Until you have the means to stimulate your circuitry in a predictable way, you can't test it. You can play with it all day, but until you have the means to subject your prototypes to a verifiable sequence of well-defined stimuli, both digital and analog, that represent the various combinations of signals to which they may be expected to respond, both singly and in combination, you haven't got the means to perform proper tests.

I personally have several old signal generators and pulse generators. I use a pulse generator to provide regular pulses, triggers, both to my circuits and to my other signal sources, and I use one or more function generators, often triggered or otherwise influenced by the pulse generator(s) or one another, to stimulate my circuitry. I use one of several counters to adjust frequencies to the desired values and I use remotely controlled multimeters, and power supplies, to monitor and control slowly changing voltages.

Pulse and function generators should, in general, be able to drive their output over a voltage range of 30 volts peak-to-peak, adjustable of course, to the required voltages, into a 50-ohm load. If they can do that, you'll never be sorry. If they can't, sooner or later you'll regret it.

You should have at least one externally triggerable function generator that can provide sine, triangle, and square waves of no more than .5% distortion over their entire frequency range, with the ability to sweep between a minimum and maximum frequency that you specify. This is absolutely necessary in proper design and verification of filters, phase-locked-loops, and many other circuits. Such a generator will be able to free-run or provide a burst of oscillating signal, or a single pulse of predetermined length, amplitude, and adjustable rising and falling slope as wel as duty cycle. It should also be able to phase-lock to an input signal, as well as allowing amplitude and frequency modulation of its output. Instruments of this type have been available for a very long time and they are essential for many types of work.

The pulse generator is very useful in that you then don't as often have to build application-specific logic or other circuitry with which to stimulate your instruments or circuit-under-test.

A high-precision (20ppm or better) frequency counter, capable of counting a decade above the highest frequency above which you intend to work, ever, is necessary so you can set center frequencies with reasonable accuracy as well as sweep limits.
A two-channel universal counter/timer is very useful, though those tend either to be quite costly or only count into the low hundreds of MHz. I personally have a couple of universal counters that count up to 225 MHz and a couple of high-frequency counters that count into the GHz range.

A colleague of mine has designed a simple DDS arrangement probably capable of tens of MHz aside from the particular DAC he uses, but it is programmable via the PIC that he's used, to set an amplitude as well as a wave table. Because he uses a very fast MCU, he can generate lots of samples per second.

This "poor-man's" function generator looks as though it has potential, but doesn't exceed the capabilities of simple XR8038, 2206, or 2207 function generators. Those are cheap and well understood. Of course, it is inherently capable of producing an arbitrary waveform, and, in the context of microcontroller testing, it could prove quite useful, given adequate output filtration and buffering, in stimulating a microcontroller with analog inputs. It looks as though, for most MCU applications, it could turn out to be quite worthwhile.

RE