Would you pick digital or analog adsr and lfos?

[skitchy]

Consider this : you are designing a synth with analog oscillators and analog filters. The design already contains a capable microcontroller for midi duties.
"Pop quiz hotshot"
Do you
a) program the adsr and lfo on the microcontroller, thereby making them digital. Its a lot easier, and you can really go nuts with the features and tweakbility.
b) build analog adsr and lfos. More work, less 'features' but true analog.

[Syncretia]

This is really a strange question.

An analog LFO/ADSR is a completely different thing to a digital LFO/ADSR. When you talk about analog, you are talk about pulsating voltages. When you talk about digital, you are talking about numbers. Only analog synths understand analog LFO's/ADSRs. Only computers understand digital LFOs/ADSRs, and even then these things are relevant only to the software you are using.

So, I don't really understand what the question is asking. But, at the end of the day, the only real difference is that a digital LFO can be synced easily, while syncing an analog LFO requires extra circuits. You could have a use a digitally controlled ADSR or LFO to modulate an analog sound but wouldn't be much difference to the actual sound.

[AdmiralQuality]

This is really a strange question.

An analog LFO/ADSR is a completely different thing to a digital LFO/ADSR. When you talk about analog, you are talk about pulsating voltages. When you talk about digital, you are talking about numbers. Only analog synths understand analog LFO's/ADSRs. Only computers understand digital LFOs/ADSRs, and even then these things are relevant only to the software you are using.

So, I don't really understand what the question is asking. But, at the end of the day, the only real difference is that a digital LFO can be synced easily, while syncing an analog LFO requires extra circuits. You could have a use a digitally controlled ADSR or LFO to modulate an analog sound but wouldn't be much difference to the actual sound.

The output of the digital LFO and ADSR are converted to analog so they can drive the appropriate analog modulation destinations.

Take a listen to the Roland JX-3P and JX-8P. There's digital envs and LFOs driving analog filters and amplifiers. The problem is the zipper noise, as well as poorly modeled envelope behavior. But with better hardware resolution and software model, it could be done so that nobody would notice the difference.

[casper007]

I'd say digital: It's easier, and you can provide more functionality!

People don't care about 100% analog,
as long as the [i]signal path[/i] is 100% analog...

[skitchy]

Yes, I'm definitely more on the side of digital ATM. The only thing that worries me is the 'zipper' noise that AdmiralQuality mentioned - I always knew it would be a problem that needed solving.

[AdmiralQuality]

Yes, I'm definitely more on the side of digital ATM. The only thing that worries me is the 'zipper' noise that AdmiralQuality mentioned - I always knew it would be a problem that needed solving.

You could solve it with a simple R/C filter *after* the digital output. A little smoothing goes a long way. Try it!

[whyterabbyt]

People don't care about 100% analog,
as long as the signal path is 100% analog...[/i]

well, that's not entirely true. some people do care. ask a large group of modular synth owners, and i think you'll find a few people who care, quite excessively. mileages, as always, vary.

[aciddose]

you probably want digital as long as your mc is capable of handling it. how much code can you fit in there? how many channels do you have and what kind of routing interface are you going to implement?

the advantages of having four or five channels and the ability to route to every parameter is that you can have an analog mod matrix. the disadvantage is going to be depending upon how high you can run the dacs and the generators on the chip.

i manage to generate 50k easily with a z8e, but it's not easy to implement this stuff with only 4k of program memory.

also i've limited myself to two channels and a serial dac. with a parallel dac, multiplexors and sample-holds you could get a lot more channels.

think about mod destinations: width, detuning, cutoff, maybe res, maybe mix levels and with a single osc you're already looking at atleast three at a minimum. (width, tune, cutoff).

the other disadvantage is that you probably will be rate-limited and probably won't be able to generate good audio-rate modulation. not because it's impossible to do, but because you won't be able to synthesize anything complex at those rates without a powerful processor.

i've been there, done that. i used both. i have analog adsr and modulator (supports audio rates to 8k), and cv is provided for pitch and cutoff. i can produce similar results in software but since i can't route to pw and the resolution is limited to 12-bits it doesn't give the same results as the analog implementations.

[skitchy]

Thanks for the information
@AdmiralQuality - in my mind an R/C filter is going to prevent fast attack and release times - is there a simple way around that problem which I'm not seeing?

[aciddose]

yes. it's called pwm.

what you probably want to do though is output your modulation signal at a high enough rate. then you won't care.

[UltraJv]

More info would help. Is it a private project? What micro contoller? There are many existing designs in kit form out there :

http://arduino.cc/playground/Main/ArduinoSynth

Ah yes, I remember you now. You got the MSP430 Launchpad. You get what you pay for

Spec :

2kB Flash, 128B RAM, 10 GPIO, 1x 16-bit timer, WDT, BOR, USI (I2C, SPI), Internal Temp Sensor, 8ch 10-bit ADC

[aciddose]

actually "you get what you pay for" is as usual, wrong.

you get what you know how to use.

the z8e series supports rates up to 50mhz, gpio in the 20s and 16k, 32k or more program memory. 4k or more ram. it's also in-circuit programmable and code can write to flash. (code can modify itself, if you're clever enough you can send new 'modules' via midi sysex and write those and call the functions!)

unfortunately this doesn't come in a kit.

personally i used the 8-pin dip 4k version because i was able to fit all the code i needed on that chip. i also run it using it's internal oscillator because that's perfectly fast enough for my purposes and doesn't result in any zipper noise or other issues at all. it also happens to have a built in uart s/r.

those are $3.50 each.

best part - it has a c compiler.

[UltraJv]

actually "you get what you pay for" is as usual, wrong.

you get what you know how to use.

the z8e series supports rates up to 50mhz, gpio in the 20s and 16k, 32k or more program memory. 4k or more ram. it's also in-circuit programmable and code can write to flash. (code can modify itself, if you're clever enough you can send new 'modules' via midi sysex and write those and call the functions!)

unfortunately this doesn't come in a kit.

personally i used the 8-pin dip 4k version because i was able to fit all the code i needed on that chip. i also run it using it's internal oscillator because that's perfectly fast enough for my purposes and doesn't result in any zipper noise or other issues at all. it also happens to have a built in uart s/r.

those are $3.50 each.

best part - it has a c compiler.

All well and good but thats not what he has. I did post the spec of the supplied devices. The kit runs at 32khz :

http://processors.wiki.ti.com/index.php ... -EXP430G2)

[AdmiralQuality]

Thanks for the information
@AdmiralQuality - in my mind an R/C filter is going to prevent fast attack and release times - is there a simple way around that problem which I'm not seeing?

That's what I use to solve the issue in software. More of a concern with LFOs and controllers than envelopes.

[aciddose]

The kit runs at 32khz

yes well, i've done all this successfully already and so i know it works fine.

here are my specs:

z8exp4k:
- 8 pin dip
- 4k program memory
- 1k ram
- s/r uart
- six gpio
- programmable interrupt controller
- watch-dog timer
- voltage brown-out protection (run special code on power-off, save to flash)
- 3.3v supply
- 5.5mhz internal oscillator

my circuit also includes:

- midi buffer circuit, opto-isolator and transistor "thru" buffer
- dual-channel serial dac, 12 bits
- opamp buffer and cv scaling circuit for 15v output
- 3.3v regulation
- serial shift register with 8 outs
- 5 sziklai pair buffers for 15v gate outputs

this all fits on a 1" x 1" board using through-hole mounting. would be smaller with smd, but i don't like prototyping with smd.

so all he really needs to do is get the chip to run at a rate like 5mhz. are you sure it doesn't support different modes and trimming?

if the limit is 32k you won't even be able to reliably receive midi.