Issue combining virtual Grounds

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Administrator
Joined 2007
Paid Member
Lets not start a riot over this ;)

The circuit I posted above (the sim) seems to meet the design brief of two separate DC coupled amplifiers running on a single rail. Each works correctly, whether independent or daisy chained together and with DC precision.

The one transistor amp above is just a classic positive ground implementation. Nothing wrong with that.
 
There is no DC offset at either output.
Of course not... you tied the inputs reference resistors to the GND node which is a black hole in LTspice. When simulating virtual gnd circuits, it's good to tie that node to the circuit through a resistor connected to the negative rail.

In any case, in this circuit proposed by Peter Pan, the rail splitter opamp in the original device is basically also driving the virtual ground of the slave device. The rail splitter opamp in the slave device is basically a nuisance, best removed. A passive rail splitter will be better. The first opamp won't have to drive 50r but much higher resistors.

We still need the passive divider to establish a rough reference if signal isn't connected but power is.
 
Last edited:
Administrator
Joined 2007
Paid Member
Thanks for your thoughts Ben. I'd have to see it drawn out to take it all in and for it to make sense... you don't just mean connecting a resistor from the ground point as it is now and taking that to the to the "neg" rail ? do you ??

The circuit should work OK for real though.
 
Re-reading my post, it comes as a bit rude. Sorry for that. And the first part is based on a bad look at your schematic. Please forget about it. Mea culpa and all that...

And I didn't see Peter Pan request in the first post (" I also want this second device to be usable with or without having its audio I/O connected to the audio I/O jacks of the first device"). Which invalidates the second point.

So I'll go bury myself somewhere... :dunno:
 
I'm hesitant to post again after my blunder but...

In the scheme proposed by Peter Pan in post 28, with the 50R resistor and the 0.1u cap, the "gnd" impedance is significantly raised. Wouldn't that introduce crosstalk issues when the slave device is operating by itself ?
 
The only place virtual grounds are used is in cmoys. Where the driven device has a fairly low impedance but the power is still negligible. It's possible to build a power amp with a VG, but it's almost never done.

Where a series of low power devices are tied together, by design the impedances are high and consequently coupling caps can be small and high quality or if DC coupling is considered necessary at all costs, then a dual-rail supply is used. The choice is: conventional dual-rail or AC coupled, not virtual ground. Where have you ever seen a preamp with a virtual ground, unless it be a headlamp doing jury service?

The OP appears totally immune to the suggestion that the device could be built without using a virtual ground, that it would be simpler, easier to understand, have a lower component count, be less problematic to implement and suffer no loss of quality. In fact I am convinced that he does not understand how to implement single-supply AC coupled opamp circuits, and this is why he clings so tenaciously to this topology.
 
Administrator
Joined 2007
Paid Member
Not if we take the virtual ground as an absolute point of reference. Noise and anything else is measurable from that point and moves about relative to that point but the ground itself is clean. Its all relative from where you measure from and from what point you decide will be a clean reference point.
 
I'm hesitant to post again after my blunder but...

In the scheme proposed by Peter Pan in post 28, with the 50R resistor and the 0.1u cap, the "gnd" impedance is significantly raised. Wouldn't that introduce crosstalk issues when the slave device is operating by itself ?

I appreciate that, but what you may have missed in my long rambling posts, is that I wanted the "slave" device to have the option of being used as a stand-alone device too. So in that case it would need its own 1/2V ground, and an active one is arguably better then a passive. This is especially true as I'm in the habit of using the 1/2V V-ground as the ground return for all circuit inputs and outputs. So the 50 ohms is a compromise.Its still better that just a passive V-Gnd, but still allows the two circuit grounds to be directly connected AND share a supply, without small ground voltage differences causing the two 1/2V Gnd op-amps to "fight". the concepts have proven themselves , so far, both on my bench, and in simulations Mooly has graciously run. But simulations and bench tests are no match for human creative thinking, nor a match for "Mr. Murphy". hence I discuss the principal so I could consider input from anyone imagining a situation where my method might cause a new problem, so i could address it.

By the way, you'll note that I seldom have bothered responding to many of Counter Cultures posts. He still does not grasp the essence of my question but apparent knows everything, and so he has proven unable to focus on any helpful responses or solutions. he does not possess your humility. This often happens when people have a lot of experience, and are no longer able to consider any new approach. Since, according to CC, my working solution is not done, though it has worked well for me for many years, his posts are not useful in this particular discussion.
 
Not if we take the virtual ground as an absolute point of reference. Noise and anything else is measurable from that point and moves about relative to that point but the ground itself is clean. Its all relative from where you measure from and from what point you decide will be a clean reference point.

I was a bit curious about those two 100uF caps at each end myself. In my original "master" circuit, I actually only had about 1uF, and had almost considered dropping it further and letting the OP amp do more of the filtering work. In fact I planned an even lower value bypass on the "slave" device, possibly just 0.1uF. The OP amp, after all, is supposed to respond quickly to load changes and maintain the 1/2V level, even without a capacitor. If anything, I'd have considered connecting the (1) feedback input on the OTHER side of the resistor to make it respond faster, but that would once again cause it to fight over any DC differential with the "master", so I'll leave that as is. But still, are these multiple 100uF capacitors you're suggesting possibly overkill?

Just for context, both of my devices will always be "driving" reasonably high-Z inputs.
 
I appreciate that, but what you may have missed in my long rambling posts, is that I wanted the "slave" device to have the option of being used as a stand-alone device too. So in that case it would need its own 1/2V ground, and an active one is arguably better then a passive.

No, actually I raised the point precisely because of the "stand alone" option. If you use that 50R resistor without using a biggish capacitor afterwards (as in Mooly's sims), the performance of the virtual gnd is significantly degraded and could lead to crosstalk issues. It's a bit difficult to say as we don't know if the "slave" device is stereo or mono, etc.

When the master and slave devices are connected, then the problem goes away as the rail splitter opamp in the master dominates (but it will have to sink most of the return currents, both from master and slave).
 
No, actually I raised the point precisely because of the "stand alone" option. If you use that 50R resistor without using a biggish capacitor afterwards (as in Mooly's sims), the performance of the virtual gnd is significantly degraded and could lead to crosstalk issues. It's a bit difficult to say as we don't know if the "slave" device is stereo or mono, etc.

When the master and slave devices are connected, then the problem goes away as the rail splitter opamp in the master dominates (but it will have to sink most of the return currents, both from master and slave).

Ah! yes.. thanks again! I appreciate that because the same thoughts crossed my mind. It is indeed a compromise. I think under the best circumstances, minimal capacitance is needed at the V-Gnd if I trust the action of the master OP-AMP, except to deal with what I'll call reasonable 'parasitic paranoia'. After all, if the "slave" device was destined to be either a slave or a stand alone, I wouldn't put that small resistor in at all. Its not a stereo device and the driven impedance is high, but there are several sources and inputs being switched around, so why risk ANY crosstalk at all? The PCB is still undergoing tweaks, and I'm getting inclined to consider a high quality Gold plated board jumper in place of (or maybe bypassing) that 50 ohm resistor. Since the slave is PRIMARILY intended as an "add on" accessory, I could just leave the jumper unconnected (but keep SOME capacitance). Should the user want to use the device for their own purposes, independent of my "master" device, it wouldn't be unreasonable to tell them to open the hood and move the jumper.
 
Administrator
Joined 2007
Paid Member
I was a bit curious about those two 100uF caps at each end myself. In my original "master" circuit, I actually only had about 1uF, and had almost considered dropping it further and letting the OP amp do more of the filtering work. In fact I planned an even lower value bypass on the "slave" device, possibly just 0.1uF. The OP amp, after all, is supposed to respond quickly to load changes and maintain the 1/2V level, even without a capacitor. If anything, I'd have considered connecting the (1) feedback input on the OTHER side of the resistor to make it respond faster, but that would once again cause it to fight over any DC differential with the "master", so I'll leave that as is. But still, are these multiple 100uF capacitors you're suggesting possibly overkill?

Just for context, both of my devices will always be "driving" reasonably high-Z inputs.

I would think 100uf is probably about right. The only place the virtual ground can go is toward either rail, be that as a slow change or hf instability or modulation caused by the load and signal. But of itself it is always unchanging... that's the bit that's hard to grasp. Because you measure and reference to that point, its everything else that moves about it. And its better (imo) to have that change slowly (bigger caps) than quickly. The opamps PSRR (power supply rejection ratio) is way better at low frequency than high.
 
... the virtual ground can go is toward either rail, be that as a slow change or hf instability or modulation caused by the load and signal. But of itself it is always unchanging... that's the bit that's hard to grasp...

You're damn right it's hard to grasp. If you can't express yourself coherently then say nothing at all. You pushed your way into this discussion, disagreed with me on an entirely spurious basis, and now this... Waffle. What is this, subjective electronics?
 
I would think 100uf is probably about right. The only place the virtual ground can go is toward either rail, be that as a slow change or hf instability or modulation caused by the load and signal. But of itself it is always unchanging... that's the bit that's hard to grasp. Because you measure and reference to that point, its everything else that moves about it. And its better (imo) to have that change slowly (bigger caps) than quickly. The opamps PSRR (power supply rejection ratio) is way better at low frequency than high.

But wait... lets imagine a simpler scenario: just one device, with one virtual ground, developed by a decent OP amp with voltage divider input, and still with a small capacitance on the input side voltage divider. Now the OP amp is a unity gain follower, so the input = the output, voltage wise. But current wise, the OP amp offers actual drive (current gain) to maintain that voltage, as best it can, in the presence of load. And in a very real sense, within the limits of the OP amps slew rate and drive capabilities, the OP amp "amplifies" the capacitance on the voltage divider (even though its curve would be linear if it did drift). So it seems to me adding so much capacitance at the output, is limiting the usefulness of that OP-AMP. Other then forcing a faster balance to 1/2V at power up, it doesn't seem to be being utilized.

Wouldn't you agree that an OP-AMP, as a virtual ground, is supposed to behave something like a +/- pair of voltage regulators. If that analogy holds, you generally don't place much capacitance on the output of a voltage regulator. As you said, the OP amp is going to be more effective at lower frequencies. That being said, the higher frequencies can be dealt with with just a small capacitor, no?
 
Administrator
Joined 2007
Paid Member
That is a perfectly valid argument and point of view.

I would say it would be worth testing that for real with a real opamp virtual ground generator. Use the "negative" rail as ground for measurement purposes (so the virtual ground generator is a just a Vcc/2 generator) and feed a signal (a small current, say -/+ 10 to 20ma) into and out of that ground and scope the residual. Then try that over say a 100 to 100kHz bandwidth.
 
I dunno if you're aware of this thread Mooly?

http://www.diyaudio.com/forums/parts/263594-opinions-pcb-layout-grounding-compromise.html

So at least one of these devices is intended as a low level switch for guitar, pedals etc. Which means that there's no requirement for large caps for DC blocking, because the impedances can be comparatively high. So low distortion is not a problem. Not difficult to achieve, not really required because it's a guitar.

He can't answer the question, why are the majority of guitar boxes capacitor coupled? But the answer is that they are single-supply. Which is what his switch should be.

I can't prevent you colluding at his ignorance, I can only deplore it.
 
I dunno if you're aware of this thread Mooly?

He can't answer the question, why are the majority of guitar boxes capacitor coupled? But the answer is that they are single-supply. Which is what his switch should be.

I can't prevent you colluding at his ignorance, I can only deplore it.

So the fact I didn't answer your question means I "can't" ? Really? For someone willing to site another's ignorannce, you sure do show yours! Did you miss my comment to another responder about your useless posts? I don't answer because you've proven any discussion with you to be mostly unproductive.

Maybe stop being a troll for a change. If someone is doing things in a different way, insisting they don't innovate is counter productive (Ironic considering your handle). Some of us are trying to innovate and find better ways to do things, rather than just cutting and pasting their way through design. Initially suggesting a different approach was helpful, but when I offer that my system has proven itself, ask for particular cases or circumstances to watch out for, or a good technical reason why the concept is flawed, you come up dry. So its YOU not addressing MY core question. And if anyone has proven they "can't" address a question or concept, it has been YOU.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.