Ideas needed to make a circuit ring - sorta

There are a lot of discussions in here about how to suppress overshoot and ringing, but here's a fun diversion--I want an analog lowpass circuit that overshoots and rings at a VERY LOW frequency.  ;D

Here's a diagram showing the desired input signal (blue) and output (red):



Vnorm is a calibrated precision positive voltage level.  Vfloor is a lower voltage level somewhere above ground (0v). The time (Tfall) it takes to decay after Toff would be approximately five seconds.  A simple RC discharge type of curve.  The turn-on transient is the kicker--it should slightly overshoot and ring at approximately 1Hz* (Fring) until it stabilizes to the level of Vnorm at some point in time (Tstable) maybe 5 seconds later.

I could probably synthesize this with a microcontroller and DAC, but I'd rather not, as the nominal voltage (Vnorm) needs to be rather accurate and I'm thinking there has to be an analog way to make this happen, since it happens in real life, albeit at much higher frequencies.   

Any ideas?

Thanks!

-Bob

*You read this right: one Hertz.

P.S. Can you guess what this is for?  ;D

But of course it doesn't happen like that if it is a simple second-order underdamped system.  The responses to a step are symmetrical.  You show a first-order response to a falling step and roughly a very underdamped second or higher order one to a rising step.  Also the rise time as drawn is faster than it would be with the ringing as shown---whether you could live with this or not I don't know.

So it's not a time-invariant linear system---but that's o.k.  ;D

I'd have two (or more) filter paths driven in parallel all the time, and simply switch between them in response to the step input.  Look at the first-order one for the negative-going transition and the underdamped higher-order one for the positive-going.  The same approach could be extended to making the rise time faster than the ringing periods, although it will be tricky to "splice" the waveforms together.

Is there a prize?

> The time (Tfall) it takes to decay after Toff would be approximately five seconds. ...turn-on transient.. should slightly overshoot and ring at approximately 1Hz* (Fring) until it stabilizes to the level of Vnorm at some point in time (Tstable) maybe 5 seconds later.

The analog circuits will NEVER decay to dead-zero. The EE will want the 63% decay for the simple curve and probably the Q or DF for the ringer.

Clearly {as bcarso beat me} you build a 1-pole low-pass and a 2-pole low-pass. And a switch to take one or the other. There are surely some elegant tricks possible, but we'd have to see the size of the motivational trophy (and some clue to voltage zone) before thinking that hard.

In this context, it is probably a synth ADSR attack/decay modulation, with a start-up wobble to fake the start of organ pipe or multi-string piano tone. Unless it is a tidal wave, sagging into the fault and then echoing across the harbor.

Pretty darn close in the guesses there, PRR. 

Vnorm is somewhere around 10 volts. Vfloor will be set by ear somewhere between 2 and 5 volts. It's a control voltage to a varactor diode in a Colpitts oscillator. At Vnorm it runs at exactly 3.998720 Mhz  Divide that by 284, then by 25 for one clue.  The other clue: think "synchronous" and "flywheel".   ;D

I got the inspiration to do this when an AC power connection to the power supply to a 12AX7A+2*6V6GT circuit I added to this device went intermittent. When I heard the distorted fade out and fade in that resulted from that, I figured I'd take the effect the rest of the way! ;D

-Bob

PRR said:

Is there a prize?

> The time (Tfall) it takes to decay after Toff would be approximately five seconds. ...turn-on transient.. should slightly overshoot and ring at approximately 1Hz* (Fring) until it stabilizes to the level of Vnorm at some point in time (Tstable) maybe 5 seconds later.

The analog circuits will NEVER decay to dead-zero. The EE will want the 63% decay for the simple curve and probably the Q or DF for the ringer.

Clearly {as bcarso beat me} you build a 1-pole low-pass and a 2-pole low-pass. And a switch to take one or the other. There are surely some elegant tricks possible, but we'd have to see the size of the motivational trophy (and some clue to voltage zone) before thinking that hard.

In this context, it is probably a synth ADSR attack/decay modulation, with a start-up wobble to fake the start of organ pipe or multi-string piano tone. Unless it is a tidal wave, sagging into the fault and then echoing across the harbor.

Click to expand...

stickjam said:

I could probably synthesize this with a microcontroller and DAC [...] It's a control voltage to a varactor diode in a Colpitts oscillator. At Vnorm it runs at exactly 3.998720 Mhz

Click to expand...

That's a lot of significant digits for something as temperature sensitive as a varicap, especially if it's tuning an LC tank (as opposed to a crystal).

If you want/need stability you could take a microcontroller and a DDS; several DDSes can be programmed to do frequency sweeps in wild and wonderful ways as that makes the sonar and instrumentation types very happy.

JDB.

AD9911.  I've used this part in my designs a few times and I'm in love with it.

Although I'm not 100% against a software solution, I'm thinking there has to be an analog electronic equivalent to this originally-mechanical system:  a switch in series with a synchronous AC motor that has a flywheel on its shaft.  While the switch is closed, the RPM's are locked to the AC line frequency.  When the switch is opened, the flywheel's inertia causes a slow decline in RPMs.  Closing the switch again, the motor's strong torque fairly quickly brings the flywheel back up to speed, however, the accellerating flywheel slightly overshoots the normal operating speed.  The motor produces a braking effects that eventually bring the system to the proper RPMs after "ringing" at a frequency determined by the mass of the flywheel and damped by the motor's torque. 

While waking up this morning I considered that maybe the true analog equivalent is a really sluggish phase-lock loop?

The audible result of this can be found after the 2:00 mark in this song: Stagnation

stickjam said:

Pretty darn close in the guesses there, PRR. 

Vnorm is somewhere around 10 volts. Vfloor will be set by ear somewhere between 2 and 5 volts. It's a control voltage to a varactor diode in a Colpitts oscillator. At Vnorm it runs at exactly 3.998720 Mhz  Divide that by 284, then by 25 for one clue.  The other clue: think "synchronous" and "flywheel".   ;D

I got the inspiration to do this when an AC power connection to the power supply to a 12AX7A+2*6V6GT circuit I added to this device went intermittent. When I heard the distorted fade out and fade in that resulted from that, I figured I'd take the effect the rest of the way! ;D

-Bob

PRR said:

Is there a prize?

> The time (Tfall) it takes to decay after Toff would be approximately five seconds. ...turn-on transient.. should slightly overshoot and ring at approximately 1Hz* (Fring) until it stabilizes to the level of Vnorm at some point in time (Tstable) maybe 5 seconds later.

The analog circuits will NEVER decay to dead-zero. The EE will want the 63% decay for the simple curve and probably the Q or DF for the ringer.

Clearly {as bcarso beat me} you build a 1-pole low-pass and a 2-pole low-pass. And a switch to take one or the other. There are surely some elegant tricks possible, but we'd have to see the size of the motivational trophy (and some clue to voltage zone) before thinking that hard.

In this context, it is probably a synth ADSR attack/decay modulation, with a start-up wobble to fake the start of organ pipe or multi-string piano tone. Unless it is a tidal wave, sagging into the fault and then echoing across the harbor.

Click to expand...

Click to expand...

Thinking about the PLL concept more, I've got a few CD4046 chips in my stash (The right Vdd supply is already there: this organ is full of CMOS).  My first "brain fart" looks something like this...



The only thing is, I have a good suspicion that it might not lock on to the input frequency with the VCO low-pass filter set to be so sluggish.  Does it need of some sort of bootstrap to keep it from locking on to some evenly divisable frequency?  I'm also wondering whether at Fin near 4Mhz, I won't get the overshoot*; that it'll crawl up to Fin and lock in too fast regardless of how slow I make the low-pass filter.

My understanding of PLLs is not great, so I'd appreciate any feedback on this, as it seems like the most obvious electronic metaphor for the motor/flywheel system.

Thanks

-Bob

*there's nothing like a normally-desirable characteristic being considered a defect. 

BTW, yes I know I could just implement some sort of pitch-bend wheel, but with the requisite center dead-zone, it's all but impossible to get the same effect.  I'd rather keep the performance control the same as the original: hit the power switch.  I already have  implemented a relay that keeps power going to the solid state circuits for up to 10 seconds after turning the organ (and the tube stage's supply) off.

Following a stupid late night idea I tried lowpassing the falling flanks of the input signal so they do not kick U1/U3 into ringing as much the rising flanks do -- with mixed results.

If the circuit is unstable enough for the values of T_stable you want (it is actually pretty hard to intentionally design something unstable without a firm grasp of control theory  ), the overshoot is too big and the falling flanks are messed up as well, so this €1-approach does not work. T_rise <  1/F_ring is not met, either.

Yes, I put together a model in Spice last night and saw that.  I used a gyrator and 'lytic to produce a 1Hz LC tank.  It wouldn't ring until I all but eliminated the lowpass filter before it, feeding it a square wave.  The overshoot was impressive (nearly 2x V_norm)

That's when I decided that maybe the purely lowpass solution wasn't the correct metaphor, thus the phase locked loop thought above.

-Bob

EZ81 said:

Following a stupid late night idea I tried lowpassing the falling flanks of the input signal so they do not kick U1/U3 into ringing as much the rising flanks do -- with mixed results.

If the circuit is unstable enough for the values of T_stable you want (it is actually pretty hard to intentionally design something unstable without a firm grasp of control theory  ), the overshoot is too big and the falling flanks are messed up as well, so this €1-approach does not work. T_rise <  1/F_ring is not met, either.

Click to expand...