Op Amp Large Signal Voltage Gain

[Frank_NH]

Hi,

I have a question about one of the parameters reported on op amp data sheets - specifically, the "Large Signal Voltage Gain" (Avol).  I was comparing the typical number for a LF353 (which is reported as 100 V/mv) with a TL072 (200 V/mv).  I know this is related to the DC gain, but some of my other reading suggests that it can impact small signal AC gain.  Now, suppose you have an op amp overdrive circuit with a large gain set by the feedback network resistors (say 500) - will Avol for a given op amp affect the op amp clipping characteristics (assuming no diodes in the feedback circuit)?

[PeterPan]

I don't really have a "good" answer for you here, but I've made lot's and lots of circuits using these readily available LF353's and TLO8x OP amps over the years, and I just want to caution you about driving the op amps all the way to clipping on their own (rather than use your own tone shapers like diode/resistor matrices). The caution is that even though these op amps are real champion work horses, you can end up with some very inconsistent results when you drive them to their limits like that. The headaches that immediately come to mind, some of which are rarer than others, include: (1) unwanted oscillations at the point where the peaks start and trail, (2) actual reversals (where one portion of one peak or another will bounce to the opposite rail), (3) a loss of all the power supply noise/ripple rejection at the peaks, which OP AMPs normally give you, and (4) just a general (and audible) inconsistency of response from one chip to another.

Anyway, just my 2¢, but  I'd seriously consider the various ways to shape your clipping without reaching the chip extremes.

[R.G.]

Occasionally someone here will ask a question that seems like it ought to have a simple answer, but doesn't. This is one such.

"Large signal voltage gain" is one of those things that have to do with much more than what the label says. It has to do with both the open loop response of the raw opamp that you can't get to and the compensation inside the opamp that you also can't get to as well as the closed loop gain you're using it at.

The raw amplifier inside the package has a high voltage gain, about the same as the large signal voltage gain down at DC. This changes as frequency rises because of the internal parasitics and such in the circuit, and that necessarily involves phase changes as well. Put feedback around the raw amplifier and chances are it will oscillate.

That's what compensation is for. The internal compensation forces the gain of the raw amplifier to roll off so that the gain is less than one before the frequency where the phase converts negative feedback to positive feedback and lets it oscillate. (You may see this called "the Nyquist Criterion" after the guy who elucidated it.) For various reasions, this involves putting a single-time constant rolloff at a low frequency so that very high DC gain gets cut down to unity by some safe high frequency. For the LM741, that was 1MHz. For more modern opamps, it can be higher, sometimes tens of MHz.

It's now stable, but the effective internal gain declines from the point where the compensation pole kicks in - often 1Hz to 10Hz! - up through unity gain at 1+ MHz. So over the audio band, the amount of gain available for feedback to hide the funny oddities inside the opamp steadily declines. Another consequence of compensation is that how fast the amplifier's output voltage can slew up or down. Compensation limits the slew rate, through some modestly complex paths.  Since a large signal, nearly the size of the power supplies, swings through more volts per microsecond than small ones, you can hit slew rate limiting before you hit high frequency rolloff. So at some frequency, the largest full-output-swing signal it can put out decreases to less than the low frequency size. For most of the opamps in current use in pedals, this happens way out at 20kHz or more, moving it out of the audio band. Still, there are some low-power and/or slow opamps that can have it happen inside the audio band.

How this impacts what happens when the raw opamp starts clipping - that is, getting too close to its power supply voltages to continue to move its output voltage - is not clear. Opamps do funny things when their inability to make the output voltage keep rising or falling, which is an effective "situational loss of gain", and the outputs get pushed to do even more. Early ones could latch up, belch fire and plastic. This was bred out of most later ones, but they can still do odd things like going into parasitic oscillation or inverting phase on their output, etc. PP's note on what could happen contains some of these. There are others.

Good, clean (and predictable!) operation under over-driven clipping is not a given in opamps. Modern ones are better than they used to be.

The relationship of large signal voltage gain to clipping behavior is not tight, but also not zero. It's kind of like does the color of your car affect how fast you'll get somewhere?  The proper answer is yes -expecially if your car color is "Arrest Me" red. 

[Frank_NH]

Thanks for both responses (it will take me a while to absorb R.G.'s reply  ).

Here's the context for my question.  I recently built an OCD clone on vero.  The design uses a dual op amp and places mosfet clippers (2n7000s) between two gain stages.  The first gain stage has a 1 Meg gain pot in the feed back loop and a 2.2K resistor (and a 68 nF cap) connecting the inverting input to ground.  There are no diodes in the feedback loop, only a 220 PF cap.  I used an LF353N op amp versus the TL082 in the original (thought it should be drop in replacement).

What I noticed was:

(1) It wants to oscillate like a banshee at high gain - this can be addressed by increasing the feedback loop cap to something like 470 pF
(2) I could take out the mosfet clippers completely from the circuit and the distortion sounded pretty much the same!  Hence, my thought was that most of the clipping was op amp clipping.  I ended up putting some LEDs in the mosfet sockets (the wiring of the mosfets permitted this), and indeed you could see them light up with the guitar signal, so something is being clipped (perhaps just the clipped signal itself from the op amp).

I have a TL072 IC which I want to try now to see if it behaves differently.  But I want to understand why the LF353N didn't work as expected.  I thought it might be due to the characteristic of the op amp clipping that must be occurring here because of the high gain in the first stage, and the only thing that really stuck out as being different between these two chips is the Large Signal Voltage Gain parameter.

[PRR]

> 100 V/mv

Is a funny way to say "100,000". (There's madness to their reason.)

You want a gain of 500. Open-loop gain of 100,000 will have error of 1 part in 100,000/500 or 1 part in 200 or 0.5%. Open-loop gain of 200,000 will have error of 1 part in 200,000/500 or 1 part in 400 or 0.25%.

You can't hear such differences. Using 5% or 2% resistors you can hardly measure the difference except same resistors around the two chips.

But more to the point: gain of 500 on any guitar signal will grossly OVER-load any amplifier, especialy on 9V or 18V supply. Just totally creamed top and bottom.

Put in 500 milliVolts. Expect 0.5V*500= 250 Volts out. But you really get like 3 Volts out. Gain is apparently 6. Now put in 20mV. Expect 0.020V*500= 10V out. Really get 3V out. Gain is apparently 150V. For any likely guitar signal, output is NOT like gain of 500, it is smooshed to constant 3V.

What you really want to know is, how hard does it smash the limits? Some smash signal like hardened steel garage walls, others gently bend like soft pine garage walls. But this is NOT a concern to "real" op-amp users, who never (hardly ever) smash their signals against the rails. Op-amp is a precision calculator, not a tool of destruction. Opamp makers may hint how close to the supply rails they will go "clean", but very few details about how it transitions from clean to clobbered.

Anyway the 100V/mV number is nearly nonsense. It can vary WIDELY. It depends on the collector impedance in the 2nd stage, which is wildly variable. Depends on chip cooking time, mask flaws, if it's time to change the gas supply, flaws in the wafer.

However above roughly 1Hz (wildly variable) it is not 100V/mV and not about collector impedance, it is about the Compensation Cap which sucks-down gain to unity at ~~10MHz. This cap seems to track input stage process variation quite well (has to, or internal compansation would be folly). So all across the audio band you use the Gain-Bandwidth Product number, not the "Avol". (Avol may have meaning in very slow or very precises computations.)

And BTW: I think genuine LF353 may be hard to find today. For "all"(?) practical purpose, the TL072 does the "same" thing with less Silicon. I would bet there are "LF353s" which are really TL072s inside.
_______________________________

> wants to oscillate like a banshee at high gain

Gain of 500 is A Lot. Murphy and Bode conspire to ensure that high-gain "amplifiers" are usually Oscillators. Great care in layout is needed to keep outputs far from inputs so sneak-back is lessened. (Which makes dual-opamp chips problematic, if both sides are worked high-gain. Audio gain of 500*500 = 250,000 in less than an inch is beg-beg-begging for oscillation.

I don't think the TL072 will be different. The specs may be a hair less, but in outright oscillation a hair less is still oscillation.

[Frank_NH]

Thanks Paul.  In my build, I experimented with reducing the gain pot to 500 K, but even there the op amp is still clipping and oscillation is occurring at max gain.  I suppose what made me wonder about all this is that this is a popular overdrive design, and clearly these issues don't appear to be a problem if using the prescribed op amp, the TL082 (I'll confirm this though...).

Looking at other designs with post op amp diode hard clipping, the MXR Distortion + has a max gain of around 200 using a 741.  The Marshall Gov'nor actually has two op amp stages (using a TL072) with lower gains in series followed by diode hard clipping.

The Proco Rat, on the other hand, has a huge gain stage, using the LM308.  Clearly op amp clipping is occurring there.  The diodes merely squash this down, but the resulting sound is pretty good for a distortion pedal.  However, I've noticed that people seem to favor the LM308 in this circuit, perhaps for it's clipping properties.

I guess the conclusion I am coming to is that you have to be more careful about op amp selection for overdrive/distortion circuits if op amp clipping is part of the design.  Some op amps are apparently better than others in this regard, and so we can't just substitute any ol' op amp.

[ashcat_lt]

There's a big difference in slew rate and a bit of difference in GBP between the Rat's 308 (with its typical external compensation cap) and something like a 072 (which is internally compensated) and that seems to make a noticeable audible difference.  It might be worth finding an externally compensated opamp and playing with that capacitor value.

I tend to agree that your oscillation issue is probably more about layout and/or wiring than the actual opamp in use.

And yeah, clipping a square wave creates a smaller square wave.  The "sides" will get steeper, and the "knee" might get sharper, but only to a point because slew rate and bandwidth will limit these.  If you're already trying to go way beyond the opamp's limit, then it's about as square as it's going to get, and the only thing that further clipping can do is make the total p2p output lower.

[boogietone]

Looking at the LM353 data sheet http://www.ti.com/lit/ds/symlink/lf353-n.pdf two things are noteworthy. The first is that the Avol could be as low as 25 and no upper limit is given. The 100 value is just typical. For any one unit, the actual value could be significantly different from another. 

The second is that much of what  RG talked about is encapsulates in figure 17, which is essentially Avol vs frequency.

[PRR]

The '308 is ten times slower than a '072. Like 0.5MHz GBP instead of >5MHz GBP.

Working at nominal gain of 500, the '308 poops-out at 0.5MHz/500= 1KHz.

Hold that thought.

ANY two wires have Capacitance between them. Capacitance is high impedance at low frequency, low impedance at high frequency.

Out past many GHz, "all" wires in a reasonable area "connect" to each other like a hundred ohms.

At the top of the audio band, two wires in the same box will act like a few MegaOhms one to the other.

This means that outputs leak back into inputs. And (with some limitations), this acts the same as a mike amp and speaker in the same room. It feeds-back, howls.

If general impedances are a few hundred K, and gains are under 10, a MegOhm of sneakage may do no harm.

But HIGH-gain circuits mean big sneakage through the same capacitances, and probable oscillation.

So HIGH-gain design is a lot about LAYOUT. Keep outputs well away from inputs. Overall and at every stage.

Since the problem is high frequency, spoiling the high-frequency gain is another tool. That 500K pot will usually need a capacitor across it. Low-GBP amps are less vicious than higher GBP amps. (Also distorted highs are just nasty-- we WANT to trim the treble before distortion.)

[boogietone]

PPR, interesting and useful observations about the stages inside an op amp. After years of just using these things, I am starting to look at understanding their innards. Thanks.