If you are ok with PC sound card, you can use single opamp, with a variable gain.
Keep a standard wave files of different frequency tones with you. Use software like soundforge, to add them. I feel this could be the easiest and cheapest way out.
Keep a standard wave files of different frequency tones with you. Use software like soundforge, to add them. I feel this could be the easiest and cheapest way out.
Hi,
Check the programmable Basic Micro Nano 8 instruction "FREQOUT". See if you can use it. You can program it to generate any frequency from 0 – 32767Hz. Down load the Basic Micro Studio Syntax Manual and check the above instruction.
Check the programmable Basic Micro Nano 8 instruction "FREQOUT". See if you can use it. You can program it to generate any frequency from 0 – 32767Hz. Down load the Basic Micro Studio Syntax Manual and check the above instruction.
Oscillators crosstalk to each other and this can cause frequency shifting and jitter. In the extreme case this becomes injection locking.
Even if they are phase locked to a common source, I have seen unexpected phase jitter. If you are going to phase lock the tones, I really cannot see why you don't just do this in software plus DAC
Could injection from one oscillator to another be solved by careful PCB layout? Is there any type of shielding you can put around traces on a board? Perhaps on a multilayer board you could have grounded copper above, below, and between signal traces.
Don't overthink it 😉
Reasonable decoupling (say, feeding each Op Amp through 10 ohms and with 10uF to ground, plus reasonably thick and short ground paths) should be enough.
You will immediately mix output signals anyway .
Proto, say, 4 oscillators and 1 submixer and see if you are happy with the idea.
Reasonable decoupling (say, feeding each Op Amp through 10 ohms and with 10uF to ground, plus reasonably thick and short ground paths) should be enough.
You will immediately mix output signals anyway .
Proto, say, 4 oscillators and 1 submixer and see if you are happy with the idea.
As the OP was not really clear as to the outcome of this project, asumptions have to be made.
1) The oscillators are spaced at about 14 each octave, 20 to 20k Hz (chromatic scale, not linear)
2) A split supply of +/- 15 to 18 VDC is used
3) Wien-bridge oscilators (low distortion) are employed, one op-amp each oscillator, sinewave output of about 15V, summed together gives about 100 mV at the summing point, add gain as needed
This is what I would do:
1) Use quad op-amps. 4 oscillators each sub-board for a total of 38 boards. Use local supply regulation and/or ICs with high CMRR
2) Select frequencies that are not multiples of each other
3) Divide the spectrum into 4 bands and have one oscillator of these on each pcb so that the oscillation frequencies are far apart from each other on each board
4) Pray
Good luck! E
1) The oscillators are spaced at about 14 each octave, 20 to 20k Hz (chromatic scale, not linear)
2) A split supply of +/- 15 to 18 VDC is used
3) Wien-bridge oscilators (low distortion) are employed, one op-amp each oscillator, sinewave output of about 15V, summed together gives about 100 mV at the summing point, add gain as needed
This is what I would do:
1) Use quad op-amps. 4 oscillators each sub-board for a total of 38 boards. Use local supply regulation and/or ICs with high CMRR
2) Select frequencies that are not multiples of each other
3) Divide the spectrum into 4 bands and have one oscillator of these on each pcb so that the oscillation frequencies are far apart from each other on each board
4) Pray
Good luck! E
Probably cost prohibitive depending on how many you need, but the XR2206CP is an option.
http://www.sparkfun.com/datasheets/Kits/XR2206_104_020808.pdf
http://www.sparkfun.com/datasheets/Kits/XR2206_104_020808.pdf
Oh my god - $84 each on Mouser! To buy 150 of them would cost twice as much money as I spent on my last car! Thanks for the recommendation though! 😀
Also, thanks Fahey and Mickey for your last two comments - A design is now under way. Because you all have been so helpful, I will certainly return for more discussion in the future.
Also, thanks Fahey and Mickey for your last two comments - A design is now under way. Because you all have been so helpful, I will certainly return for more discussion in the future.
Ha... good thing mouser isn't the only place to buy things. That price is for an eval board anyway.
You can get them for $1.75 if you look hard enough.
The XR2206, from what I recall, generates a triangle wave and uses (perhaps several stages of) diode shaping to get a sine wave, which will still leave distortion surely over one percent. It should be just about all odd harmonics, if that helps. Various oscillator designs using a moderate-priced opamp or one or two transistors (as have already been discussed) can easily give better performance at lower cost.
[1] you will never have to worry about all-sum being 3 volts or anything else that large. Analog oscillators - unless phase-loop type (which ideally sync to a crystal clock) - drift all over the place, and really can't be set with frequency accuracy better than the tempco of the components and their accuracy. I.e. ~2 to 5%
[2] because of [1], you won't have any ability to control phase locking of the signals. Even if you use phase-locked loop synthesis, you'll STILL not really have much control over whether two independent frequencies are going to be phase or antiphase (or between) locked.
[3] Then there's the case where you get "another theory", and want to change all the frequencies of all the oscillators. Going to be putting TRIMPOTS in there, buddy? You'll need that to control things - and even so, the dynamic range of adjustment is only a few octaves for given capacitor values.
[4] The "summing thing" will need trimpots too. Might as well just buckle in and use a summing precision op-amp for all 150 channels. Grounding the + input, and putting summing pots on all inputs, to the - input ... with 1:1 gain feedback, basically drives the - pin toward ground. So much so that its effective impedence is measured in milliohms. Thus, there is almost no crosstalk between all summing inputs.
[5] If you're particularly "analog inspired" and "digital averse" ... then you may want to avoid using phase-locked loop topologies. No matter how they're cast as bridges between the analog and digital domain (which they absolutely are!), they're still half-digital, and an amazing amount of digital "leakage" gets onto the output waveform. Its basically inevitable. So, to get rid of the digital cruft ... you'll have to introduce some tuned filtering to clean up those sinusoids. More components.
[6] So ... lets get down to brass tacks. How many components per oscillator?
osc: 2 transistors + 2 caps + 3 resistors + 1 pot
filt: 1 resistor, 1 capacitor (not adjustable)
sum: 1 pot, 1 resistor, 1 coupling capacitor
So that makes 2Q, 4C, 5R, 2P ... times 150 = 300 transistors, 600 capacitors, 750 resistors and 300 trimpots.
The "output" is trivial by comparison. 1 OpAmp, a couple resistors and an output coupling capacitor. Power supply ... probably 25 milliwatts per oscillator or so, and 100 milliwatts for the summing amp. Any LEDs? 100 + 150 * 25 + misc = 4,000 milliwatts, or 4 watts. Well, the power requirement isn't bad. Could do that with rechargable batteries.
Personally, I think insistence on an all-analog solution is just silly. Far better would be to read up on the DAC and data buffer of your Mac or PC, and how to send stuff to it in real-time. You might not have to look further than "how to create a WAV file from EXCEL spreadsheet data". Because if you create a WAV file, then there's nothing further to figure out: play back the darn thing, rely on the analog filtering of the PC DAC to accurately reconstruct smooth analog waveforms ... and love the fact that you can synthesize, literally, any arbitrary sum of component frequencies, even if that isn't quite in real-time. WAV files "store the result of many billions of CPU cycles of computation".
Note: I'm a total analog head. Yet ... having once built a 200 oscillator synthesizer (it was supposed to make for a good organ unit. Never worked out anywhere near as good as I hoped for it.), I recommend "Go Digital, lad!" I wouldn't recommend digital if your oscillator count was less than a dozen or two. That can be done in hardware and is fun. Anything more than 3 dozen ... is just foolish in 2013 to do in straight analog.
GoatGuy
[2] because of [1], you won't have any ability to control phase locking of the signals. Even if you use phase-locked loop synthesis, you'll STILL not really have much control over whether two independent frequencies are going to be phase or antiphase (or between) locked.
[3] Then there's the case where you get "another theory", and want to change all the frequencies of all the oscillators. Going to be putting TRIMPOTS in there, buddy? You'll need that to control things - and even so, the dynamic range of adjustment is only a few octaves for given capacitor values.
[4] The "summing thing" will need trimpots too. Might as well just buckle in and use a summing precision op-amp for all 150 channels. Grounding the + input, and putting summing pots on all inputs, to the - input ... with 1:1 gain feedback, basically drives the - pin toward ground. So much so that its effective impedence is measured in milliohms. Thus, there is almost no crosstalk between all summing inputs.
[5] If you're particularly "analog inspired" and "digital averse" ... then you may want to avoid using phase-locked loop topologies. No matter how they're cast as bridges between the analog and digital domain (which they absolutely are!), they're still half-digital, and an amazing amount of digital "leakage" gets onto the output waveform. Its basically inevitable. So, to get rid of the digital cruft ... you'll have to introduce some tuned filtering to clean up those sinusoids. More components.
[6] So ... lets get down to brass tacks. How many components per oscillator?
osc: 2 transistors + 2 caps + 3 resistors + 1 pot
filt: 1 resistor, 1 capacitor (not adjustable)
sum: 1 pot, 1 resistor, 1 coupling capacitor
So that makes 2Q, 4C, 5R, 2P ... times 150 = 300 transistors, 600 capacitors, 750 resistors and 300 trimpots.
The "output" is trivial by comparison. 1 OpAmp, a couple resistors and an output coupling capacitor. Power supply ... probably 25 milliwatts per oscillator or so, and 100 milliwatts for the summing amp. Any LEDs? 100 + 150 * 25 + misc = 4,000 milliwatts, or 4 watts. Well, the power requirement isn't bad. Could do that with rechargable batteries.
Personally, I think insistence on an all-analog solution is just silly. Far better would be to read up on the DAC and data buffer of your Mac or PC, and how to send stuff to it in real-time. You might not have to look further than "how to create a WAV file from EXCEL spreadsheet data". Because if you create a WAV file, then there's nothing further to figure out: play back the darn thing, rely on the analog filtering of the PC DAC to accurately reconstruct smooth analog waveforms ... and love the fact that you can synthesize, literally, any arbitrary sum of component frequencies, even if that isn't quite in real-time. WAV files "store the result of many billions of CPU cycles of computation".
Note: I'm a total analog head. Yet ... having once built a 200 oscillator synthesizer (it was supposed to make for a good organ unit. Never worked out anywhere near as good as I hoped for it.), I recommend "Go Digital, lad!" I wouldn't recommend digital if your oscillator count was less than a dozen or two. That can be done in hardware and is fun. Anything more than 3 dozen ... is just foolish in 2013 to do in straight analog.
GoatGuy
- Status
- Not open for further replies.