I (Tom Gardner) have been creating circuits for mumble years, using wirewrap, IDC and standard PCB technology with plate-thru hole (PTH) components. I've recently been forced to use surface mount devices (SMD), and was concerned that it would be too difficult for an amateur. I was wrong. On the off-chance it inspires others, here's what I've used, what worked and what didn't.

The tl;dr version is DesignSparkPCB, DirtyPCBs, OSH Stencils, 179C SnPb paste, magnifying visor, *8 lenses, reflow sand in skillet, soldering iron. My current design's principal parameters are:

  • 0805 and 0603 resistors and capacitors
  • 8 pin SOIC integrated circuits, 1.3mm pitch with 0.6mm pads and 0.7mm gap
  • fine pitch 8 pin integrated circuits, 0.65mm pitch with 0.3mm pads and 0.35mm gap
  • fine pitch 100-way mezzanine connector, two rows of 50 pads 0.8mm pitch with 0.5mm pads an 0.3mm gap

My experiences and techniques are an alternative those documented in Info On Custom PCB Sources and Solder Reflow Oven.

Thanks are due to all those that have helped me, including but not limited to Ian, Arthur, Russ, Nathan, Simon.

PCB Design

I've used the free-as-in-beer DesignSparkPCB commercial product from RS. I didn't choose Eagle because of its irritating limitations (size, pins etc). I didn't try KiCAD, and have no opinion about it.

DesignSparkPCB provides:

  • a large library of components via RS. In reality the RS library was missing several components I needed, but I've found it very easy to create components (both schematic and PCB). Hence the RS library hasn't been as beneficial as I originally thought
  • decent schematic and PCB layout editors, with forward/backward changes and reasonable design rule checking

I originally ran DesignSparkPCB in WinXP, but now run it under wine-1.4 in xubuntu 12.04. Less than once a day it crashes, which is acceptable. N.B. I haven't tried the 3D modelling, because that's uninteresting.

To check the gerbers, I use

I made the component pads longer than specified in the data sheets in order that hand soldering might be easier, but:

  • for the non-leaded components such as 0805 resistors it might mean that too much solder paste sets up thermal stresses; in future I will be less pessimistic
  • for leaded components, particularly the FMC connector, the slot in the paste stencil was larger than necessary and may have contributed to the stencilling problem mentioned below. On the other hand it helped with the Schmartboard technique. In future I will make them shorter, but still longer than the minimum

I spaced the components by 1mm on all sides, so that I could get an iron to all of the pads. That was unnecessarily pessimistic for 0805, 0603 and IC components - I would use 0.5mm in future, provided that they aren't in the “shadow” of tall components.

Summary: all the design tools work well.

PCB Manufacture

I've used the traditional laser toner mechanism for:

  • through hole components on single and double sided boards, with acceptable results
  • gash test boards when experimenting with SMD components; the results are surprisingly good and would be acceptable in an emergency

For double sided boards, I've used DirtyPCBs principally for cost, but also to have a benchmark against which I can evaluate a “real” PCB manufacturer. The principal features are:

  • 10off 5cm*5cm 2 layer is $14 (i.e £8-9) including sea shipping
  • (10 off 10cm*10cm 4 layer is $50)
  • FEDEX/UPS costs an extra $17 (and I suspect you might get stung for import duties and handling charges)
  • order placed to PCBs received: 21 days

all of which is remarkably good.

Comments on the DirtyPCBs board:

  • the silkscreen is sufficiently legible but is not a continuous line; this is unsurprising since I pushed their process to the limit (0.15mm width)
  • the solder mask can be offset by 0.2mm (board too small for scale problems to be evident); whether this is a problem depends on the PCB's feature size. Solder mask on top of a pad can be easily removed by scraping with a jeweller's screwdriver
  • one board had an unacceptable 1mm diameter blob of unetched copper, bridging two connections. If it had occurred on the inner layer of a 4 layer board, it would have been invisible and uncorrectable.
  • another board had several spurious 1mm diameter circular holes in the solder resist. In the wrong places, this would probably allow solder bridges

There is no way that board should have passed a “100% etest”. I believe the etest only checks for continuity on the basis that plated thru holes are the most likely defects - but it does mean a completely unetched board would pass! So, black mark for quality against DirtyPCBs, but was I unlucky or are other manufacturers actually better?

A blog about one person's opinion of Seeed vs ITead Studio vs OSH Park.

PCB Assembly

I didn't want to be constrained by the time it takes to apply solder paste, position components, and reflow in the HackSpace oven, so I looked for ways I could do the whole process at home.

I used lead-tin solder paste melting point 179C, so as to minimise the chance of scorching components, 25g, £20.

Placing Solder Paste

Using a polyimide paste stencil from OSH stencils 5cm*5cm $14. Order placed to order received: 14 days (including their noticing I'd made a cockup and flipping the order at no extra cost).

  • works well for discrete components and “large” ics e.g. SOIC outline with 2.0*0.6mm pads separated by 0.7mm
  • just about works for “small” ics e.g. TSSOP outline with 1.5*0.3mm pads separated by 0.3mm
  • fails with the large FMC mezzanine connector with 100 3.0*0.5mm pads separated by 0.3mm; the film lifts and solder paste pools underneath some of the pads (other pads are perfect). Maybe it would work with more practice on my part (and I have an idea about how to improve my technique). But in the end I just wanted the damn board made; my remaining life is too short as it is!

N.B. I cannot fault the OSH Stencil - the problems are due to the PCB/component design and my poor technique. Their service is good as well. I would use them again.

Extruding paste through a 0.5mm needle just about works, but there's difficulty

  • controlling the speed
  • “chopping” off the right amount
  • needle getting blocked too often (I used 30AWG wirewrap wire to clear the needle)

I'm sure it would work well with an automated machine, but I haven't got one.

Using a jeweller's screwdriver works surprisingly well. It is the best way for the FMC mezzanine connector: just puddle the solder along the edge of the contacts (i.e. not under) and don't worry about bridges across the solder resist.

Also see the wiki page on smt_stencils

Placing components

I tried several techniques:

Overall I find the tweezers most useful.

Vision

Placing components:

  • +2D to +3D £5 reading glasses from Sainsburys, optionally together and optionally at the same time as
  • a £10/£20 magnifying visor, widely available under a number of names

Inspecting joints needs more magnification:

  • I sometimes use a *8 surface contact magnifier which also includes mm/thou rulers (0.2mm, 10thou resolution), or
  • a handheld *8 lens
  • I didn't bother with a microscope (neither USB nor analogue), since I'm unconvinced they would offer me any advantages

Soldering

For hand soldering, a temperature-controlled iron with a 0.5mm tip. I used the cheapest iron I could find, so I don't know the temperature, but it wasn't set at maximum. It is possible to solder all components by hand, which enables a board's function to be tested bit-by-bit, but it would be tedious for a second board.

Be aware that touching ceramic components with a soldering iron can cause cracking through thermal stresses, and MLCCs have “exotic” dielectrics that can change at high temperature.

I also used a technique inspired by SparkFun's reflow skillet method. Through hole components are left until later. I initially intended to use my slow cooker, but didn't reach the necessary temperature quickly enough, so instead:

  • from Wilkinson, a £5 saucepan
    • stainless steel not non-stick Teflon, since PTFE can be rather unpleasant if it gets too hot
    • with a glass lid so I can observe the solder paste melting. The idea is the lid probably keeps the air above a PCB warmer, thus potentially heating up components faster and minimising strain due to temperature differences
    • 2mm of sand in the bottom, to diffuse the heat and act as a buffer
  • method, which needs calibration using sacrificial trial boards, just as you would with any other method:
    • large SMD components (e.g. tantalum capacitors) that might move too readily can be restrained by roughly tacking them in place using a soldering iron with a 0.5mm tip
    • loosely hook 30AWG wirewrap wire into each corner mounting hole, and use the wires to lower the PCB into the saucepan. The wires are sufficiently flexible that pressing down the lid deforms the wires thus keeping a reasonable seal
    • warm pan (inc sand) on lowest gas setting, and lid on
    • every 30s remove the lid and use non-contact thermometer to measure sand's surface temperature
    • when sand surface rises to >140C, put the PCB on the sand, replace lid, continue to monitor temperature
    • if necessary increase gas for 30s to rapidly increase temperature to just below the solder's melting point
    • watch solder paste through lid, continue to monitor temperature
    • increasing the gas a notch causes the temperature to stabilise at around 220-230C
    • after appropriate time, remove PCB; the sand will stay warm for a long time - probably too long a time
    • through-hole components can now be hand-soldered

Soldering the large FMC connector works well with a technique inspired by SchmartBoards:

  • hold the connector in place with polyimide tape, which withstands the temperature
  • using a jewellers screwdriver, add solder paste around outside (not under) and across the solder resist
  • tack one corner pin in place with a soldering iron, then mechanically drag the opposite corner into the right place and tack it down
  • run soldering iron along solder pads and between two pins to drag paste/solder along solder pad against connector leads. Ensure both leads are heated both times so solder reflows under leads. Having extra-long solder pads helps
  • there is a visible flux residue, not apparent with the saucepan reflow technique

Soldering very fine-pitch ICs is similar except that

  • the paste is placed first, then the IC positioned
  • the IC is held in place by pressing down with a jeweller's screwdriver while one corner is tacked
  • the IC is dragged and the opposite corner tacked; obviously “dragging” is only possible for ICs with flexible leads

Rework

I remove excess solder using a copper braid that I impregnate with liquid flux just before use. That seems to work, but one pad/track did begin to lift after I heated it up too much.

To remove a single component, surround other components with polyimide/kapton tape, and use a hot air gun. If positioned correctly, the tape will deflect the hot air away from surrounding components.

It is possible to reflow a second time, apparently without problems.

Here's an extraordinarily comprehensive professional guide to circuit board repair and rework.

Summary

Currently it is a two phase process:

  • place solder paste on large pads using a stencil, add components, tack heavy components in place with soldering iron, finally reflow in saucepan
  • use screwdriver to add solder paste for fine-pitch components, then soldering iron

which is satisfactory but inelegant.

Next time I'll try using the stencil and manually adding the paste for the fine components, then reflowing everything in the saucepan.

  • resources/creating_pcbs_with_surface_mount_devices
  • Last modified: 11 months ago
  • by tggzzz