Is there such a thing as a best OP-AMP

[zombiwoof]

> I prefer the Jfet-input LM833s....  

Did they change it? The 2003 LM833 datasheet shows PNP inputs.

You missed the smiley face after - it was a bit of a joke.  But that would be my best op amp if they made them.  For audio. Maybe  

> the Burr-Brown OPA series chips

Guitar-topic forums?



The OPA chips will drive heavy loads with more zeros in the THD spec. That impresses a lot of folks. They are probably -cleaner-, "more transparent", which is useful in Hi-Fi and in studio applications. But much guitar electronics is about "banging-up" the too-clean tone of a naked steel string.

Some of the OPA series are very high-strung, fussy about bypassing and layout. 741 through TL072 are much more tolerant of simple unsophisticated construction.

Let's not forget that they are $4+ a pop!   Enough to dissuade me.   I have one from my pre-DIY experiments, actually.   It never hurts to order one and have one on hand to hear the specs for yourself...if your lotto ticket from last night was a winner.  

I bought a couple of Burr-Brown OPA2134PA from Small Bear for about $3 each, and I've seen them in the $2+ range from some Ebay sellers.  I haven't gotten a chance to try them yet, but I've seen them mentioned by pedal modders as an upgrade to the 4558 type in some pedals, for being low noise and smoother.  In a one-off pedal, I don't think $3 is too much to try something different, but I can see if someone is making pedals in large numbers to sell that the cost would be a consideration.

Al

[JDoyle]

In my humble opinion, the best op amp is a discrete one.

Probably.  But fitting all those parts onto a daughterboard and then stuffing a couple of those suckers into a GGG or Tonepad layout is gonna be a bitch!   

Truer words have probably never been spoken! 

Not to mention that your definition of "best" is forced to include needing poor specs on anything to do with matching, such as temp drift, input bias drift, DC offset, input error voltages, and so on. Those things are intimately related to on-chip device matching and thermal coupling.

Not necessarily - only if you use single discrete transistors.

If you build it with duals like the LM394 or SSM2210/20, or arrays like the LM/CA3046, CA3083, CA3146, CA3183, any of the THAT arrays, even a CA3086 and a CD4007 (Small Bear has both), one can not only adequately (much more than adequately for FX use) avoid any of the issues you point out, but at the same time, eliminate the added noise of having every single transistor built on the same substrate and necessarily isolated from one another via reversed biased diodes, use high quality PNP devices, or even germanium transistors if that floats your boat, and design the circuit to YOUR specifications instead of being slave to the circuit design of someone who has to ensure that the chip's specs are good enough to satisfy numerous applications (with audio not very high on the list) for it ever to have a chance of becoming a commercial product.

In the end, you really only need the long tail pair input transistors to be matched and thermally coupled. It is best if any current mirrors you happen to use are matched as well, but not absolutely necessary, it depends on their function, like when used as the LTP load. Even IC op amps normally include offset adjust pins so in the best case possible, matching isn't perfect, though thermal coupling is obviously best when the transistors are on the same chip.

After all, nearly every hifi power amplifier available is simply a big, high dissipation op amp.

Regards,

Jay Doyle

[Wonderdog]

In my humble opinion, the best op amp is a discrete one.

Probably.  But fitting all those parts onto a daughterboard and then stuffing a couple of those suckers into a GGG or Tonepad layout is gonna be a bitch!   

Truer words have probably never been spoken! 

I don't know, I think discretion is the better part of valor

[R.G.]

If you build it with duals like the LM394 or SSM2210/20, or arrays like the LM/CA3046, CA3083, CA3146, CA3183, any of the THAT arrays, even a CA3086 and a CD4007 (Small Bear has both), one can not only adequately (much more than adequately for FX use) avoid any of the issues you point out,

(a) matched duals *are* ICs. Check the TI patents. 
(b) "adequately" and "much more than adequately for FX use" is another way of saying

your definition of "best" is forced to include needing poor specs on anything to do with matching, such as temp drift, input bias drift, DC offset, input error voltages, and so on. Those things are intimately related to on-chip device matching and thermal coupling.

(c) talk to Bob Widlar's ghost or read some of Widlar's history for things other than the input pair needing matched.

but at the same time, eliminate the added noise of having every single transistor built on the same substrate and necessarily isolated from one another via reversed biased diodes, use high quality PNP devices, or even germanium transistors if that floats your boat,

Yep, there are all kinds of things that float all kinds of boats. Again, it kinda depends on what "best" is, doesn't it? If one defines "best" as excluding ICs, then no IC can be "best", can it?    But you are right - only a discrete Opamp can be "best" - if that's what you demand as "best". I could dig through the points, but I'm kinda down on typing for the day. Ask again if you want clarification.

and design the circuit to YOUR specifications instead of being slave to the circuit design of someone who has to ensure that the chip's specs are good enough to satisfy numerous applications (with audio not very high on the list) for it ever to have a chance of becoming a commercial product.

"... and I did it MY waaaay!" was always high on my list too. But that varies from person to person. Some people just want it to be small, or work, or whatever else is in their own personal "best".

In the end, you really only need the long tail pair input transistors to be matched and thermally coupled. It is best if any current mirrors you happen to use are matched as well, but not absolutely necessary, it depends on their function, like when used as the LTP load. Even IC op amps normally include offset adjust pins so in the best case possible, matching isn't perfect, though thermal coupling is obviously best when the transistors are on the same chip.
[/quote]

After all, nearly every hifi power amplifier available is simply a big, high dissipation op amp.

Yep - and ones which avoid any drift issues by limiting the bandwidth way above DC, or trying to band aid it by servoing.

I'm not trying to be difficult. Just saying "Free your mind," as Morpheus was so fond of saying.

[TELEFUNKON]

So basically the best op-amp is the one that works best in a certain circuit........ Most of the ones I see mentioned cost next to nothing anyways!

ohh! but the shipping fees ! 

[R.G.]

So basically the best op-amp is the one that works best in a certain circuit........ Most of the ones I see mentioned cost next to nothing anyways!

ohh! but the shipping fees ! 

Leading to Keen's Fifteenth Law: never order just one unless you're SURE you will never need another one or unless the second one is ruinously expensive.

The rationale is that you're much more likely to need another one of something you've used before than you are to use something you've never used before, even if only for repairs.

[Paul Marossy]

Leading to Keen's Fifteenth Law: never order just one unless you're SURE you will never need another one or unless the second one is ruinously expensive.

The rationale is that you're much more likely to need another one of something you've used before than you are to use something you've never used before, even if only for repairs.

If it's affordable, I always order a few more than I need. That way I have some extras in case I mess up one or need one for a repair or something.

[mikemaddux]

Look for where the price beaks happen...and order a pack from futurlec to get you started with an assortment

[soggybag]

I suspect that if you sell effects there is a best op-amp and your effects are using it. If you make effects for yourself, there is no best op-amp because you are constantly testing op-amps and can never find one that does everything.

[JDoyle]

(a) matched duals *are* ICs. Check the TI patents. 

Good point - especially in the case of the LM394 types which are a bunch of of transistors paralled. However, I do think you would agree that we are getting awfully close to a semantic argument over the term 'IC' and that there is a wide gap between an IC op amp and matched duals or arrays. It isn't the 'IC' part I dislike, but the fact that ALL of the components are integrated on the SAME chip, creating the noise/poor PNPs/etc. issues. I guess it is my lack of clarity in my statement or a poor use of the word 'discrete'...

(c) talk to Bob Widlar's ghost or read some of Widlar's history for things other than the input pair needing matched.

I have read a lot of the history, Widlar's papers, interviews, etc. And I agree with you - when all of the parts are integrated into the same chip, which is the specific situation Widlar was researching/discovering/inventing/creating. Going discrete changes the rules/requirements. Because IC design requires the use of ratios of part values and for all practical purposes eliminates the ability to use specific values - IC design by it's nature requires matched parts; not so in a discrete design.

When you break each section of the circuit out into discrete circuit blocks (LTP-->VAS-->Buffer Out), you no longer have a need to match transistors anywhere but the input pair and it's current mirror load, which you are no longer required to use as going discrete removes the restriction on resistors and capacitors that integrating the circuit places on the design. Though this is not a recommendation for NOT using them, after all, in an op amp, the more gain the better. The VAS and output buffer can be Class A, as dissipation and current consumption is no longer the worry it is in an IC opamp; need not have their circuit characteristics based upon any other portion of the overall circuit - their CCS loads, if used, need not be based on a multiple of the LTP's tail CCS. Going discrete breaks the components out from under the worry of developing thermal gradients in the IC and the matching issues that causes; etc.

Yep - and ones which avoid any drift issues by limiting the bandwidth way above DC, or trying to band aid it by servoing.

Which is the same situation that we have here with FX, the situation I was considering when recommending going discrete. My guitar can't output DC, my amp can't reproduce it - even if I could hear it; so there isn't any need for my op amp to reproduce it, which additionally virtually eliminates any worry about drift, and if it doesn't, the fact that in an FX circuit I'm almost certainly designing for 9V/0V operation, requiring input and output caps as well as one in the negative feedback loop, which will in itself remove worry about drift; and if it never clips, offset won't matter much either...

I'm not trying to be difficult either - just trying to better explain my position and definition of 'best' God knows making eight solder joints to something that takes up less real estate than a postage stamp, not to mention can give you two for the price of one, is MUCH better 99% of the time than the work, set-up, and real estate involved with a discrete design!

I just happen to think they sound better in the specific case of the audio path of an FX circuit.

Regards,

Jay Doyle

[R.G.]

I just happen to think they sound better in the specific case of the audio path of an FX circuit.

Well, why'd'n't you say so!? 
I'll support that one! I completely agree that you perceive discrete as sounding better in that context, and prefer them.

I myself have proven to my personal satisfaction that *I* can't even distinguish, let alone prefer, discrete opamps from at least some IC opamps every time I've set up a test that is sufficiently immune to my biases. And I've seen the same result from other people in the same setup. But it's an arduous test, necessitating as it does that I run a large number of trials of comparisons where I am carefully guarded from knowing a priori which of IC/not IC I'm listening to. I guess putting it that way, I've proven that at least some people can't hear the difference.

However, no amount of such testing can ever prove that there does not exist someone who CAN tell the difference; it's logically impossible to prove a negative.

Preference is very personal.

[JDoyle]

I just happen to think they sound better in the specific case of the audio path of an FX circuit.

Well, why'd'n't you say so!? 
I'll support that one! I completely agree that you perceive discrete as sounding better in that context, and prefer them.

I myself have proven to my personal satisfaction that *I* can't even distinguish, let alone prefer, discrete opamps from at least some IC opamps every time I've set up a test that is sufficiently immune to my biases. And I've seen the same result from other people in the same setup. But it's an arduous test, necessitating as it does that I run a large number of trials of comparisons where I am carefully guarded from knowing a priori which of IC/not IC I'm listening to. I guess putting it that way, I've proven that at least some people can't hear the difference.

However, no amount of such testing can ever prove that there does not exist someone who CAN tell the difference; it's logically impossible to prove a negative.

Preference is very personal.

Oh I will completely concede that my IDEAS of why a discrete version will sound better most likely outweigh any differences I hear - if I do at all! Preference is very personal and bias is nearly impossible to eliminate.

But when it comes to clipping - I do think a discrete can be made to sound better - especially if JFETs are used as CCS loads where when nearing clipping they necessarily pass from saturation mode into triode mode and the consequent Rds increase 'softens' the clipping characteristic. At least that is what I think...

[Paul Marossy]

Oh I will completely concede that my IDEAS of why a discrete version will sound better most likely outweigh any differences I hear - if I do at all! Preference is very personal and bias is nearly impossible to eliminate.

Which is basically the point I was trying to make earlier, that our notion that this chip is better than that one colors what we think we are hearing.

Just for fun once, on a scope I compared the clipping on a BJT opamp called a "JRC4558" and a FET opamp called a "TL072". The JRC4558 waveforms looked more sharp and jagged than the ones from the TL072. But for some inexplicable reason, in the circuit it was being tested in, the JRC4558 sounded slightly better to me even though the scope suggested that it ought to sound harsher. I suspect this might be due to the harmonic distortion happening in one vs. the other but I have no way to verify that. I really wish I had a full fledged spectrum analyzer so I could see specifically what was happening with the harmonics.

Anyway, many times when I analyze something on a scope, what I see contradicts what I imagine it should look like just by hearing it. My only criteria for an opamp in a DIY disortion/OD circuit is what it sounds like, not the specs... I mean they all pretty much have the bandwidth needed, etc. Well, that and how much self-noise it has -- which I actually place more importance on. But once again, the circuit design itself often determines a lot of how the opamp itself colors the sound and/or clips the signal. And therefore, I come back to the same conclusion that any suitable dual opamp in a Tube Screamer for example will sound roughly the same because it's only part of a larger picture.

If I was designing something for NASA, of course I would have a completely different set of criteria.

[brett]

Hi
These op-amp comparison threads are 99.9% delusional.  A real waste of time.  The emperor has no clothes !
There's simply no evidence for real preferences (except for hiss).
And if there's no evidence, I would like to think we don't waste our time on it.  Otherwise we might as well start discussing the validity of Ohm's Law or the effect of the Moon on electron flow in JFETs (yes, there's an effect, but can anyone hear it?).

cheers
PS I wonder what motivates this pursuit of holy grails? (we often act a bit like audiophools).   

[Paul Marossy]

There's simply no evidence for real preferences (except for hiss).

My viewpoint exactly. But you probably can find some evidence for making an argument for one type vs another if you could see what was happening with the harmonics. But whether or not someone can actually truly hear that is what I question.

PS I wonder what motivates this pursuit of holy grails? (we often act a bit like audiophools).  

It's because of the hype about certain things. Like a vintage Tube Screamer with an "original" JRC4558 in it is better than any other Tube Screamers. Or for the same reasons why an original Vox Clyde McCoy has sold for up to $1000 on ebay. People like to jump on bandwagons, you know.  

[Mark Hammer]

There are optimal and sub-optimal op-amps for any given application, and that's about it.

Are the various TI Bi-Fet op-amps "good"?  Yeah, pretty much.  Does the bandwidth-preservation aspect of a really high input impedance matter when I'm making a mixer, compressor, or EQ?  Damn straight it does.  Does it matter when I'm making a guitar-oriented overdrive?  Not at all.  In fact I'll probably have to figure out how I'm going to shave off all that top end so that I get a smoother sound.  In this latter application, what serves as a weakness of a particular chip in a different application can serve as a strength.

Case in point.  Anderton's Octave-Doubling Fuzz uses a pair of dual op-amps, and uses one op-amp stage as an inverting input stage.  My immediate reaction is "Don't use it in inverting mode! That has a lower input impedance!  You'll lose all your bandwidth."  Silly me, though.  Anyone who has ever used an octave doubling fuzz knows full well that one of the first things you do is go to your neck pickup and turn the treble down on the guitar for best octaving.  So, of what earthly use is either:a) an op-amp with a high input impedance and big slew rate, or b) an input stage with a higher input impedance?  Here, cheap and cheezy is a "better" op-amp for the application.

[alparent]

So basically the best op-amp is the one that works best in a certain circuit........so get a selection, ask for counsel to many friends (this forum  )  and try them out! Most of the ones I see mentioned cost next to nothing anyways!

Like I was saying! 

[JDoyle]

PS I wonder what motivates this pursuit of holy grails? (we often act a bit like audiophools).

There is a HUGE difference between even the most delusional guitarist, say Eric Johnson, and the HiFi fool who spends thousands on speaker cables -

In the guitarist's case, they are WITHIN the feedback loop of the creation of sound, whereas the HiFi fool is not.

Even if you can't hear the difference between op amps, and even if everyone else you ask can't either - how can you say that one op amp doesn't effect the RESPONSE of the entire rig and thereby improve the quality of the sound to the guitar player's ears because the FEEL of the way the entire rig responds is better with one op amp than with the other?

We all know that effects are just one small part of a large system - with the biggest part of that system being ourselves, and I would contend that the guitar player is the MOST important aspect of the overall tone, and everything that effects the guitar player has an effect on the tone, with the actual sound out of the amp being but one part of many.

Regards,

Jay Doyle

[Paul Marossy]

There is a HUGE difference between even the most delusional guitarist, say Eric Johnson, and the HiFi fool who spends thousands on speaker cables -

Yep, there is.

Even if you can't hear the difference between op amps, and even if everyone else you ask can't either - how can you say that one op amp doesn't effect the RESPONSE of the entire rig and thereby improve the quality of the sound to the guitar player's ears because the FEEL of the way the entire rig responds is better with one op amp than with the other?

There might be some validity to that, but it's very hard to prove conclusively. People are going to notice the noise of a circuit way before they will notice the response with one opamp vs. another, I mean pretty much any opamp has a high enough slew rate, etc. for your average OD pedal. They might be able to pick up on small differences in harmonics, though. If there IS a difference between a certain opamp vs. another in a certain circuit, I have a hunch it's because of how that device treats the harmonics.

We all know that effects are just one small part of a large system - with the biggest part of that system being ourselves, and I would contend that the guitar player is the MOST important aspect of the overall tone, and everything that effects the guitar player has an effect on the tone, with the actual sound out of the amp being but one part of many.

Yeah, and if the guitar player believes that his vintage $600 Tube Screamer with the "original" JRC4558 chip in it sounds better than the new ones with the exact same components in it, then to him that will affect his tone - in his own mind and attitude. I still maintain that our preconceived notions that this is better than that is still a lot stronger than we think. Especially if what you believe if reinforced by other people, and those who are "authoritive" about the subject.

[amonteiro]

There is something very important about op amp designs that is: the characteristics of the circuit are theoretically independent of the op amp itself. Its behaviour will depend fundamentally on the surrounding components.

So, the input impedance of a circuit depends on the configuration, not on the input impedance of the op amp itself. With an op amp with 100KOhms input impedance you can make a pre-amp with 1MOhm or 1 Ohm input impedance.

To chose an op amp there are some parameters that can be more important to consider than the rest:
- Minimum supply voltage - many op amps start only at +/- 5V (or 10V), so they may not work with 9V; even so, most of these work below the specified supply;
- Consumption - if you plan to work on battery, it may be important; my belief if that nowadays most users have power supplies and also many pedals do not even have place for a battery;
- Noise - the higher the gain, the noisier the output; old op amps like the TL082 have equivalent input noises like 18nV/sqrt(Hz), while the RC4558 has 12, the LM833 and NE5532 have about 5, and the LM4562 has 2.7 (the last one is expensive and has a latch up problem when used with high resistors at the inputs!).
LM833 and NE5532 from some manufacturers work from 5V while from others start at 10V!

Best regards,
Antonio