One thing I havent got yet is an audio oscillator - it's a "one of these days" projects. 
Anyway, it seems to me that if we restrict ourselves to audio frequencies then a DDS oscillator might not be too hard for a diy project.
For those who have not come across this technique, imagine you have a singe sinewave cycle mapped into an eprom that is 65536 addresses long. If you step through these addesses one at a time and pump out the data into an A/D then you get a good sinewave. The faster you clock through the eprom the higher the frequency of the sinewave. A bit of low pass filtering and you are there.
Now think of it this way instead . If you instead have a =fixed= clock frequency of 65536 Hz then you get a sinewave of 1Hz. Now change things so you step through every *second* eprom address - you get right through the addresses in half the time so you get 2x the frequency sinewave. Every 3rd address = 3x the frequency etc. If the amount of addresses stepped through does not exactly divide into 65536 it doesn't matter. It just rolls over to a number greater than zero and then steps through a different set of equally spaced addresses. Eventually it will repeat itself.
One of the really good things about this method is you use enough high pass filtering for the highest frequency you are going to use. As you go lower in frequency the waveform doesn't deteriorate because the sampling *rate* remains the same but as the output frequency is lower it means there is more detail put into the waveform. Less filtering needed anyway. If the sampling rate is far enough above the output frequency, you will get a flat frequency response. Crystal locked discrete steps too, if you want.
Have to think some more about this.
Anyway, it seems to me that if we restrict ourselves to audio frequencies then a DDS oscillator might not be too hard for a diy project. For those who have not come across this technique, imagine you have a singe sinewave cycle mapped into an eprom that is 65536 addresses long. If you step through these addesses one at a time and pump out the data into an A/D then you get a good sinewave. The faster you clock through the eprom the higher the frequency of the sinewave. A bit of low pass filtering and you are there.
Now think of it this way instead . If you instead have a =fixed= clock frequency of 65536 Hz then you get a sinewave of 1Hz. Now change things so you step through every *second* eprom address - you get right through the addresses in half the time so you get 2x the frequency sinewave. Every 3rd address = 3x the frequency etc. If the amount of addresses stepped through does not exactly divide into 65536 it doesn't matter. It just rolls over to a number greater than zero and then steps through a different set of equally spaced addresses. Eventually it will repeat itself.
One of the really good things about this method is you use enough high pass filtering for the highest frequency you are going to use. As you go lower in frequency the waveform doesn't deteriorate because the sampling *rate* remains the same but as the output frequency is lower it means there is more detail put into the waveform. Less filtering needed anyway. If the sampling rate is far enough above the output frequency, you will get a flat frequency response. Crystal locked discrete steps too, if you want.
Have to think some more about this.
Graham, there are several interesting app notes on this for the Analog Devices integrated DDS signal generators, with on-chip synchronous DACs IIRC. I once had a similar idea, using a microcontroller and a DDS chip and very little more, OK a display and some rotary encoders to set freq etc.
Certainly doable, but I never found the time to put it in practice.
Jan Didden.
Certainly doable, but I never found the time to put it in practice.
Jan Didden.
Hi Graham
You would not even need to do this in hardware at audio frequencies. Nowadays even PC processors have enough power to do this.
But IF you want dedicated hardware you could use a DSP to generate the waveform, using a sinewave-approximation algorithm. If the amplitude resolution doesn't have to be too high then there are even chips available that include all the main blocks needed: Accumulator, wavetable and D/A
(keywords: NCO - "numerical controlled oscillator" or DDS - "direct digital synthesis")
If you need a quick-and-dirty solution you can use the following ( in German only unfortunately):
http://www.audio-software.de/downloads.html
There was also a free generator software by Neutrik but I have to find that one again. I am not sure though if that could directly output audio via soundcard or if you'd have to burn it on CD.
Regards
Charles
You would not even need to do this in hardware at audio frequencies. Nowadays even PC processors have enough power to do this.
But IF you want dedicated hardware you could use a DSP to generate the waveform, using a sinewave-approximation algorithm. If the amplitude resolution doesn't have to be too high then there are even chips available that include all the main blocks needed: Accumulator, wavetable and D/A
(keywords: NCO - "numerical controlled oscillator" or DDS - "direct digital synthesis")
If you need a quick-and-dirty solution you can use the following ( in German only unfortunately):
http://www.audio-software.de/downloads.html
There was also a free generator software by Neutrik but I have to find that one again. I am not sure though if that could directly output audio via soundcard or if you'd have to burn it on CD.
Regards
Charles
Which is fine if you have a "nowadays" PC.
P200MMX here. 
I wanted to avoid using a soundcard because I have never had much luck finding on that had a flat output below 30Hz. Even with really big coupling caps added on. Must be a software thing. Besides, timewise, nothing beats rotating a pot or similar to find a resonance peak, IMHO.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.