Sorry if this gets long... I don't have much talent for brief explanations.
I have a circuit that uses several low voltage rail to rail op amps , in a 5V single ended design. In that circuit I made a virtual ground from a spare op amp, configured as a simple unity gain non inverting follower with its + input fed by a simple voltage divider. I also use a small cap between the midpoint of the divider and 0V. So the output of this op amp is my circuit wide V-gnd... plain and simple... no added series resistor, no shunt capacitor. All other op amps that need a ground reference do so though large resistances (100K or 1 meg), and in all situations where an output of any OP amp might ever be shunted close to A-Gnd, there is a 1K resistor (this avoids any issues with small offset voltages fighting for dominance). The circuit works flawlessly and is quiet as a mouse and for better or worse, all the grounds on the ins and outs of this circuit are all connected directly to the virtual A-Gnd. I don't know if anyone else does that but I did. As long as this device has its own independent wall transformer supply, it should not matter what reference point becomes ground. And of course all I/O points have a decoupling capacitor anyway, and all gnd runs meet up at a single point. .
But now I have a dilemma. I have created a new external "add on" device and I'd like to power it from an "AUX 5V" output jack I put on the first device. This 2nd device will also have some op amps of its own, and so will also need a virtual Gnd. 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. So now I have two units powered by the same supply, each needing an independent V-Gnd. Obviously I can't assume the DC ground voltage from 2 independent V-GNDs will be absolutely identical. So what I'm thinking of doing, is have the V-Gnd in the second device look like this V-Gnd variation...
Lets say R3 is maybe just 50 ohms. If the 2nd device is operated independently, this V-Gnd circuit should be pretty close to a true reference. I'm honestly not sure if doing so would require an extra capacitor for C2, but it probably couldn't hurt. If all the resistive OP-AMP circuit connections to this ground are up above and beyond 100K as I did in the first circuit, any cross contamination of signal audio between these circuits should be very very small, though not "zero", as in the first device, where there is neither an R3 or C2. So what I'm hoping to gain through that small sacrifice is that if the grounds of the two devices do get connected together, any tiny difference between the two analog grounds will be quietly dissipated across R3. Even an unimaginably high 1/2 volt error between the V-Gnd of the 2 devices should only result in a tolerably small 10mA dropped across R3. (I've never seen more than a 10mV difference on the bench so far). But in that case the more direct V-Gnd from the first device (that has no such resistor) will "prevail".
Now on the bench, so far, I've not seen any issues doing this. But is there some terrible gotcha I'm missing here or a better approach I should take? Any thoughts or opinions would be hugely appreciated!
I have a circuit that uses several low voltage rail to rail op amps , in a 5V single ended design. In that circuit I made a virtual ground from a spare op amp, configured as a simple unity gain non inverting follower with its + input fed by a simple voltage divider. I also use a small cap between the midpoint of the divider and 0V. So the output of this op amp is my circuit wide V-gnd... plain and simple... no added series resistor, no shunt capacitor. All other op amps that need a ground reference do so though large resistances (100K or 1 meg), and in all situations where an output of any OP amp might ever be shunted close to A-Gnd, there is a 1K resistor (this avoids any issues with small offset voltages fighting for dominance). The circuit works flawlessly and is quiet as a mouse and for better or worse, all the grounds on the ins and outs of this circuit are all connected directly to the virtual A-Gnd. I don't know if anyone else does that but I did. As long as this device has its own independent wall transformer supply, it should not matter what reference point becomes ground. And of course all I/O points have a decoupling capacitor anyway, and all gnd runs meet up at a single point. .
But now I have a dilemma. I have created a new external "add on" device and I'd like to power it from an "AUX 5V" output jack I put on the first device. This 2nd device will also have some op amps of its own, and so will also need a virtual Gnd. 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. So now I have two units powered by the same supply, each needing an independent V-Gnd. Obviously I can't assume the DC ground voltage from 2 independent V-GNDs will be absolutely identical. So what I'm thinking of doing, is have the V-Gnd in the second device look like this V-Gnd variation...
An externally hosted image should be here but it was not working when we last tested it.
Lets say R3 is maybe just 50 ohms. If the 2nd device is operated independently, this V-Gnd circuit should be pretty close to a true reference. I'm honestly not sure if doing so would require an extra capacitor for C2, but it probably couldn't hurt. If all the resistive OP-AMP circuit connections to this ground are up above and beyond 100K as I did in the first circuit, any cross contamination of signal audio between these circuits should be very very small, though not "zero", as in the first device, where there is neither an R3 or C2. So what I'm hoping to gain through that small sacrifice is that if the grounds of the two devices do get connected together, any tiny difference between the two analog grounds will be quietly dissipated across R3. Even an unimaginably high 1/2 volt error between the V-Gnd of the 2 devices should only result in a tolerably small 10mA dropped across R3. (I've never seen more than a 10mV difference on the bench so far). But in that case the more direct V-Gnd from the first device (that has no such resistor) will "prevail".
Now on the bench, so far, I've not seen any issues doing this. But is there some terrible gotcha I'm missing here or a better approach I should take? Any thoughts or opinions would be hugely appreciated!
I think you want too much.
'all I/O points have a decoupling capacitor.' Wazzat mean? Is the circuit AC coupled or DC coupled? If the I/O are AC coupled, you don't need a virtual ground.
Post the circuit.
'all I/O points have a decoupling capacitor.' Wazzat mean? Is the circuit AC coupled or DC coupled? If the I/O are AC coupled, you don't need a virtual ground.
Post the circuit.
What I mean is, all ground referenced signals to/from the outside world are AC coupled. But because each circuit has an OP amp generated V-gnd, and they are going to share the same DC supply, I need to know if my scheme for joining the two virtual grounds would be troublesome.the signal is AC coupled, but the grounds are going to get connected together.
Sorry... the whole circuit is a bit huge, and I should have posted it before going off on this topic. I guess I was hoping to confine the discussion, for now, to issues of joining two virtual grounds.
If the signal is AC coupled, you don't need to return it to ground, you can return it to either supply rail. That's the whole point of AC coupling, it frees the output from whatever DC bias (1/2 Vcc) you have set on the opamps, so you can connect the return to any point which is at AC ground, which both power rails are...
You don't understand. I know the signal is AC coupled, and I'm not talking about the signal at all. I'm talking about grounds, which in may case are developed by rail split OP amp circuits. I think the important point you are missing is that in my devices, the grounds on all the audio input/output jacks are connected directly to the virtual ground. Grounds are not AC coupled, in any circuit I've ever seen. So they are all sitting at 2.5VDC.
So let me simplify my original question. Look at the Virtual Ground (1/2V) circuit I posted, and imagine that it is used for all the the grounding in one device with several OP AMPs. Now assume another device that has a similar Virtual Ground circuit, minus R3 and C2. That device also has OP amps, ground referenced to this other virtual ground. So what happens if both devices are run off the same supply, AND the two virtual grounds get cross connected by cables? If both Virtual grounds were absolutely 2.5000000 VDC, there might not be an issue. But the two will probably be slightly different. So I'm asking if the 50 ohm resistor I've shown for R2 will be sufficient to eliminate problems caused by any small difference between the V-grounds.
No, no. YOU don't understand.
The power supply connects the power rails. It has a very low impedance. For this reason both power rails are effectively @ AC ground.
If the signal is DC coupled, then it has a DC potential, and the return from the load must be connected to a point at the same DC potential as the output, or DC current will flow.
If, however, the output is connected thru a capacitor, then the load return can be to any point which is at AC ground (although the exact choice will have a big influence on amplifier performance.) Look at any (AC coupled) amplifier with a single-sided PSU, the load return is almost always made to the negative rail, which coincidentally is connected to ground.
You either need AC coupling, or a virtual ground, but you don't need both. Which is why I'm pushing you on this point, because it's a lacuna in your knowledge of circuit theory. Brush up on this before you go launching off into more ambitious fields.
The power supply connects the power rails. It has a very low impedance. For this reason both power rails are effectively @ AC ground.
If the signal is DC coupled, then it has a DC potential, and the return from the load must be connected to a point at the same DC potential as the output, or DC current will flow.
If, however, the output is connected thru a capacitor, then the load return can be to any point which is at AC ground (although the exact choice will have a big influence on amplifier performance.) Look at any (AC coupled) amplifier with a single-sided PSU, the load return is almost always made to the negative rail, which coincidentally is connected to ground.
You either need AC coupling, or a virtual ground, but you don't need both. Which is why I'm pushing you on this point, because it's a lacuna in your knowledge of circuit theory. Brush up on this before you go launching off into more ambitious fields.
Last edited:
Well regardless of how ambitious YOUR own fields of endeavor are, you will do much better at it when you are able read and comprehend what is being described before filling in the gaps with your own assumptions, assuming the person you're talking to is lacking in understanding, and claiming superiority. As for me, I have been involved with electronic design for over 45 years, both analog and digital, solid state and vacuum tubes, digital analog and mixed signal, and have designed and built tons of amplifiers from scratch. I understand what you are describing very well, but you do NOT understand what I'm asking. Obviously you can't help me here because YOU are so focused on what you THINK I'm explaining, that you are completely unable to stop assuming and SEE what I am ACTUALLY explaining..
If you want to be helpful, look at my original explanation, look at exactly what I am describing. and SEE where it differs from the assumptions you are forming. Consider for a moment that you may indeed be missing the point. Until you see it, AND drop your arrogance, please don't respond anymore.
It is frustrating when I'm unable to convey a specific situation. Then to be told over and over that I don't understand when I know the person telling me so does not see my point, with a final statement like "Brush up on this before you go launching off into more ambitious fields" just got my goat. But still, I apologize to everyone. CC did his best to to explain something to me with good intentions, and had I taken the time to exemplify my question with at least a simplified circuit drawing, I could have saved everyone some time.
I'll have to be content to let my bench confirmations do for now and hope for the best. Thanks for trying to help me CC. Again I apologize.
I'll have to be content to let my bench confirmations do for now and hope for the best. Thanks for trying to help me CC. Again I apologize.
Why not just make one circuit that develops your Vmid (2.5V) voltage, and ship that around to any outboard modules that need it through the IO connector? I might be glossing over something, but it's sort of like having two or more clocks in a system - they'll never be in sync, so why not just use one clock and have all devices reference that? Or, in your case, one Vmid reference, and transmit it to remote devices through the cabling?
Sure, there are issues with this approach, and you can completely avoid them using AC coupled signals. This would allow each remote device to have its own Vmid generator and they could still transmit signals between the subsystems.
Still, if you can use odd cabling and connectors, you could use only one Vmid reference and DC couple everything. But, the fact that every amplifier has a DC offset sorta makes this not so useful in practice, but depending upon the circuit details, maybe AC coupling is really tough and this is easier. You get to decide!
Sure, there are issues with this approach, and you can completely avoid them using AC coupled signals. This would allow each remote device to have its own Vmid generator and they could still transmit signals between the subsystems.
Still, if you can use odd cabling and connectors, you could use only one Vmid reference and DC couple everything. But, the fact that every amplifier has a DC offset sorta makes this not so useful in practice, but depending upon the circuit details, maybe AC coupling is really tough and this is easier. You get to decide!
Yes, but...
It's the ifs and buts that make the reply long-winded, and they're why I wanted to see the OP's circuit.
If you're going to make the circuit AC coupled, then you don't need a virtual ground at all.
The whole point of the virtual ground is that it permits DC coupling. The thing everybody wants to avoid is capacitors in the signal path. If you're going to accept capacitors in the signal path, then you might as well go the whole hog and design a proper single-supply AC coupled circuit, and save yourself an opamp and a load of complication.
Or use a 3-pin connector and pass the virtual ground from module to module.
But no, the OP wants to have his cake and eat it. He wants to stick with the connector he's got and clag on a further layer of performance degradation so he can congratulate himself on how clever he's been.
Last edited:
Hey thanks Monte. I actually am letting the main circuit "ship" its V-gnd to the second, as you're suggesting. So the second device indeed gets 5V and 0V through the power connector to the first device, and does fine with the 2.5V V-gnd (shipped, as you say) through the short connecting audio cable. Normally I'd just maybe add an extra capacitor there in the 2nd device, between 0V and the incoming 2.5. The reason this became an issue for me is because even though the second device is intended as an add on to the first, I see practical usage for it as a stand alone device. So it would be nice if it could somehow have its own "backup" 2.5V V-gnd to allow this possibility. Hence my 2nd V-gnd in devcie #2, with the added resistor.
By the way, AC coupling in the tradition sense, where the analog I/O are simply referenced to 0V (with 0V being the only exported ground), would have indeed eliminated this issue. The second device would have still required a V-gnd 1/2V point to operate off a singe ended supply, but I wouldn't have had the worry about the two devices being cross connected. The reason I use V-Gnd for all references is because I have found that it produces much lower noise, especially when it comes to power supply noise rejection, then the traditional approach.
Last edited:
Lets go back to post #1
You have a 5 volt powered opamp circuit that has been configured to be a -/+2.5 volt supply with a virtual ground. All good so far.
You want to add a second similar type circuit that also has to be "free standing" in its own right but also sometimes connected with the first.
One possible solution might be to make the second virtual ground with a trimmer in one arm of the reference resistors. We are talking fine adjustment so perhaps a 100k resistors and a 470 ohm series pot. Adjust so that the difference between virtual grounds is 0.000 volts. You might need a parallel high value resistor across the other arm in order to get a plus and minus swing with the adjustment.
Its been mentioned that if the circuits are AC coupled then the signal returns can be rail referenced. I'm not so sure that's a good idea... or at least the terminology needs setting out clearly. Once you have decided a design is a virtual ground one, then you can't return signal related feeds to the rails even if they are nominally all at the same impedance. Slight rail disturbances due to the opamp loading under dynamic conditions will reflect into those rail returns and totally wreck and distortion figures. The virtual ground is the absolute reference... nothing else... everything else has the ability to move relative to that point.
A third wire is another option, and not just a means of carrying the ground between modules. You could use it to feed the output of the first virtual ground into the input of the second (it would over ride the high impedance of the reference resistors. Just an idea.
You have a 5 volt powered opamp circuit that has been configured to be a -/+2.5 volt supply with a virtual ground. All good so far.
You want to add a second similar type circuit that also has to be "free standing" in its own right but also sometimes connected with the first.
One possible solution might be to make the second virtual ground with a trimmer in one arm of the reference resistors. We are talking fine adjustment so perhaps a 100k resistors and a 470 ohm series pot. Adjust so that the difference between virtual grounds is 0.000 volts. You might need a parallel high value resistor across the other arm in order to get a plus and minus swing with the adjustment.
Its been mentioned that if the circuits are AC coupled then the signal returns can be rail referenced. I'm not so sure that's a good idea... or at least the terminology needs setting out clearly. Once you have decided a design is a virtual ground one, then you can't return signal related feeds to the rails even if they are nominally all at the same impedance. Slight rail disturbances due to the opamp loading under dynamic conditions will reflect into those rail returns and totally wreck and distortion figures. The virtual ground is the absolute reference... nothing else... everything else has the ability to move relative to that point.
A third wire is another option, and not just a means of carrying the ground between modules. You could use it to feed the output of the first virtual ground into the input of the second (it would over ride the high impedance of the reference resistors. Just an idea.
Well here's the thing... If the two V-Gnds really had to be 100.0000% identical in order to cross connect, I'd probably be better of letting the V-Gnd in device #2 be a simple pair of resistors and a cap. An OP-Amp V-gnd is obviously much lower impedance and better on several levels, but I don't have much confidence that the second V-Gnd could be fine tuned to exactly meet the firs, AND STAY that way. So this is why I thought to use a second Op-Amp based V-Gnd, whose impedance was compromises only by 50 ohms. The 50 ohs, I'd hoped, would harmlessly dissipate any small difference, so that I wouldn't have to hope to "null" out the difference. It seems to work on the bench, but if there is a potential for some hidden instability there I'd be more inclined to add a switch or an on board jumper to put the 2nd device V-Gnd in or out of circuit. I just hate to add that concern to the end user.
One of the issues I face with device #2 is that aside from its signal interface to my device #1, It allows up to 8 other external device to be "patched in" through other jacks. Those external devices can not conflict with mine from a DC point of view because they all are mandated to run off their own independent supplies. So if my "2.5V 1/2V gnd" gets connected to another devices 9V gnd, and still another devices 0V or -12V gnd, as long as all the supplies are isolated there should be no conflict. But the possibility does exist that many of those external and largely "undefined" devices (we're talking about musicians effects pedals here by the way), will have internal capacitors on their inputs and outputs. So some AC coupling may be inevitable, not for the interaction between my devices, but in the I/O to other "unspecified" devices.
[/QUOTE]
That did cross my mind too, but it sounds like trouble, and I'll tell you why. Before you spoke of fine tuning two V-Gnds to be identical with a trim pot. But one thing you can't tune out os offset voltage, which even the best OP amp will have. So if My local V-Gnd is developed by an OP amp, and it is also the signal ground, then sooner or later the two signal grounds of my modules are going to have to meet. Now if a third wire were used to try to influence that 2nd V-Gnd on the voltage divider side of the OP-AMP, there's still that pesky offset voltage.
But now again, going back to the original solution I've been kicking around, do YOU see a problem letting such a small V-Gnd voltage difference just exist across my proposed 50ohms? So far I've seen this difference be in the order of 10mV, which would only cause about 200 uA of unwanted current. I haven't measured any increased noise so far in the bench test, but audibly it all sounds very quiet, even with my test amplifier cranked.
Sounds really complicated. At this point, why not just go to a ±2.5V supply and be simple? Or, +5V, a Vmid, and AC couple.
I could see problems with ground returns shared among a lot of circuitry though that Vmid reference. To avoid that, you'd need to treat it like a real reference, not a ground. If it was used as a real reference, then each sub-module would have to have its own buffer to handle a circuit's ground currents, but that would also allow you to easily do the internal / external Vmid reference 'switching'.
Not sure how the noise works out better. There has to be at least one reference / regulator somewhere in each approach, so the noise from that is inherent, but can be made very low. Reservoir caps and RC filters can control the rest, as well as interference.
I use amplifiers with power supply rejection, so I can put the necessary regulator noise to a supply rail and use a real ground. Still, I am using a Vmid generator for a 3.3V microprocessor system. It's very useful for me to be able to set one voltage as the midpoint for the A/D converters, and also use that as a fake ground in 3.3V analog circuits that feed the A/D. Avoiding coupling caps is a good thing in the SMT world, so using a Vmid as a ground reference is better for me than AC coupling.
I could see problems with ground returns shared among a lot of circuitry though that Vmid reference. To avoid that, you'd need to treat it like a real reference, not a ground. If it was used as a real reference, then each sub-module would have to have its own buffer to handle a circuit's ground currents, but that would also allow you to easily do the internal / external Vmid reference 'switching'.
Not sure how the noise works out better. There has to be at least one reference / regulator somewhere in each approach, so the noise from that is inherent, but can be made very low. Reservoir caps and RC filters can control the rest, as well as interference.
I use amplifiers with power supply rejection, so I can put the necessary regulator noise to a supply rail and use a real ground. Still, I am using a Vmid generator for a 3.3V microprocessor system. It's very useful for me to be able to set one voltage as the midpoint for the A/D converters, and also use that as a fake ground in 3.3V analog circuits that feed the A/D. Avoiding coupling caps is a good thing in the SMT world, so using a Vmid as a ground reference is better for me than AC coupling.
You can do what you want, but is there a fly in the ointment?
Yes.
That's the question you wanted answered isn't it?
You have to AC couple the feedback networks in the circuits (to ground) as well as the signals (at the I/O). Otherwise you get DC in the feedback (if the grounds aren't exactly the same) and it creates offset voltages, which at the very least cost headroom. As I've pointed out several times now, once you've gone to the trouble of doing that, you don't need the virtual ground any more.
Yes.
That's the question you wanted answered isn't it?
You have to AC couple the feedback networks in the circuits (to ground) as well as the signals (at the I/O). Otherwise you get DC in the feedback (if the grounds aren't exactly the same) and it creates offset voltages, which at the very least cost headroom. As I've pointed out several times now, once you've gone to the trouble of doing that, you don't need the virtual ground any more.
Look at the LM3886 datasheet, you'll see that in the single-supply version (AC coupled) the negative rail is connected to ground.
I should point out that the first device is already "done". And by "done" I mean really done! PCB boards cut, devices are in labeled packages, and there is a small base of installed users/customers. This second device is the result of test market requests I'm trying to satisfy. So this means that device #1 is what it is, but device #2 has room for design mods, because its still being bench tested.
Now as it turns out, device #1 had an AUX 5 VDC out. It was originally provided as a convenient way to recharge a LiPo battery in a totally different unit, a wireless controller for device #1, which I've left out of the discussion for simplicities sake. Device #1 also has several audio ins and outs, and the signal grounds on all all these ins and outs *IS* 1/2 V, with respect to the internal (and up to now) isolated 5V supply. It is, by design, meant for direct coupling.
So then the request for this device #2 came along, which will also handle audio from some other sources. It would also be convenient if it could be powered from that AUX 5V I already have on device #1. The concern comes from the fact that at least one of the audio outputs on devcie #2 MAY be connected to an audio in of device # 1.
Now both these devices use similar 5V rail to rail OP amps. (LME49721s FYI, not sure where the LM3886 came into discussion). So I have a situation where IF device #2 is indeed powered from device #1, AND the two are also connected via audio cables, 3 voltages will arrive at device #2 from device #1: 5V, 0V and 2.5V. So the issue I'm concerned about is that regardless of whether the signal wires are AC coupled, DC coupled, or not even connected, there is still this second device which must develop its own 2.5V V-reference in case it is operated stand alone, but needs a simple way to make sure slightly different 2.5V references don't fight with each other. Again, I'm not talking about the AC or direct coupling of the signal. I'm focusing on making sure the ground references don't collide in the LIKELY case that the two 2.5V references are not 100% identical. I DON'T think you're suggesting to AC couple the ground references, right?
So full circle... you are saying that my 50 ohm solution, to avoid a fight between the two 2.5V reference sources will work, BUT that I have to AC couple the feedback networks to ground, as well as the I/O. This is that part I have not been able to make you see. Even if you take the signal handling out of the picture, hopefully you see now that it is the ground reference between the devices which I want to coerce into playing "nice" together.
So now, if there were a differential between the two 2.5V gnds, similar to the 10mV differential I've been seeing, why would that compromise my headroom? I'm only dealing with line level audo that seldom peaks much beyond 100 mV (or 200 mV p-p) and I'm using 5vdc rail-rail OP amps, referenced to 2.5V. I think that's a reasonable headroom that could tolerate a 10mV loss , would you agree? As to feedback networks, it may be significant (perhaps a saving grace) that all the signal passing OP amps in this circuit (both device #1 and #2) are all simple unity gain buffers, with their + inputs referenced (biased in case of an AC coupled source) to V-Gnd (2.5V), always through about a meg ohm of resistance.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.