[sdiy] droop of sustain level in analogue ADSR?

[rsdio at audiobanshee.com]

Thanks, Tim. I enjoyed the write-up and graphic illustration of what you found.

Brian


On Nov 15, 2014, at 12:42 PM, Tim Stinchcombe <tim102 at timstinchcombe.co.uk> wrote:
>> Hi, had a rather intense discussion with a colleague today.
>> He claimed there are analogue ADSR-designs which suffer from 
>> a voltage droop in the sustain stage. Much like a 
>> sample&hold. I didn't exactly believe that. Does anyone know 
>> of such a design? thanks, Michael.
> 
> Ages ago I had wondered what the long-term sustain behaviour the Doepfer
> A-140 would be, as in sustain the main cap discharges through the decay pot
> and the emitter of a PNP transistor whose base is held at the sustain level
> (wiper of a pot to the 12V rail), and it always struck me it would take a
> long time for the transistor to turn off, as the differential between the
> emitter and base very slowly decreases. So without further ado I set mine up
> and logged the sustain voltage over about 20 minutes - here's a plot of the
> result:
> 
> http://www.timstinchcombe.co.uk/synth/140_sustain_droop.gif
> 
> This was using the 'H' time setting (main cap is 100u); A, D, R all fully
> CCW; sustain set arbitrarily around 3V, and at the resolution of my
> Picoscope, it didn't really pick up the initial spike of the envelope.
> 
> Green trace is the 'full' envelope (against the right axis), showing that
> the voltage is slowly decaying. The red is an expanded vertical scale (left
> axis), showing greater detail - at some point I suppose that the envelope
> does actually drop below a level which causes the PNP to turn off (possibly
> one of the wobbles, but I didn't wait for the 140 to warm up, nor indeed the
> Picoscope, so there could be any manner of temperature drift effects in
> there...). One contributor to the droop will be the input bias current of
> the TL084 OA buffering the cap: the blue line is a straight-line fit to the
> end of the curve, and equates to about 5nA per second from the cap - this is
> a good deal more than the typical input bias current for the TL084, so other
> leakages are apparently present.
> 
> Whether this sort of droop is going to cause a problem probably depends on
> what time-frame any particular user is interested in.
> 
> Tim