Is this legit LM3886 balanced option?

[valtra103]

Hello. I hope some of the more experienced people would help me with my design.
I am trying to design LM3886 With balanced input.

[IMGDEAD]https://i.xomf.com/wkdjd.png[/IMGDEAD]

Is this circuit good. I don't have LM3886 model in Spice, so simulation is almost useless with random opamp in it's place.
One thing i am confused is what minimum gain LM3886 needs in this config, to function properly
Thanks in advance

 

[JMFahey]

i hope will give me silent 6ch amplifier in single enclosure, without need of ground separating resistors or star grounds,

Balanced input is a way to minimize interference (hum/Buzz) picked up in LONG lines carrying weak signals, has absolutely no influence on grounding problems with multiple power amps in a single chassis.

I fear you'll still have to follow proper grounding rules.

 

[valtra103]

Balanced input is a way to minimize interference (hum/Buzz) picked up in LONG lines carrying weak signals, has absolutely no influence on grounding problems with multiple power amps in a single chassis.

I fear you'll still have to follow proper grounding rules.

Balanced removes ground loops. How it can have no influence?

 

[tomchr]

Is this circuit good. I don't have LM3886 model in Spice, so simulation is almost useless with random opamp in it's place.

TINA-TI is a circuit simulator available from TI. It is compatible with the LM3886 model that you can find on the LM3886 Product Page on ti.com. Look in the Software & Tools section for the models. That'll allow you to simulate everything related to stability and circuit performance. The only thing not covered by the TI LM3886 model is supply interaction and, I think, the mute function.

Your circuit looks good at first glance, though. I suggest adding 180 pF across the inputs of the LM3886 and appropriate compensation to deal with the overshoot in the transient response that you get from adding that 180 pF cap.
The LM3886 tends to show a little instability as the output voltage approaches the negative rail. The 180 pF deals with that, but causes overshoot in the transient response. An RC across the feedback resistor usually get things back in line.

I think you had a question about decoupling as well. You can see my take on it on my Taming the LM3886 pages.

One thing i am confused is what minimum gain LM3886 needs in this config, to function properly

The minimum gain of the LM3886 is 20 dB (10x), which is what you have, so life is good. If you need more gain, you can use an instrumentation amplifier front-end.

Tom

 

[valtra103]

TINA-TI is a circuit simulator available from TI
Tom

I forgot about it, thank you. And it's free, I thought it is not. Downloading now

I think you had a question about decoupling as well. You can see my take on it on my Taming the LM3886 pages.
Tom

Thank you. It's very nice that you posted in my thread. I actualy have read all of your page multiple times. Your website is inspiration for this project along with G-Word whitepaper by B. Putzeys. I just mentioned decoupling, so people won't start mentioning that I have none. I will do everything as close to your recomendations as I can.

The LM3886 tends to show a little instability as the output voltage approaches the negative rail. The 180 pF deals with that, but causes overshoot in the transient response. An RC across the feedback resistor usually get things back in line.
Tom

I will add those parts.

 

[valtra103]

I added parts. Is this correct?

 

[AndrewT]

The differential to single ended stage MUST very closely match all components around it if it is to achieve high CMRR (Common Mode Rejection Ratios).
This requires that any component added across R7 to be exactly matched by the same added component across R6.

The 3886 stability margins are not at their best when the gain is 10times (+20dB).
Look at the stability margins in the plot in the datasheet. The 3886 performs better when gain is ~+28dB. This tends to be confirmed by the many posters who comment on audio quality vs gain.
You can adjust the stability margins for any opamp using a noise gain correction. (I think this is what Tomchr is referring to). This technique works with the 3886 and can be used to bring the +20dB gain stability margins up to match those that are available @ +28dB. Your 180pF with an extra resistor is a part of that noise gain adjustment.

With some careful designwork you should be able to get the 3886 to work well at even lower gain than +20dB.

I think R10 <<10r, maybe around 1r0 to 4r7.
With V3 @ 1.6Vac, the 4ohm load will be @ 64W maximum power from the +20dB chipamp. A lower gain may be an advantage.

 

[valtra103]

This requires that any component added across R7 to be exactly matched by the same added component across R6

Thank you for reply. If I understand correctly I should add RC network across R6, with as close nominals as possible.

3886 and can be used to bring the +20dB gain stability margins up to match those that are available @ +28dB. Your 180pF with an extra resistor is a part of that noise gain adjustment.

Are you saying that, if I increase gain, those components are not needed anymore (180pF and R12,C4) ? Because I could increase the gain to 26dB for example. This amp will be powered by active crossover / preamp in separate box. So then I would just need less gain in preamp. My complete system will need to max out power at 0.5-0.7V input. So I would not have excess gain even with 26dB at power amp.

That 4ohm load in circuit is mistake. It will be 8ohms.

 

[AndrewT]

Compare the stability margins for the two gains of +20dB and +28dB (or +30dB, it's nearly the same).
There seems to be a consensus that the 3886 suits the higher stability margins of +28dB, or +30dB.

You can get similar stability margins using lower gain if you change the noise gain.
You can also alter the stability margins by using noise gain adjustment.
Some of tomchr impressive results are probably due to carefully placed components that are improving/changing the stability margins. But he has kept his actual implementation secret. He has given a lot of clues, especially in "taming".

I used it in a +20dB implementation for a computer speaker, but I took a guess at values and it came out OK, rather than "designed" !

Here's a couple of papers showing the potential of noise gain adjustment.

 

[AndrewT]

Gain and pre-amp voltage levels.
With a gain of +20dB in the power amplifier, then for an output of 66W into 8ohms one needs a maximum input of ~2.3Vac
That is roughly the maximum output of a CD player and many other digital Sources.
Feeding a CDP through a unity gain buffer/volume control and into a +20dB power amplifier just hits the 50W to 66W maximum of a 3886 chipamp.
Theoretically any more gain is just wasted and results in extra noise.
i.e +0dB in the pre-amp and +20dB in the chipamp/power stage is about optimum for many of our digital Sources.

 

[valtra103]

Gain and pre-amp voltage levels.
With a gain of +20dB in the power amplifier, then for an output of 66W into 8ohms one needs a maximum input of ~2.3Vac
That is roughly the maximum output of a CD player and many other digital Sources.
Feeding a CDP through a unity gain buffer/volume control and into a +20dB power amplifier just hits the 50W to 66W maximum of a 3886 chipamp.
Theoretically any more gain is just wasted and results in extra noise.
i.e +0dB in the pre-amp and +20dB in the chipamp/power stage is about optimum for many of our digital Sources.

My source is only 0.7V.

 

[valtra103]

Compare the stability margins for the two gains of +20dB and +28dB (or +30dB, it's nearly the same).
There seems to be a consensus that the 3886 suits the higher stability margins of +28dB, or +30dB.

You can get similar stability margins using lower gain if you change the noise gain.
You can also alter the stability margins by using noise gain adjustment.
Some of tomchr impressive results are probably due to carefully placed components that are improving/changing the stability margins. But he has kept his actual implementation secret. He has given a lot of clues, especially in "taming".

I used it in a +20dB implementation for a computer speaker, but I took a guess at values and it came out OK, rather than "designed" !

Here's a couple of papers showing the potential of noise gain adjustment.

Sorry I missed this message. Thanks for PDFs, i will read them. But 20dB gain is too low for me anyway. So could I omit those parts by increasing gain to lets say 26dB?

 

[AndrewT]

Then use B.Putzeys' balanced impedance volume control as your pre-amp.
It has enormous excess gain for lower voltage level sources.

 

[valtra103]

Then use B.Putzeys' balanced impedance volume control as your pre-amp.
It has enormous excess gain for lower voltage level sources.

But why it is bad to increase LM3886 gain to 26dB? Higher stability then, so maybe I don't ever need those extra parts for noise gain increasing? Maybe it would be stable without them at 26dB? All I want to know if LM3886 with higher gain does not need those parts anymore

What I mean, maybe this would be already stable:
[IMGDEAD]https://i.xomf.com/vxvrh.png[/IMGDEAD]

For some reason i can't install Tina Ti... I want to try to simulate it.

 

[AndrewT]

+26dB has better margins than +20dB gain.

But those higher margins are usually not enough.
Look at the output Zobel. It is a stability enhancing component.
Look at the input filtering. It too is a stability enhancing component.
The 3886 NEEDs help. Provide all that it needs to get good performance.

For a 0.7Vac maximum input and with +0dB preamp plus +28.6dB chipamp (1k:27k) you get a maximum output of ~45W
You could go to 1k:30k (+29.5dB) or even 1k:33k (+30.4dB) to get closer to 60W from 0.7Vac

I have tried three times on two different computers and still can't install Tina !

 

[valtra103]

+26dB has better margins than +20dB gain.
Look at the input filtering. It too is a stability enhancing component.
!

How I could add input filtering to this configuration? I know how to add it to normal non inverting configuration, but i don't know how to do it with this.

For a 0.7Vac maximum input and with +0dB preamp plus +28.6dB chipamp (1k:27k) you get a maximum output of ~45W
You could go to 1k:30k (+29.5dB) or even 1k:33k (+30.4dB) to get closer to 60W from 0.7Vac
!

My idea actualy is to have 26dB at LM3886 and 6dB in preamp. I think that runing higher signal level through active crossover is better for signal to noise, so I plan to have 6dB of gain just before crossover. And then 26dB at LM3886.

I have tried three times on two different computers and still can't install Tina !
!

I managed to install it just now. My computer showed unknown file for Tina download. After some thinking I opened it with Winrar, and inside was setup. I will try to do simulations, when i figure out how to use it. LM3886 pins are not even labeled

 

[JMFahey]

Balanced removes ground loops. How it can have no influence?

That's a blanket statement if I ever saw one, and applies when signal is being sent from "here" to "there" so you don't add extra hum along the path or from different grounds used forn preamp and power amp; but if your 6 amps in a box are poorly grounded and humming (which is your original problem), it won't help you with that.

Hope you get the difference.

FWIW you have already been suggested how to do things right, including star grounding , but somehow you seem to despise good advice.

Getting into long discussions about differential input minutiae will not help the main problem: 6 amplifiers in the same chassis sharing the same power supply.

 

[valtra103]

Getting into long discussions about differential input minutiae will not help the main problem: 6 amplifiers in the same chassis sharing the same power supply.

This is thread about differential inputs.
I just want help designing Balanced input LM3886. It is main reason of this thread. I want my 6ch amp with balanced inputs AND correct grounding. Grounding is not problem. Balanced implementation is what I intended to ask. Sorry for unclear post. I edited it to be straight to point.

FWIW you have already been suggested how to do things right

And I am thankful for everyone for sugestions. And I am trying to implement that. The thing is that sugestions vary a little so I am trying to clear up things. Tom sugested cap between inputs. Andrew sugested RC between inputs. So I am just trying to learn and implement sugestions correctly.

 

[valtra103]

Now I have two possible implementations, if I understand correctly. And I am puzzled which way to go.

1.
[IMGDEAD]https://i.xomf.com/xqhgf.png[/IMGDEAD]

2.
[IMGDEAD]https://i.xomf.com/qyjwj.png[/IMGDEAD]

 

[tomchr]

Sorry I missed this message. Thanks for PDFs, i will read them. But 20dB gain is too low for me anyway. So could I omit those parts by increasing gain to lets say 26dB?

No. You still need Cc = 180 pF; Rf2 = 20 kΩ; Cf = 47 pF.

But why it is bad to increase LM3886 gain to 26dB?

Increasing the gain to 26 dB means 6 dB lower loop gain available -> 6 dB worse on most parameters. Andrew is not correct that the LM3886 will operate at compromised stability at 20 dB gain and reducing the loop gain as proposed in the .pdf by Bob Pease, will reduce performance as I described two sentences back.

You'll get better performance if you leave the LM3886 with a gain of 20 dB and put the remaining 6 dB up front in the instrumentation amp front-end. This is because any modern audio op-amp will have better THD and noise than the LM3886, hence, applying the max amount of loop gain possible to the LM3886 will result in the best performance.

What I mean, maybe this would be already stable:
[IMGDEAD]https://i.xomf.com/vxvrh.png[/IMGDEAD]

It'll be stable as long as the output voltage is away from the supply rails. Once the output approaches the negative rail, you'll get oscillation unless you add the stability components.

As Andrew correctly points out, R4-R7 need to be pairwise matched. I.e. R4/R6 = R5/R7 for optimum CMRR. If you use 0.1 % resistors, good layout, and good thermal matching, you should be able to get down near 60 dB CMRR. Considering that resistors R6 and R7 have much higher dissipated power than R4, R5 that may be tough to achieve.

Also: R3 should be 0 Ω.

For some reason i can't install Tina Ti... I want to try to simulate it.

Contact the applications team via the E2E forum on TI's website. They're paid to help you get it going.

Tom