Basic opamp based preamp

[Stormrider]

R4 needs to be 15K for the linear volume control to work. It makes it act like a log pot

ESP - A Better Volume Control

Either way, shouldn't a compensation resistor be in series with signal to pin 3, not going to ground?

Adding 1K from pins 1 to 2 didn't make a difference.

Should that BW limiting cap go from pin 3 to ground?

[Mooly]

Stormrider... just some more thoughts on it all,
You have found for yourself that the DC offset is a real issue with the 5532 as predicted. Another important point I overlooked... that same offset voltage and bias input current will make the volume control noisy in operation.

With no output caps, R7 is not required.

The noise added by the resistors is absolutely negligable in a circuit like this.

When you measure DC offsets make sure the input to the preamp is shorted to ground. If it's floating and picking up noise etc then that may give the impression of a DC shift as the volume is turned.

The balanced line driver stage doesn't cancel DC offset and will actually add it's own offset contribution to the final output. Again using FET opamps this is negligable.

[stratus46]

Originally Posted by Stormrider

Tried dropping it to 10K, no change in offset voltage.



Tried this, and even though it makes perfect sense, no change.

If I drop R2, the 100K, down to 22K the dc offset on pin 1 drops to ~5mVdc but that will make my input impedance way too low.

What are you using it with that 22K is way too low? Tube gear maybe, solid state unlikely.

G²

[Mooly]

Adding a resistor that's equal in value to R2 from the output to inverting input (instead of direct link) will minimize offset but thats not ideal from a noise pickup point of view.
The input bias current of - and + inputs have to be the same for minimum offset with bipolar devices.

The answer is to use FET opamps or if you are intent on bipolar, use single devices and use the offset null facility.

Actually my preferred option would be to use the opamps in a virtual earth configuration to minimise common mode effects. Signal polarity would be maintained.

[AndrewT]

if you feel you need a little gain then put that stage first instead of the unity gain buffer.
If you insist on DC coupling then learn how to eliminate the offsets of every stage. Don't let them ripple through.
If you do go with DC coupling then add a DC servo to correct the little bit of offset that is left and will vary with temperature. Finally for a DC coupling pre-amp you must include an output mute that is triggered by an output offset detector. You don't want an output offset that gets multiplied and passed to your speaker, because the source or the preamp goes faulty and the servo is incapable of servoing out a big offset.

An alternative is to AC couple the whole preamp. Good polypropylene capacitors of affordable sizes can be used without affecting the audio band if you select resistances/impedances appropriately.

The input buffer is best located at the source equipment, not at the input of the preamp. If the sources cannot drive the cables then sort the sources. A buffer at the front of the pre-amp does not sort the sources.

[Stormrider]

Thanks for the help everyone. I am going to breadboard a DC servo and test it out, but this is a new(er) schematic:

I calculated C2 using the formula C=1/(2*pi*6800*200Khz)

[Mooly]

Everyone will have their own ideas on this.
You are not understanding the offset issue and how it works. R3 isn't needed for a FET device... with a bipolar opamp you can't equalise the input bias currents as long as the pot wiper is DC coupled.
Short out R3 and the design is OK using OPA2134 etc. Some may argue that C1/R2 is a little limiting for extreme LF performance and increasing R2 to 220k may be better.

I'm surprised the original uses a 100k pot. Stray capacitance/input capacitance of U1B means the HF response varies slightly with volume setting. I would use 10k here and alter R4 accordingly.

It's fine... any DC offset with an OPA2134 will be way to small to cause any issues.

Servos have their place, but I don't think it's here really. They add extra complications and add another "pole" into the LF response.

[Stormrider]

Originally Posted by Mooly

Everyone will have their own ideas on this.
You are not understanding the offset issue and how it works. R3 isn't needed for a FET device... with a bipolar opamp you can't equalise the input bias currents as long as the pot wiper is DC coupled.
Short out R3 and the design is OK using OPA2134 etc. Some may argue that C1/R2 is a little limiting for extreme LF performance and increasing R2 to 220k may be better.

I'm surprised the original uses a 100k pot. Stray capacitance/input capacitance of U1B means the HF response varies slightly with volume setting. I would use 10k here and alter R4 accordingly.

It's fine... any DC offset with an OPA2134 will be way to small to cause any issues.

Servos have their place, but I don't think it's here really. They add extra complications and add another "pole" into the LF response.

I calculated a 1.6hz Fc for the C1/R2 combo. F=1/(2*pi*RC)

That extra 15K is a mistake; not sure why I put it in there.

[Stormrider]

Ok, how's this look?



Dropping the value of the pot to 10K would make R4 a 1.5K resistor to keep the ratio the same. Is that too much of a load for the first stage?

[Mooly]

The OPA2134 is rated OK for driving 600ohm loads (to within a couple of volts of the supplies) so no problems there at all.
The input RC... well some will argue for lower
It's not just the reduction in output but the phase shift that's introduced too... but again it's fine. With a cap at the output the same applies depending of course on the load the final stage works into. Some power amps may be quite low input impedance.
If you are thinking of feeding a power amp directly from this then some reliable relay delay on the speaker output may be worthwhile... the switch on thump could be pretty large
You could also always add a normally closed relay across the output of the preamp that opens after a couple of seconds... that's easy to arrange.