easy function generator

[patroclus]

Is there an easy way to build a fyunction generator??

I heard about IC 8038 (square, sine and triangle, up to 200Khz)
I also thought on this simple scheme:

Storing the waveforms digitally encoded in a ROM, and then clock a counter, taking the 8 bit data bus out of the ROM to a Digital/Analog converter.

What do you think?? It would let you use wide frecuencies (up to several Mhz), lots of waveforms of any shape, and could be controlled by an OP amp to get milivolts to several volts output.

 

[dipal_z]

I think its good idea. Use of any programmable logic has one major advantage over discrete solution that is you don’t have to change any hardware for minor changes (like tuning the shape of waveform) and major changes (adding new waveforms).

But you need to put the filter at the output of the ADC to get the perfect shape of output waveform and it might be difficult to design a filter which will work properly over wide frequency range.

Also you can try using microcontroller along with ROM which will give you more flexibility and control.

 

[Nigel Goodwin]

It's a commonly used method, the output filter really needs to be a variable one, which alters as the frequency changes.

You might like to look at **broken link removed** which is very simple and based on a PIC.

 

[patroclus]

so the output of the D/A driven by the ROM, without filter, won't do the job??
I know, twithout filter the waveform would be some kind like "stepping" waveform... but, giving enough points?

 

[Nigel Goodwin]

so the output of the D/A driven by the ROM, without filter, won't do the job??
I know, twithout filter the waveform would be some kind like "stepping" waveform... but, giving enough points?

It depends how good you want it to be?, you might also reconsider your 'good for several MHz' - assuming you use 8 bit resolution?, then to get 1MHz out you need to clock it at 256MHz, which is rather too high for the components involved.

 

[patroclus]

if using parallel 8 bits D/A converter, for 1Mhz signal, you should use a clock of X*samples per period of the signal Mhz.
As for the filtering, I suppose that if you don't filter, a sine wave would have lots of other frecuencies appart from the fundamental, due to the fast changes...

When as transistor input, used to measure gain at output, is this very bad thing?? it affects transistor response from a "pure" sine wave??