4
\$\begingroup\$

Even today we have breadboard and strip board on which one can make "quick and dirty" circuits and also test prototypes. However, we have now moved into an era of predominantly surface mount components many of which work at very high frequencies.

If one wants to make a circuit one may first need to design some simpler prototypes of fully understand how the complex off the shelf ICs work and determine their performance, sometimes we may find that the datasheets are wrong too. In this process if we find that a connection was wrong or the chip was fried, we can rewire/replace them easily when we have a strip board or veroboard, we can always use a breakout board for surface mount devices to generate pins from them.

However, for circuits that work and high speeds, doing so is not possible anymore. How does one follow the prototype stage in this case? It seems that the process will become longer and more expensive and rather inconvenient if PCBs have to be respinned many times.

\$\endgroup\$
  • \$\begingroup\$ I want to make a device that shall process audio signals. It shall use ADCs and DACs that will need 10s of MHz from a single oscillator. The same shall be fed to the PIC32 micro controller also. The data converters need analogue supply, digital supply and reference voltage and they are surface mount as well. It does not look simple as a personal project but I still need to do it because I want to \$\endgroup\$ – quantum231 Jun 5 '15 at 17:05
  • \$\begingroup\$ If you want to evaluate a specific IC, the manufacturer will typically sell an evaluation board with all the right hookups in order for you to test the device directly. Otherwise if you want to test/breadboard a circuit, 2 layer PCB's are cheap to throw. If you know the general type of circuit you want to test, I would layout a 2 layer board (maybe with just a plane on the back) with a few parallel designs on it, proper IC footprints with grounding for any high speed stuff and maybe a few bus lines, use 0ohm jumper resistors for bridging traces. \$\endgroup\$ – crasic Jun 5 '15 at 17:26
  • \$\begingroup\$ Also, breadboards are capacitative and slow, but you can do limited breadboarding with SMD IC's using SMD-SIP adapters like this \$\endgroup\$ – crasic Jun 5 '15 at 17:36
  • \$\begingroup\$ Is there a way to find out the capacitive load that breadboard/veroboard tracks can give? \$\endgroup\$ – quantum231 Jun 5 '15 at 22:38
7
\$\begingroup\$

There are techniques that can be used for prototyping. If you're prototyping a whole system you're probably not doing it right- but to test smallish bits of analogish circuitry, it's practical. It's good to go to a PCB layout early, but not necessarily as the first step.

Here's one hacked together circuit by a fellow I happen to know John Larkin- (he's since moved to gold-plated for the boards)

enter image description here

And another (the high-speed section is kept very small)

enter image description here

This is done with shears, dental burrs etc. The ground planes under everything mean that the circuits are fairly quiet. You can also stack breakout boards for SMT chips on the ground planes (you can buy them or have your own made).

\$\endgroup\$
  • 1
    \$\begingroup\$ For those wondering, you can buy the SMD breakouts here: qrpme.com/?p=product&id=MEP. I had the same thing in my answer but failed to get the upvotes :) \$\endgroup\$ – crocboy Jun 5 '15 at 19:11
  • 1
    \$\begingroup\$ Crikey, have an upvote then. \$\endgroup\$ – Spehro Pefhany Jun 5 '15 at 19:12
  • \$\begingroup\$ In all fairness your answer has prettier pictures :) \$\endgroup\$ – crocboy Jun 5 '15 at 19:14
  • \$\begingroup\$ wow.this looks, well errr gruesome. Anyway, any breakout board will connect to a connector that has 90 degree bend is it not? A header pin soldered onto the board. This will cause impedance discontinuity and thus reflections, is that not true? \$\endgroup\$ – quantum231 Jun 5 '15 at 22:30
  • \$\begingroup\$ That's true, of course not every part will likely be high speed. You can see the discontinuity from a bend in a pcb trace easily on a TDR. \$\endgroup\$ – Spehro Pefhany Jun 5 '15 at 22:47
6
\$\begingroup\$

Well once you start designing real things you need to do it the real way :) At a professional level, and even for home projects we spin boards for proto all the time. You're not going to get clean power, clean signaling from a wired up breadboard. As an example I recently made something very similar as a side project PIC32 and some audio circuitry. The 4-layer bare boards cost me about $200 or less and I bought a stencil too. Then hotplate re-flowed the whole thing, including some very tiny QFN components.

It's much easier to make a board and blue wire your mistakes then it is to blue wire a whole board and try to decipher the rats nest of problems in front of you :)

Another approach of course is to buy a pic32 eval board, and an eval board or breakout board from say sparkfun for your ADC. Then connect those together as a prototype. For tiny simple circuits I sometimes use surfboards (check digi for those).

Otherwise many people have come before you looking for ways to make cheap pcb boards faster for in house proto. There's everything from etching it yourself, to cnc routing, and now even conductive ink printing. In a case like yours I find just shipping it out and waiting 5 days is the most convenient. Of course you can't get around the time it takes to design a schematic and layout the board. But carefully designing your schematic should be your first step anyway :)

If you find that the expense of these boards is too much for you or your project I'd suggest trying to simulate as much as possible, especially spice your analog input section. Review your design early and often, and have someone else review it as well before sending it of to be built.

\$\endgroup\$
  • \$\begingroup\$ 4 layer board hmm. If I use a breakout board that needs 10s of MHz of signals, I think that the connector will cause high speed effects and lead to poor performance. I am using off the shelf components, what do I need to do SPICE with? What is blue wire? \$\endgroup\$ – quantum231 Jun 5 '15 at 22:27
  • \$\begingroup\$ 10s of MHz is really not that fast normal connectors should be fine. For spice I use pspice but a lot of people use ltspice which is free. \$\endgroup\$ – Some Hardware Guy Jun 5 '15 at 22:29
  • \$\begingroup\$ Since I have an audio application, what is the worst than can happen if I do not isolate the analogue and digital grounds? Will the sound become illegible? \$\endgroup\$ – quantum231 Jun 5 '15 at 22:33
5
\$\begingroup\$

I'm in the same boat as you - I want to be able to quickly prototype high-speed circuits, similar to the ease and speed of a breadboard.

Unfortunately, I think PCB's are pretty much the only way to do things. They're expensive and take time to design/make/order, so this obviously isn't ideal. You just have to look at the requirements of a high-speed circuit to realize that it isn't feasible to "prototype" in the conventional sense:

  • Device footprints - only PCBs give you the ability to mount small-pitch parts properly, with a soldermask and such. Higher-speed parts lead to more difficult pagckages, like BGA
  • Ground plane requirements - high-speed circuits must have proper grounding, usually in the form of a plane. In addition, your audio circuit example is a mixed-signal device, meaning the grounds need to be isolated to work properly.
  • Trace lengths - Usually clock sources need to be as close as possible to the relevant IC. This is also true of data buses and other signal paths that require strong integrity.

There may be one option for you, one that I have used in the past. It's called Manhattan construction. This technique is commonly used by RF tinkerers to build circuits. Basically, you build everything over a piece of copper-PCB, which acts as a ground plane. You connect discrete components by directly soldering leads together. Here is a good post documenting the subject (you can also find more online). You can then buy these pads for attaching SMD IC's. I used it to build a simple DDS circuit that had a 50MHz clock and interfaced to an FPGA. It worked pretty well, but you usually have to use through-hole components, which gives you some noise.

I'm like you; always looking for a way to prototype my high-speed designs fast and cheap. But due to the nature of these circuits, it's difficult to do without a proper PCB.

\$\endgroup\$
  • \$\begingroup\$ I know that I will need analogue and digital ground as 2 separate grounds connected at one point. There are 2 questions, what happens to the ground plane if I have mix of through hole and surface mount components on a 2 sided board (2) I have 10s of Mhz Oscillator and interface signals between the MCU and the data converters, to connect the data converters with the oscillator and the MCU means having a thicker bridge between the analogue and digital ground planes underneath the board. But isn't this what we want to prevent all along by having "separate grounds"? \$\endgroup\$ – quantum231 Jun 5 '15 at 22:21
  • \$\begingroup\$ You can do manhattan like stuff with smd just fine, its only going to be quite fiddly and very messy... \$\endgroup\$ – PlasmaHH Jun 5 '15 at 22:39
  • \$\begingroup\$ Your ground plane doesn't have to be continuous - if you have through hole parts, put the ground plane around the traces on the bottom, and the top if you can. You'll end up with ground planes on top and bottom, which you should connect with vias. Not sure about your second question, I'd need to see schematics. Usually it's hard to separate ground planes on devices like data converters that handle mixed signals. \$\endgroup\$ – crocboy Jun 5 '15 at 22:49
  • \$\begingroup\$ @crocboy, the analogue and digital grounds are connected at a single point right? when signals on the top layer are routed such that they cross the boundary between the analogue and digital grounds underneath the board, the two grounds have to be connected to create a bridge, this is done so that the signals routed on the top have a continuous ground plane underneath them. My question was ultimately related to the physical width of this bridge. \$\endgroup\$ – quantum231 Jun 7 '15 at 13:11
  • \$\begingroup\$ Ok, I see now. Ground planes get really tricky especially with ADC/DAC because they require mixed signals in a single package \$\endgroup\$ – crocboy Jun 7 '15 at 13:17

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.