Inverting LM3886

[sghr220]

Hi all 🙂,

I am trying to build a stereo LM3886 based amplifier using the inverting configuration. The schematic below is for one channel.

The amplifier board will run from a +-30V regulated supply with 30000uF per rail.
The audio input will be fed to both channels by a 50K ohm dual potentiometer.

Questions:

1- What is the recommended value for C7 (bypassing the +ve terminal to ground)?
2- Are there any additional components that i am omitting from the circuit (besides the R//L output network)?
3- Where should the 220pF capacitor (RF filter) be connected, between the +ve and -ve terminals or between the -ve and ground terminals?

An externally hosted image should be here but it was not working when we last tested it.

​

Best Regards.​

 

[infinia]

the gain is only -2 ( not recommended , unstable ) and the input impedance <10K , you'll probably need to add another op-amp to buffer the input. and might need to use a whole new circuit. it's been done before.

 

[sghr220]

the gain is only -2 ( not recommended , unstable )

Sorry that's my mistake R1=1K not 10K so the gain is -20V/V.

the input impedance <10K , you'll probably need to add another op-amp to buffer the input. and might need to use a whole new circuit. it's been done before.

The input impedance is determined by R1 which in this case is 1K so that should be low for the source to drive right? What about the input potentiometer before the input cap? and Why is these values recommended in the LM4780 datasheet (apart from being an honest mistake by National's engineers)?

Best regards.

 

[Robert Kesh]

the gain is only -2 .

I make it -200K/10k = -20

Sorry that's my mistake R1=1K not 10K

That would make it -200

 

[sreten]

Hi,

A very good question is why are you using the inverting configuration.
I cannot see any point to using it. Especially with a dual 50K pot.

rgds, sreten.

 

[sghr220]

I make it -200K/10k = -20


That would make it -200

Sorry for the confusion, i updated the posted schematic to:

Rf = 200k
Ri = 10k
Rb (R7) = 10k

So the gain is -20 V/V and the input impedance = 10k approximately. 🙂

Now for my questions:

1- What is the recommended value for C7 (bypassing the +ve terminal to ground)?
2- Are there any additional components that i am omitting from the circuit (besides the R//L output network)?
3- Where should the 220pF capacitor (RF filter) be connected, between the +ve and -ve terminals or between the -ve and ground terminals?

Best regards.

 

[sghr220]

Hi,

A very good question is why are you using the inverting configuration.
I cannot see any point to using it. Especially with a dual 50K pot.

rgds, sreten.

* Better THD (no common mode distortion in the input stage).
* No electrolytic capacitor in the signal path (Ci).
* I want to try it as the title of the topic suggest. 🙂

1- What is the recommended value for C7 (bypassing the +ve terminal to ground)?
2- Are there any additional components that i am omitting from the circuit (besides the R//L output network)?
3- Where should the 220pF capacitor (RF filter) be connected, between the +ve and -ve terminals or between the -ve and ground terminals?
4- What is the recommended value for the dual pot?

Best regards.

 

[sghr220]

Should i dispense with the input pot completely and control the source instead?

 

[sgrossklass]

I am not sure how well the LM3886 works with relatively high feedback Rs like that. Plus, there's the problem with the 50k pot and its worst-case 12k5 output impedance (a 10k would be a better match for a 10k input impedance).

Both issues could be addressed by using an opamp buffer. (If you want to stick with the 50k pot, pick a part with low input impedance distortion and moderate current noise. Decent output driving capability much recommended.) Then you could reduce the LM3886 input resistor to, say, 2k2.

Your resistor R2 should match the DC resistance that the other input sees, i.e. (R1 + R3) || R4. BTW, the use of R3 is debatable - it is only needed to reduce the much higher input impedance of the noninverting circuit. With R3 removed, R2 = R4. (With a preceding DC-coupled opamp buffer, it's R1||R4.) You need to do this in order to keep DC offset low.

C7 should form a near-short at audio frequencies, in order to avoid introducing extra resistor noise via R2. Try 10µ or so. Should be a bipolar.

The RF filter cap should go before the buffer, IMO. You can reserve a space for the cap between the inputs of the LM3886 but I'm not sure how useful it would be. If in doubt, determine cellphone sensitivity empirically.

 

[sghr220]

I am not sure how well the LM3886 works with relatively high feedback Rs like that. Plus, there's the problem with the 50k pot and its worst-case 12k5 output impedance (a 10k would be a better match for a 10k input impedance).

Thorsten's original inverted gainclone used a feedback resistor Rf of 220K and input resistor of 10k so i am guessing there'll be no problem there.

I have the 10k dual pot but i am leaning towards not using an input attenuator of any kind and control the source level instead.

Your resistor R2 should match the DC resistance that the other input sees, i.e. (R1 + R3) || R4. BTW, the use of R3 is debatable - it is only needed to reduce the much higher input impedance of the noninverting circuit. With R3 removed, R2 = R4. (With a preceding DC-coupled opamp buffer, it's R1||R4.) You need to do this in order to keep DC offset low.

I am planning to use a 220k trimmer from Vin+ of the amplifier to the signal ground to adjust the DC offset value.

C7 should form a near-short at audio frequencies, in order to avoid introducing extra resistor noise via R2. Try 10µ or so. Should be a bipolar.

Is there a formula to calculate the value of this capacitor?

The RF filter cap should go before the buffer, IMO. You can reserve a space for the cap between the inputs of the LM3886 but I'm not sure how useful it would be. If in doubt, determine cellphone sensitivity empirically.

Should the RF filter cap (300pF) connected between the amplifier terminals or between the -ve terminal and the ground?

I am not really sure i understand the "cellphone" bit completely. 😱

Thank you for your much appreciated input. 🙂

Best regards.

 

[sghr220]

Another question...

In the schematic below the feedback loop is taken from after the output resistance (0R22) Why is that? and what is the purpose of this resistance anyway?

An externally hosted image should be here but it was not working when we last tested it.

Best regards.
​

 

[ashok]

Usually a resistance at the output is to decouple the output from any capacitive load which might lead to instability. Not quite so sure why the NFB is taken after that resistor. Could be only stability related but the only thing I can see is that the output impedance will probably be higher than the output stage impedance in this case. Someone must have the correct answer.
Also if the resistor is within the NFB loop then the output impedance seen by the load will again be very low and lower than 0.22E + Zout ( with NFB taken before the resistor). So maybe 0.22E is enough to decouple the output stage from a capacitive load and including it into the NFB reduces it's effect on Zout .... ?

 

[sghr220]

Usually a resistance at the output is to decouple the output from any capacitive load which might lead to instability. Not quite so sure why the NFB is taken after that resistor. Could be only stability related but the only thing I can see is that the output impedance will probably be higher than the output stage impedance in this case. Someone must have the correct answer.
Also if the resistor is within the NFB loop then the output impedance seen by the load will again be very low and lower than 0.22E + Zout ( with NFB taken before the resistor). So maybe 0.22E is enough to decouple the output stage from a capacitive load and including it into the NFB reduces it's effect on Zout .... ?

Thanks Ashok for sharing your thoughts 🙂... What do you think the value of C7 should be? are there any formula to calculate it?

Best regards.

 

[sghr220]

Hi all,
Below is a revised version of the circuit, still can't decide on the following:

1- Is the value of C7 suitable (1uF)?
2- Is it better to use a trimmer for R2 (10K)?
3- The feedback loop is taken after the 0.22 R, Is there a problem with that? and what should be the rating of this resistor?

An externally hosted image should be here but it was not working when we last tested it.

Best regards.
​

 

[Mooly]

Bob Cordell has a circuit for an inverting LM3886 in his book, which uses an OPA2604 opamp as a front end to give DC precision.

In your circuit above the feedback should be before the 0.22. The 10k and 1uf aren't really doing much tbh and can be linked out or you could use the node as a point to inject a DC current to trim offset to zero.

 

[sghr220]

Bob Cordell has a circuit for an inverting LM3886 in his book, which uses an OPA2604 opamp as a front end to give DC precision.

In your circuit above the feedback should be before the 0.22. The 10k and 1uf aren't really doing much tbh and can be linked out or you could use the node as a point to inject a DC current to trim offset to zero.

Thanks Mooly for your input, i really want to keep it simple without compromising the sound quality (more on that later) so wouldn't the 10k trimmer be able to reduce the DC offset to a minimum?

Here's a quote from decibel dungeon:

UPDATE 9th June 2003

Today, I removed the 18K and 22K resistors that went from pins 7 and 8 to signal ground. I put them there to reduce the DC offset (the original circuit had a plain wire between pin 7 and signal ground) of 28mV, to near zero. I now have 216K between the non-inverting input and signal ground, the DC offset is still zero but the magic sound of the original set up has returned. It is quite clearly a much better sounding arrangement!

I am aware of Bob Cordell complementary input stage and the JFET input stage but i really don't like all that to a power amplifier circuit that's supposed to be simple (three resistor amp) in the first place. That in my humble opinion negates the point.🙂

Best regards.

 

[sghr220]

In your circuit above the feedback should be before the 0.22.

Isn't this 0.22 R an attempt to isolate both the output stage and the feedback takeoff point from capacitive loads?

 

[tomchr]

The Zobel network (R5, C6) needs to be on the "inside" of the 0R22. Put it right up agains the IC pin. I suggest using 33 nF and 10R rather than the values proposed by National Semiconductor. 100 nF zobel cap will dissipate a lot of power in the resistor.

C8 is shorted in your schematic. Remove it.

I do realize you have 3300 uF on each rail, but you'll need something smaller in addition to the 100 nF and 3300 uF. Seriously! Put a 10 uF tantalum by the IC pins.
The reason for this is the internal parasitic inductance (ESL) of the 3300 uF cap. The point of the bypass caps is to have a low supply impedance. The impedance of a cap is high at DC but decreases with increasing frequency - until it hits the self-resonance frequency (SRF) of the cap. For your cap that probably happens around 100 kHz - 1 MHz (my guess). The SRF of the 100 nF cap is much higher but its impedance at 1 MHz isn't low enough to ensure a low supply impedance. That's why you need the 10 uF in there as well. I hope this makes sense.

1) C7 should be sized such that the pole formed by C7 and R2 falls well below the audio band. So, 1/(2piR2C7) should be 2 Hz or lower.
2) Don't use a trimmer for R2. Use 10k||200k if you're concerned about DC offset.
3) The feedback should be AFTER the 0R22 just as you've drawn it. You can calculate the power rating of the resistor from the worst case swing and load. That would be the highest voltage swing and the lowest load resistance. Use a resistor with 3~4x the max dissipated power. Otherwise, it'll get screaming hot. The 0R22 is intended to keep the amp stable with capacitive load. I don't think it's the best way for an audio amp, but I guess it'll work well enough.

Bob Cordell's circuit does not improve the DC performance of the LM3886. It's an inverter with a DC servo. Cordell uses an inverting op-amp and an op-amp buffer to invert the phase and provide a high input impedance.

~Tom

 

[sghr220]

That's a great answer, thanks Tom.🙂

The Zobel network (R5, C6) needs to be on the "inside" of the 0R22. Put it right up agains the IC pin. I suggest using 33 nF and 10R rather than the values proposed by National Semiconductor. 100 nF zobel cap will dissipate a lot of power in the resistor.

Sorry about that and i will consider the values you suggested.

C8 is shorted in your schematic. Remove it.

No it's not shorted its between the positive and negative supply pins of the LM3886.

I do realize you have 3300 uF on each rail, but you'll need something smaller in addition to the 100 nF and 3300 uF. Seriously! Put a 10 uF tantalum by the IC pins.
The reason for this is the internal parasitic inductance (ESL) of the 3300 uF cap. The point of the bypass caps is to have a low supply impedance. The impedance of a cap is high at DC but decreases with increasing frequency - until it hits the self-resonance frequency (SRF) of the cap. For your cap that probably happens around 100 kHz - 1 MHz (my guess). The SRF of the 100 nF cap is much higher but its impedance at 1 MHz isn't low enough to ensure a low supply impedance. That's why you need the 10 uF in there as well. I hope this makes sense.

Thanks for the detailed explanation but would a 2.2uF tantalum do as it's hard to get a 10uF rated at 35V and above?

1) C7 should be sized such that the pole formed by C7 and R2 falls well below the audio band. So, 1/(2piR2C7) should be 2 Hz or lower.
2) Don't use a trimmer for R2. Use 10k||200k if you're concerned about DC offset.
3) The feedback should be AFTER the 0R22 just as you've drawn it. You can calculate the power rating of the resistor from the worst case swing and load. That would be the highest voltage swing and the lowest load resistance. Use a resistor with 3~4x the max dissipated power. Otherwise, it'll get screaming hot. The 0R22 is intended to keep the amp stable with capacitive load. I don't think it's the best way for an audio amp, but I guess it'll work well enough.

Bob Cordell's circuit does not improve the DC performance of the LM3886. It's an inverter with a DC servo. Cordell uses an inverting op-amp and an op-amp buffer to invert the phase and provide a high input impedance.

~Tom

1- C7 should be 10uF for a pole @ 1.59 Hz.
2- Trimmer will give more flexibility i guess?
3- Thanks for confirming that 🙂, would a 10W resistor do for 28V supply driving a 4 Ohm speaker?

Best regards.

 

[sghr220]

I do realize you have 3300 uF on each rail, but you'll need something smaller in addition to the 100 nF and 3300 uF. Seriously! Put a 10 uF tantalum by the IC pins.
The reason for this is the internal parasitic inductance (ESL) of the 3300 uF cap. The point of the bypass caps is to have a low supply impedance. The impedance of a cap is high at DC but decreases with increasing frequency - until it hits the self-resonance frequency (SRF) of the cap. For your cap that probably happens around 100 kHz - 1 MHz (my guess). The SRF of the 100 nF cap is much higher but its impedance at 1 MHz isn't low enough to ensure a low supply impedance. That's why you need the 10 uF in there as well. I hope this makes sense.
~Tom

That's really interesting and then some....

SMALLER IS SOMETIMES BETTER
What we need is a capacitor with a higher self-resonant frequency. And the truth about capacitors is this: for a given capacitor type, smaller capacitor values generally have higher self resonant frequencies.
As I show below, the reason why a smaller value capacitor appears to be better is not because it has lower L, but because it has lower C. We need good decoupling, not a low value for C x L, which means a low value for C.

So according to this information should i use say 1000uF per rail given that the power supply has enough capacitance?

Best regards.