Digital-analog synth

[KammutierSpule]

Hallo there! First post here!

I've starting reading about DIY synths and I'm plan play with my own. I've found nice pages around web, i'm studding the schematics and so on.
I've some ideas of what I want and I'd like know if you would like share and give me any sugestions for this:

Firstly, I'm not play doing a real analog synth, but some hybrid digital/analog control.

I'm planning use a microcontroler to control and play with analog part.
With PIC I can get: 10 MIPS, PWM (8..10bits), ADC >50ksps 10bits, comparators, IO ports, and so on.

I dont like much the idea of using a DAC because I dont have so uch CPU power to play with and to all in the CPU, but I can use some kind of "resistor DAC" to control some levels, maybe pwm to control some LFO voltages, etc
What's your opinion about "resistor DACs"?

I would like my circuit will not be too much complex, because I plan then add paralell voices, replicating the circuit, but I dont want it so simple because I would like play with lot of parameters and haver rich sounds

I'm not interested also in audophile quality assegurance, firstly I want just some nice noise

The main goal using a microcontroler is to let the micro produce some kind of sound, or play wish diferent waveforms, maybe selecting them or guide them (waveguide?) and minimise the use of external components. In the schematics I'me learning in web, they are more focus for
all real< analog synths:

What do you sugest I learning and looking for? What ideas do you think I can explore using the microcontroler and analog?

I've found about 1 ou 2 projects using microcontrolers, but they seam too simple to mee, just "tri,saw,sqr" waves->Filter->envelop

Idea: If I use two separate "8bit or more resistor DAC" (resistor so we can haver different level because resistor tolerance) to play "two voices" an then put it to filter,efx,etc... will output listened sound less "8bit" ?!
what about use two 12-bit DAC?

Thanks in advance for sugestions!!
Mario "KammutierSpule" Luzeiro

[blue hell]

I've seen some LFO implementations done on a PIC and some years ago I tried to do some simple audio stuff on a PIC12, your plans may be a bit ambitious, the AVR sems more suitable, otoh there is DSPIC and maybe a PIC24 or 32 might be useful ... some quick links :

AVR synth info : http://www.elby-designs.com/avrsynth/avrsyn-about.htm

And a PIC based LFO can be found at : http://www.tomwiltshire.co.uk/sdiy/

[blue hell]

Oh and there is the soundgin of course

[JovianPyx]

As Mr. Blue said, the AVR appears to be more capable as a CPU for this purpose. And the AVRsynth is already developed - you can start with that as a known working project and learn from the code and then modify or even completely rewrite it.

However, the AVRsynth is fairly well maxed out, so don't expect it to be much more than a monosynth with 2 oscillators.

I apologize for this next mental tic of mine: Consider an FPGA, at least eventually... Very powerful in comparison to the AVR, it rivals the power of DSP chips. It is, however, a very different design paradigm, but it is not unlearnable (after all, I did and I don't think I'm a friggin' genius ) It does require a bit of reading though.

[KammutierSpule]

I already have look to AVRsynth and some other PIC LDO's, LFO is easy to be done at some point, but I want more. LFO's will be usefull at some point, so it will be need to be implemented anyway. I'm now more concerned in "VCO" and "VCF" part integration and control.

I know AVR are good uC(and now there are real beasts ARM's) , but I can easy get samples from microchip. An usualy PIC can give about 10MIPs, it not so few (about >200 intructions for sample @ 44.1Khz), maybe I will look in dsPICs (30MIPS).

I'm thinking make some kind of "opensource" project, so, easly phisicall program the uC and get parts will be a must.

I think I'll find some how.. to make somthing like re-synth an analog circuit and use for it less cpu power.
Example of my idea: having a VCO controled by PWM os something to get the right frequency and use control IO's with resistors to control gain in the output wave, I supose in that I'll control the shap of output wave.

Mario

[JovianPyx]

Do I read this right that you're doing a hybrid synth - where some bits are digital (LFOs) and other bits are analog (VCO) ?

Depending on what exactly you want to put into a PIC, or several PICs, certainly LFOs will work, also envelope generators.

Filters can require a bit more horsepower, but are not impossible, the AVRsynth is a complete synth, with a filter running in an AVR. A slower CPU like a PIC could do a filter - but be warned, I believe most PICs use 8 bit data words and you may want to use wider words for things like filters and NCOs. 8 bits is considered low-fi, but you may like that. A PIC can do 16 bit arithmetic, but you sacrifice clock cycles to do it. Remember that filters use multiply operations which are very slow programmed in PIC. The AVR or the dsPIC sound like better ideas for what you want to do.

[KammutierSpule]

[blue hell]

[JovianPyx]

[blue hell]

Yup 16x16->32 and even 16x16->16 still take relatively much time combining the partial mult's. MAC's still take time, and I'd want saturating and signed arithmetic for audio stuff.

But this is going OT a bit I guess

[jksuperstar]

From what you describe, I think you might not like using resistor DACs, and use either a real DAC, PWM, or most likely, a digital potentiometer, like analog devices sells. Then, you can take clearly defined analog VCO's or filters, and replace the pots or CV ins with your digitally controlled voltages, while minimizing extra logic or buffering.

A serial interface to any of these (DAC, digital pot, etc) is typically built-in to the microcontroller. Very simple protocol, and virtually no overhead in your code.

[KammutierSpule]

[blue hell]

As a side note ... I've found SPI easier to implement than I2C on PICs, both with and without using the built in HW support.

[jksuperstar]

Don't forget, I2C is a registered trademark of Philips, so some other manufacturers have an interface that's compatible, but they just call it a "Two Wire Interface" or TWI. Just take a look through your specs for the processor you want to use, and try to get the gist of the interfaces purpose. It's not that difficult, you'll learn a bit in the process, and it'll open up a lot of pins on your microcontroller to have things like switches, encoders, pots, and LEDs to control or see what your synth is doing.

[carld]

I've actually been working on a synth based on Arduino, written in C++. I'm using a pretty crude R2R ladder DAC with pretty good results, considering that my 2Rs aren't even 2*R.

Right now it's 8 bits wide, as I wanted to save some of the pins for inputs, volume control, etc. It could be expanded reasonably easily, tho.

Features:
- Wavetables (I've got sine, octaves, thirds and a major chord implemented)
- Square, Triangle, Sawtooth
- Sequencer, Appegiator
- MIDI support (ie Sequencer takes MIDI notes and params - in/out will be easy to add)
- Detune
- Transpose

Written, not working:
- Polyphony
- ADSR Envelope Generator

Planned:
- LFO
- Filters (may be tough)
- Other Envelopes
- Noise (white easy, others easy if I can figure out frequency filtering)
- Ring Modulator

It was sounding pretty nice, when I was banging out the bits directly, but I want to add polyphony, which requires time-slicing the output, and I've introduced a bug or a timing issue in the process. Once I have it back in shape, I can post the code. I also have some pictures of the breadboard, but it's exactly what it sounds like, an Arduino with a couple of pins connected to a pot, and the rest going to the R2R ladder.

As far as the performance of the AVR goes, it is pretty good, but tight I would say. Using Arduino makes things a little more convenient for me (I'm a software engineer, playing with hardware, so just soldering together the system was a big step for me ).

The code is pretty clean and tight, and I think will be useful for building *parts* of a complete synth. I'm really designing it as a framework for playing around with the different ideas. For example, I have been thinking as you are about having the real high-frequency sound work being done off-chip, so the AVR just has to do the low-frequency thinking, UI, etc.

Others have mentioned using a digital pot to control hardware oscillators, which is a cool idea that I want to play with. I've also seen designs using looping FIFOs outputting to DACs, which fits my wavetable approach nicely. This also could allow for larger tables, completely off the AVR, perhaps with external EEPROM or even removable flash that the AVR never even has to see, other than to initiate/coordinate loading of the FIFOs.

Maybe we could work together on something...

[KammutierSpule]

[carld]

Interesting. This is one of the reasons I'm thinking about going to a more high-powered platform, like a DSP or faster processor, to do things like FFT for filtering and other effects that would be harder to do in hardware. I mean, the cool thing about analog is the purity of the sound, the downside is that it doesn't scale well. Digital seems to be getting to the point where some serious number crunching can go on at 44kHz (and higher) for multiple channels. But for more modest processors, one or two channels is probably the most you can do easily, which doesn't give much scale. So, better to use these processors to control and inject information...

[KammutierSpule]

[carld]

I haven't actually counted (although I plan to now), but roughing out the math says (16MIPS/44 kN/sec =~ 363 inst/N). It's a fair amount to work with, even allowing for overhead for timer interrupts, etc., but anything requiring floating point math or other involved calculations needs to be done over several samples, during initialization, etc., which can be problematic.

[KammutierSpule]

[carld]

[KammutierSpule]