Hi,

I'm building this buffer. Today when I tested it playing the strings very loud it got a little distorted.
So I think that maybe the biasing isn't that well designed. I got 1.1V on the opamp +input.
I know I should lower the value of R8, but I don't want to "lose" input impedance. Is there any criteria to follow in this case?

Thanks

I believe you're having problems with input bias current.

Although ideal opamps have infinite input impedance and require zero input bias current, this is not true of real opamps. It *matters* what that opamp you have covered up is. If it's an NE5532, it will take a lot of input current because the input devices are bipolar transistors. If it's a JFET input like the TL07x series, it will take almost - not quite - zero input current.

You really have two choices. (1) reduce your 10M input resistor so the input bias current does not make the input voltage sag so much, or (2) use two resistor in series instead of that 10M, perhaps two 100k's, and bootstrap the junction of the two 100Ks to the output to increase the impedance to AC signal while leaving it low for DC.

Quote from: Labaris on May 23, 2013, 10:16:30 PM
I got 1.1V on the opamp +input.

That is because your voltmeter is pulling the voltage down. The true voltage may be OK in reality. Triple check for a mis-wired connection elsewhere (but as RG hinted, make sure you're using a FET opamp like the TL07x).

Quote from: merlinb on May 24, 2013, 04:50:33 AM

Quote from: Labaris on May 23, 2013, 10:16:30 PM
I got 1.1V on the opamp +input.

That is because your voltmeter is pulling the voltage down. The true voltage may be OK in reality. Triple check for a mis-wired connection elsewhere (but as RG hinted, make sure you're using a FET opamp like the TL07x).

So if he puts the DVM on the opamp output (pin1) instead - it should closely reflect whatever DC level is on the +ve (input) pin, with no DVM loading issues to worry about.

Quote from: R.G. on May 23, 2013, 10:59:01 PM

You really have two choices. (1) reduce your 10M input resistor so the input bias current does not make the input voltage sag so much, or (2) use two resistor in series instead of that 10M, perhaps two 100k's, and bootstrap the junction of the two 100Ks to the output to increase the impedance to AC signal while leaving it low for DC.

One other option would be to tweak (decrease) the value of R22 (& leave R23 as it is)...measure the DC output on pin1 of the opamp until the most appropriate (lower) value of R22 has the output at about 4.5V (in other words shift the voltage of VB up to conteract the voltage drop across your 10M resistor.....your schematic only shows one stage, so it's an option here....but if you've several stages...obviously tweaking VB will impact the other stages too!)

If you leaving the buffer in the signal path all the time I would

Remove R7 and make R8 a 4.7meg or 5.1meg this would be the close to same input resistance, if you are not going to switch the circuit in and out you don't need R7.

Use a fet input opamp as others posted.

You often don't need much more than 1meg input  for common passive pickups.

Not posting how you are going to use this circuit makes it difficult to give circuit adjustment advice.

You are missing an input series connected resistor between R8 and the + input.  I can not see the output part of the circuit the 1/2 power supply part is blocking it.  Do you have a series output resistor?

Quote from: Gurner on May 24, 2013, 05:06:59 AM
One other option would be to tweak (decrease) the value of R22 (& leave R23 as it is)...measure the DC output on pin1 of the opamp until the most appropriate (lower) value of R22 has the output at about 4.5V (in other words shift the voltage of VB up to conteract the voltage drop across your 10M resistor.....your schematic only shows one stage, so it's an option here....but if you've several stages...obviously tweaking VB will impact the other stages too!)

Yep, you can do that. I cheated - I didn't tell him that because it's a fragile solution with lots of drift issues and is dependent on the device plugged into the opamp socket.

Thanks for all the replies.

1. I was using an NE5532 but the design has a TL072 (don't have any FET opamp right now). Maybe that's all the problem...
2. I'll measure DC on the output, thanks for the tip.
3. I don'd really understand the bootstrapping technique that R.G. showed
4. I need R7 there, because it's part of a switchable L-pad built between the input jack and the buffer input.
5. Why should I need and input series resistor on pin 3?
6. The design includes a series output resistor before the decoupling cap.

A NE5532 and 10 meg are not a good mix.

http://www.diystompboxes.com/smfforum/index.php?topic=93793.0
http://www.diystompboxes.com/smfforum/index.php?topic=93330.0

Quote from: Labaris on May 24, 2013, 07:22:39 PM
1. I was using an NE5532 but the design has a TL072 (don't have any FET opamp right now). Maybe that's all the problem...

The input bias current specified for the On Semi NE5532 is typically 200na, and max of 800ma. 200nA times 10M ohms is 2V. So a typical NE5532 could be expected to sag by 2V from the Vbias point the series resistors generate. Your meter (they are often 10M, but not always) would cause even more sag measuring in parallel at the input. That's for an average device. A worst-case one with 800nA

Quote3. I don'd really understand the bootstrapping technique that R.G. showed

It adds a bit of the output back to the middle of the bias resistor. Since the output is very nearly the same size as the input signal, it forces the bias voltage up when the signal is going up, and down when the signal is going down. This, in effect moves the input "bias" voltage in the same direction as the input is trying to go, and means there is a much lower voltage across half the input bias resistor. So the input signal doesn't have to provide so much current. It is effectively seeing a higher input resistance. This has to be done with care, because it is a form of positive feedback, but it can help in many cases.

Note that the spec sheet for the NE5532 also notes that the input impedance of just the 5532 by itself is "typically" 300k, it can be as low as 30k. Bootstrapping can't help that much without getting very near oscillation.

The NE5532 is good for many things, but it's not a high input impedance device.

Quote
6. The design includes a series output resistor before the decoupling cap.

Good. That will make subbing in a TLO7x OK with output capacitive loading.

Quote from: Gus on May 24, 2013, 08:14:07 PM
A NE5532 and 10 meg are not a good mix.

http://www.diystompboxes.com/smfforum/index.php?topic=93793.0
http://www.diystompboxes.com/smfforum/index.php?topic=93330.0

Thanks for the info.
-I get what's bootstrapping is about now, but adding four more components to the board is impossible (these are two buffers that share the input stage).
-The input series resistor on +input is for protection... against what?
-Maybe having DC offset of 2V (for example) isn't that bad if we consider that a buffer has unity gain. Am I wrong?

Quote from: R.G. on May 24, 2013, 09:05:02 PM

Quote from: Labaris on May 24, 2013, 07:22:39 PM
1. I was using an NE5532 but the design has a TL072 (don't have any FET opamp right now). Maybe that's all the problem...

The input bias current specified for the On Semi NE5532 is typically 200na, and max of 800ma. 200nA times 10M ohms is 2V. So a typical NE5532 could be expected to sag by 2V from the Vbias point the series resistors generate. Your meter (they are often 10M, but not always) would cause even more sag measuring in parallel at the input. That's for an average device. A worst-case one with 800nA

Quote3. I don'd really understand the bootstrapping technique that R.G. showed

It adds a bit of the output back to the middle of the bias resistor. Since the output is very nearly the same size as the input signal, it forces the bias voltage up when the signal is going up, and down when the signal is going down. This, in effect moves the input "bias" voltage in the same direction as the input is trying to go, and means there is a much lower voltage across half the input bias resistor. So the input signal doesn't have to provide so much current. It is effectively seeing a higher input resistance. This has to be done with care, because it is a form of positive feedback, but it can help in many cases.

Note that the spec sheet for the NE5532 also notes that the input impedance of just the 5532 by itself is "typically" 300k, it can be as low as 30k. Bootstrapping can't help that much without getting very near oscillation.

The NE5532 is good for many things, but it's not a high input impedance device.

Quote
6. The design includes a series output resistor before the decoupling cap.

Good. That will make subbing in a TLO7x OK with output capacitive loading.

Thanks. Very clear now.
And I never thought that NE5532 would be a so bad choice for this job.
The output resistor was taken from your on-board buffer. How do you pick the value oh that resistor?

Quote from: Labaris on May 24, 2013, 09:18:01 PM
The output resistor was taken from your on-board buffer. How do you pick the value oh that resistor?

It has to be big enough to stop the oscillation and small enough so that with the anticipated loading it doesn't divide down the signal noticeably. There may be a perfect value for a given circuit, but I don't know how to find it. Something between 22 ohms and 1K will work. Pick one. Mostly, they'll all work.