Basic, high fidelity buffer

[jul059]

Hello everyone,

I am designing a high fidelity guitar buffer. Maybe overkill, but it is also a learning experience.



The circuit is inspired by a very famous buffer, you may recognize it. I would appreciate your opinion, specifically on :

R4 : I would like it to be lower, but simulations show that the buffer becomes an amplifier at very high frequencies ( >100 kHz) if the resistor is below ~4.5k. Any idea on why that is? I will also try to use 1w metal film resistor for R4 to reduce resistor noise since it is in the signal path (http://www.aikenamps.com/index.php/resistor-types-does-it-matter). Logical?

C7 : Highest film capacitor value I can easily get to reduce low frequency attenuation.

C1 : opamp supply bypass cap. don't know if it will do much... But Rod Elliott http://sound.westhost.com/dwopa.htm says it reduces oscillations in opamps. wire should be as short as possible and as close to the opamp pins as possible.

C9 : I will use 2x1uF film caps in parallel. Again, I want to reduce low frequency attenuation, but I don't want to use electrolytic because of poor audio performance in signal path.

R6 : Really, I have no idea what a good value for that is...

R2 : Here, most opamp buffers on the internet use a 1Meg resistor. My simulations shows that the lower the value, the less the buffer will act as a unity gain and will attenuate all frequencies. So why not 10Meg? Thermal noise issues?

C8 : "copied" from the famous buffer. Don't know much about what it does.

Finally, R1 and R3 : I guess the idea is to strike a good balance between current flowing through R1 + R3 and added thermal noise. With the values shown, current in R1 is about 1.4 mA. Very reasonable.

Thanks a lot

P. S. : obviously, V2 is the guitar input.

[knutolai]

Have a read at these articles. They will probably answer some of your questions regarding frequency roll-off and impedance.
http://www.geofex.com/circuits/what_are_all_those_parts_for.htm
http://www.muzique.com/lab/imp.htm

[midwayfair]

C8 is just a filter cap for Vb (4.5V). It helps reduce power supply noise. You should have one on the +9V rail as well

Not sure what "famous" buffer you're referring to. This is a textbook circuit.

Your frequenecy cutoff for C9 is 0.7Hz. If you made it 1uF, it would be 1.5Hz. If you made it 1uF and plugged the buffer into a mixer with 10K input impedance, the frequency cutoff would be 17Hz. If you plug it into a guitar amplifier, the input impedance is 1M, so you could go as low as like 2K for R6 before you would hear any change for a guitar. I think you can see from this that there's absolutely no reason for the added expense of two 1uF caps in parallel on the output.

C7 is also complete overkill by a factor of 1000. 1 NANO Farrad is sufficient (15Hz cutoff) for a 10M input impedance.

[merlinb]

R4 : I would like it to be lower, but simulations show that the buffer becomes an amplifier at very high frequencies ( >100 kHz) if the resistor is below ~4.5k. Any idea on why that is?

Capacitance between the +ve and -ve pins leading to positive feedback at HF. You can get rid of this by adding a shunt capacitor from the +ve input to ground, creating a low-pass filter that kills RF. I'd try 100pF as a start. It does rather depend on what you think your source impedance will be, of course.

C9 : I will use 2x1uF film caps in parallel. Again, I want to reduce low frequency attenuation, but I don't want to use electrolytic because of poor audio performance in signal path.

Actually, electrolytics give exemplary audio performance provided they are big enough. In other words, you have to set the cutoff frequency very low. A 100uF cap wth a fairly high voltage rating (63V or more) will produce zero measurable distortion in this application.

R6 : Really, I have no idea what a good value for that is...

It's fine.

R2 : Here, most opamp buffers on the internet use a 1Meg resistor. My simulations shows that the lower the value, the less the buffer will act as a unity gain and will attenuate all frequencies. So why not 10Meg? Thermal noise issues?

Sure, against a 4.7k input resistor, a 1Meg shunt resistor will give you 0.037dB more attenuation compared to 10Meg....


1oMeg could give you problems with over-sensitivity to input hum or even offset voltage. Your opamp could have as much as 10nA input bias current, which would lead to 100mV across 10Meg. Not a big deal I suppose, bit not worth having for a 1Meg resistor, IMO.

C8 : "copied" from the famous buffer. Don't know much about what it does.

Smoothes out any noise from the PSU from being fed to the +ve input.

Finally, R1 and R3 : I guess the idea is to strike a good balance between current flowing through R1 + R3 and added thermal noise.

Nope, because their noise is filtered out by C8. You can use 1Meg resistor if you like, providing it's a FET opamp (I think yours is).

[merlinb]

C7 is also complete overkill by a factor of 1000. 1 NANO Farrad is sufficient (15Hz cutoff) for a 10M input impedance.

Don't forget he specified hifi. In hifi we don't work to -3dB cut-offs, we work to -1dB, or better yet 0.01dB.

You also need a series output resistor to buffer the opamp from cable capacitance. Leaving this out is a rookie error. 47 ohms should be fine.

[jul059]

C7 is also complete overkill by a factor of 1000. 1 NANO Farrad is sufficient (15Hz cutoff) for a 10M input impedance.

Don't forget he specified hifi. In hifi we don't work to -3dB cut-offs, we work to -1dB, or better yet 0.01dB.

You also need a series output resistor to buffer the opamp from cable capacitance. Leaving this out is a rookie error. 47 ohms should be fine.

Why are the resistors at the input/output important?

[JustinFun]

Not sure what "famous" buffer you're referring to. This is a textbook circuit.

And which much desired boutique pedal was renowned for its buffer until it was traced and found to be a simple textbook opamp circuit?

[merlinb]

Why are the resistors at the input/output important?

They prevent oscillation due to cable capacitance.

[samhay]

Why are the resistors at the input/output important?

What Merlin said. But also, the input resistor provides some protection from e.g. static. Some/most op-amps don't like having their outputs grounded, which can happen if you short the output by standing on a cable or plugging in to something you shouldn't. An output resistor prevents this from becoming a problem.

[PRR]

> R4: I would like it to be lower

Only in simulation. In real life, you *want* some resistance between the precious chip and static-zap, or radio stations.

Is this for audio? Then minor wobble past 10KHz is a small detail.

Is this for *guitar*?? Then total system hiss is (should be) dominated by pickup+pot resistance. Possibly as low as 5K in bass. Notably over 50K in treble where we will hear hiss. A good-quality (carbon-film is fine) resistor of 10K gives very acceptable hiss, good protection against bad inputs (speaker-plug accidents), static zap, and some improvement in radio rejection.

This is in line with the hiss graph on page 7 of the LT1469 data-sheet.

> buffer becomes an amplifier at very high frequencies ( >100 kHz) if the resistor is below ~4.5k. Any idea on why that is?

Page 9 of the LT1469 data-sheet.

Note particularly the bumpy curves bottom-left of page 9 of the LT1469 data-sheet. Your stimulator has zero load capacitance. Your real-world will have a cable, 30pFd per foot, or 300pFd for a little 10-foot cord. 3,000pFd if you send your sound the long way around the cocktail lounge to a PA board. Looks to me like a '1469 feeding a long cable nakedly may become a police-band radio transmitter. (OK, the police don't use that band any more, but somebody must.)

> C7 : Highest film capacitor value I can easily get to reduce low frequency attenuation.

0.016Hz?? Or 0.06Hz for -0.5dB?

Yeah, I would like to have an income 100 times bigger than my living expenses, and the cap is easier to get than such an income, but it's awful generous/greedy.

> C1 : opamp supply bypass cap. don't know if it will do much...

It is essential.

> R2 : Here, most opamp buffers on the internet use a 1Meg resistor. My simulations shows that the lower the value, the less the buffer will act as a unity gain and will attenuate all frequencies. So why not 10Meg? Thermal noise issues?

"Attenuate" how much? 0.04dB? Do you know how small that is? Do you know that you could find 0.04dB make-up gain anywhere in a typical audio system? Yes, do not throw-away good signal, but don't fret the small scrapes-- it's not your 2015 Ferrari.

What is wrong with 10Meg? Read the LT1469 datasheet, Input Bias Current. (This part is funny, Ib- is trimmed but Ib+ is not.) "Typical" 10nA in 10Meg causes 0.1 Volt of DC offset. You better plan on the "MAX" case, 0.4V DC offset.

Does that matter? With a bipolar supply the general idea is to have outputs _AT_ DC Zero to avoid/reduce coupling caps. Tenth-Volt offsets would be Really Bad.

In this case of a single 9V supply, you are forced to bias-up "halfway" between the supply rails. Nominal 4.5V. However 4.1V or 4.9V will work just about the same. So the 10Meg appears to be acceptable, if still much bigger than needed (for loading, attenuation, or bass response). Or is that a 28V supply? Then 14V, 13.6V, 14.4V.... what's the difference? As slim as none.

Hiss will go down as R2 goes up, but insignificantly once it is much greater than the ~~50K of the pickup/guitar.

> C8 Don't know much about what it does.

Possibly something to ponder. What if the power supply has crap on it? (Since it probably comes from an AC wall-outlet, it probably will.) Crap divides across the two 10K, then between R2 and the source, and some amount of crap goes right into the amp and out. C8 against 10K||10K is a high-cut, about 1Hz (0.677Hz), which will cut 120Hz crap about 0.7/120 or like 177:1 or 45dB, and higher crap even more.

{edit--- bleepin censor!!}