Need sugestions about virtual ground or symetric supply for OPAMP

[wuemura]

Hello guys!

I was talking with audioguru some time back about my hafler circuit (schematic attached), it was clipping and i did know exactly why, we exchange some ideas and i choose to let it go.

I've never used the center channel and when i did the quality was very bad, the overall sound got even worse since i've added 2 more TL074, so i redesign the center channel block, but after building the prototype clipping was worst than before in all channels.

So it was not the signal, it was not gain but i've find out that the clipping problem was related with the virtual ground, i was using a pair of 10K resistors to make a voltage divider and every time the sound clipped i got oscilations at that point. Doing my research i've found this information:
http://headwize.com/projects/showfile.php?file=opamp_prj.htm

There may be times, though, when a virtually-grounded dual supply has a tendency to "rail" when the resistor-type voltage divider cannot maintain the virtual ground at 1/2 Vcc. Such cases may occur when the opamp draws too much current or input signal (for example, a high boost equalizer) pushes the opamp into heavy clipping and power supply is unable to recover.

Changing the R12/13 to 4.7K did a huge increase in sound quality, audio now was cristal clear again, so i thought.

Here is a audio sample (8Mb, sorry if is too big):
http://www.sendspace.com/file/l79kos

If you hear it with headphones you will notice the audio clipping, not that much, i recomend that you play with your audio card balance to hear what each channel is doing, left channel is the center channel or voice and right channel is ambience. Analyzing the virtual ground i get a very smal oscilation when the audio clip. Good, now i know what is going on with the circuit and why is clipping.

I've found a circuit in a old magazine that use a CA3140 and a few components, the question is, this circuit can or could make the virtual ground to be more stable?
Or should i consider design a full symetric supply based on SG3524?

My goal is to keep this small because there is alot of components already, and if the CA3140 solution work out there is no need for me to use the SG3524 that will need a dozen more components.

Thank you guys!

 

[esp1]

hi,
Have a look at www.analog.com site and search for 'phantom ground'.

Application note: AN581

 

[wuemura]

Hello esp1, how are you?

Thank you for your reply
The aplication note AN581 (Download here) show a similar circuit with the CA3140.

But i really worry about is if this buffered virtual ground will or could hold a 5Vpp RCA signal to prevent audio clipping.

At this sample:
http://www.sendspace.com/file/3fxm5w (2.06Mb)

I use a 2.3vpp signal with the configuration of the schematic, using the 4.7K resistor voltage divider, and as the sample show this is the circuit limit, anything higher than that start to have small clipping.

Thank you.

 

[audioguru2]

Hi Wellington,
I am glad that you discovered the cause of your distortion. ;D

Your circuit doesn't need the very high power virtural ground circuit with the old CA3140 opamp and high current darlington transistors. An ordinary opamp by itself and made into a follower as shown by Analog Devices has an extremely low output impedance to drive a virtural ground.

I don't believe it. Analog devices show the battery symbol backwards on their applications note.

 

[wuemura]

Hello audioguru!
How are you, you big dog  (or duck) ;D

Thank you, it was not an easy task and i wont let it go that easy
I love learning this stuff. 8)

So, what you think about the new circuit, it was a brain burning task, you have more experience than me in that area.

tell me something, if i do the circuit by Analog Devices, do i need to keep biasing the opamp inputs?

Thank you.

 

[audioguru2]

Hi Wellington,
The opamp is just a follower of the resistor divider that has a capacitor to ground. Some opamps oscillate if the output has a capacitor load.

I hear an odd form of distortion on the ambiance channel of your Troubled Waters clip. It sounds like buzzing when the brass horn instruments play loud. The center channel sounds very clear.

 

[wuemura]

Hi audioguru,

So your sugestion is to increase the resistor values to 100K, decrease the capacitors values and remove the 100K bias resistor?

But there is a problem, if i remove the 100K bias resistor i will short the input channels messing up with the effect, right?

The buzzing distortion at the brass horn instruments was in the original CD, including the "buzz and Hiss" noise.

I will make the changes and see what it does.

Thank you.

 

[wuemura]

Hi audioguru,

Changing that R12/13 from 4.7K to 100K makes the sound worse, at the center channel for example, Elvis voice start to crack up soon as he start singing, i used the same 2.3Vpp signal, when we get near to the end all channels has a huge distortion like a guitar overdrive.

The capacitor load at the output (22n) is needed because of the power amp, otherwise the tweeter start to click and pop out, maybe i could use a bigger ones like 100n.

The condition becomes normal if i use again 4.7K in R12/13, but keep the changes in the eletrolitic capacitors.

 

[wuemura]

Doing frequency analysis response, changing the eletrolitic capacitor from 220uF to 22uF i get a decrease at bass response, sound looks like a dry high frequency and a dull midle to bass response, changing back to 220uF gives a more natural flat sound

 

[wuemura]

I will design the "phanton ground" using a TL072, do you think that this will be enough to make the virtual ground more stable and avoid the distortion?

 

[audioguru2]

The very low output resistance of my virtual ground opamp feeds all the bias resistors for the circuit.

That was Elvis Presley? From 1955 when they had recording equipment with severe distortion? No wonder I heard the distortion.

The 22nF capacitors overload the output of the opamps if the frequency is higher than 18kHz. 100nF capacitors would cause overloading at frequencies above only 4kHz.

I used two 100k resistors and a 22uF capacitor because the capacitor is mostly charged in 1.1 seconds and is fully charge in 5.5 seconds. 220uF will take 55 seconds to fully charge.

Use a TL071 single opamp or one opamp from a TL072 dual opamp to make the virtual ground (phantom ground) circuit. Its output resistance will be very low, next to nothing. Make sure it doesn't feed a capacitor to ground.

 

[wuemura]

Hi audioguru,

Yes, it was Elvis Presley, the recording was made live at the show from the sound board at 08/12/1970, the original has that distortion and noise, the Buzz, Hum and Shhh.

I did not know that the 22nF could overload the opamp but i tought it wold atenuate the high frequency around 17.5Khz acording with my tests, so i could remove this capacitors, change the circuit to see how things work out, if the tweeters start to pop out again i could try 10nF instead.

One thing that i have also noticed is the turn on time, using a pair of 100K resistors make the sound crack when the circuit is turned on using 22uF/100uF, changing to 220uF/100uF it gets around 3.5 seconds for  the opamp start operating at full. Changing from the two 100K to 4.7K and 220uF/100uF the turn on time is less than a second, and changing the capacitors to 22uF/100uF it makes the turn on even faster but it decrease the bass response.

I will use the TL072 because i have a bunch of it, i will adapt the "Phanton Ground" without the capacitors to the ground, build a new prototype and i post here the results, ok?

Thank you

 

[audioguru2]

The 0.1uF filter capacitor in the Analog Devices phantom ground circuit is much too low.
They use a 110k resistor for their opamp but it isn't needed with the FET inputs of a TL072 opamp.

Use the Virtual Ground circuit I posted.

 

[wuemura]

Ok, i will made the changes with your circuit, just give me a minute to work this out.

 

[wuemura]

So, moment of truth.

I've made a few tests with exacly same components as your virtual ground design, it was almost. I pushed the circuit limits by the way, so instead of 2.3Vpp signal i use a 5.0Vpp signal to push it to the limit and to meet the specifications that i will work with. Using this signal i get distortion at high notes and fluctuations over virtual ground point, this is good because it will work with signals from about 1Vpp to 3.3Vpp.

So i start to play with the resistor dividers, Increase the two 100K to something like 330K i get a bad distortion, getting this value to about 48K it will work nice but not at the level that i need, sounds very good but it will distort with some specific conditions from the source. Using 22K get's better but still distorting a little bit. Lowering to 4.7K if there is a high voice for example the voice sound os ambience do a strange vibrato effect in some very specific type of music.

6.8K is the exact value that will give me a rock stable virtual ground, 10K work also, but 6.2K is how low we can get with that high RCA level.

Using the headphone it sounds perfect clear, but in very specific conditions and paying atention there is a low level not easy to notice some kind of "granulation fade out" for very short period, again, only in very specific conditions.

Since i'm using a TL072 that is a dual opamp, i will try to double buffer the virtual ground and see what it does.

 

[audioguru2]

Hi Wellington,
The junction of the two resistors at the input of the virtual ground opamp is a filtered voltage divider of the supply voltage. Maybe your supply voltage is fluctuating?

 

[wuemura]

Hi audioguru,

I'm using a comercial PSU (regulated at 12V/1A), i know that is is very stable, and i'm sure that the problem is the virtual ground, it has to be stable at 1/2 Vcc or i will get the distortion problems, messing up with the values shows that.

But your virtual ground solution works very well!

I'm doing some tests here with a dozen CD's that i have and the distortion and clipping are gone for good. I did change some values because using a double buffered virtual ground change some aspects of the circuit, what was good because all the CD's that i've tested worked out fine.

The design that i came out is attached in this message and belive me, is good, the audio output got a little louder than before using 2 virtual ground buffers, so i've changed the circuit input resistors from 51K to 62K.

The solution that you gave work fine with standard RCA levels like 1Vpp, i did a small headphone amp using your virtual ground and and work it out fine (using a simple TL072 like the cmoy one), no noise or distortion, if some one try to build a pocket headphone amp this is the way to go. With the signals that my circuit will work with the attached schematic work best.

I did not made further tests with the values but R1 and R2 can be 12K.

I will record and upload a new sample, this one has a guitar at the back, this is one of the songs that had the distortion at the high tones of this guitar, now is tight.

 

[wuemura]

So, here is the sample with the double buffer configuration and all the changes.
http://www.sendspace.com/file/0syxro (2Mb)

The distortion was noticed at the beggining when there is something that sound like a pipe organ, moust of the time there was a distortion. But with the new virtual ground it sounded great now.

 

[audioguru2]

It sounds good now.
How do you like the Hafler surround sound effect?