andales said:
"It has become a hobby of mine to build the insides of a ICs coz for me, it is the only way to understand how ICs work. Long time ago I built a decoder without using decoder ICs (74LS48 for example) , instead I used logic gates ICs to build a decoder inorder to better understand why 74LS48 can display 0-9. I have built a counter using flip flops without using some counter ICs, with the same reason as stated earlier"
There's a huge difference between building a basic logic device out of individual logic gates, and developing a complex analog circuit that gives you specific waveforms and works across a broad frequency range. Digital is pretty easy, as long as your design isn't extremely flawed it should work pretty well at quite a range of speeds/frequencies. With analog (particularly in a function generator), you need linearity, and achieving that across a very broad range of frequencies is not so easy.
If you're truly crazy enough to attempt this, you ought to forget about doing it all at once and work on it in stages. For starters, you might want to try to build a decent op-amp, and a decent oscillator where you can reliably change the frequency over a huge range, so you can use them as building blocks if you ever get around to attempting the system as a whole.
But avoiding the use of op-amps, for example, is crazy. You say you used logic gate IC's in your previous logic designs, well why didn't you use discrete transistors instead? Probably because that would have resulted in a huge board, and not have even worked well anyway. The same could be said about op-amps. If you download the datasheet for the LM741, a very old and simple op-amp, you'll see about 20 transistors in it. If you really want to learn how they work, why not just try building one, playing with it, and then just use real op-amps in your other design to save the trouble. It's certain to work poorly, as you won't have the transistor matching you would in a single IC, not even counting all the other factors working against you, but you might be able to learn something from it. The problems of linearity and dynamic range and frequency response and all that are things that have already been dealt with by the people who designed op-amp IC's...
It's one thing to reinvent the wheel, it's quite another to reinvent an entire car. You said you built a logic decoder, well, if you later decided to build a more complex logic design that incorporated logic decoders, would you actually re-build your own elaborate logic design multiple times, or would you just go ahead and use commercial logic decoder IC's, now that you know how they work?
If you choose the former, well, I think you're nuts. If you choose the latter, then why not do the same with this function generator project? If you want to know how a certain part works, go ahead and build one, but refusing to use any IC's just to force yourself to do everything completely from scratch truly is masochistic, and pretty pointless, as you're sure to end up with a very poorly-working finished product despite spending FAR, FAR more time building it.
Now, since I get the feeling you are not going to give up on this easily, before you ask further questions about the details of a function generator, why don't you go download the datasheets for some commercial function generator IC's and look at how THEY get the job done. For instance, the datasheet for the old MAX038 has a block diagram of the whole system that is very useful. There are still other details, like how to generate sine/triangle/sawtooth waves, most of which you can also answer for yourself just by searching google and/or the forums.
At the very least, you should do your research so you know what you're getting into... Your basic questions about oscillators suggest that you don't yet realize what you're in for.