Fall, Rise and Staircase up/down waveform generator?

[anotherjim]

Looking at this again...
Oscillator. Minimise currents by using as large a speed control resistance you can get. 1M pot is a good choice. Prevent excessive speed with a series resistor in line with the pot. No point going too fast. 1/10th the pot value is a good choice but might be best anywhere 10k to 100k.
For the speeds you want, where is the speed pot set? If possible, reduce the 1u integrator cap. Compared to an LFO, the sequencer clock will run faster. x8 faster for 8 steps so you ought to be able to have that cap smaller.
To minimise clock ticks, you want the integrator speed set by the smallest C and the largest R that get the speed range you want.
Choose a low power opamp. TL062 is a good choice.
Don't leave out the power supply bypass capacitors. We don't show these in circuit ideas/snippets for clarity. Any IC making or using square waves or pulses needs ceramic bypass caps at the power pins.

Sequencer.
I missed this one. Pin 13 should be tied to 0v. As drawn with the 100k in between (not needed) it just might work, but only due to internal propagation delay and can't be reliable. Pin 13 is actually a Clock Inhibit as its positive logic - it does what it's called with a logic 1 applied. I suppose they call it Clock Enable as abbreviating Clock Inhibit confuses with Clock In.

[strungout]

I made a mistake on the schematic, pin 13 on the 4017 is indeed grounded.

I made some more changes:
-tied all digital grounds back to the DC jack (as close as possible on my BB).
-2M rate pot in series with a 47k. Got the capacitor down to 82n. Satisfied with the speed range.
-changed the pulse generator to a TL022. Was using an LM358.
-9V regulator to feed both the TL022 and 4017.
-100n bypass ceramic caps on TL022 and 4017.
-100r and 100u filter for the TL022 and 4017.

With no discernable improvement. I don't know what else to try. And I'm afraid adding the extra 4 stages to the phaser will just add more noise...

[ElectricDruid]

How much noise is *going in* to the phaser? Could the source be *before* the circuit?

One thing about any sort of cyclical effect (even if that's a repeating pattern of steps) is that it makes any background noise very obvious. Any hum or hiss going into a phaser or flanger becomes a very obvious "WHOOSHWHOOSHWHOOSH" on the way out.

It certainly sounds like you've tried all the tricks as far as reducing LFO clock ticking goes. So yeah, I agree, more stages will probably make it worse. Time for a 571 or noise gate?!?

[Rob Strand]

With no discernable improvement. I don't know what else to try. And I'm afraid adding the extra 4 stages to the phaser will just add more noise...

Making those changes is part of the way forward.  There's *many* possibles sources for these issues and by doing your best you remove a lot of maybes from the list.   Without doing that you have a list of maybe's and no way to pry-off the cause.

There's has to be a problem still lurking there.

The wire from the output of IC1 to the rate pot is a source of noise.  You want to keep that wire short and away from any audio.  Also if you put resistor in series with the rate pot to limit the maximum rate the resistor is best place right at the output of IC1, then to the rate pot.   You actually don't want *any* LFO signals or wires near any audio.

As far as narrowing down the tick issue.  For debugging it's a good idea to disconnect the LFO from the rest of the circuit.  Just set-up that fixed trimpot + cap idea feeding where the LFO connects.   With the LFO not connect we can't blame any noise and ticks on the modulation process.   The problem has to come from the power, grounds or couple through the air.

There's a few techniques for finding issues:
- listen to points is the circuit, even the digital ones and power using an audio probe idea.

- A very simple technique is to touch points or parts in the circuit with your finger.
  This is a hack technique but it's extremely successful in weeding out complex audio problems.
  The idea is to find patterns in cause and effect.
  Touching points which are audio might identify points which let in the noise.
  Touching points which are digital makes the digital signal couple more easily into the audio.
  You can also touch audio points with one finger and digit points with another.

- Simplify the circuit.  Strip out parts of the circuit to see what parts give rise to
  certain problems.   For example on the phaser, disconnect the output of the phase-shift section
  from the output mixer.  Or disconnect output to input connection between the phaser stages.   
  Best to ground opened input.
 
  Start at the output end.  Reconnect the break then do the same thing one stage before.
  A caveat is breaks and ground input mustn't stuff-up the DC biasing of the circuit - you can connect
  a cap big from the input to ground or Vref.  You might even find connecting to ground through the cap
  does something different to connecting to Vref - that's a sign of a problem with Vref or ground. 

  Try to take note if things get better, worse or stay the same.

[anotherjim]

Maybe a little point is I think you should have a pull-down resistor on the input to the buffer IC3. At least x10 of a sequence pot value. What value are those pots?
IC 3 & 4 are an LM358?

[strungout]

Rob: I'll try those suggestions. I'll also rearrange the breaboard a bit, keep the sensitive parts further away from the ticking stuff. When I build this, it'll be in a big box with three boards: the phaser + extra stages and the sequencer. I should be able to keep the two parts well away from each other.


Jim: The controls are 10k. I used a 100k pull-down resistor at IC3's input. Think it helped a bit. Less hiss. IC1-2 were the LM358, replaced by the TL022. I was using a 4558 for IC-3-4, but I changed it to a TL072.

Anyway, gotta go tinker.

[Rob Strand]

I'll also rearrange the breaboard a bit,

If your breadboard has a metal base try connecting the metal to circuit ground.  Sometimes it can stop a lot of weird interference and noise.  It rarely does any harm.

[anotherjim]

TL072 could be the worst choice for the job. The pot outputs voltage range from 0v to 9-0.7v (the sequence diodes volt drop). This exceeds the input range of the opamp and it can't output that range either. It's also an unnecessarily fast opamp.
Since you have LM358, I don't understand why not try it?
The 358 isn't quite perfect. It can input and output right down to 0v but at the upper end, it limits at +supply-1.5v. This is still better than the TL072 and can be fixed near enough by adding a diode in series with the sequencer diode commoned output so there's x2 diode volt drops into IC4.
An additional worry with the TL072 with voltages near the limits of its input and output range is that the filter won't work properly.

[ElectricDruid]

An additional worry with the TL072 with voltages near the limits of its input and output range is that the filter won't work properly.

+1 agree. I'Ve run into this feeding a 0-5V signal from a PIC through a TL072 filter running on 0-9V. The poor op-amp can't handle stuff that close to the ground rail and goes wappy.

[anotherjim]

And another little thing with the Filter stage - do you need the negative feedback resistors? A simple wire from output to -input should suffice?

[PRR]

> do you need the negative feedback resistors?

They are a factor in the filter design. The brain-pain to re-compute the filter as unity gain is never worth it in parts cost. I used to do that for complex filters where normalizing to unity gain was worth the slip-stick wear (or charging the TI).

[anotherjim]

Rf/Rin = 1k/100k. Is the filter x1.01 gain fussy?

[Rob Strand]

The given filter should not have gain resistors.  It's a unity gain design.

Non-inverting filters are quite sensitive to the gain.  Even  a 1% gain increase will raise the Q, making the frequency response more peaked and increasing the overshoot.  For the current filter the gain causes the response to peak about 0.5dB above what it supposed to be (surprised?).

You can compensate by decreasing R2 value but that's fixing the wrong problem!

Anyway, I mentioned the gain resistors earlier in the thread.

[PRR]

> Rf/Rin = 1k/100k. Is the filter x1.01 gain fussy?

Did not see that. Correct, no filter built in this crowd can be that fussy. Must be an un-rounding error/slop.

[strungout]

Jim: Actually the 1k/100k has been removed. I was trying stuff out cause I didn't know what to do with the inverting input, but you told be to just tie it in a feedback loop.


I tried the LM358. No-go. Causes this weird 'blip' at the start of the sequence. I'm getting the feeling I should use a rail-to-rail op-amp. Any suggestions?

Re-arranging the board didn't do much, if anything, noise-wise. No metal base.

I isolated the phaser from the rest of the circuit, using a 100k pot to control it, and it is noisy on its own. I have a built Ross phaser, from tonepad pcbs and it is also noisy, but a bit less.

In that phaser, there are four 10k to ground for each stage. When I touch the grounded lead, the noise is significantly reduced. Might be a problem there. Haven't found one yet. http://www.tonepad.com/getFile.asp?id=99

I'll go work on it some more.

Current schematic:

[anotherjim]

Have you lost the pulldown resistor from the sequence diodes? You have to have one or else the selected diodes are trying to forward bias into the high input impedance of the opamp which has no reference. Is that why the LM358 misbehaved?

[Rob Strand]

Re-arranging the board didn't do much, if anything, noise-wise. No metal base.

There has to be electronics design issue somewhere.

I tried the LM358. No-go. Causes this weird 'blip' at the start of the sequence. I'm getting the feeling I should use a rail-to-rail op-amp. Any suggestions?

Have you lost the pulldown resistor from the sequence diodes? You have to have one or else the selected diodes are trying to forward bias into the high input impedance of the opamp which has no reference. Is that why the LM358 misbehaved?

I think Jim's on the right track.   Having the opamp inputs float or go to zero is going to be an issue.   Also the LM358 needs some sort of path to ground for it's bias current.

A second issue is even though the LM358 can handle it's input going to ground, the circuit is operating as a buffer.   The LM358 output cannot swing *all* the way to ground.

Can I suggest placing say 47k to ground on the input of the first LM358 then add say a 1M from the LM358 input to the +9V rail.   That will bias everything up.  Like you might be able to change the 47k to 22k or something but let's start somewhere out of harms way.

I isolated the phaser from the rest of the circuit, using a 100k pot to control it, and it is noisy on its own. I have a built Ross phaser, from tonepad pcbs and it is also noisy, but a bit less.

In that phaser, there are four 10k to ground for each stage. When I touch the grounded lead, the noise is significantly reduced. Might be a problem there. Haven't found one yet. http://www.tonepad.com/getFile.asp?id=99

That at least lets you know what noise to expect from the phaser section itself.   Yes, there may a problem lurking in there.

How are you generating Vref?

[anotherjim]

As a buffer, the LM358 output goes to about 0.8v with input 0v, at least mine do and used as amplifiers do swing down to 0v. I suppose it's something to do with the -input bias requirement? Anyway, do you care about 0.8v?

All the readily available rail to rail opamps I'm familiar with are CMOS and are actually only 3/4 rail to rail. The inputs are limited in positive swing - see datasheets. TLC272, TLC2262, LMC662.
Without an input pull-down resistor, these CMOS parts will be worse than anything else you've tried. Very high input impedance = no forward bias current for the selected step diode except for the reverse leakage current of the other 7 diodes.

[ElectricDruid]

Anyway, do you care about 0.8v?

For a sequencer output, you probably do. As a Pitch CV, that's almost ten semitones. In a synth circuit though, we'd have a bipolar supply and this "hitting ground" problem wouldn't arise.

Here, it may not matter at all. Could we stick resistors top and bottom on each pot to keep the pot output within the range the op-amp can output? It'll just offset the phase frequency, but there's loads of ways to compensate for that.

[PRR]

All the readily available rail to rail opamps I'm familiar with are CMOS and are actually only 3/4 rail to rail. The inputs are limited in positive swing....

LM324 output will pull-down to <0.1V fine. It's pure BJT, not that I care.

Yes, at 1V/Oct that is more than a semitone, and yes it loses 1.4V up top.

Inputs will pull a hair below V- before control is lost.