How do you do Product and Device selection?

[obiwanjacobi]

When starting a new project - after the initial idea has taken root - how do you select your products and devices?

Does creating the schematic and selecting the devices happen at the same time?
Or do you search for devices first, you think might be useful and create a schematic around them? Then do some tweaking.

And

How to quickly find the (type of) device you need?
How do you even know a certain type of device exists?

Reason I ask, is because this always takes soooo much time I loose interest in my own projects...

Thoughts?

[Wilksey]

It depends if it is something similar to what you have produced before of if it is a completely new type of system.

If it is similar to one done before, then you should already kind of know what you want and you will spend less time, I always research what the latest and greatest product is to see if they have any cost and performance benefit over a previous one, for example, a lower power consuming part but holds the same functionality.

If it is a new product and you have no clue, a quick google search will tell you if such a device exists and what the parameters of that device is, in most cases, online suppliers, RS, Farnell (E14), Digikey, Mouser, etc, will have parametric searches which you can utilise, though not 100% accurate all of the time, i.e. "show me items in stock only" and it shows you several "awaiting delivery", which is not really "in stock"!

I find also having a few component distributors on your search list beneficial as they can tell you and do the "leg work" for you.

I will always do component selection and a basic pin designation planner for each part if I can, i.e. uC's, modules, it helps when making the schematic to do it more efficiently in my opinion.

Sometimes it can take a long time, but the most part it is quite straight forward, for example, for my last project it took me 2 days to find all the components, make the symbols in the schematic capture program and do pin designation lists for the components, 80% were new parts, but a LDO is an LDO and a switcher is a switcher at the end of the day, most have similar or the same components, perhaps just different values, so that also makes it simpler if you spend a bit of time thinking about if there is a pin compatible version of a previous device you also don't need to worry about footprints and the like.

Footprints can be a pain in the arse, especially if they use some odd ball package, but I tend to chose parts with similar or the same footprints if I can, and if the cost is of no relevance, i.e. 1-10p difference in cost is not a massive issue if it makes the layout simpler (for me anyway).

[DimitriP]

I'm pretty sure the answers vary depending on what kind of "project" and "devices" we are talking about.
As for "knowing", nowdays searches with the right keywords reveal wonders.

Including finding out your project has already been done by someone else .
Which can be a good thing. Or bad

Maybe you can give us some examples of devices you looked for and took too long to find ...

[tszaboo]

When starting a new project - after the initial idea has taken root - how do you select your products and devices?

Does creating the schematic and selecting the devices happen at the same time?

Yes, pretty much.

Or do you search for devices first, you think might be useful and create a schematic around them? Then do some tweaking.

The part you are describing is called architecture planning. It is often ignored integral part of any project, where the most and biggest mistakes can happen.

And

How to quickly find the (type of) device you need?

Depends. I can select an opamp in 5 seconds if neccessary, or I can spend days to select the correct transistor. All depends on how many of the product is going to be made, or how hard the task is.

How do you even know a certain type of device exists?

Well, every (not busy) week I spend few hours looking at new components and components I havent used before, no matter how unrelated they seem. Well there used to be component engineers, but if there isnt one at a company, the designer has to do it, right?

Reason I ask, is because this always takes soooo much time I loose interest in my own projects...

Component selection can be really long process, but it is very important. I feel sad, when I see hobby projects with completely inappropriate selection, where a single one can decrease the overall performance or just make the entire thing fail. Rail to rail opamps are the prime example.
I think it is also a lot of fun.

[obiwanjacobi]

Projects vary from scratch builds in Audio/MIDI to connecting Arduino-style pre-fab boards. Devices are parts as well as those pre-fab boards.

The other day I needed to go from 5V to +/-12V. I knew about DC-DC converters but I couldn't find how to get the -12V. I thought of taking two +12V ones and stacking them, until someone (here on the forum) pointed me to the invertors... I looked for a good 3 hours.

Also selecting the right MCU is always "fun". How do you how much flash/ram you'll need!? I always select on IO support and take the largest one ... I just recently found out that there are hybrid controllers that have FPGA's on board - how cool is that! I always wondered about that and it already exists. (not that I will use it any time soon, I have no clue about programming FPGAs). But perhaps - one day...

I understand it is my lack of knowledge that is at the basis of this. I often do not know the correct (english) names for parts so searching the web is ... cumbersome. I spent a good 1.5h just finding those flat cable plugs (IDC) because I did not know the name. I ended up asking farnel support... 

I recently needed (and still do) a flat-flex connector to connect a small resistive touch pad (no lcd - just 4 connections). I still have no idea. I have tried to measure the 'pin' pitch but don't know the correct word for those connectors so search results are all over the place.

The list goes on... 

Anyway thanks for the answers. It confirms some of my thoughts.

[T3sl4co1l]

Top-down and bottom-up design.

Diagram what system-level components your project needs:

Inputs
Outputs
Comm channels
Converters (signals or protocols)
Power
Amplification, conversion, inversion, etc.
Controllers
Management (on/off logic, safety or protection, etc.)
Other brains, brawns

Then step down a level and define what needs to implement those blocks: interface protocols, signal levels, bus sizes and speeds, converter/controller chips, MCUs (including capability, size, speed, cost and familiarity), digital logic, and good old fashioned analog circuitry.

Along the way, you might come up with some neat ideas for the very lowest levels of implementations.  A tight inner loop (in C or asm?) for a key function.  A precision analog circuit (e.g., voltage/current limiting; calculating power; measuring phase shift; etc.).

Elaborate on those.  If you like discrete or chip level analog, expand and draw as much as you like.  Get familiar with the circuit and its low level operation.  If you aren't sure about operation, try breadboarding little pieces.  Or maybe a block at a time, or the complete system.  Make sure to test (mentally or physically) against different inputs, especially unexpectedly large or nonsensical levels (ESD, EFT or surge transients being a prime cause of destruction, and EMI/RFI being a prime source of errors or malfunction in improperly hardened systems).

At the same time, fill in with the functions from the upper levels.  Implement those in more detail if you like, or fully low level as above.

Eventually, after many passes and iterations, you'll have had a lot of back-and-forth, laterally between the responsibilities of each block, and up and down between the high-level and low-level design.  Probably you'll wrap up a lot of discrete functions into handy chips you've found for the purpose, or decided to roll them into the MCU or something.  You'll have massaged the total design into something pretty reasonable, certainly not perfect (there can be no such thing!), but good enough to go on with.  At this time, you'll have boards ready to order, and you can move on to the prototyping stage.

Tim

[obiwanjacobi]

I think I already do some of that. I think in building blocks that I later fill in. I also build and test like that - per block as much as possible. Learned the hard way - you only have to make so many (large) projects in one go that do not work... 

I am currently working on a MIDI hub/switch/mapper and have already prototyped (on perf board - not bread board) the hardware for the MIDI interface and tested that with my Arduino and in that process written the MIDI code I wanted (Arduino Libs are 'suboptimal'  ). Now I am focusing on the MCU and what the user interface will be like. But this was a project where I knew enough in advance to have a reasonable understanding of what I had to do - also a pretty simple project.

To be honest, my projects never go beyond the perf-board phase - I never make real PCBs for them... 

Thanks for the structured explanation.