Op amp suggestions for battery powered buffer

[wayfaerer]

Hi! I'd like to make an op amp based buffer for the beginning of my signal chain, and am curious if you have any suggestions for an op amp given that it will be battery powered. This is what I'm hoping for, in order of importance (most to least):

• Inaudible noise
• No/minimal coloration
• Reasonably low power consumption
• Can operate even when the 9V supply drops, let's say to around 7V or so

Ordinarily I would use a TL071, and may still. I've never had a problem operating a TL071 with a 9V power supply even though the datasheet lists its minimum recommended supply voltage as +/- 5V. But I'd like to keep the buffer working even if the battery voltage drops below 9V, which might be pushing it.

I noticed the TL081 has a +/- 3.5V minimum supply voltage. I know it's not marketed as low noise, but not sure if the noise would be audible. I'm not very good at understanding datasheets.

Lastly, I've used an LM358 (dual op amp) before, which operates down to around 5V, but I found that it was noisy. However, it may have been a cheap knockoff batch of 358's, so I'm not sure if the noise I experienced would be present if I bought from a reputable seller. Also, I'm not sure about power consumption. It says "low power" but not sure if they're referring to low power supply requirements or low power consumption.

Any other good options I'm missing? Cost is not an issue.

Obvious solution: socket it and try it out. I will definitely do this. Still looking for opinions. Thanks!

[PRR]

The TL07x has been used in millions of 9V pedals.

Below 7V total an internal bias falls out of range and the amp won't meet all specs. TI would love to sell you other opamps aimed at low-V power so has not been motivated to go-back and document the loss of performance. But in small audio most everybody's experience is that the TL07x works fine at 9V and even down to 7V.

> socket it and try it out.

Wise words.

[ElectricDruid]

Definitely socket it and try it out.

In the past, when I was thinking only of battery operation, I've experimented with the TL061 and TL081 as well as the TL071. They're optimised for different parameters. The 061 is low current, so makes your batteries last longer, but isn't quite as quiet - not so you'd notice much though, especially in a buffer. The 081 was the original of which the 071 was the "low noise" version. I also used to play with the LF441 series (dual is 442, quad is 444). These are low power chips again, but have pretty good specs nonetheless and might be worth a look.

That said, all of these suggestions are things that have been around for 20+ years, so there may well be much more sophisticated things available that knock them into a %^&*ed hat. You might pay for the extra performance though...

HTH,
Tom

[wayfaerer]

I keep reading about all these great modern options... so I go to buy them and find that they aren't available in a through-hole package. I know I could get a surface mount adapter board, but I'm working with limited space. Bleh.

[R.G.]

Does it have to be an opamp?

A while back I designed discrete buffer specifically for long battery life as an inboard buffer for a guitar.

Check out:

http://www.geofex.com/FX_images/Onboard_Preamp.pdf

[wayfaerer]

Ha, I read that page earlier today, but I didn't understand the circuit. I'm hoping for a tiny buffer with just a few components. Something like the ones on the "Basic Buffers" page of Jack Orman's site.

No it doesn't have to be an op amp. I've just used them a lot more than transistors, which is why they are my go-to.

[Rob Strand]

Ha, I read that page earlier today, but I didn't understand the circuit. I'm hoping for a tiny buffer with just a few components. Something like the ones on the "Basic Buffers" page of Jack Orman's site.

That shouldn't stop you using it.    What you gain is low power.   If less parts is more important then use a TL071 with the Vr network on Jack's site.   There is a minimum amount of baggage you need for any of the circuits:  Vr network, input network R's & C's, output network R's and C's.  You could use a low power opamp but then you will get more noise.   Only you can decide how to trade off all the requirements.

[PRR]

> I didn't understand the circuit.

There's just three basic clumps.


The core action is an Emitter Follower. For good input impedance we need a current gain over 10,000, and one single BJT can't do that. It is a two transistor compound. The Darlington is well known. The Sziklai NPN-PNP is similar but can pull-up a bit better, and at 6V supply it may matter.

The pull-down action is a current source. This could all be replaced with a 30K resistor, but it would not pull-down as far.

The emitter follower needs to be DC biased near half supply. This is common in many 9V pedals, especially op-amps. Because only a small amount of bias current is needed, this uses big R and small C relative to the 10K+10uFd parts we use in less thrifty designs.

"???" is not a White Cathode Follower connection, though it looks like one. It may reduce response at RF or bypass battery noise.... R.G. would have to address that.

[Rob Strand]

"" is not a White Cathode Follower connection, though it looks like one. It may reduce response at RF or bypass battery noise.... R.G. would have to address that.

I noticed the same thing.   I haven't tried to simulate it.

The pull-down action is a current source. This could all be replaced with a 30K resistor, but it would not pull-down as far.

A resistor won't drive a capacitive load (the upper transistor acts like peak detector).    To "simply" the ckt  I'd keep the transistor at the output and replace the left current source transistor with a voltage divider.

[wayfaerer]

Thank you for the explanation, I really appreciate it and it makes more sense to me now. While it sinks in, though, do you have any intuition on how it might perform in terms of power consumption, noise, and accuracy of an unamplified guitar signal? Compared to, say, a TL071 op amp buffer. I realize the circuit is designed for low power, but I have no idea how low we're talking.

Also, I see that it's optimized for 6V--would it work equally well from 9V down to 6V?

Thanks again!

[PRR]

> would it work equally well from 9V down to 6V?

Says in the PDF.

I see it does not specify supply current. I work it out as 100uA-120uA, which is say 0.12mA. Compare with TL071.

[PRR]

> A resistor won't drive a capacitive load (the upper transistor acts like peak detector).

Sure it will! Just not as well.

30K pull-down on 300pFd (10 foot cable) does pretty well to 17KHz. Guitar is not likely to have even 1.7KHz as high as the general level 100Hz-1KHz. So 30K would do OK with much more than 300pFd capacitance.

Here's simmed 4KHz into 500pFd. You can just see (not hear) the resistor's pull-down weakness.



Ah, but let's try 5,000pFd (over 100 feet of cable). Here the R-loaded (Red) actually makes *prettier* waves. Why? At peak up-swing the 98uA resistor is up to 188uA! So it starts down real good. The 100uA CCS (green) is always 100uA, and falls on a straight line. As we pass through "zero" (2.6V the way my models sat) the R-loaded's curvature makes it fall behind the I-loaded form.

[Rob Strand]

Sure it will! Just not as well.

True.   IIRC before it gets to the peak detector "take-off" the cap load cause Vbe to go up and that makes the distortion go up  (probably not much difference in this case between R and current source.)

Here the R-loaded (Red) actually makes *prettier* waves. Why? At peak up-swing the 98uA resistor is up to 188uA! So it starts down real good. The 100uA CCS (green) is always 100uA, and falls on a straight line. As we pass through "zero" (2.6V the way my models sat) the R-loaded's curvature makes it fall behind the I-loaded form.

Yes, there's not much in it.  For some reason I thought the current source was a bit better than that (hmmm).
The White follower, with cap linking the top and bottom, should be *much* better.

Not to distract too much from the thread (and this *is not* a suggested alternative to RG's ckt.).  Ages ago I had a buffer with a capacitive load which suffered badly from "peak detection".   In the end I just lowered the RE value and put up with the extra current.    However, one alternative circuit was a parallel buffer.   It gives strong pull-up and pull down without "slew-limit" and doesn't get caught-up with biasing push-pull stages.   It's a funky circuit.  Notice the base current flows in one transistor and out the other so the bias point at the bases sits close to the centre.

I hope I drew it correctly,

[PRR]

I've seen video gear which used your 2-Q design dozens of times. Fine line-driver.

[Rob Strand]

Yes, there's not much in it.  For some reason I thought the current source was a bit better than that (hmmm).
.

I think is it:   
For the resistor the slope is 1/(RC), and the current source the slope is I/C.
Since R = VE / I, the resistor slope is 1/(RC) = I / (VE * C),  so the resistor is worse by a factor of 1/VE.
It also means the resistor fairs worse at higher voltages (for the same operating current).

Hope I haven't screwed up.
see fix below

[Rob Strand]

I've seen video gear which used your 2-Q design dozens of times. Fine line-driver.

Cool.  Thanks for letting me know because I've never seen it in anything.  I've seen single Q's with low REs, like 560 ohms, and chips.

The input impedance is similar to the single nowhere nears as high as RG's Sziklai.

[pinkjimiphoton]

look up ca3140ez. real clean and quiet,  will run down to about 5 volts, won't clip even if it goes a 1/2 volt under the negative rail.

https://www.renesas.com/us/en/www/doc/datasheet/ca3140-a.pdf

[Rixen]

https://docs.google.com/spreadsheets/d/1VjHWtAaqLJBZi7CKDw5SlzluXgqv__cc42lV0K1jMvQ/edit?usp=sharing

Have a look at the TL061 operation down to 4 V and 0.2mA per channel. Higher 1/f noise than TL074 though.
Also the TL94x series- main problem is the high bias currents which can be a problem with high input R in high gain devices

I'm always on the lookout for new devices to add to the table linked above, thanks for the ref to CA3140ez Jimi, I'll incude it in my next revision.

[PRR]

> Since R = VE / I, the resistor slope is

Ah, but what is "V"?

For idle current (battery drain), the idle point.

At peak positive, almost *twice* that! So at the start of the pull-down it does better than a current source of the same battery drain.

At the bottom, it loses current and speed; but (for the particular point I happened to pick) it never gets worse than CCS until both are bottomed-out. And it is sine-like rather than triangle slope, maybe more musical.

These are extreme cases: 4KHz and 5,000pFd loading. It is clear to me that this will always dirt-up in the 100Hz-1KHz body of the music, not the upper harmonics.

[Rob Strand]

Ah, but what is "V"?

It was VE = V subscript E = quiescent emitter voltage.

Hope I haven't screwed up.

I knew there was a risk posting something techy today with my foggy head!     

Fixed:
For the resistor the *initial* slope is V_E/(RC),  where V_E = the quiescent emitter voltage.
Since R = V_E / I,  *initial* slope = I /C.
For the current source the slope is I/C.

At the bottom, it loses current and speed; but (for the particular point I happened to pick) it never gets worse than CCS until both are bottomed-out. And it is sine-like rather than triangle slope, maybe more musical.

So yes what you said is correct: initial slope is the same, but after that the resistor is worse off (by factor exp(t/(RC)) ).

From a HiFi perspective I always thought the current source holds up better.   Maybe it's this case:  a HF signal riding on a LF signal.  When the output swing is near the negative rail the resistor case can't track the slope.

For the peak detect case the resistor case may well sound better.  Still, it's an angry waveform and it would be best not to get there in the first place.

These are extreme cases: 4KHz and 5,000pFd loading. It is clear to me that this will always dirt-up in the 100Hz-1KHz body of the music, not the upper harmonics.

Yes, the usual 9V + 10k emitter resistor case has the potential to fall over when you get to extremes.  It's not something people consider at all.    An opamp will hold-up.     I guess not all buffers are alike.      An interesting thing is the White follower clips asymmetrically, so does that sound better or worse than a push-pull buffer.