DIYstompboxes.com

Notice â€" Highly Subjective Op Amp Evaluation by lo-tech Guitarist

This saga started out with a modified version of Gus’ Muff Fuzz #1 or Gary’s Mockman.  Both add a 470K/1meg resistor across the diodes, as a mod to stop gating.  Since the Muff Fuzz runs one op amp into the other which has diodes in the feedback loop, it seemed like a good test bed.  If you are going to use dual op amps, you might as well use both of them!  I was going to leave it alone, but I couldn’t.  I was getting too much noise, but really liked the sound.  This is the first distortion with diodes in the loop I have tried, and I like the smooth rich distortion.  The diodes in the circuit such as GE's, SI's and LED's make a much more noticeable difference than the op amps.  I think I can hear a difference when using a Jfets or Mosfets as clipping diodes.

I initially tweaked the diodes and found that using LED’s raised the output, but the noise did not increase, plus the LED’s had a nice chunk to them.  The down side is that the distortion level is lower, so I used an MPF102 Jfet for the other diode which has a slightly higher clipping threshold than a single SI (I tried a Jfet/Mosfet combo, but it was too fuzzy for my taste, the Jfet/SI combo is nice, I’ll probably need some switches). This got more distortion, plus more even ordered harmonics due to the asymmetrical distortion produced by dissimilar clipping thresholds.  The LED is 1.8v and the Jfet .7v.  I hypothesize that having diodes with higher clipping thresholds provides the op amp with more opportunity to clip, whereas using lower threshold diodes results in the diodes clipping more and the op amp less.  I may just be dreaming.

As I sorted thru the op amps I had, thanks to one of the “elders” of the forum, I was also watching to see which op amps others were using.  I started requesting samples from various manufactures and found that Texas Instruments had most of what I wanted and was vary gracious and quick in providing their samples for the test.  I also searched thru their list of op amps and found others to try.  (Thanks TI.com)  

I needed some consistent way of measuring the 15 or so op amps I had.  Each time I thought I heard something different; I would later be unable to duplicate the findings.  Most of my observations have question marks by them. It's difficult to hear differences between 2 seemingly different designs of op amps like a 4558 and TLC2272 cmos.  It's more difficult to hear a difference between 2 op amps of the same basic design.

I came up with 2 tests besides the basic and highly subjective listening test.  I have found that one evening something sounds great and the next day it doesn’t.  

First was the plug it in and see how much “Noise” it made test.  I had members of the family assist and was surprised to find that some noise, even though the same perceived volume was judged as harsher than others.  Smooth and flowing vs. harsh and irritating seemed to be the difference.

The second test was the plug it in and see how quickly and how much “Feedback” it got while the guitar rested on the bed opposite the amp test.  I think this relates to gain.  The 4558 was adjusted to slowly develop feedback and with no changes to the set up different op amps were substituted.  Another unexpected finding was that the CMOS op amps would feedback on a lower harmonic than the others.  Does this mean more gain at certain frequencies?  Were the harmonics odd vs. even???

In all tests the TI RC4558 served as the constant with which the others were grouped.  I had to group them in clusters to make testing reasonable.  I use a standard group that included the 4558, a Jfet input group which includes the TL072, a CMOS group which includes the rail to rail op amps and the high performance group that includes the Burr Browns.  Each time the 4558 was used to compare the others in the group.  The results of the tests are below.

If you don’t stand in exactly the same spot when you test, that has more impact on the sound than different op amps do.

I did find that most of the op amps with high gain were also noisy.  This would seem to be typical, the OP275 being tops in both categories.  I think the Muff Fuzz circuit with 1 meg resistors in the loops seems subject to noisiness.  Lying on top of my SS amp didn’t help either.  I’m sure it will be quieter once I get it into a junction box.  Still obvious differences existed between op amps.

The TI RC4558 is not a bad choice and I can see why it is commonly used.  When maxed it breaks up nicely and it isn’t nosier than some others with the same “gain”.  The noise was judged as being harsher than some others though.  At times it seems to have sort of a swirl to the distortion that is a little different, but at other times it doesn’t.  2 commonly recommended op amps, the LF353 and NE5532 do sound sweet, but have less gain than the 4558 and less bass.  That can be remedied.

Distortion:  For me that is what it is all about.  The 4558 has plenty and maybe the 4559 has more.  The higher gain op amps don’t necessarily break up more or have higher levels of distortion and sustain.  Some seem to stay clearer, perhaps due to there design for hi-fi circuits/slew rate.  Some sound more “robust” or have more balls than others.  Some may emphasis different harmonics, but again the diodes have much more noticeable effects on the nature/character of the distortion.  You don’t have to have your family/friends come in and help you hear the difference.  If you don’t stand in exactly the same spot when you test, that has more impact on the sound than different op amps do.

I have concluded that most of the difference between the op amps was due to more or less gain, distortion, highs and/or lows.  These are all things that can be addressed with different caps or different gain settings.  This seems to account for most of the differences I could hear repeatedly.  The ones I thought had sweeter smoother highs, turned out to have less gain and/or bass, which smoothed out the highs and could be accomplished by using smaller input caps.  I found that I could use a larger input cap with some op amps that would sound too woolly with others or some diode combinations worked better for the same reason.  The cmos op amp’s that go rail to rail seem to be a bit woollier than some others which works better with smaller caps or SI diodes.  The Jfet/Mosfet’s as diodes worked better with less wooly op amp cap combo’s.  This is based on my particular taste and could differ from person to person.

Output:  I haven’t turned the output pot very far at all, so I’m assuming that they all have plenty of drive for a tube front end, although this version doesn’t have a gain control and doesn’t get really clean.

It's fun searching, even if it is subjective and tone is pretty much that way.   My double picking and jazz scales are much better than before I started.  

OP AMP’s

Bi-Polar Op Amps
RC4558   Old standby good gain and distortion, some noise, Mojo???
RC4559   Turbo version of 4558, cool   
NE5532   Very Quiet, sweet highs, less gain???
LM833   Nice w/big bottom??

Jfet Inputs
TL052      Originally sounded superior, but now like the rest
TL072      Another std.  sounds good less mid’s than  4558
LF353      frequently recommended, sounds good, quiet/less gain
LF412      Little stronger than LM353

High-Performance
OP275   Lots of gain and noise.  Probably fine enclosed.  The hi-fi
 -  -  -      Op amps seem to be slightly smoother/creamer sounding???
OPA3140   Quieter than 2604 probably less gain, nice
OPA2604   Almost as much gain as 275 but quieter may not break up as much

Cmos â€" Rail to Rail
CA3260   Mosfet in Cmos out spec.s sound cool
TLC2262   Cmos may sound fatter w/sweeter highs??? I will probably use
TLC2272   one just in case.  Usually lower noise.
TLV2372   Sounds nice

:roll:

Gee, your fingers must be tired  :lol:

Thanks for the opamp taste test, it's interesting info. That's interesting about the OP275, I used it to good effect in some of my guitar effects. It's touted as being the first opamp to use the Butler front end, etc., but it's a little overrated. The NE5532/5534 is a pretty good all-around performer, too. I have even had good results using a lowly 741 and 1458. But that may be due to how I set my controls - I usually have the gain at max. or close to and the volume not much louder than the bypassed signal. I'm sure the noise would be very apparent with the volume at high levels, though.

That's some nice work, WG.

Quite a test you've done there! I'm always happy to hear I'm not the only one who likes the NE5532. I dont think my ears are good enough to hear the difference between different brands of the same chip. Like you said, it stands out more in certain pedals than others.  Kind of a reminder that we should all pitch in on testing stuff instead of waiting for someone else to expend all their time, effort, and money for our benefit. On the other hand, it's fun! Thanks!

Great taste test WG! 8) :D

Nice work, from another 5532 head out here in effects land. ;)

Take care,
-Peter

Great testing.  Informative.
(Hey, count me "in" as another 5332 user.  They're cheap and easy to get, as well as giving high sound quality)

Thanks for the reviews!

This is good information. I'm sure well have our favorites with regards to op amps.

:shock:  I'll need to read all of that when my attention span is a little longer!

Seriously,  shows a true regard for the community.

Good detective work, WGTP.  Useful info in the style of the old forum.  Much appreciated.  Now I can stop lusting after those exotic op amps.

Interesting
 I tend to like the lf353 and ne5532.  To my ears they seem to have more gain.  I think sometime grit is heard as gain.

 My ear,brain hear the ne5532 having more gain.  

 
Gus

I'm sure my testing method are subject to some valid criticism, but that is the best way I could think of with limited resources and knowledge to test the little beasts.  I too was suprised that some of the more commonly recommended devices feedback at a lower level than the 4558 and even more suprised that the cmos feed back on lower harmonics than the others.  I hadn't even considered that as a test criteria.  They may have more hi end roll off???  Compared to the 2604, the cmos sound less brilliant.  I know the 5534 made my rat sound better compared to a TL070 and the 353 sounds good as well.

More feedback could be a problem at high volume levels and may not be desireable.

I tried the "feedback" test again tonight with interesting results.  First pass I used an SI/mosfet diodes and the results from low to high:  353, 5532, 4558, 2272, 2372, 2134, 2604.

Then I tried it with 2 LED's for diodes and the results were different.  The difference in feedback levels were much closer and the 2 CMOS op amps, the 2272 and 2372 both feedback on lower harmonics, AS DID THE 353??? The 2372 sort of tried both harmonics before settling on the lower one.  The order from low to high:  4558, 2372, 353, 2272, 5532, 2134, 2604.

I find that most amazing that the 2272 and 2372 both feedback on lower harmonics. I wonder if they're just better at lower frequencies and those just dominate?

Great info! :D 8)

Thanks again!

-Peter

I had seen some stuff about the OPA2604 a while back and when I made a mouser order, I tried ordering one. It cross-ref'ed to a TSH22, they were about a buck and some change each. I don't know if they are as good as the Burr-Brown chips, but for some reason today I had time to kill so I started swapping out chips in my pedal, and they each sounded very similar, and I was separating the 'good' ones from the mediocre ones so I could narrow it down. The TSH22 was the last one I tried and once I plugged it in, I knew I didn't want to listen to the other ones again.
For some reason, this one 'felt' good. It responded to my playing in a way the others didn't, including JRC4558, JRC4559, RC4558 and 4559 (TI), TL072.
This may not be the case in all pedals because my circuit has an unbuffered front end. Something like a tube screamer with a buffer at the input may interfere with this 'response' effect, or it may not, I don't know.
Anyway, you're right about the 'creamy' part, it has a very smooth high end to it. It has a nice character to the tone I can't quite describe.
I guess now I have to try out the BB OPA2604 and see if it gets even better, but I'm pretty happy for now.

If I remember right, the OPA2604 type opamp has a very wide bandwidth of ~10mHz. In a solid state circuit, this is necessary for more high order harmonics, which sounds smoother. And it should also sound a little more like a tube circuit.

Quote from: Paul MarossyIf I remember right, the OPA2604 type opamp has a very wide bandwidth of ~10mHz. In a solid state circuit, this is necessary for more high order harmonics, which sounds smoother. And it should also sound a little more like a tube circuit.

Well, 10MHz is a wide bandwidth and it means that it will certainly have full gain throughout the aural spectrum, but we only need to go out to 20kiloHz as that is the limit of human hearing.

I also don't know how it would make it sound more like a tube circuit, could you explain Paul.

Thanks,

Jay

Well, maybe more properly said, "it could" make a circuit sound more tube-ish.

Read this article: http://www.diyguitarist.com/PDF_Files/TubeEmulation.pdf

Most opamps have a gain-bandwidth of around 3mHz. A 12AX7 has a gain-bandwidth on the order of about 8mHz. The OPA2604 is around 10mHz.

I interpret that, in terms of harmonics, as being a lot closer to a preamp tube than the typical opamp. Therefore, what is interpreted as "sounding smoother" is probably because one can hear the high order harmonics that we wouldn't normally hear from a generic opamp. What I'm not sure of is how strong the second order harmonics would be compared to the third order. That's what would make it sound very tube-ish - a very strong 2nd order harmonic with the 3rd, along with a little compression, which may or may not be present in a distortion circuit. It seems like they all compress the signal at least a little bit.

Tubes also produce far more of the high order harmonics than SS devices do.

Just my opinion, of course.  8)

Quote from: Paul MarossyWell, maybe more properly said, "it could" make a circuit sound more tube-ish.

Read this article: http://www.diyguitarist.com/PDF_Files/TubeEmulation.pdf

Ok, I read it and although seemingly well researched, I find it a bit questionable. Most specifically this sentence:

After adjusting the tube bias, the input and output gains and the emulator grid-to-plate capacitance, the waveforms for a variety of input levels and frequencies show a good match at various overdrives, frequencies and loads

I read this to mean: "After I changed everything around, I was able to come up with a particular configuration with the tube that matched the circuit I was using, thus proving my point."

Quote from: Paul MarossyMost opamps have a gain-bandwidth of around 3mHz. A 12AX7 has a gain-bandwidth on the order of about 8mHz. The OPA2604 is around 10mHz.

And a 2N3904 has a GBP of 300MHz  :D  and there are reasons for the difference between discrete single stage transistor/tube circuits and op amps. The phase shift in an opamp is a complex thing, not so in a transistor/tube stage. An opamp has cascaded gain stages each producing their own rolloff due to the Miller capacitances in each stage. A discrete amp only has one to worry about.

Quote from: Paul MarossyI interpret that, in terms of harmonics, as being a lot closer to a preamp tube than the typical opamp. Therefore, what is interpreted as "sounding smoother" is probably because one can hear the high order harmonics that we wouldn't normally hear from a generic opamp.

A higher GBP doesn't mean that you can hear the harmonics better only that those harmonics can be amplifed to be louder. For example, assuming a total open loop (no feedback) gain of 100db, a 1MHz GBP can amplify a 20kHz signal (the limit of human hearing, give or take, so the highest harmonic we have to worry about) by 20db, where as a 10MHz GBP can amplify it by 40db; it is still there in the 1MHz GBP, just not as loud. But that isn't all that important as we are worried about the production of harmonics by the distortion of the stage, in which case both the 1MHz and the 10MHz can produce them with ease.

In any case, higher order harmonics are what contributes to the "harsher" sound of a transistor amplifier clipping. The difference lies in the tubes natural response to squash the signal instead of hard clipping like a transistor. To quote Boscorelli:

In a squarewave, harmonic energy exceeds that of the fundamental, with much of the energy clustered above the 5th harmonic. In a squashed wave, harmonics roll of rapidly, less than 2% of total energy falling above the 5th harmonic.

Quote from: Paul MarossyWhat I'm not sure of is how strong the second order harmonics would be compared to the third order. That's what would make it sound very tube-ish - a very strong 2nd order harmonic with the 3rd, along with a little compression, which may or may not be present in a distortion circuit. It seems like they all compress the signal at least a little bit.

Right, it is all of these factors, along with the shifting bias point and squashing before hard clipping, that make the tube sound. IMO, a tube is a tube is a tube  :D . You can't emulate it with SS, so why try? I just try to make SS sound good.

Quote from: Paul MarossyTubes also produce far more of the high order harmonics than SS devices do.

Actually I think it is the opposite, thus the "harsh" sound of transistor clipping.

Quote from: Paul MarossyJust my opinion, of course.  8)

Me too.  :D

Hmmm... points well taken.

But, from everything I have read thus far, and the curves I have seen comparing tubes to transistors, tubes produce a lot more high order harmonics, odd and even order - but more even than odd. Transistors produce a strong third order harmonic, and everything is else is small in comparison. Odd order harmonics sound harsh to the human ear, and even more pleasant, right? If it were not for the strong 2nd order harmonic, tubes would probably sound a lot more harsh. If you really, really clip a signal with a transistor you basically end up with a square wave with slightly rounded corners. Would a tube react the same way?

Perhaps I am mistaken comparing transistor devices to opamps? Although opamps contain a bazillion transistors and diodes....

Anyhow, that's just they way I see it.  8)

Just my 2 (euro)cents...

To those who like NE5532 opamps and therefore suggest to use them: be aware! They almost double the current draw in my Son Of Screamer! (6.3ma when idle, instead of 3.4ma using a TI4558)

Giulio

You guys know way more about this stuff than I do.  I figure if nothing else, the list of models that work would be helpful.  The differences between op amps are pretty subtle, yet at times they seem to jump out at you, it's weird.

Most op amps have caps in the loop that roll off the highs and/or low input impedance that lowers the highs from the guitar.  So, does this self restrict the bandwidth.  (I thouhg bandwidth had to do with how much of the stage a band took up.)  

I wonder how the different parameters effect the distortion.  bandwidth, slew rate (I think that R.G. speculated that might have to do with recover from clipping), rail to rail capabilities (for both input and output)(which seems to be related to cmos op amps), etc.

The cmos amps has repeatedly feedback at the lower harmonics.  I was surprised that the 353 started it too.

The other Burr-Brown I tried the opa2340 seems a little quieter with less gain, but it may break up more than the opa2604.

Do you all think the "feedback" test relates to op amp gain?

Of course I haven't tried different brands of the these same chips.

I also haven't tried my tests at high volumes.

Final installment of the op amp evaluation

I didn’t really find new magic in this bunch either.  I really like this circuit (Gus’ modded Muff Fuzz #1/Gary’s Mockman) and all the op amps sound good in it.  It seems to maintain sustain even when you turn it down and the distortion decreases.

New concept - “Distortion Reluctance” - the 4558 doesn’t have much, but others seem more reluctant to break up.  This seems to be the thing that may separate the 4558 from the others.  Since distortion is the point, low DR would seem to be a good thing, as long as it’s the kind you want.  I know some don’t like op amp distortion characteristics, but this sounds as good, or better than of the others I’ve built.  

Late Entries

TLV2262   Cmos crunches up nice, quiet lower gain mid-rangey
TLE2072   Excelsior upgrade to TL0xx series, highest slew rate evaluated, high noise/gain, bright, smooth highs, cool
OPA2350   Burr-Brown CMOS, high slew rate, average gain, OK   
OPA2228   Burr-Brown  similar to 2604, not as much gain

I put an Burr-Brown OPA2134 in my SD-1/808(analogman mod) and boy did it wake up and become more full-range sounding, almost, but not, metally. Sweet.

Regards

RDV

I have finally settled on the OPA2134 for the Mutant Muff too.  It seems to breakup more than the OPA2604.  The OPA2350 is also a BB and CMOS, but not as much gain, it woulds work well for situations when MAX drive isn't needed.

I'd like to try some different OA's in a new DS-1. Has anyone found any suitable subs for it's inline chip? I keep hearing there's a 4558 and 5532 version of it.

Is there a schematic available for this ckt?

I bumped the other thread, but i should have bumped this thread instead - so i did !
;D

FLASHBACK - I was reading the document I wrote in '04 and still have pretty much the same perspective.  Alot of the differences between op amps can be minimized, adjusted, corrected, addressed, etc. by chaning the resistors in the feedback loop or to ground for gain adjustment.  The cap in the feedback loop or elsewhere can be moded to address some high frequency issues.  The input cap can be adjusted for bass or the cap/resistor to ground off the feedback loop.  The CLIPPERS used have a MUCH more detectable effect than op amps.  The difference between 2 SI's and 2 LED's doesn't require back and forth A/Bing to hear.  Ge's, SI's, LED's, Jfets, Mosfets (body and mosfet) all are pretty easy to hear.  As Mark, et al have said it is probably the clipping threshold that mostly accounts for the noticable difference, but transfer function seems to play a role as well. 

Certainly try several different op amps to get a feel for it.  I think some probably work well in some distortons and others better in others.
A diode in the feedback loop clipper with an inverting input op amp may not reveal the differences as well as diodes to ground.  A non-diode clipper like a Black Cat might be the best way to hear differences, as the distortion is all coming from the op amp.

Here is the circuit I used with some clipping variations.  http://www.aronnelson.com/gallery/v/WGTP/AlteredMuff.jpg.html?g2_imageViewsIndex=1

IIRC Jack at AMZ indicated that putting a small resistance in series with the diodes minimizes the difference between op amps.  I "think" it reduces the "shock" to the op amp from the clipping.   

The more I mess with this stuff, the more I think it is about pre/post eq, as R.G. and others have been telling us for a while now.  :icon_cool:

If you plot the current per voltage for diodes, you find that they all have some amount of roundness where they start to really conduct. They don't suddenly change from fully off to fully on. The roundness is different for different diodes, as you'd expect by the differences in their device physics.

But what really matters is the relative size of the signal you put across the diode versus the size of that rounded area.

Silicon has a "knee" that's about from 0.45 to 0.65V for "average" junctions, of which there are none, but it makes a good illustration. That 0.2V difference in diode voltage takes it from almost off to almost fully on. If you use signals below about 25mV, you can use the silicon diode knee as a variable resistor and have distortion under a percent or so. That is, a diode doesn't distort tiny signals to any noticeable degree AT ALL if the signal is small compared to the size of the knee, no matter where in the diode characteristic it's biased.

Thomas Organ Vox amps used four silicon diodes as a variable pass resistor in their tremolo - it's one of the better sounding tremolos in a solid state amp.

As you make the signal bigger compared to the knee, the distortion becomes first noticeable, then big. Eventually the signal gets so large that the knee only encompases part of the signal. This is the typical back to back diode pair. The signal is biased at 0V on the diode characteristic, and there is no distortion because the diode knee isn't reached until the signal hits 0.45V or so, and then that part of the signal pushes the diode into its knee, and the part of the signal bigger than that gets clipped.

When the signal get big enough to try to push out the top of the knee, the diode can no longer change to smaller incremental resistance, so the signal is flat-topped there. Signals that just push into the knee, or barely through it are softly distorted.

Now what happens when you put a signal that is 10V p-p unclipped into a diode clipping pair? It zips through the diode knee so fast into hard clipping that there is no noticeable knee region at all - you get a nice square corner in the clipped signal. And how is that different from a signal that "wants" to be 100V p-p? It isn't. Both are indistinguishably clipped to razor sharp corners.

Somewhere between the signal being about the same as the diode knee voltage and maybe 3-4 times the knee voltage you go from soft clipping to hard clipping and get into diminishing returns. It's the relative size of what the signal is (or would otherwise be) without clipping versus the size of the diode knee that matters.

A series resistor changes things. A diode goes from a big resistance, maybe 100's of k to M's before it turns on to a few ohms when it's fully on. If you put a resistor in series with the diode, the diode still does the same thing, but the sum of the diode and the added resistor cannot get smaller than the added resistor. You've limited how hard the composite diode/resistor can turn on and limit the signal. So a resistor in series with a diode will force the resulting signal to never fully flat line on top, no matter how hard you drive it. At big signals, the diode becomes a switch between high resistance for low voltages to the added resistance for voltages more than the diode knee.  You can sit and play with adding different resistor values to each diode, using a variable one, adding pots, caps, etc ad nauseum. Each tweaks with the resistance one or both diodes get to when you trip them over their knees. Likewise, subbing in different materials changes the base voltage before the knee starts and the extent of the knee.

But it's the relative size of the signal (or what it would be if it didn't get limited first), which translates into the vertical speed in volts/uS of the signal traversing the voltage through the diode knee that determines how sharp the diode knee looks to the signal. That's why you can make diode/resistor shapers for bigger synth-style signals. The diode/resistor nets are concocting a big, loose, "knee" that changes resistance over a big range. That's why the amplitude of a triangle put into a sine-shaper network is critical - it relies on the soft clipping of the knee region to round it off.

And then there's pre/post EQ...   :icon_biggrin: