Another big opamp listening test

[Mooly]

I'm sure 😉

(it will probably be another hour or so before I look in again... looking forward to it)

 

[PMA]

The poll has been closed, so it is time to disclose the opamps under test.

aa (6 votes) - Tesla MA1458, equivalent of LM1458 (dual LM741), MA1458

bb (13 votes) - National Semiconductor LM4562

cc (8 votes) - SGS Thomson TL072

dd (7 votes) - Burr Brown OPA2134

More about test conditions and measurements performed can be found on my web page Another big opamp listening test

------------------------------

Thanks to everyone who listened to the samples, voted in the poll and who posted in this thread. The results are interesting and the most votes came for the opamp with best parameters, bb, LM4562.

 

[Mooly]

Wow, that's really interesting.

I'm reading your linked article where you say:

Otherwise there is no audible difference between them, though in uncontrolled tests performed by opamp swapping there were seemingly significant audible differences.

I think that is something many of us can relate to. When we are the ones in control, doing the swapping etc then we 'think' we hear the differences.

What do you make of the results ? Are you surprised the 1458 turned in a seemingly near identical subjective performance to all the others.

As I mentioned, I felt the hf 'shimmer' was different on bb although I voted for aa. Under perfect listening conditions it was possible to differentiate them although I also tried another test set up using my germanium SE Class A headphone amp and I consistently got a 40 to 95% 'you are guessing' result. Listening direct with headphones into the laptop is where it seemed it was possible to differentiate between them.

So a great test Pavel 🙂

What does this tell us with implications for desired slew rate and so on 😉

 

[C37]

Thanks for setting this up. As with xrk's blind test i had difficulty to pick one clear winner but kept going back to bb, ( no particular audio set here, i'm at work so pc to focal spirit one s headphones which are a tad on the bass heavy side).

I secretly hoped there was a TL072 in the mix as I'm wondering if it would not be the case upgrading them in the active Xo of my tannoys. Maybe i'd go for the LM4562, but wow, i want to hear this back again on a proper set up now.

Mmmh. Confusing

 

[PMA]

Karl, I think that the quite good result for the MA1458 may be for the reason of quite low signal level (168mV = full scale). So there was no push on slew rate. On the other hand, lower signal level means that the noise of the 1458 resulted in the worst measured S/N with this opamp and this opamp also adds some mains line artifacts, though very low in level. The noise of the MA1458 was probably masked by the noise of the music track itself.

I think that the result of the test is self-explanatory. Opamps are not an evil, and in case they are properly used and we are not biased, we can hardly find any sonic character of them, especially in a circuit like unity gain buffer. Still it makes sense to use low noise parts with low distortion and reasonably high slew rate.

 

[scott wurcer]

although I voted for aa.

What does this tell us with implications for desired slew rate and so on 😉

Not surprised JC told me years ago that he and Dave Wilson put the 741 right up in the pack when blind "rolling" in his active crossover. And what, no CMOS op-amp for the bass 😉?

 

[Mooly]

Ah, that is a low level for sure, but that said it is what it is and we can say under those conditions the differences were minimal.

So still a valid result and observation in that it helps set a baseline.

 

[Mooly]

Not surprised JC told me years ago that he and Dave Wilson put the 741 right up in the pack when blind "rolling" in his active crossover. And what, no CMOS op-amp for the bass 😉?

Hearing that makes me feel a little better 🙂

 

[PMA]

Pity John did not participate in the test 😉

 

[dedegogo]

aa = was the one that I kept coming back to.
bb = this seemed the odd one out.
cc = a little lifeless.
dd = constrained.

I voted for aa and yet it was bb that seemed to have the best hf detail and shimmer.

Mooly, you can really trust your ears.

Another member (Craig?) was also correct that opamp is a waste of time, at least with these clips. True, this is against normal measurement. But if you trust your ears, the measurements mean nothing.

I believe that power supply is critical for opamp circuit. But i have never understand why some PS with good measurement do not sound good.

Another key component which determines the sound result is the compendation cap, in this case theinput filter. 470pf is imo too big for some opamp.If 4562 needs this cap, it doesnt mean that TL072 (dd) should also have the cap. It is not fair.

Compensation cap is the determinant factor in discrete amplifiers. Why not in opamps.

OPA2134 is hard to make right. Instead of input filter, a cap across feedback R is my solution.

 

[Markw4]

Interesting. This is why I wondered about the source and the data converters. It's quite possible that if the source track would have sounded dull and dry, that a little distortion could add some harmonics that would make it sound subjectively better.

I have to admit I didn't work very hard trying to listen to the files, and only seriously tried comparing aa and bb, as they seemed the most different. But I only did two sittings in test mode, and seemed to keep picking bb when I thought it was aa. That is, somehow my brain seemed to be choosing the one it preferred, and overriding my conscious effort to use more system 2 reasoning to choose.

Why did I try to use system 2 to choose rather than just trusting system 1? Because I couldn't help it, since I can't observe how my system one works, I just didn't feel comfortable trusting it. This is what I think anyway. There is a bit of guessing and construction on my part.

 

[Markw4]

In addition, I wanted to ask Pavel after the test was over what he thought the test was intended to show, if anything.

So, now that he did say, I have to say I am concerned because the test files were 16-bit apparently because there was a high noise floor, making any lower bits useless. But, if we want to see if people can differentiate single opamps, why would we bury the details in noise? Why not use a very hi-rez, 24-bit, low noise selection of source material, so that if distortion is added it will modify the sound of things like musical instruments, since we kind of know what they usually sound like and can tell if they are wrong. Very hard to do the same thing with 16-bit noise. That Mooly could do as well as he did is quite remarkable, it seems to me.

 

[jan.didden]

I find it significant that bb got most of the votes. Unfortunately, very early in the test there were some statements that it was 'the best' from members and we really don't know whether this bias did play a role. This is unfortunate, otherwise the result could be quite valuable.

Jan

 

[vacuphile]

Maybe i'd go for the LM4562, but wow, i want to hear this back again on a proper set up

Beware: the 4562 is a fast opamp that easily picks up EMR. You may need to take precautions or it will pick up the telephone faster than you do.

 

[jan.didden]

Another key component which determines the sound result is the compendation cap, .

As far as I know, none of these opamps needs (and I assume did not have) a compensation cap.
An input filter rolls of the high frequencies in the input signal and does nothing to the opamp.
I assume that it was the same for all cases.
PMA can you confirm?

Edit - I checked the website, it is as I thought.

Jan

 

[jan.didden]

it will pick up the telephone faster than you do.

😀

 

[C37]

Beware: the 4562 is a fast opamp that easily picks up EMR. You may need to take precautions or it will pick up the telephone faster than you do.

Thanks for warning. More so as it's already picking the phone as is (Plate amp) wide open at the back. Time to consider a full case... Sorry for O/T

 

[PMA]

I find it significant that bb got most of the votes. Unfortunately, very early in the test there were some statements that it was 'the best' from members and we really don't know whether this bias did play a role. This is unfortunate, otherwise the result could be quite valuable.

Jan

I completely agree with your post, Jan. On the other hand, it is an open web test and we have to count with some disturbances. IMO, the test result and even the discussion has been better than I have expected.

 

[PMA]

Thanks for warning. More so as it's already picking the phone as is (Plate amp) wide open at the back. Time to consider a full case... Sorry for O/T

The 4562 PCB must be definitely put into shielded metal box. Please see my note and warning at the end of the page
Another big opamp listening test

I also have measurements without and with top Al cover (of the box), you would not believe the difference. On the other hand, when shielded well, this opamp is sonically excellent, which just reflects its parameters 😉.

 

[dedegogo]

Why did I try to use system 2 to choose rather than just trusting system 1? Because I couldn't help it,

Mark, why do you seem to feel guilty? Which opamp did you prefer? Imo, aa bb or cc, each of them can be preferred. But im sorry, preferring dd doesnt make sense. It doesnt mean that OPA2134 is bad but this is not the right circuit for it. It can be very good when the cct is right for it.