking to Dan and describing the issues, he said it would be addressed)
Dec 2 - Run 3 shipped.
![](http://retroactive.be/lsu3_001_run3_zoom.jpg)
Plating and drill alignment is finally addressed. Drills are still upsized and causing severe breakout.
OSHPark uses about 3-4 different fabs for the 4layer batch. Whichever one you get is up to the phase of the moon. Each one is identifiable due to characteristics like LPI silk, copper overetch, low DPI silk, etc...
The moral of the story: Don't use 10 mil drills with them, and prepare for delays. Plating fails have happened before, a friend wrote up a failure analysis with SEM imagery and sent to the fab. That fab is still experience plating failures and still being used by the service.
One such failure courtesy azonenberg
![](http://retroactive.be/azon_oshfail.jpg)
SEM comparison of plating (good is first, bad is second)
![](http://retroactive.be/azon_oshfail2.jpg)
You can read his full failure analysis pdf
here Laen and Dan of OSHPark are both great guys and a pleasure to talk to. I will still order 4layer pcbs in the future, but not important ones.
Mitch of Hackvana can also be talked to directly via email or IRC. I will be ordering from him in the future as well.
I hope this was objective and gave you an idea of what to expect from these two services.
Links -
OSHPark
http://oshpark.com/Hackvana
http://www.hackvana.com https://docs.google.com/document/d/1p6FH25ltGpzJQ5_8fbflDukqEKghiEcpuhJpngth2Is/edit