A Project Management Question

[German_EE]

Suppose that you are building a small project that contains a single PCB but the circuit can be easily divided into five separate sections. Do you:

A Build the whole thing in one go and debug at first turn on if required

B Build it section by section, checking that a section works before moving on to the next one

Please explain your choice

[Chriss]

I would tak A.

Sent from my SM-J500F using Tapatalk

[xani]

If circuit is neatly divided into sections, why not, it saves a lot of time if something is wrong. Even on small circuits I at least mount power supply part first and test that, having a bad part or sth burn something important would be a shame

[capt bullshot]

Depends on my confidence with the project.

[Chriss]

Actually it depends from:
- how complex is the circuit
- how experienced is the person who build it
- how precise is the pcb made...

Sent from my SM-J500F using Tapatalk

[MosherIV]

Both, best of both world.
Design each sub circuit so that they can be joined with 0\$\Omega\$ link.
Same with powering each sub circuit.

Now, you build the whole board but leave the links out.

When debugging, add the links as you need them.

[hendorog]

B. Because I generally don't know everything at the start, and usually the circuit isn't the complete system.

By doing it in small sections I can use what I have learned about the entire system as I progress.
If I do it in one hit then I have locked in the design based on my knowledge at the start.

Also means firmware and software can be written for each part as they are built, which makes that faster - instead of waiting for all of the electronics to be working first.

If I wanted it all on a single board as a final product, then I'd start that process once I was happy there were no (or hopefully not many) surprises left.

[SiliconWizard]

It... depends.

If the different sections use expensive components, and there are dependencies between those sections, going for B seems more adequate. Otherwise you could waste a lof of cash in case one section doesn't work as expected.

OTOH, if the manufacturing time of the PCB is very long (can be if you're using very specific technologies), and/or the assembly (if it can't be hand-assembled) has a very long yield time as well, A may be the better option.

To best decide, do a compared cost and time analysis including the dependencies between sections if it applies, and a rough estimated probability of each section to be a failure (depending on complexity) and then decide based on this. This is actually basic project management stuff.

Now if there is no real dependency between sections or if the parts/PCB are not going to be expensive, or not going to take significant time to manufacture, there isn't much need to bother. In that case, do whatever you're most comfortable with.

A typical "6-sigma" approach would likely tell you to do A, but in case some parts can be excessively expensive, to mitigate the risk, possibly implement more than just one version of each section (for the sections that you know are more likely to cause issues), so in case one of the sections has bugs/problems, you have a fallback alternative and limit the risk of getting an unusable board. You get the idea. The main point with this approach would be to gain time, which can be precious. The extreme case would be when designing a new IC. Having wafers manufactured can be so expensive and so long that taking this approach is often the only reasonable option.

[German_EE]

An example of why this may be important. You're building an oscillator and a buffer and when it's built you find that there's no output, an investigation shows a short to ground on the connection between the oscillator and the buffer. Is the short at the oscillator end or the buffer end? If you had built the oscillator first  you could have debugged this stage then moved on. Before building either stage it also might have been a good idea to build and test the supply regulators.

However, doing things in stages takes time, and to some time is money.

[ejeffrey]

To me that isn't a great example.  If the short is a design flaw once you identify it you just locate it in the design.  On the other hand if it is an assembly problem like a solder bridge, building submodules won't necessarily reveal it since it is likely somewhat random and also the chances of it happening probably depend on the entire board.

For a project described as "small" I would almost always do it in one go an make sure to have good test points and possibly 0 ohm links as MosherIV suggests.

[SiliconWizard]

An example of why this may be important. You're building an oscillator and a buffer and when it's built you find that there's no output, an investigation shows a short to ground on the connection between the oscillator and the buffer. Is the short at the oscillator end or the buffer end? If you had built the oscillator first  you could have debugged this stage then moved on. Before building either stage it also might have been a good idea to build and test the supply regulators.

As MosherIV hinted, and in the same vein as I also suggested, in that case separate the sections so they can be tested independently. Add test points liberally as well on sections that are critical.

In your example, instead of directly connecting the oscillator and buffer, do not. Make it possible to connect them externally. But you will also be able to test each section separately. Nothing lost.
Also do not connect power supplies directly if you have doubts. Make it optional. If there's a problem with the supplies, you can always power the rest with external supplies and move on.

So in your example, A is a perfect approach unless there are other considerations such as very expensive parts for sections you're still not sure of. Again, do some preliminary analysis.

Again, building the sections in stages would be justified if taking the above approach is costly. If not, working in stages will not only be longer, it will also cost more (having to design/manufacture several PCBs instead of one, etc.)

[Chriss]

Again. ... it depends... :-)
- no enough info about the project
- no info about the skill of the op
- no circuit
...

What's the point of the question of the op?

Sent from my SM-J500F using Tapatalk

[BrandonH1987]

Hmm .... Interesting questions, can I get more feedback about certifications?

[nali]

Depends, but I'm usually also inclined to go the 0R link if space and signal integrity aren't an issue. An added bonus is they often make convenient test points and/or a takeoff point for bodge wires if needed.

[MK14]

WARNING: THREAD IS A 1.5 YEAR NECRO