Op Amp Buffer

[SparkyColdfire]

Hi all,

New guy here.  I wanted to say this is a great site, and I've been lurking for quite a while soaking up all this great information. I have a question that I hope someone can help me with.

I want to build a small pre amp and attenuator so that I can boost my instrument signal up to line level for some rack gear, then bring it back down to instrument level again to go through the rest of my pedals and amp.  I built a Tillman jfet booster and that was a fun little project, but didn't quite have the gain I was looking for (although I liked the way it sounded), so now I'm playing with op amps.  My original thought was to build an input buffer stage followed by a gain stage for the transparent boost.  I realize I could probably combine these into one non-inverting op amp stage, but for learning sake I have broken the project up into two chunks so that I can play with each individually and study the pros and cons of different designs for each part. 

The input buffer is desired to have at least 1 Meg input impedance, unity gain (a given for a buffer) and flat frequency response across the guitar spectrum. I would also like to minimize noise since I will be boosting in the next stage.  I breadboarded the three non-inverting single-supply designs shown in this picture. 



The first works great, but the other two do not.  The second one had less than unity gain (I could audibly notice a drop in volume) and seemed to roll off some of the low frequencies relative to the bypassed signal.  The third one didn't work at all and when examining the voltage levels at the inputs and outputs of the op amp, it looked like the output was pretty close to the + rail.  My problem is the first one has the biggest resistors and will theoretically be the noisiest. The last one has no resistors at the front end and relies solely on the op amp input impedance, which sounds pretty desirable.  The middle one is in-between.  None of these are my designs (I'm not there yet), but rather they are cobbled together from a variety of published sources. 

So my question is... Should either of the  bottom two designs work?  What are your thoughts on these?  Could the breadboard implementation of these be the reason for their failure (thinking of stray noise that you would not get in a proper enclosure).  Do you have any other buffer circuit suggestions? Thanks for any help you can provide.

[GibsonGM]

Look here!
http://www.muzique.com/lab/buffers.htm

[amptramp]

The third one cannot work - there is no return for the non-inverting input current.  Depending on whether the op amp input has PNP or NPN inputs, the output will go to one rail or the other.  CMOS inputs will not work properly but the leakage through the input capacitor may allow it to work, although badly.

There should be little difference between the first and second versions in noise or performance.  The second version takes more current, which may be responsible for the change in characteristics that you noted.  Since the input resistance is the same for the first two versions (well, 5K higher for the second), the noise performance should be the same.  A pair of 2M resistors in parallel gives 1M and has the same noise properties.

This assumes your input is referenced to ground.

[Minion]

You don"t really need a Buffer stage and a Gain stage , a gain stage is also a Buffer stage so you might as well just use a single opamp as a Buffer and for Gain ....

[SparkyColdfire]

Thank you for all your comments.

Yes, I had seen that link to the various buffers and I have also read the TI guide "Op Amps for Everyone".  There are also some good ap notes from Microchip and other chip makers that I have read through.  I also found an interesting article called "Demystifying single-supply op-amp design"  The thing is, sometimes the information appears to be contradictory, and sometimes the circuits are simplified, or intended for other applications.  The tricky part for me as a beginner is understanding where the simplifications or assumptions are and why some of the circuits are built the way they are.  I.e. why is that resistor 4.7k?  Oh, that reminds me, I also read the interesting article called "What are all those parts for?" All very good reading material for us noobs.

Last night I rebuilt the second circuit l showed above and it sounded just like the first one. So I must have made a mistake the first time I assembled it.  Thanks for pointing that out.

I am surprised that you say both the first and second should have the same amount of thermal noise.  If that is the case, I'll use the first one since it draws less current and has one less part.  I see now that the third one was bad from the start.

So now that I have a decent buffer, I decided to work up a basic non-inverting amplifier.  Here is what I think I'll build next.  As Minion pointed out, this is an extension of the non-inverting buffer I started with.



Please let me know if you see any potential pitfalls.  Some questions I have about this amplifier... Will still have an input impedance of a Meg Ohm?  I think so, but I'm not sure if any of the input signal makes it's way to ground through the feedback network?  Also, since I want the gain to be variable from 1 to 10 (via the feedback pot) I scaled Rf and Rg such that there would be no need for a resistor to minimize the bias current at the non-inverting input.  Did I do this correctly?  If I wanted to scale them down, say to 10k and a 100k pot (in an effort to reduce noise) what would be the impact? 

Thanks again.  I'm really learning a lot, but there is always more to learn. 

[Minion]

The Circuit looks Fine but I personally would use lower gain set resistors , Like 10k and a 100k pot or even 1k and a 10k pot , The higher the resistor value the more resistor noise .....

Cheers

[R.G.]

I am surprised that you say both the first and second should have the same amount of thermal noise.  If that is the case, I'll use the first one since it draws less current and has one less part. 

That is not the case, if you will add a capacitor from the midpoint of the bias divider to ground. This eats the thermal noise from the bias network, reduces the voltage across the 1M bias resistor, and makes the thermal noise about as low as you can get it for that high an input impedance. Ott refers to this style of biasing as "noiseless biasing", in contrast to the thermal noise of the first circuit.

Please let me know if you see any potential pitfalls.  Some questions I have about this amplifier... Will still have an input impedance of a Meg Ohm?

No. In fact, the input impedance is so low that it will hardly work at all. That's because the input signal goes through that input 0.01uF, then through the 10uF to ground. You can't bypass the junction of those two 2M's if you do it this way. Back to circuit #2, with the 10K/10K/1M divider network. That gives you 1M input plus low thermal noise.

I'm not sure if any of the input signal makes it's way to ground through the feedback network? 

No, it doesn't, but that's no help because of the capacitor shunt from above.

Also, since I want the gain to be variable from 1 to 10 (via the feedback pot) I scaled Rf and Rg such that there would be no need for a resistor to minimize the bias current at the non-inverting input.  Did I do this correctly?

No. The resistance between the (-) input and ground is effectively the 100K in parallel with the feedback pot in whatever setting it sits. When the feedback pot is set to 0, the impedance to ground is effectively zero since the output is effectively ground, by superposition. And the DC impedance to ground from the + input is 1M. So there is a big DC bias current imbalance at best, and a huge one at worst.

... which doesn't matter at all in this circuit.  Impedance to ground for DC bias reasons only matters in circuits which have either massive gain (this doesn't) or need great DC accuracy. This circuit only amplifies AC, and you don't much care if the output is 20-30mV off of the perfect bias point. So - it's not balanced, but it doesn't matter.

If I wanted to scale them down, say to 10k and a 100k pot (in an effort to reduce noise) what would be the impact? 

None, for the reasons I mention.

Further, it's probably overkill to be worrying about thermal noise so much with such low gains. The thermal noise would be lower, but the opamp itself will make a bigger difference than changing the resistances. And if you use a carbon composition pot, the cheapest and most available kind, the pot itself will have more excess noise than anything else in the circuit. Which, again, may not matter because of the low gain.

I applaud your efforts to learn. You're asking pertinent questions, but questions that are more advanced than your understanding of the basic circuits. Dig in, learn it all!

Go man!

[SparkyColdfire]

Thank you so much for taking the time to write such a complete response.

I had to read it a few times but I think I get it.  You right, I'm realy at a basic level but attempting to move up quickly, perhaps too quickly.  I'll keep studying, and try to think through my concepts better.  I see how I basically shorted my input to ground in that last schematic.  No, not what I wanted to do at all.  

So I drew this up.  See if I implemented all your suggestions correctly.  R1, R2 divider for ref voltage, 1Meg R3 for my input impedance, C1 for removing ac from power supply, C2 for additionally stabilizing the reference voltage, R4 and pot to get my 1 to 10 gain, and coupling caps at the ends. 



Is RF noise any threat to this, (do I need a low pass up front to filter it) or is that also one of those things that I need not worry about at this gain?

I still need to understand more about op amps so I'll keep at them, thanks again for the help.

[SparkyColdfire]

I built this and it looks like it works fine.  The schematic above didn't work though because R4 shouldn't go to real ground, it should go to the reference voltage as shown in this drawing.



One thing I noticed is I dont have nearly as much headroom as I would have thought.  Perhaps this is due to the op amp I'm using, an LF353.  When I play some blues on my single coil I can crank up the gain to 10 and I get a nice clean boost. However when I play chords on my passive humbucker, I can only turn up the gain to approximately a factor of 3.5 before I start hearing some distortion. So I recorded some samples with my looper and then repeatedly looped them into this preamp and slowly turned up the gain until I noticed just the slightest distortion.  Then I measured the AC signal coming out.  I got .71 volts average, or .79 v RMS, or 1.1 volts peak.  This is with a 9v battery that put my reference voltage right at 4.5 volts.  Wouldn't you guys expect to be able to boost to at least 2 volts peak without clipping?  Can you suggest a more applicable op amp, or is this just a practical limitation of all op amps?  Thanks for the help.

[cloudscapes]

what software are you using to draw out your schematics? is it cut-and-paste?

[SparkyColdfire]

Yes, it's just cut and paste.  I spent several hours trying a variety of free or demo versions of SW tools, and none of them did what I wanted.  So I spent another hour or two and drew a nice variety of the basic components then rotated them all in the 4 primary directions.  I also drew some common labels.  Now when I want to buld a schematic, I just bring up my pallet of parts and plop them down where I want and connect them with straight lines.