Virtual Ground Power supply

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
So I am building a project with a CMOY headphone amplifier with a DS1802 Digital pot for volume control and CD4052 mux for source selection.

For power I am using a AC-AC wallwart with rectifier and voltage regulator.

Now the Cmoy uses a resistive voltage divider to achieve its dual supply rails (in my case i have +/- 8V). All of this works perfectly fine and sounds great.

The problem occurs when I attempt to power a microcontroller from the +8V (Right now I'm prototyping using an arduino uno). As expected the load from the uC is much larger than anything else and the supply rails become very unbalanced (from +8 to like +3-4V).

So I need a bit of help fixing this error as cheaply as possible as well as conserving some board space. I've read tangent's article on virtual grounds but am still at a bit of a loss.

Thanks for any help, if you need a schematic I can try and draw one up.
 

SgtWookie

Joined Jul 17, 2007
22,230
There is a page on virtual grounds here:
http://tangentsoft.net/elec/vgrounds.html
You could use the same circuit as the one where the opamp is labeled OPA with an L2722; which is a low-dropout 1A dual opamp.

But, you sort of say that your power rails go from 8v down to 4v - if that's the case, you need a power supply that is rated for more output current.
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
Yea when I power the arduino from +8V and the virtual ground the rails get unbalanced. They go from +/- 8V to like +3/-13.

The AC/AC wallwart I have is rated for 500mA. AFAIK nothing except the micro draws that much current.

Thank you for the op-amp rec, Ive read tangent article a couple times but I dont have enough experience to find the correct part for my particular application.
 

Kermit2

Joined Feb 5, 2010
4,162
a regulator on the positive rail only will unbalance the power supply, so the simple fix would be to not only increase your power supply current sourcing capacity but also add a resistive load to the negative half to restore your balance/virtual ground point.
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
okay sgtWookie's recommendation works perfectly thank you. Rail voltages keep perfectly balanced. I just wired up a small button to test volume control (the remote I coded for is not with me atm) everything works excellently with absolutely zero noise.

Theres a few things Im worried about now.

Voltage regulator gets much hotter now. I do have a small heatsink on it and I do realize the reason it gets hotter, just wondering if I should be worried.

Secondly, the new opamp also gets hot. I can keep my finger on it for like a minute before it gets uncomfortable. Wondering if thats normal as well, none of my other ICs even get warm.
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
Okay so power op-amp get really hot pretty fast when I hook up the uC otherwise it doesn't.

Putting a 1K ohm resistor on the output (like in tangents' schematic) makes the circuit behave like it was before (unbalanced when powering the micro).

Lastly I was wondering what i should do for the unused op-amp channel. I dont leaving them floating is a good idea but I dont know whether to ground them or what.

Sorry for all these questions, this is what happens when you undertake such a project without knowing that much. However this is the reason I do this and go to university. Thanks for all the help. I'm going to begin drawing up a schematic.
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
Yes many times in fact. The schematic with the op-amp is almost identical to the one posted on tangents' site. Its what Im going by right now.

I'm pretty limited on space right now (the op-amp circuit was pretty much an exact fit). I dont really want to use transistors because I honestly dont really understand them (though I have this really fat book on my desk that seems to be all about them but thats for next week when classes start).

Am I missing something obvious?
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
I quickly drew up a schematic so hopefully it'll be of some help. I apologize if its not up to standards but I have zero experience with such things.

I omitted the micro controller because I feel its connections are obvious. Its powered from +8V and GND. Its SPI pins are connected to the SPI on the DS1802. Two other pins are used to control the mux's A/B control inputs.

It also will receive input from an IR receiver and control an RGB LED. However these two I generally leave off when hardware testing.


Please understand that I do read and take all of your recommendations. Maybe I'm missing something but you said to use that op-amp with this schematic. I have done all of that.

I fear that the problem is related to the current sourcing ability of the wall transformer I used (500mA). I have begun a search for a new one although they all seem to be around $10+S/H. Pretty expensive at this point so I'd like to be sure of such things. I'm also not sure how to test such a thing. I do know that ammeters are to be put in series but where. I've also heard to put a resistor in place and measure voltage drop across it to commute current.

Sadly the most I know of circuits is from a basic DC theory class. Please do not be too harsh on my ignorance. Thank you for all your help.
 

Attachments

SgtWookie

Joined Jul 17, 2007
22,230
I apologize; I kind of snapped at you and you didn't deserve that.

What's been happening fairly regularly is that either people simply don't follow advice, or actually try to argue against what I'm trying to point out, and I wind up having to explain it three times before they finally understand or will agree to try it. After this has happened dozens upon dozens of times, one tends to develop a somewhat pessimistic outlook. That's just so you hopefully can understand why I sounded rather peeved.

Let's start with your regulator R12 and R11. 120 Ohms is good for R12, as that guarantees the output will be within manufacturer's tolerances for voltage regulation, as 1.25v/120 Ohms = 10.25mA.

So then for R11, 10.25mA*1,500 Ohms = 15.375V, + Vref (1.25v) = 16.625v out. Are you good with that? 8.3125v per rail?
What is the input supply voltage across C4?

Then let's look at the R1/R13 divider. Vreg(out) / (220k*2) ~= 16.5v / 440k = 37.5uA. For voltage dividers, I like to keep the current somewhere between 0.1mA and 2mA. If you go too high, you're wasting power. If you go too low, you can wind up with noise and/or instability. The L2722, being a dual power opamp, might need more current to keep the inputs happy - so let's settle on 1mA. 16.5v/0.001A = 16.5k Ohms. So, two 8.2k Ohm resistors will give just over 1mA current through the divider.

What's with R14, R15, and then VR1 and VR2? Those are symbols for something like a TL431 shunt regulator. Is that what you're doing for the +2.5/-2.5v out?

It's really odd that the opamp is heating up, as there really isn't anything to load it - unless the diagram is not what you show. If you have the uC's H1 connected to GND, and H1 is not set to INPUT or high impedance, you will cause the uC to "fight" the power opamp.

I have no clue how much power your uC is supposed to be drawing, but stepping down from >18v to 5v using linear regulators is a really inefficient way to do it; most of the power is going to be dissipated as heat in the regulator.
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
I understand why you two made those remarks. It happens regularly when you are a veteran of some kind of specialized topic forum. I just wanted to make my positions and intentions clear.

~8V rails is perfectly fine with me. Like I said, aside from the op-amp heating up every works and sounds great. I hear no noise even with the pot turned high.

AC IN is from an AC-AC wall transformer. Its output is spec'd at 24V- 500mA.

Are you saying I should lower the 220K resistors (R1/R13) to 8.2K?

Yes you're exactly right, its a TL431 shunt regulator. The ds1802 has a max supply voltage of 5V and its input must fall between those limits. Therefore since my input is audio I used that to make my supply rails +/- 2.5V.

I honestly copied that straight from another schematic so I have no idea on optimal resistor values.


For the microcontroller I'm prototyping using an arduino uno and the final product will probably be an arduino pro mini. Makes it easy and I've always wanted to play with one.

On the UNO I'm supply the VIN pin with the +8V rail and GND with the virtual ground. The micro runs on +5V and VIN is regulated to that. Link is to the uno's site incase you're unfamiliar with it.

http://arduino.cc/en/Main/ArduinoBoardUno

I'm not sure what you mean by H1. The H1 on my schematic refers to the DS1802 Digital Potentiometer. I apologize its not labeled, I couldn't find an eagle part for it so I had to quickly make one. H1 refers to the high side of the pot, L1 the low side, and W1 being the wiper. Heres it's datasheet..

http://www.maxim-ic.com/datasheet/index.mvp/id/2778
 

SgtWookie

Joined Jul 17, 2007
22,230
OK. If the UNO isn't connected, does the L2722 and your LM317 still get hot?

I'm wondering if perhaps the L2722 is oscillating at high frequency. You can dampen that by inserting a Bucherot cell between the output and the negative rail. A Bucherot cell is a low-value resistor and a cap (around 1uF to 10uF) in series.

Did you do anything with the 2nd L2722 opamp channel? If not, you should tie the inverting input to it's output, and tie the non-inverting input to the voltage divider. Don't tie the two outputs together, or they will "fight" each other - to their flaming death. Ask me how I know this. :rolleyes:
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
No with the uno unconnected the opamp does not get hot. In fact I tested that earlier; as soon as you connect the UNO the opamp begins to heat up, disconnecting the uno and the opamp immediately cools off.

I have not done anything with the second channel, I asked earlier what to do with that since I figured leaving it floating wasnt a great idea. When you say tie it to the voltage divider do you mean the midpoint between the two resistors. That is, the + side of the first opamp channel?

I will try the Bucherot cell idea. Though I wont have access to my stuff till tomorrow evening, I do have some 330ohm resistors, is that low enough?
 

SgtWookie

Joined Jul 17, 2007
22,230
No with the uno unconnected the opamp does not get hot. In fact I tested that earlier; as soon as you connect the UNO the opamp begins to heat up, disconnecting the uno and the opamp immediately cools off.
OK, then your UNO is really pulling some current. You really need a separate plug supply for it. You know, you could stop by a Goodwill store, they usually have dozens of them sitting around. If you could find one between 8v and 12v, that would do it.

I have not done anything with the second channel, I asked earlier what to do with that since I figured leaving it floating wasnt a great idea. When you say tie it to the voltage divider do you mean the midpoint between the two resistors. That is, the + side of the first opamp channel?
Yes. And yes, I did mean for you to replace the 220k resistors with 8.2k resistors (ideally) - but you could use anything between 4.7k and 10k and it would work just about as well.

I will try the Bucherot cell idea. Though I wont have access to my stuff till tomorrow evening, I do have some 330ohm resistors, is that low enough?
Well, by "low value resistors", I meant between around 3 Ohms and 20 Ohms - so no, 330 Ohms is too high.
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
While I do have another DC power supply I could use for the UNO, I seriously would rather find another solution.

Two plugs for one little box seems like alot, I rather not have two power cords for just this one thing. Can I just get a different higher output supply?

Secondly if I power the UNO separately won't I have an issue with logic levels. The UNO's output will be referenced to its ground while the stuff its interfacing is referenced to the virtual ground which will be at +8 in relation to ground, correct?

How can I measure the current draw from the UNO? Do I just put a meter in series with its supply pin?

Too bad about that resistor thing. I dont think I have anything that low on hand. I don't want to make any my orders unless I'm sure it'll help.
 

Thread Starter

jtavrisov

Joined Jun 2, 2011
29
Any other solutions you guys can come up with it? I'm sorry for being impatient but I pretty much have only one more week to finish this before it gets troublesome.
 

SgtWookie

Joined Jul 17, 2007
22,230
While I do have another DC power supply I could use for the UNO, I seriously would rather find another solution.

Two plugs for one little box seems like alot, I rather not have two power cords for just this one thing. Can I just get a different higher output supply?
The problem appears to be power dissipation in your 16.5v regulator and L2722 caused by a rather heavy load from your UNO. If you want these power dissipation problems to go away, you will either need to go to a DC-DC switching converter solution, or use another supply.

Secondly if I power the UNO separately won't I have an issue with logic levels. The UNO's output will be referenced to its ground while the stuff its interfacing is referenced to the virtual ground which will be at +8 in relation to ground, correct?
You could connect the UNO supply ground to the virtual ground, as long as at least one of the supplies is isolated from earth ground.

How can I measure the current draw from the UNO? Do I just put a meter in series with its supply pin?
You may exceed the capacity of your meter to read the current. If you attempt this, use the 10A scale. It would be best to measure the voltage across a 1 Ohm resistor that was connected in series with the supply, but apparently you don't have any such low-value resistors.

Too bad about that resistor thing. I dont think I have anything that low on hand. I don't want to make any my orders unless I'm sure it'll help.
Unless you have an O-scope on hand, it will be difficult to tell whether it is oscillating or not.
 
Top