Op amp providing virtual ground

[Rob Strand]

OK I see, it's just a misunderstanding.

The input side is high impedance and I was told that this helps to reject noise from power supplies.

It's not the high impedance that helps reject noise.   It's the presence of the capacitor to ground on the voltage divider.  That applies to both the simple resistor base Vref circuit and the opamp based Vref circuit.  The opamp Vref circuit is just a buffered version of the resistor version.

Using higher impedance on the Vref *voltage divider* lets you use smaller caps to get the same amount of supply filtering.  The side effect of that is the higher impedance makes it more subject to loading on Vref, and also noise due to loading on Vref - how the noise gets there is very circuit specific but it is cause by the circuit itself.

When you analyze Vref circuits important to separate the effects of DC loads on Vref and AC loads on Vref.

Not all circuits load Vref so you could see 1M resistor in some cases.   For Boss pedals they power the buffers from Vref circuit, and Vref goes to many places in the circuit, so a value of 10k is better.

The thing about most pedals is the *ac* loads on Vref are small so putting a cap on Vref will make the Vref impedance low enough to prevent AC signal connected at one point on Vref getting into the audio of other points connected to Vref.

The output impedance is low so you can use lower value resistors to VREF to cut down on noise as well. I'm not at all considering the circuit being used in an audio path, only as a VREF virtual ground on a single ended supply.

The low impedance itself doesn't lower noise.   However, there are cases when it can help reduce noise.   If you have a circuit which is loading down Vref.

as ElectricDruid says, most pedals are trying to stay dirt-cheap and don't want an extra opamp. even so there are lots of pedal designs that use that circuit. but they usually aren't drive pedals, since those rarely have a noticeable current draw. certain pedals with LFOs can really benefit from the buffer Vbias though.

It's not always about being cheap.  It's about stopping problems in specific cases.  Adding an opamp based Vref to noise critical audio circuit is actually bad as the Vref opamp generates noise and that noise injects directly into the audio which reduces the signal to noise by 3dB.   I would so far to say an opamp based Vref is a bad design choice for low noise circuit. RG and I have discussed opamp based Vrefs many times on the forum.   You can remove some noise but only at the expense of making the Vref impedance go up - which might not be desirable for some circuits.

If we go back to the resistor based Vref circuit the Vref cap is only effective for audio frequencies.   If you connect low frequency AC load to Vref like an LFO that cap in very ineffective in keeping Vref constant as a result you hear ticks and thumps in the audio.  In fact the it acts more like a DC load on Vref.  What happens is the load on Vref causes the Vref voltage to rise and fall.     One solution, which is used by Boss, is to use a separate Vref for the LFO.   Any small Vref variations are OK for the LFO.   It's a good solution as it keeps the ticks and thumps away from the audio altogether.

The biggest distinction between the resistor base Vref and the opamp based Vref is the *DC* output impedance.  If you did want to put the audio and LFO on the same Vref then opamp based Vref would definitely help reduce the ticks or thumps as opamp output impedance is low and we aren't relying on the big Vref cap to keep Vref constant.

If you look a some compressor designs they draw high current from Vref and the opamp based Vref can help here too.

What if you have low noise circuit and high AC and/or DC loads on Vref?

IMHO the best solution is to use an opamp Vref for the loaded circuit and tap off Vref *before* the opamp for the low noise circuit.   You still need big enough cap on the resistor side.   That is the best of both worlds.   The only thing you need to consider is that the opamp Vref could be offset a few mV from the resistor side Vref - not many audio circuits will care anyway.

RG and I discussed other aspects of the opamp Vref in old posts.

[CheapPedalCollector]

Thanks Rob that is a great explanation and I understand now (I think).

So I could do as you say and maybe fix the ticking/thumping in a phaser circuit like the phase 45/90 by using a buffer for the LFO and the audio off the divider?

Would it offer any help for power supply oscillation issues? What about voltage doubler/charge pump noise?

[Rob Strand]

Thanks Rob that is a great explanation and I understand now (I think).

So I could do as you say and maybe fix the ticking/thumping in a phaser circuit like the phase 45/90 by using a buffer for the LFO and the audio off the divider?

Would it offer any help for power supply oscillation issues? What about voltage doubler/charge pump noise?

It will only fixed the ticks if the problem is caused from Vref.  There's many other causes, like noise on the power supply itself, and that can be cause by the LFO putting noise on the power supply.

Some of this stuff is like a leaky boat.   There's plenty of places for it to leak but the only time it doesn't sink is when *all* the holes are plugged up.   If you don't know where the holes are the problem is finding where the hole(s) is (are).   Sometimes not that obvious as there are multiple causes.  You can only create a list of suspects and work through them.

[amptramp]

One thing to point out is that the op amp usually cannot drive an electrolytic capacitor and its output impedance increases with frequency, so if you want to filte the output, you need a resistor between the op amp and the cap.  If it is not filtered, the op amp noise appears at the output and since it is set for unity gain, if you have 2 mV of noise referred to the input then you will have 2 mV at the output.

[antonis]

What Ron said..

https://www.allaboutcircuits.com/technical-articles/how-to-drive-large-capacitive-loads-op-amp-circuit/

https://www.analog.com/en/analog-dialogue/articles/ask-the-applications-engineer-25.html

https://www.analog.com/en/analog-dialogue/articles/techniques-to-avoid-instability-capacitive-loading.html

https://training.ti.com/ti-precision-labs-op-amps-stability-capacitive-loads

https://ww1.microchip.com/downloads/en/appnotes/00884a.pdf

https://www.st.com/resource/en/application_note/an2653-operational-amplifier-stability-compensation-methods-for-capacitive-loading-applied-to-ts507-stmicroelectronics.pdf

[Rob Strand]

One thing to point out is that the op amp usually cannot drive an electrolytic capacitor and its output impedance increases with frequency, so if you want to filte the output, you need a resistor between the op amp and the cap.  If it is not filtered, the op amp noise appears at the output and since it is set for unity gain, if you have 2 mV of noise referred to the input then you will have 2 mV at the output.

Those were some of the things that came up in the old threads with RG and myself.

[pokus]

Thanks for all of that kind of great feedback.

I've looked at the Boss BD2 schematic and noticed the 47u cap at the output of the vref providing op amp.
As some mentioned that driving a capacitive load directly from the op amp isn't a great idea, I wonder what could be a good value for the resistor in series with the op amp out to the cap to ground?
Assumed that my circuit connected to this is uncritical for output impedance up to let's say 1k ohm. Would a 220 ohm resistor and a 22u cap be a proper choice?

[FSFX]

As some mentioned that driving a capacitive load directly from the op amp isn't a great idea.

In this case the op amp is acting as a DC amplifier. Most op amps have a finite output impedance which provides current limiting when initially charging the capacitor.

[merlinb]

I've looked at the Boss BD2 schematic and noticed the 47u cap at the output of the vref providing op amp.
I wonder what could be a good value for the resistor in series with the op amp out to the cap to ground?

Do not use a cap at the output of the opamp, the opamp's output is already better than any cap. The cap makes things worse. Boss was wrong too.

[antonis]

Boss was wrong too.

They probably had too many caps on the self..
(can't see any reason for C3 & C5 existence before and after Q2/C4 capacitance multiplier..)

[PRR]

Do not use a cap at the output of the opamp, the opamp's output is already better than any cap. The cap makes things worse. Boss was wrong too.

Yes. As you say, it is not about the opamp being harmed. It is that there are resistors and capacitors everywhere, too many R&C makes an oscillator, but the opamp designer very-carefully gave you a part which doesn't oscillate, IF you don't add more R&C to it.

Let's see a picture. Here is our old friend '741. Other opamps may be "better" in some way but most don't work "different".


Bottom circuit, green dot and trace, is the open-loop (no feedback) frequency response. It is flat(ish) from deep DC to about 10Hz. Then an internal RC kicks-in and it rolls-off "one pole", 6dB per octave. 90deg phase-shift over most of this range. At 1MHz a second RC (probably an output device) kicks-in another second RC, so above that it rolls-off 12dB/octave approaching 180deg phase-shift.

Top circuit, blue dot and trace, shows '741 with unity gain NFB. The 10Hz pole is effectively slid-up to 1MHz. Closed-loop response is flat to 1MHz. Then the second pole kicks-in and response drops at 12dB/octave approaching 180deg phase-shift. But the gain falls faster than phase-shift rises so there is only a mild bump.

Center circuit, red dot and trace. I have added a somewhat pedal-typical 220r+470uFd inside the loop, "to filter crap". Two-pole. Now we approach 180deg phase shift at 1,100Hz, opamp still has gain of 900, we get a BIG bump-up resonance smack in the middle of the audio range. The hiss in the system is amplified over 100X, an impure mosquito whine. (The exact frequency is very variable, depending; but parts near these values will do this thing.)

It won't always be this bad. Capacitor ESR is our friend!! A fat can from 1981 may hardly sing (or filter), while new large ceramics are like a steel flute: don't damp hardly at all.

[Rob Strand]

Yes. As you say, it is not about the opamp being harmed. It is that there are resistors and capacitors everywhere, too many R&C makes an oscillator, but the opamp designer very-carefully gave you a part which doesn't oscillate, IF you don't add more R&C to it.

The standard way to deal with the oscillation issue is by bypass the output feedback at high frequencies using an RC network around the opamp.   This applies to your "red" dot example.

There's plenty of articles on opamps capacitive loads, however this one does a good job for the type of circuit we have here,

https://vdocuments.mx/hello-and-welcome-to-part-six-of-the-ti-precision-labs-on-.html?page=1

I'm pretty sure RG and I discussed this in the past.

[dgeez99]

An added comment re: using an ordinary opamp for rail splitting, as opposed to a purpose-built virtual-ground device such as the TLE2426.  The concern here is on stability, particularly with regard to the opamp's ability to drive a capacitive load - so you really should look carefully at the load impedance this opamp would see and check that against its specifications. 

Note that even the TLE2426 datasheet specifies an instability region with regard to load capacitance, which can be readily overcome by adding extra capacitance.  It's important for this discussion to note that this approach doesn't always work with all ordinary opamps.

Here are two short articles on this that may be useful on the topic, I picked these because they're pretty simple - #2 is from an engineer at TI, the company that makes the TLE2426..

1. https://www.electronicsweekly.com/market-sectors/power/design-ideas-practical-ways-split-voltage-rail-2015-05/

2. https://www.electronicsweekly.com/market-sectors/power/ten-best-ways-split-voltage-rail-2015-02/

Cheers!