Embedded system, from prototype to final device

[Mustang1945]

Hi, i am from Ecuador, South America. I always had been interested in embedded systems. I make a few interesting proyects in the university, but always as prototypes, you know, the protoboard with a lot of wires or the simple PCB with a lot of wires.
But now i am interested in go to next level, the next step, because sometimes your idea is good and somebody wants to buy it. But no one would buy a prototype that after 3 months of working (maybe less), it fails. Here in my country its trending to buy different modules then you integrate all them with a microcontroller. For example you buy a rain sensor, a soil moisture sensor and a PIC, and make something. If something gets broken you just replace that module.
Do you think is a good idea to build an embedded systems like that ?
Or would it be better to integrate all of them in the same PCB ?'
And the main question, what are the recommended steps to go further
beyon a prototype and make a good electronic product?

Please also recommend me a good book to learn C for beginners. We learn just the basics in the University and i have been learning by myself but i think that a book would be the best option.

Thanks.

[temtronic]

One thing I've done in the past 2 decades is to finally just use 1 or 2 PICs. Couple of years ago I settled on using the PIC18F46K22 as 'my' PIC. It has lots of memory, 2 hardware UARTS, lots of other peripherals, high speed (64 MHx, internal clock) and comes in a 40 pin DIP ! It also will run at either 5 v or 3v, a BIG bonus considering all the 'neat' devices like SD cards, GPS, etc. are usually 3 volts.
While it is way more PIC than most projects require, I've built my own library of functions that I know work. I see a lot of threads here about 'new' PICs and while they look great, the compiler may not have them, some bits in a register may be wrong or other 'issues'.
If you can design a single PCB using say the 46K22 and use it for several projects, the cost goes down AND programming time as well, as you KNOW the PIC.
These days most people focus on the chip cost NOT the R&D cost. That new PIC may have a great feature BUT if it takes you all week, the R&D time adds up to a LOT of money. Sometimes it's better to spend an extra $1 for the bigger PIC. Consider $100/hr for R&D, if you 'wasted' an hour finding out that bit 7 of an SFR is supposed to be a 0 not a 1, that $100 could have bought 100 extra big PICs that you KNOW work.
Unless you're doing production runs of over 100 units, I don't see the cost/time benefit of using a small PIC. I also prefer BIG enclosures not tiny ones. My standard box is roughly 6 by 8 by 2 inches, with a 4 by 6 PCB. It allows lots of room for components, easy access and, well, looks good!

[PCM programmer]