4
\$\begingroup\$

Is it necessary to go straight to ordering a professionally made PCBs when one wants to test a surface mount component circuit involving high rates of change of voltage (dv/dt) and current (di/dt)? That is, this question refers prototyping SMD PCBs when one must minimize interconnect length to avoid parasitic inductance. Does this requirement make readily available breakout boards useless? (http://www.futurlec.com/SMD_Adapters.shtml). I've etched my own PCBs, but, I have found that soldering fine pitch parts without a soldermask difficult. Have people found this to be a viable option even with leadless packages with 0.5mm pad spacing and a big thermal pad such as the QFN?

For a particular example I refer you to this question: What is causing large oscillations in my DC/DC boost converter? Is this ground bounce or some other effect? This was my first attempt at SMD circuit, a DC/DC converter, and in dealing with parasitics. It was based around a small QFN and I couldn't think of any way to prototype the circuit other than to go straight to getting the board done professionally. A very tight layout was indeed critical to getting the board to work. I found that my layout wasn't sufficiently tight and I will need to make another board revision.

Did I have a better option to getting the board done professionally? I am asking in case there is another option I don't know of.

Does anyone try to solder 30 gauge wire wrap wire onto the small pads and wire the chip up somehow?

\$\endgroup\$
  • \$\begingroup\$ Theres a net.rumor about someone dead-bugging a BGA FPGA (ie, upside down, soldering discrete wires to the balls) but that was probably for low-frequency use given the wire inductance. Having found discrete wires a pain for anything beyond 8 pins due to the domino-effect of mistakes, I prefer toner transfer PCBs and careful soldering / use of braid and inspection. For the QFN power pads, probably a large through hole underneath, maybe manually backfilled with thick wire or some kind of strap since you won't have hole plating. \$\endgroup\$ – Chris Stratton Dec 8 '11 at 21:41
  • \$\begingroup\$ @ChrisStratton, I think you have it reversed. Open Circuits talks about using it for prototyping RF. I also seem to remember Dave Jones talking about using it for a GSM device with a headphone jack. opencircuits.com/Dead_bug_style \$\endgroup\$ – Joshua Dec 9 '11 at 1:00
  • 1
    \$\begingroup\$ @Joshua - dead bug is great for RF with physically small IC's, but treating a BGA FPGA (a much larger die and package than most RF parts, with high contact density likely necessitating long, fine wires) that way would likely lead to too high a lead impedance for reliable operation with fast signal transitions, particularly on the supply pins. So it's probably confined to using the FPGA well below its maximum clock, with slower I/O than it would be capable of on a well designed multilayer board. \$\endgroup\$ – Chris Stratton Dec 9 '11 at 4:26
  • \$\begingroup\$ @ChrisStratton, thanks for clearing that up. That makes a lot of sense. In general, higher density parts require thinner wires, which have higher resistance, so the parasitic capacitances and inductances high a higher impact on high frequency components. Never had to work directly with RF, but you never know when this information will become useful. \$\endgroup\$ – Joshua Dec 9 '11 at 18:16
8
\$\begingroup\$

What's your work environment? Mentioning toner transfer makes me think you're a hobbyist (which is fine), but as a hobbyist you're doing this because it's fun. Your time takes on a different value, and your budget outlook is quite different.

As a professional, I build circuit boards because it makes money for my employer. I'm paid fairly well, and it's not economically sensible for me to mess around with toner transfer and trying to solder to that board. I take my time and try to do it right the first time, send the boards out for manufacture, and move on to other projects. When the boards get back, I send them through the reflow oven or have a tech solder them up (the former is easier with soldermask, the latter is easier with silkscreen and soldermask) and test. If it works, great! If it doesn't, I revise the board accordingly and try again. Usually, the board works the first time, but if not, I revise it and send it out again.

Making a toner transfer board (or, at my workplace, a board cut out with a PCB router) is valuable when there's a major time crunch and you'd rather spend extra time to make sure that your prototype for the prototype works, rather than counting on the real prototype working the first time. I'm not going to sell or mass-manufacture routed boards, and they're laid out fundamentally differently than professionally made boards:

  • Vias are free on professional boards, and difficult, large, and time-consuming on self-made boards
  • Soldering is much more difficult. Keepaways, plane spacing, and thermals all behave very differently without soldermask. I'll work to make soldering easy on a self-made board, but lay out a professional board differently.
  • Trace/space is smaller on a professional board. This could lead to major layout differences on some boards. Especially with high-frequency signals, moving things closer together can change impedances and cause problems.
  • Some parts simply can't be soldered effectively on toner-transfer boards. 144-pin QFPs, QFN and BGA parts, and other tight layouts are far, far easier with soldermask.

In most cases, it's a better investment to send out for a few samples of the final product and wait for shipping than to do a toner transfer board as a prototype. If you enjoy doing toner transfer stuff, enjoy getting better at soldering, and your time isn't a part of your budget (hint: It isn't, even if you're a hobbyist - you have limited time too), then toner transfer makes some sense. If not, just get the real thing.

\$\endgroup\$
  • 3
    \$\begingroup\$ I think the real role for toner transfer is when you need the result (or something approximating it) today in order to be able to continue the project. \$\endgroup\$ – Chris Stratton Dec 8 '11 at 22:35
  • 3
    \$\begingroup\$ @Chris: I hear that sometimes, but needing someting today is almost always a result of bad planning. Learn to plan better and you don't have to deal with compromises to get quickness. \$\endgroup\$ – Olin Lathrop Dec 8 '11 at 22:45
  • 4
    \$\begingroup\$ @OlinLathrop - such planning problems often come from last minute mandates or lack of communication from outside the engineering department or at least outside the project team, even when the engineers have asked specific questions in advance to head off such problems. And then there's trying to get authorization for the premium on a quicker than usual board turnaround. More schedule slip comes from institutionally bad interface patterns between people than individual people putting hours into something specifically effective that they weren't supposed to need to do. \$\endgroup\$ – Chris Stratton Dec 8 '11 at 22:55
  • 1
    \$\begingroup\$ @Kevin Vermeer: Well, I have build electronic devices and firmware during my professional career for my employer. But, I am primarily a mechanical engineer and electronics hobbyist with delusional hopes of doing more electronics design on a professional basis. ;) I like your response because you addressed it from both a professional and hobbyist's viewpoint. \$\endgroup\$ – Dave.Mech.Eng Dec 10 '11 at 21:06
  • \$\begingroup\$ Hand soldering QFN's is a challenge and likely to be unreliable even on a dedicated adapter pcb. You will need a stencil do it properly (you can put the paste with a rubber pad). And this is another extra cost. For best reliability a real PCB and a professional smd mounting service with a pick and place machine may be necessary. These days, some parts got so tiny, it's not for humans anymore. \$\endgroup\$ – Fredled Feb 6 at 23:28
7
\$\begingroup\$

I think that in the general case the answer is you have to use a real PCB. Something where parasitic inductance and capacitance is critical won't behave the same on a breadboard. Fortunately, getting real PCBs made is much more accessible, faster, and cheaper than it used to be. For example, I use Gold Phoenix for most prototypes. For $120 you get how ever many boards fit in 100 square inches, two layers, plated thru holes, soldermask on both sides and silkscreen on one side, and electrical testing. That's cheaper then futzing around with lesser alternatives. I usually receive boards 1 1/2 weeks after sending the gerber files. That's about right for putting the BOM together and ordering whatever parts you don't have in stock. While you're waiting for the boards and parts to show up, you go on to another project.

\$\endgroup\$
  • \$\begingroup\$ Thanks! Great and helpful answer. Kevin's was also a good answer. I'm going to credit Kevin's response with the "right" answer for this one since he addressed a wider audience, i.e. both professionals and hobbyists. However, I encourage everyone to read both answers. There should be a way on the stack exchange to choose multiple correct answers! There are cases, such as this, where multiple distinct points of view are valid. \$\endgroup\$ – Dave.Mech.Eng Dec 10 '11 at 21:14

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.