The Trouble with Output Stages - 1 kΩ Resistors

[cardiad]

This post is going to describe what my problem is with output stages.
I will also propose a solution to this alleged problem.
I am very anxious to know what other synth-DIY enthusiasts have to say in regard to this issue.


I've noticed in many circuits, there is a 1 kΩ resistor between op amps and their output jacks.
I assumed this was to limit the current that could possibly be sourced or sunk by the output op amp.
This seems very important when working with 3.5mm or 1/4" jacks where the outputs often get shorted to ground while connecting and disconnecting patch cables.
I know I certainly wouldn't want to burn out my op amps from simply plugging and unplugging patch cables!
The 1 kΩ output resistor seems like a good idea at first glance.

Here are some examples of the 1 kΩ resistor on output stages:

Ken Stone's DC Mixer has a 1 kΩ resistor on its output:
http://cgs.synth.net/


Ray Wilson's Sequencer has 1 kΩ resistors on its outputs as well:
http://www.musicfromouterspace.com/forums.html?MAINTAB=SYNTHDIY&PROJARG=YOURANALOGSYNTH/page4.php?b=b&VPW=1910&VPH=803


The issue I have with this type of output is the following:
When you have an input module (or several input modules) being driving by an output module, their combined input impedance will set up a voltage divider with the 1 kΩ resistor from the output module.
The effective output voltage is attenuated by this voltage divider.
In other words, the magnitude of the signal is less than it should be.

One obvious solution to this is to remove the 1 kΩ resistor on the output of a module.
However, this is not an option if you want to protect your circuit from sourcing/sinking too much current.

Another solution is to make all inputs on every modules buffered with unity-gain op amp circuits that have really high impedance (on the order of mega ohms or higher).
But I've looked around quite a bit and I don't have not seen anyone meticulously putting unity-gain buffers on all of their inputs.
Most people seem to have roughly 100 kΩ of input impedance on most inputs, and 1 kΩ of output impedance on most outputs.

The following schematic shows what I am describing:


For Option 1, (the traditional 1 kΩ current limiting setup) the effective signal amplitude at the input of Module B is roughly 99% of its original value.

Moreover, if module A was driving more input modules (say, B, C, D, and E) the effective signal amplitude at those four modules would now only be 96.2% of the intended signal amplitude
(This is assuming each module has a 100 kΩ input impedance).
The effective output voltage signal will be attenuated further with each additional input module that is connected to it.

For audio signals (sine waves, triange waves, square waves) this might not be that bad. It might go completely unnoticed.
But I start to worry when I think about control voltages being attenuated.

Using Option 1 would make it hard to drive multiple modules with good CV tracking (1V/oct).
When a CV is at 0V, there is no issue, but when the CV output is at 4 V, a single input module can attenuate that to roughly 3.96 V.
If we are assuming a volt-per-octave scale, that would mean that the input module would see a CV that is half a semitone lower than it was intended.
When four input modules are attached (all with 100 kΩ input impedance) a 4 V CV will look like a 3.85 V CV - now nearly a whole tone lower than what was intended!


What I propose is moving the feedback path around. See the following image (Option 2)



For Option 2, There would be very little error in the output.
This is because the output stage is essentially setup as a non-inverting amplifier.
The op amp will use its very high gain to make sure the output voltage (at the jack) is the same as the input voltage (at the non-inverting op amp input)

Using this configuration might make it necessary to add an extra op amp for the output circuit on some modules.
But overall, it seems like simply moving the feedback path can greatly improve performance of the output stage.

Here are some things I've considered, but chose not to discuss as I feel they are non-issues:
- The gain of the op amp is set by the input impedance of the modules it is driving.
- The gain of the op amp will momentarily spike when the output is shorted to ground.
- A small value capacitor (100 pf) may be added between the output and the negative input to reduce oscillation in the feedback path, but most likely is not necessary.


I am very interested to hear what other synth-DIYers have to say about this output stage configuration.
Are there any obvious problems with this that I'm missing?
It is hard for me to believe that this output configuration hasn't been considered by others...


Thank you for reading.
Jensen

[blue hell]

cardiad

The 1k resistor should be in the feedback path, yes, or when the opamp is short circuit protected (to at least ground - and many are (like the popular TL07x and TL08x. although there is a restriction on shorting more than on out at the same time)) just leave it out.

[cardiad]

[elmegil]

A pragmatic instance of the problem:

I bought an Esplora Arduino -- built on a board like a game controller, with a joystick and a diamond of buttons -- and used it to create a simple controller for a synth. I used an unbuffered digital output for gate.

I think you see where that went. Almost the very first time I plugged and unplugged a cord....

I hacked my way around the problem, but my esplora is down one digital pin.

SO

I combined several things done wrong: no buffering, no output protection.

I have not seen the common 1K output resistor (usually outside the feedback loop) creating any problems in my mostly hand-built synth, when using it with 100K input resistors. 1/100 is 1% voltage drop from the divider aspect, and audio signals in particular aren't really that touchy. For CV....I could see that being a problem sometimes, but so far it hasn't turned up as an issue.


Edit:

Ok, naive boy will rescind his commentary above, other than the anecdote about what happens with no protection at all

Just read the Muff's variation on this, and the link on to the discussion of output resistors from a couple years ago, and I see the problem with CV. In fact I now have to revise a project I was working on to address this...

[Cynosure]

This is a very timely post. There was someone just last week having trouble with a MIDI to trigger PCB. For some reason, the PCB he purchased did not have any buffers on the outputs. The digital pins were connected directly to the outputs. The outputs weren't triggering the other modules properly. This could have ended similar to elmegil's project, but he was lucky and adding in TL074 buffers resolved the triggering problems.

I tend to error on the side of caution and add buffering and resistors to everything. However, I would really appreciate it if someone here on the forums who knows better could create a sort of buffering and protection guide. Something that shows the best ways to buffer, protect and remove bias for the following:

1. Audio Output
2. Audio Input
3. CV Output
4. CV Input
5. Trigger/Logic Output
6. Trigger/Logic Input

[elmegil]

Dr Sketch & Etch posted the following diagram in this thread on Muff's:

http://www.muffwiggler.com/forum/topic-64576.html

With some explanation. Basically these are his "go to" circuits for the four functions shown.

[Neil Johnson]

First off, welcome to the OP.

Now, the output resistor. Historically it was put there to allow a crude form of mixing simply by jamming multiple outputs together. Crude indeed.

For op-amp protection 1k is way overkill, and most op-amps these days have full output protection already.

More importantly the resistor separates the op-amp output pin from capacitive loading effects from cables which could cause the op-amp to oscillate. For that you don't need 1k, maybe 47-100 Ohms would be enough.

However, for CV outputs you want something much much lower, maybe 1 Ohm or less, and for that you put the resistor inside the feedback loop as shown above. But that simply trades one problem for another. To solve it properly you need an additional capacitor between the output pin and the -Ve input pin, typically 22-100pF. I did a post on MW some time ago which showed the effect of various output configurations and it was clear that the small capacitor is needed as well.

Neil