TDA2822 doesn't play nice with op-amp buffer

[Ksander]

Dear all,

A while back, I made a simple TDA2822 amplifier, based on the ruby amp and the TDA datasheet (basically the attached schematic, which I quickly cobbled together). Although it worked reasonably well, the 'tone control' that I made was nonsense (not pictured), and I recently decided to redo it.

As an upgrade, I wanted to use a dual op-amp (tried TL072 and NE5532) with one side as a buffer, and the other using some -yet undecided- tone control circuit. However, even with only the buffer, this makes the speaker just buzz loudly, and I get no guitar sound. Bypassing the buffer, directly inserting the guitar signal in the TDA does work fine (but gives me no tone control).

The op-amp buffer is set up as follows:
- the guitar signal going into a 100nF cap,
- then there is a 100K-100K voltage divider to raise the signal to 4.5v
- this goes into the non-inverting input
- the inverting input and output are tied together
- the output is sent through a capacitor (tried 100nF, 1uF, 10 uF, 100uF) to the TDA input.

So basically, I just replaced the jfet buffer with an op-amp buffer. The TDA part of the circuit is unchanged. I also tried placing 100R, 1K, 10K and 100K resistors in series between the buffer output and TDA input, but this made no difference.

The power lines have 10uF and 100nF capacitors between them as you see often. The buzzing is present with and without them.

Any ideas why the op-amp buffer does not work, and what I could do to get it to work?

[antonis]

Is the other half of dual op-amp left "floating" (unconnected somehow)..??

[fryingpan]

Two things:

- 100K // 100K is far too low an impedance for guitar. You should aim at 1M (some circuits present as far down as 300K and work fine, if a bit muted on the treble, but the "standard" is 1M).
- some opamps do not like their output and inverting input tied just like that. Try adding a resistor in between.

[Ksander]

Thanks for pointing this out  However, when i probe the output of the op-amp, i get a fine signal, so that can't be the reason why the power amp oscillates/buzzes?

[Ksander]

Is the other half of dual op-amp left "floating" (unconnected somehow)..??

No, i tied the non-inverting input to ground and the inverting and output together.

[PRR]

i tied the non-inverting input to ground and the inverting and output together.

On a single-ended supply? Do you see how that might not work?

[ElectricDruid]

The simple 100K/100K divider for the Vref is going to give you a rather low input impedance of around 50K. You'd be better off with 47K/47K with a 47u in parallel with the lower resistor to make the Vref, and then connect the +ve input to that with 1Meg.

And +1 what Paul said - connect the unused op-amp's input to Vref, not Ground. Ground only works on a bipolar power supply.

[Ksander]

I got it like this now, with your advice. It works much better, in that I get sound. However, with the pot above some 25% there is horrible distortion and popping. This did not happen with the jfet buffer. Any ideas on why this may be?

Edit: could it have to do with the circuit being on a breadboard? The circuit with the jfet (j113) is soldered on a solderable breadboard and this drives the speaker loudly without issues. As a test, I copied the jfet circuit on the breadboard, and then it also gets noisy with the pot beyond some low threshold value.

[Ksander]

That was it. I soldered it up, and it works well with the proposed suggestions. Thanks!

It does need a chunky psu tho

[fryingpan]

Did you use a NE5532? Input impedance is only 300K typical. A TL072 would work better. Even bootstrapping the NE5532, IIRC from Douglas Self's measurements, would get you worse noise than with the TL072, admittedly not a quiet opamp.

Since you already have a JFET, why not try Rod Elliott's JFET-opamp guitar preamp?

https://sound-au.com/articles/jfet-design.htm#s6

Careful that there is a typo and he calls two different resistors R3 (one is a trimpot, the other is a 1M or maybe 100K resistor, if you want to try it with a bipolar-input opamp). Also, for some reason all the hi-fis have become hi-ifs.

[m4268588]

Input impedance is only 300K typical.

[PRR]

NE5532? Input impedance is only 300K

Open loop.

What happens closed-loop?

[Ksander]

What happens closed-loop?

I don't really know what to look for in the datasheet, but I don't see anything on closed loop resistance. Regardless, and/or coincidentally, I ended up using the TL072 because the NE5532 sounded noisier.

The jfet circuit I had before was from the ruby amp, and is more or less the same as jfet buffers on muzique.com. The ESP looks more advanced. The reason I didn't go that way is that I build stuff by daisychaining functional blocks that others have come up with, and I found some simple buffer and tone-control circuits using op-amps that seemed suitable for what I wanted.

[fryingpan]

NE5532? Input impedance is only 300K

Open loop.

What happens closed-loop?

Yes, input impedance should go up (a lot) thanks to feedback. Nevertheless, a lot of designs (eg. Tubescreamer) use a discrete buffer before a 4558 (another bipolar opamp) and not many designs, if any, use a discrete buffer before a TL07x, for instance. Art least for noise considerations.

[fryingpan]

What happens closed-loop?

I don't really know what to look for in the datasheet, but I don't see anything on closed loop resistance. Regardless, and/or coincidentally, I ended up using the TL072 because the NE5532 sounded noisier.

The jfet circuit I had before was from the ruby amp, and is more or less the same as jfet buffers on muzique.com. The ESP looks more advanced. The reason I didn't go that way is that I build stuff by daisychaining functional blocks that others have come up with, and I found some simple buffer and tone-control circuits using op-amps that seemed suitable for what I wanted.

It's not much more advanced, really. It's just "fancier". It has high THD, all 2nd harmonic, and a lot of potential gain.

[antonis]

I don't really know what to look for in the datasheet, but I don't see anything on closed loop resistance.

It's increased by (1 + A_o*β), where A_o = open-loop gain and β = feedback fraction (inverse of what's left after subtracting unity from closed-loop gain)..

Another way to see it is:  Z_{in(closed-loop)} = Z_{in(open-loop)} * (A_o + G) / G, where G = closed-loop gain..

[fryingpan]

Then why use bootstrapping in the first circuit?

https://sound-au.com/project202.htm

According to the formula, assuming even a gain of ten (20dB), and assuming that an audio opamp can easily have 100dB open loop gain (the 4580 states 90dB minimum, 110dB typical):

Zin = 300000 * (100000 + 10) / 10 =
approx. 3Gohms. Even the lowest capacitance piezo pickup may require, say, 100Mohms load impedance for full bandwidth.

(If you use the above formula with 90dB open loop gain, approx. 30000x, then it still works out at 900Mohms).

[antonis]

Then why use bootstrapping in the first circuit?

https://sound-au.com/project202.htm

According to the formula, assuming even a gain of ten (20dB), and assuming that an audio opamp can easily have 100dB open loop gain (the 4580 states 90dB minimum, 110dB typical):

Zin = 300000 * (100000 + 10) / 10 =
approx. 3Gohms. Even the lowest capacitance piezo pickup may require, say, 100Mohms load impedance for full bandwidth.

(If you use the above formula with 90dB open loop gain, approx. 30000x, then it still works out at 900Mohms).

I suggest you a good cup of coffee, fryingpan..   

Closed-loop input impedance is severely dominated by bias configuration..

[fryingpan]

Then why use bootstrapping in the first circuit?

https://sound-au.com/project202.htm

According to the formula, assuming even a gain of ten (20dB), and assuming that an audio opamp can easily have 100dB open loop gain (the 4580 states 90dB minimum, 110dB typical):

Zin = 300000 * (100000 + 10) / 10 =
approx. 3Gohms. Even the lowest capacitance piezo pickup may require, say, 100Mohms load impedance for full bandwidth.

(If you use the above formula with 90dB open loop gain, approx. 30000x, then it still works out at 900Mohms).

I suggest you a good cup of coffee, fryingpan..   

Closed-loop input impedance is severely dominated by bias configuration..

Yes, but you could even use a 100Mohm bias resistor (you can actually find them) and most of the noise will be shunted to ground. Bootstrapping itself reduces Johnson noise but increases the impact of Johnson noise because it's "in the circuit".

[fryingpan]

Then why use bootstrapping in the first circuit?

https://sound-au.com/project202.htm

According to the formula, assuming even a gain of ten (20dB), and assuming that an audio opamp can easily have 100dB open loop gain (the 4580 states 90dB minimum, 110dB typical):

Zin = 300000 * (100000 + 10) / 10 =
approx. 3Gohms. Even the lowest capacitance piezo pickup may require, say, 100Mohms load impedance for full bandwidth.

(If you use the above formula with 90dB open loop gain, approx. 30000x, then it still works out at 900Mohms).

I suggest you a good cup of coffee, fryingpan..   

Closed-loop input impedance is severely dominated by bias configuration..

Yes, but you could even use a 100Mohm bias resistor (you can actually find them) and most of the noise will be shunted to ground. Bootstrapping itself reduces Johnson noise but increases the impact of Johnson noise because it's "in the circuit".

Of course, the problem at that point is that a large resistance to ground leads to huge voltage offset  (but that could potentially be true with even 1Mohm -> 1uA bias current * 1Mohm = 1V). In fact I think that for such a use case (target impedance higher than, say, 100K) you really want to use a FET-input opamp...